STATE OF MINNESOTA AND BLUE CROSS AND BLUE SHIELD OF MINNESOTA,

PLAINTIFFS,
 

V.
 

PHILIP MORRIS, INC., ET. AL.,

DEFENDANTS.
 
 

TOPIC: TRIAL TRANSCRIPT

TRANSCRIPT OF PROCEEDINGS

DOCKET-NUMBER: C1-94-8565

VENUE: Minnesota District Court, Second Judicial District, Ramsey County.

YEAR: February 26, 1998

A.M. Session
 

JUDGE: Hon. Judge Kenneth J. Fitzpatrick, Chief Judge
 

TEXT:
 

THE CLERK: All rise. Ramsey County District Court is again in session, the Honorable Kenneth J. Fitzpatrick now presiding.

(Jury enters the courtroom.)
 

THE CLERK: Please be seated.

THE COURT: Good morning.

(Collective "Good morning.")

(Witness sworn.)

THE CLERK: Please state your name and spell your last name for the record.

THE WITNESS: Timothy Wyant, W-y-a-n-t.

THE CLERK: Thank you. Please have a seat.

THE COURT: Counsel.

MR. HAMLIN: Good morning.

(Collective "Good morning.")

MR. HAMLIN: For the record, Your Honor, plaintiffs call Dr. Timothy Wyant.

TIMOTHY S. WYANT called as a witness, being first duly sworn, was examined and testified as follows:

BY MR. HAMLIN:

Q. Good morning, Dr. Wyant.

A. Good morning.

Q. Dr. Wyant, can you tell me what your current position is.

A. I'm an independent statistical consultant and the principal of essentially myself as a consulting firm called Decipher.

Q. And where is that firm located?

A. That's in Tacoma Park, Maryland.
 

Q. Can you give me a description of your business.

A. I provide statistical consulting in areas that require building of statistical models, analyzing and processing and compilation of -- of statistical databases for statistical models for a variety of clients, federal government agencies, corporations, non-profit trusts, and sometimes litigations.

Q. Can you give me an idea of some of the clients for whom you have worked.

A. Sure. I've worked for federal agencies, for the Health Care Finance Administration, Inspector General's Office, for the Occupational Safety and Health Administration; non-profit trusts I've worked for include the UNR Asbestos Trust, the Dalkon Shield Claimants Trust; litigations I've worked on for the variety of law firms representing both plaintiffs and insurance companies; and I've worked for a number of auto insurers, Nationwide, USAA, GEICO, and companies of that sort.

Q. Can you tell us about your educational background. Where did you go to college?

A. I got a bachelor's degree in mathematics from Oberlin College in 1970, and a doctorate in biostatistics from Johns Hopkins in 1979.

Q. Can you tell us about the curriculum for the doctorate that you received in biostatistics at Johns Hopkins.

*2 A. That curriculum involves many courses in statistical modeling and


 

database building and sample surveys, design and analysis of such surveys, and the applications of models and surveys to public health problems, problems of estimating mortality rates, clinical trials, and problems of that sort.

Q. How long is the course of study to get the Ph.D.?

A. Well for me it was about six years.

Q. And did you write a thesis?

A. Yes, I did.

Q. Can you tell us about the thesis.

A. Thesis had to do with statistical method or tool called confidence interval, and it had to do in particular with applications of confidence intervals in situations like length of stay at heart attack wards in hospitals.

Q. Now I want to turn now to your employment. Were you employed at the U.S. Geological Survey from 1976 through 1979?

A. Yes, I was.

Q. Can you first tell us what the U.S. Geological Survey is.

A. Well the U.S. Geological Survey is a federal agency. It does a lot of mapping for the country, a lot of the cartography. You've probably seen their topo maps. But they're also responsible for a lot of research and study of the geologic and -- and the water resources of the country in terms of determining, for example, oil reserves, determining likelihoods of floods and earthquakes, and looking at water quality, monitoring stations that send data up to satellites and bring them back into Reston, Virginia.

Q. Can you tell us what you did at the U.S. Geological Survey?

A. I was a mathematical statistician at the Geological Survey, and my job there was to collaborate with geologists and hydrologists and all the other research scientists at the Survey who needed assistance with statistical modeling in their work.

One of my main projects for a couple of years was building statistical models that would analyze the likely risk of oil spills from federal beach sales offshore. The federal government owned a lot of land under the ocean just off the -- off the beaches and were in the process in the seventies of selling those drilling rights to some of the oil companies, but before they could do that they had to file an environmental impact statement. So we would collect weather data from Texas towers to get hourly winds for thousands and thousands of hours, current data from National Oceanic and Atmospheric Administration and wildlife data for various sensitive sites, and data from the industry on tanker spill rates, accident rates, and the rates at which offshore platforms failed, and combined all this information together to build a statistical simulation model that would show what the likelihood was of oil spills at different places in the lease area and whether or not those spills were likely to come ashore and impact beaches, wildlife refuges, or what have you.

Q. You used the term "data" several times there, Dr. Wyant. Can you tell me what you mean by that?

A. Sure. We mean computerized databases, and in this case of millions of -- of weather records. Every half hour off Nantucket, for example, for several years, what direction is the wind coming from and how fast is it, and every half hour you collect that, so in the simulation models you can reproduce what the wind looks like over as many year period as you want to by running those computer tapes containing all that information through your statistical model.

*3 And all these other data sources were similar in that. For example, you have a long record of spill -- spills reported to the U.S. Coast Guard off of ships or offshore drilling platforms, and those again would be coded and entered into the computer and then compiled so that we could calculate the probability of spills occurring given certain scenarios of tanker routes and so on.

Q. Dr. Wyant, were you pursuing your Ph.D. at Johns Hopkins while you were working at the U.S. Geological Survey?

A. Yes, I was.

Q. And you received your Ph.D. in 1979; correct?

A. That's correct.

Q. And did you then leave the U.S. Geological Survey?

A. Yes, I did.

Q. And what did you do?

A. I was an independent statistical consultant for approximately one to two years. I worked on cases before the Federal Energy Regulatory Commission, natural gas supplies west of the Mississippi. These were cases involving the El Paso Natural Gas Corporation.

I also worked on some validation screening programs for the Department of Energy that had to track flows of oil into and out of the country and had enormous computerized databases of -- of every offloading of ships coming into the country of oil or ships leaving the country, and all of this had to be tracked and reconciled because there were different tax rates on the -- on these different sources of oil.

Q. How long were you an independent consultant?

A. Approximately one to two years.

Q. Then what did you do?

A. Then I went to work for one of my clients, Econometric Research, Incorporated, in Washington, and that's a firm that does statistical modeling and also builds computerized databases to support such modeling. And at that firm I -- at one time or another I was senior statistician, vice-president, and director of computer operations.

The computer operation had the responsibility for the whole firm of building computerized databases to support statistical modeling efforts.

Q. Can you give us some idea of the work that you did while at Econometric Research.

A. I did a great deal of work on employment discrimination cases which involved taking computerized personnel data systems and looking at records that related to disproportionate hiring, firing, or what have you, of women or minorities. Also did consulting and statistical modeling for corporations. For one example, ARCO, working in collaboration with weather scientists and naval architects, we built a model, a statistical model to look at the rate at which operations would be affected by ice packs. They were planning to build ice- breaking tankers to go in and collect some of the oil in the North Slope of Alaska that couldn't be piped down through the pipeline, and one of their big concerns was the extent to which they were going to be shut down during the winter and the extent to which, if they built any fixed platforms offshore, that those would be affected by ice ridges building up and impacting on them in the winter.

*4 So again we collected data from aircraft over- flights of frequency of ice ridges in the area, and in collaboration with these other scientists developed some basically design specifications for fixed platforms or for tanker traffic in that area of -- north of Alaska.

Also at Econometric I worked on litigations involving some product liabilities for products that were found to cause injury or disease in -- in people that used them.
 

Q. How long were you at Econometric?

A. About six to seven years.

Q. Then what did you do?

A. I left Econometric Research and became an independent statistical consultant, and that's where I am today.

Q. Now Dr. Wyant, in your experience as a biostatistician, have you worked with large databases?

A. Oh, yes.

Q. And have you worked with statistical modeling?

A. Yes.

Q. And have you worked with other experts in a team approach?

A. Constantly.

Q. Let's talk about a few of the significant assignments in your professional career.

You mentioned the Dalkon Shield litigation. First of all, can you tell us what that is.

A. Well the Dalkon Shield was a -- a product that was found to cause disease and injury among women. It was manufactured by the A. H. Robbins Company in Richmond, Virginia, in the 1970s, an intrauterine contraceptive device. And after it was found to cause disease and injury, there were hundreds of thousands of claims filed against the company and the company declared bankruptcy.

To resolve the bankruptcy, the court had to establish some value for all of these claims. This was difficult because you couldn't really get the value of each claim without doing a lot of investigation. It was entirely impractical to go through all the 300,000, and it was also difficult because these claims involved -- some of them were worth nothing, some of them were very minimal claims of a thousand dollars or so, but some of them went up to three or four million dollars, those would be for cerebral palsy kids, birth-defect cases, and the claims involved lifetime care in intensive care facilities, and so without knowing how many of those three-, four-million-dollar claims there were and how many of the zeroes among the 300,000, it was difficult to set any total amount on the -- the value of these 300,000 claims.

Court established a process where experts like myself from - - from the various parties -- I worked for the plaintiffs' committee, and our first job was to create a statistical database so that we could estimate the total value of these 300,000 plus claims, and we built that database by taking a sample of those claims, collecting some information from the sample, and then using some other information that we had from another statistical database to estimate the value of the claims in the sample, and from that we could project a value for the 300,000 claims.

My role in that process was designing the sample surveys and the structure of the surveys and how many people to look at, and also to work in collaboration with doctors who were experts in these kinds of diseases to work out how to build the database by coding medical records from the sample using ICD-9 codes into a computer database that ultimately had between two and three million records in it that described the 6,000 claims we pulled as a sample, and based on statistical models that were presented to the court of that database, the court established two and a half billion dollars as the total value of these 300,000 claims.

*5 Q. How long did you work for the Dalkon Shield litigation or trust? A. Well what I just told you about was -- was the litigation which had to be resolved for the company to come out of bankruptcy.

After that there was a non-profit trust set up, and that was the Dalkon Shield Claimants Trust, that was separate from the litigation, and it was a trust set up just to disburse this money, to take these 300,000 claims, and they had this two and a half billion dollars, and when it was gone, the trust was over. But the trust had kind of the same problem that the court had in that they had 300,000 claims to deal with -- or at this point it was 200,000 because some had been disallowed by the court -- and they still didn't know exactly what the total value was. They now had a database and some estimates, but their problem was that they had to make an offer to each person. But the person could turn down the offer and go to trial, and then the trust would have to pay for all the costs of the trial and then pay for any judgment if the person won.

So the trust first of all had to pay values that were fair and make sure, for example, that a legitimate cerebral palsy birth- defect case got enough money so the kid was taken care of for their life, but they couldn't -- if they paid too much to everyone, then they might run out of money at the end of the line, and that would certainly not be fair, and if they paid too little, then people would turn down the offers, go to court, and then the trust would end up spending money on trials and lawyers instead of paying to the claimants. So it had a real tough decision in terms of setting up exactly what the payment levels were for each claimant coming in the door of the trust.

Now I was retained to use the database as we had built during the litigation and to build some additional statistical models and refine them as -- as the trust went on and started paying clients so -- so it could get more information so that it could set payments at a level that were, first of all, fair, and second of all would encourage people to be reasonable enough that people would take their offers, but not so high that they would run out of money. And it was my job to use these statistical models in collaboration with a doctors panel and other experts that we brought in to build an evaluation system that would pay these individual claims and then keep forecasting, you know, have we got enough money left or not? Do we have to make a change so that, you know, we can be sure that if the last person in line on these 200,000 is a cerebral palsy claim, that we're not going to be there with -- with an empty wallet? And that -- that process -- actually I'm still retained by the Dalkon Trust. It took about eight years to pay these claims.

They're down now from over 200,000 to about 40 active claims that are in some process of litigation as we sit here.

Q. You mentioned ICD-9 codes. Can you tell us what you mean by that? We've heard about that from other witnesses, but I'd like to hear your understanding.

*6 A. Sure. ICD-9 codes are a classification system for classifying diseases. It's a coding system, and it makes it -- use of this coding system makes it much easier to study diseases using computerized databases and statistical models because it's a way of specifically categorizing diseases in a way that's easy for computers to deal with. And those are the typical way, for example, almost all medical billing is done in this country with a medical insurer, auto insurer covering injuries from auto accidents, whoever.

When they see a bill from a doctor, it's usually coded with ICD-9 codes to say what the injury is that this bill was for, and that's pretty much the way all the billing systems work.

Q. In your work in the Dalkon Shield litigation and with the Dalkon Shield Trust, did you have occasion to use and review ICD-9 codes?

A. Yes.

Q. Can you briefly describe the team approach that you alluded to with respect to your -- your work on the Dalkon Shield Litigation Trust.

A. Sure. To make sure that our statistical models were providing fair and reasonable values and taking into account the kind of factors that -- that influence the value of these claims, we worked with a team of gynecologists and infectious disease experts to make sure that -- that, to the best extent possible, we could go into these computer databases and -- and look at the, for example, the ICD-9 codes there to make sure that we were looking at the correct codes for these diseases and interpreting them correctly and putting them into our statistical models in a reasonable way.

Q. Now did you also work for the Centers for Disease Control?

A. Yes, I did.

Q. Can you tell us what that organization is -- what that organization is.

A. Well the Centers for Disease Control in Atlanta, or with CDC, they monitor and track disease rates in the U.S. and study causes of disease. The division that I worked with was the International Division, which does similar projects overseas, and in particular I worked on some African projects, providing statistical modeling assistance to medical professionals there who were studying childhood diseases in African clinics. And also one study I worked on involved looking at the factors that -- that influence a cholera epidemic spread around the shores of Lake Tanganyika, and particularly into some towns and villages in Berundi.

Q. How long were you involved in that work?

A. Couple of years.

Q. And when was that?

A. That was in the early 1990s. I can't remember exactly. Probably '92.

Q. You mentioned other medical professionals. Can you explain that. Who did you work with?

A. Well again I worked with experts at the CDC who knew the characteristics of these diseases and what factors were worth looking at, what factors were likely to be important or not important, and how to interpret the way these factors intermingled and combined to change the risk of cholera, for example, in the study I was just talking about.

*7 Q. And why did you work with these other health professionals?

A. Well I'm a biostatistician and I'm kind of a numbers guy on these projects, and although I have some background in working on public health problems, I wouldn't really feel comfortable trying to work on one of these projects without someone who knew the basic science and knew the biology and knew something about how the disease processes work.

Q. Did you also work for the Health Care Financing Administration?

A. Yes, I did.

Q. What is that organization?

A. Health Care Financing Administration is a -- is a federal agency that -- their main roles are to oversee the Medicare and the federal part of the Medicaid programs. I worked for the officer -- in the Office of the Inspector General. My project for them involved allegations that blacks were waiting longer for kidney transplants than whites, and that was -- if that was true, that was clearly against the rules and regulations. And -- and another part of their function was overseeing the -- the organ sharing in the country, and when there are kidney donors, who gets kidneys and other organs.

But people in the system and in the transplant centers said that, well, there really wasn't any difference, that the problem was in the statistics, that -- that they weren't really comparing blacks to whites by comparing like blacks to like whites if there were different blood types and antigens, the kind of things in the immune system that make a person a better fit to get a kidney or not, and they said that if you -- if you control for or adjusted for the fact that blacks had characteristics that made them less likely to be a good kidney recipient, that all this difference between blacks and whites would go away.

So our job -- and I say "our." I was working with people who knew about kidney transplants and the biology of the immune system, when they're likely to be rejected and -- and so on and so forth. We worked up a statistical model to take into account all of these factors that they felt really might explain this difference, and the statistical model took in information from hundreds of thousands -- or tens of thousands, anyway, of kidney transplants over a several-year period to look at -- and also the people waiting on the recipient lists, to look at these various medical factors that the experts said were important to look at in seeing whether you're a likely recipient or not, whether, if you need a kidney, you're likely to get one fast because there is a lot of people out there like you, or whether you may just have to wait because it's the luck of the draw. And once we did take all these factors into account, it did appear that -- that the difference between black and white waiting time wasn't as big as it was when first reported, but it still persisted even after comparing blacks and whites on a like to like basis for all of these antigens and blood types and these other -- other factors that -- that influence whether or not you're going to be a lucky recipient and get a kidney fast, or an unlucky one and may have to wait a long time.

*8 Q. And did you do -- or strike that.

Did you work with large databases in that project?

A. Yes.

Q. Can you tell us a little bit about that.

A. Well again, these -- these databases involved people on waiting lists for kidneys throughout the country and databases of people who had received kidneys over the last several years, and all of these are tracked in United Organ Sharing Headquarters in Richmond, Virginia, and kept on computers there.

Q. And did you do statistical modeling on that project?

A. Yes. That was the way in which we --

Instead of simply comparing average waiting times of blacks to whites, we broke down the problem so we could, first of all, essentially stratify by these different factors, compare like blacks to like whites in terms of their likelihood of receiving an organ fast, and then compile all the results into a final analysis of -- of -- of the issue of whether the blacks do seem to be waiting longer for their kidneys than the whites.

Q. How long did this work take?

A. Again, that work spread over about a two-year period, and that was, again, in the early '90s. I couldn't remember the exact dates.

Q. Have you also worked as a consultant for the city of New York?

A. Yes, I have. That project involved emergency medical transport services. The city of New York has about a million calls a year for ambulances, and a couple of things that -- like in all ambulance systems -- they're concerned about were response time, how fast could they get an ambulance to the site once there was a call to a dispatcher, and the other was what kind of ambulance got there. Because as in most systems, New York runs what are called ALS and BLS units, Advanced Life Supports and Basic Life Supports, and Advanced Life Supports are staffed with paramedics and able to take on more complicated heart attack situations and whatnot. And first of all, they wanted to improve their response time, but one of their problems was that -- that they had two things kind of working in opposition, because if they just sent the ALS units out on any call, they could lower their response time a little bit, but then if a serious accident came in, the ALS unit was already, you know, enroute to a trauma center or something and wasn't available.

So we took data from the New York City Emergency Medical Services System, these million calls a year they were getting, and -- and that involved a database of -- I forget exactly, but of approximately 10 million records of all the sites ambulance -- basically the logs of the ambulances during -- during the day, where they were, when they were called, how long it took to get to scene, how long it took to leave the scene, how long it took to get to the trauma center and where the locations were in all the boroughs of New York. And then in conjunction with some experts who build dispatch systems and run them in various cities, and other experts who -- I worked with a guy in particular who started out as an emergency med tech and then became eventually the vice- president of a company that works on consolidating ambulance services for big municipalities. Working with them we built statistical models that gave the city some options in terms of how it could -- certain things it could do to increase response time, to improve it -- or to decrease response time, to improve its response times, and try to make sure there was a greater likelihood at the same time that there was an Advanced Life Support unit available when a call that -- that needed an ALS unit came in.

*9 Q. When did you do this work?

A. That would have been about the mid-1990s. Probably, '94, '95.

Q. And how long were you involved in this project?

A. That was about a year.

Q. Have you also been engaged as a consultant in several litigations regarding asbestos?

A. Yes, I have.

Q. Can you tell us about that.

A. Well those litigations, again, involve products that are known to cause disease, asbestos fibers in -- particularly in construction and insulating materials, and particularly back in the forties, fifties and sixties, people that worked with that, again like with the Dalkon Shield, there are a variety of litigations out there against some of the manufacturers and miners of asbestos products for seeking compensation for illnesses and disease caused by asbestos. And again in these cases there are now typically hundreds of thousands of claims, and they range a great deal in -- in terms of -- of how many dollars. There are fatal diseases and there are non-fatal diseases that don't really cause much lung dysfunction. And in all of these litigations they face a problem saying, well, what is the total value of these 200,000 or 300,000 or 100,000 claims? And there's an additional difficulty there because a lot of these are cancers that have a long latency period, sometimes 50 years, so a lot of the cancers haven't occurred yet. And when a lot of these litigations where they set up, often, a trust like the Dalkon Shield Trust I talked about, the trust has to work out some rules for dealing with all the cancers that will keep occurring for a number of years.

And so, again, I worked on those cases with empidemiologists and built statistical models by first building databases of tens and hundreds of thousands of -- of claims that are at least on record now, and also using other data on occupational mixes in the past when a lot of the exposure occurred in order to estimate not only what the value of these claims are now, but how many will occur in the future and when they'll occur, five years, 10 years down the line, or whatever.

Q. Have you also published papers in peer-reviewed journals?

A. Yes.

Q. Can you tell us about that.

A. I've published two articles in peer-reviewed journals or volumes of invited chapters in the area. One was an article on sample survey design when I was with the U.S. Geological Survey, had to do with irrigation water use, and the other was a paper on the kidney transplant analysis that I described a couple minutes ago.

Q. And are you also a member of several professional organizations?

A. Yes.

Q. Are you a member of the American Statistical Association?

A. Yes.

Q. Are you also a member of the Biometric Society?

A. Yes.

Q. Were you retained by the state and Blue Cross Blue Shield of Minnesota in this matter in 1994?

A. That's correct.

Q. Are you prepared today to offer an opinion about the health-care costs paid by the state of Minnesota and Blue Cross Blue Shield of Minnesota to treat diseases and conditions caused by smoking, made worse by smoking, or made more expensive to treat by smoking?

*10 A. Yes.

Q. And is the time period covered by your opinions from 1978 through 1996?

A. Yes.

Q. Have you reviewed the testimony of Dr. Jonathan Samet in this case?

A. Yes, I have.

Q. Are you relying on that testimony?

A. In part, yes.

Q. Have you reviewed the testimony of Professor Scott Zeger in this case?
 

A. Yes, I have.

Q. Are you relying in part on that testimony?

A. Yes.

Q. Now have you worked with others in preparing your opinions in this case?

A. Yes, I have.

Q. Who have you worked with?

A. Dr. John Samet, Professor Scott Zeger, and also Dr. Len Miller at the University of California at Berkeley.

Q. And what was your role in preparing these opinions?

A. Well we built a statistical model to estimate the smoking-attributable expenditures for the plaintiffs in this case during this time period, and I was involved from beginning to end in design and building and testing of that statistical model, and I was also involved greatly in the collection and assembling of data from the state of Minnesota and from Blue Cross Blue Shield with which to drive that statistical model.

Q. Now the statistical model that you're referring to is the refined model referred to by Dr. Zeger?

A. That's correct.

Q. Did you also have any involvement in working on or developing the core model referenced by Dr. Zeger?

A. Yes, I worked on that as well.
Q. Now how long have you worked on this project?

A. Since about the beginning of 1995, so that's three years I guess.

Q. Have you had meetings and discussions with Drs. Samet, Zeger and Miller?

A. Many meetings.

Q. Now as part of your responsibilities, did you collect claims records from the state and Blue Cross Blue Shield of Minnesota?

A. Yes, I did.

Q. Could you first tell us what you mean by a "claims record."

A. Well a claims record is essentially a -- it's a computerized doctor bill. When a doctor or a hospital in Minnesota sends a bill in to Blue Cross or -- or to Medicaid, it gets entered on a computer, or sometimes these days it's already computerized before it gets there, and that's the way it gets into the system for review and for payment of the bill, is getting into that computerized system. So when -- when I say a claims record, that means going back to that computer system and looking at whatever was entered into the computer for that doctor or hospital bill at some time over this period.

Q. From what programs did you obtain the state claims records?

A. Well the state claims records came from two programs, Minnesota Medicaid and Minnesota General Assistance Medical Care, GAMC.

Q. Did you familiarize yourself with the Medicaid program?

A. Yes, I did.

Q. Did you review documents, including claims records?

A. Yes.

Q. Did you talk to state administrators of that program?

A. Yes. I talked to a variety of administrators of the plaintiffs' programs, administrators of the database system, programmers and programmer analysts in the database system, and other people as well who were experts in various programs.

*11 Q. And did you also review a Medicaid Sourcebook?

A. Yes. That's one of the documents I looked at.

Q. Let me direct your attention to Trial Exhibit 26049. You've got that in front of you. It's in the red rope. If you could take that out. Perhaps you could set the red rope aside there.

Can you identify that document?

A. Yes. That's a Medicaid Sourcebook prepared by the Congressional Research Service, 1993 update.

Q. And you reviewed this document?

A. Yes.

Q. And is it a reliable public report?

A. I believe so.

Q. And does it set forth a description of the Medicaid program?

A. Yes.

MR. HAMLIN: Your Honor, we offer Trial Exhibit 26049 under Rule 803(8). It's a public record.

MR. BIERSTEKER: No objection on that basis, Your Honor.

THE COURT: Court will receive 26049.

BY MR. HAMLIN:

Q. Dr. Wyant, could you give us a brief description of the Medicaid program.

A. Well the Medicaid program is basically a way to pay for medical care for generally poor and low-income people in Minnesota. It's jointly funded by the federal government and the state of Minnesota. And there are certain categories of people that it covers: blind, aged, disabled, and people in families, particularly people that qualify for Aid to Families with Dependent Children, again on Medicaid. So there are certain restrictions about getting on to the program, but in general all the people that do get on are in need of some financial assistance to pay for medical care.

Q. Did you also familiarize yourself with the General Assistance Medicare -- or Medical Care program?

A. Yes, I did.

Q. And did you review documents?

A. Yes.

Q. Including claims records?

A. Yes, I did.

Q. Did you talk to administrators of the program?

A. Yes. Again, administrators and people familiar with the computer databases and -- and other experts who -- who knew something about the workings of that program.

Q. Let me now direct your attention to Trial Exhibit 26050, which should be in that manila folder next to you.

A. Yes.

Q. Can you identify this document?

A. Yes. That's one of the annual reports from the Department of Human Services of Minnesota and Minnesota General Assistance Medical Care.

Q. And was this one of the documents that you reviewed?

A. Yes.

Q. And do you find it to be a reliable public report?

A. Yes.

Q. And does it set forth information regarding the General Assistance Medical Care program?

A. Yes.

MR. HAMLIN: Your Honor, plaintiffs offer Trial Exhibit 26050 under 803(8) as a public record.

MR. BIERSTEKER: No objection, Your Honor.
 

THE COURT: Court will receive 26050.

BY MR. HAMLIN:

Q. Could you give us a brief description of the General Assistance Medical Care program.

A. Again it's like Medicaid, a program for paying for medical care for people who for one reason or another are unable to pay for it themselves. And generally it tries to cover people who don't fit into one of the slots that Medicaid would cover, and so it covers people who, for example, might be poor but don't meet one of the criteria like being a head of household in a family qualifying for Aid to Families with Dependent Children. For example, single men are more likely to qualify for GAMC assistance than they are for Medicaid assistance, for -- for one example.

*12 Q. Have you also familiarized yourself with Blue Cross Blue Shield of Minnesota health care plans?

A. Yes, I have.

Q. And how did you go about doing that?

A. Again, I went out to Blue Cross and met with administrators of the programs, actuaries out there whose -- whose job it is to provide reports and analysis of -- of their programs to Blue Cross management, and met with administrators, database professionals and computer programmers.

Q. How many meetings did you have with Blue Cross administrators?
 

A. Countless meetings.

Q. Over how long a period of time?

A. Over two years.

Q. And did you also review Blue Cross claims records?

A. Yes, I did.

Q. Did you restrict yourself to any of the health care plans in your review?

A. Yes. We looked only at what are called the rated fee-for- service plans, and these were plans of Blue Cross -- basically they would make an arrangement with an employer or perhaps a union to cover all the employees, and it's those kinds of plans that we were looking at.

Q. Did you look at any individual policies?

A. No.

Q. Do you have an exhibit regarding the number of records collected from the state of Minnesota and from Blue Cross Blue Shield of Minnesota and the number of people with smoking- attributable diseases?

A. Yes.

Q. Can you turn to Trial Exhibit 30202.

A. (Coughing) Excuse me.

Q. Is that the exhibit?

A. Yes, it is.
 

Q. Was that prepared at your direction?

A. Yes, it was.

MR. HAMLIN: Your Honor, plaintiffs offer Trial Exhibit 30202 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 30202 for illustrative purposes.

BY MR. HAMLIN:

Q. Dr. Wyant, could you first of all tell us the title of that exhibit.

A. This exhibit is called "Billing Records Analyzed in the Statistical Model."

Q. And could you tell us what is on that exhibit.

A. This is just a count of the number of records in the computer files that we assembled to work with in building our statistical model. Again, these are basically computerized doctors' bills over this time period of 1978 to 1996. And for the state of Minnesota we assembled 224 million records, and for Blue Cross Blue Shield of Minnesota we looked through 60 million records, and in these records we found more than 90,000 people who were being treated for one of the major smoking-attributable diseases.

Q. And how did you determine the 90,000 Minnesotans who were treated for major smoking-attributable diseases in these records?

A. We used ICD-9 codes and some rules that were developed by Dr. Samet.
 

Q. Now can you tell us generally what was in these claims records, first of all from the state of Minnesota.

A. Well the records have, as I say, ICD-9 codes; that is, they have disease information, they have what the state of Minnesota paid in these claims, and they also have identification information about the person in terms of their age, their gender, and often some other information about them as well.

*13 Q. Can you tell us about the claims records of Blue Cross Blue Shield of Minnesota.

A. Again, same kind of information appears there, particularly disease information and how many dollars were spent by Blue Cross Blue Shield to pay for medical care in these various computerized bills.

Q. Now do you have a block diagram of the statistical model and -- and data sources that were used in this case?

A. Yes.

Q. Could you turn to Trial Exhibit 30182, please. Do you have that?

A. Yes, I do.

Q. Is that the block diagram?

A. Yes.

Q. And was that prepared at your direction?

A. Yes, it was.

MR. HAMLIN: Your Honor, plaintiffs offer Trial Exhibit 30182 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 30182 for illustrative purposes.

BY MR. HAMLIN:

Q. Dr. Wyant, could you first tell us the -- the title of this exhibit.

A. This is called "The Statistical Model."

Q. And what statistical model does that refer to?

A. That's the model that Dr. Zeger referred to as the refined model, and it's the model that I built in collaboration with him and Dr. Samet and Dr. Miller.

Q. Can you describe for us what's on that exhibit.

A. Well we have here some empty boxes that we'll fill in in a minute. The boxes just indicate different data sources like the claims records we were just talking about that feed into the statistical analysis to give us a result.

Q. Now what was the purpose of the collection of all of this data?

A. Well the purpose of the collection of this data was to have sufficient information to provide a reliable estimate of smoking-attributable expenditures, so in particular we were looking for -- because Dr. Samet laid out the foundation for us in epidemiology of -- of looking at smoking as causing disease resulting in dollars, particularly looking for smoking, we were looking for disease, and we were looking for dollars in these various data sources.

Q. So you were looking for smoking.

A. Yes.

Q. And you were looking for disease.

A. Yes.

Q. And you were looking for cost.

A. Expenditures, yes.

Q. And what did Dr. Samet tell us about this relationship?

A. Well we're relying on his expertise on smoking causing disease and also on disease resulting in costs.

Q. Now why is it that we need this information for the statistical model?

A. Well if the statistical model is going to be an accurate and reliable representation of a particular population on Minnesota Medicaid or Minnesota Blue Cross Blue Shield, then we have to collect information that's pertinent to those programs.

Q. Let me put the diagram back up on the easel. That's Trial Exhibit 30182.

What I want to turn now to is the billing information. I'm going to place on the diagram an attachment, "Billing Records."

Now when did you begin collecting billing records from the state of Minnesota?

A. It was about the beginning of 1995.

*14 Q. And these billing records were from what programs?

A. For the state it was for Medicaid and for GAMC.

Q. Where did you go to collect these records?

A. Well, we went to the state agencies, the Department of Human Services, again talked to their people, and we identified various computer records that they had available. In particular for the state they looked at a set of computer tapes called the Medicaid Statistical Information System, or the MSIS tapes.

Q. And did you have meetings with the state administrators and state employees about these records?

A. Oh, yes. Many, many.

Q. Over how long a period of time?

A. Over two years.

Q. What if any access did you have to the state of Minnesota computers?

A. Well, some of their programmers gave us assistance, but also we had full access to the system, so either myself or programmers who worked for me would spend many days on site looking at these computerized records or also accessing the computer from our own offices.

Q. Now is there identification information on these individual claims records?

A. Yes, there is. But one of the parts of the analysis from the beginning for us is that we didn't want to see any identifying information and completely preserve the confidentiality of the patients and doctors in this system, so one of the first things we did was meet with the programmers at Medicaid and GAMC and developed the plan that they would pre-screen any of these computer data tapes that we were getting so that either the identifying information -- certainly names and addresses but also things like birth dates -- were wiped off the tapes before we saw it, and then I.D. numbers like Social Security numbers were all recoded into some nonsense code so that we could see that this was one person in the system, but we had no way of knowing, you know, what their I.D. number was for Medicaid or what their Social Security number was.

Q. Can you give us a list of some of the demographics that appear on these claims records for the state.

A. Well again we get age and we get sex, gender, in these claims records we also have race, education and marital status and maybe some other information as well. I'm not -- I don't remember at this time.

Q. You also have -- excuse me.

You also see disease information in those records; right?

A. Oh, yes. And then besides the information about the people, there is the complete billing history of their medical events while they were being covered by Medicaid and GAMC. And this would include diseases, it would include type of service; that is, whether it was a hospital or doctor's visit, and it would include the dollars spent by the state to pay for these services, and it would include the dates at which the person went to the hospital or went to the doctor's office.

Q. What if any checks are made of these claims records to ensure accuracy?

A. Well on the tapes we're looking at, the MSIS tapes, those are -- those are claims that are actually paid, so they've already been through various state audit systems in the course of their normal business of -- of checking medical bills, but in addition these particular tapes are made to -- for the federal government, they're made to the specifications of the Health Care Finance Administration that takes these tapes and uses them for statistical analysis of Medicaid programs in all the states, and the Health Care Finance Administration applies its own checks on these tapes. In fact, there are several hundred checks applied or that are required by the Health Care Finance Administration, and until a tape submitted from the state to them passes all these checks, it's not accepted by HCFA or the Health Care Financing Administration, and it's sent back to be redone until they can get it -- get it right, if -- if in fact they don't pass on the first go-around. So before we get it, there is already many layers of checking that are done on these data.

*15 Q. Now you mentioned MSIS.

A. Yes.

Q. Can you tell us what that stands for?

A. Again that's called Medical Statistical Information System, and it just means this set of tapes that are prepared in the normal course of business during the year by Medicaid for reporting to the federal government what's going on inside the state program.

Q. What if any accuracy checks did you yourself do of this data?

A. We did a variety of additional checks. As one example, the Health Care Finance Administration gets these tapes and runs certain reports off them, and so we wrote our own programs to produce essentially the same report and then compared that with the results of what these other programmers had done at the Health Care Finance Administration. And then we also spent time on the telephone with Health Care Finance Administration programmers and administrators who deal with these tapes to make sure that our understanding of -- of what was reasonable in terms of handling these tapes was correct.

Q. What if any review did you conduct of these claims records with your colleagues, including Dr. Samet?

A. Another thing we did early on was to, once we compiled a number of these tapes, was to put them together and print out complete claims histories for randomly selected people in the system so that you could see the progress of diseases at certain dates and then going to the doctor's office and -- and getting certain kinds of treatment, and then going to the hospital, getting out of the hospital, other diseases. And went over many of these with Dr. Samet to make sure that -- that A, all of our programs seemed to be working in a reasonable way and the data was as you would expect both as an epidemiologist and a clinical doctor, and to make sure that we had a good understanding of -- of the statistical data, what was there and -- and how it looked and how it worked in terms of the things we needed to be aware of in using this data as the basis for a statistical model.

Q. Let's turn now to the Blue Cross Blue Shield of Minnesota claims records.

THE COURT: Counsel.

Q. When did you begin --

THE COURT: Counsel.

MR. HAMLIN: Yes.

THE COURT: Do you wish to take a short recess at this time?

MR. HAMLIN: Fine, Your Honor.

THE COURT: Okay.

THE CLERK: Court stands in recess.

(Recess taken.)

THE CLERK: All rise. Court is again in session.

(Jury enters the courtroom.)

THE CLERK: Please be seated.

THE COURT: Counsel.

MR. HAMLIN: Thank you, Your Honor.

BY MR. HAMLIN:

Q. Dr. Wyant, do the claims records for the state of Minnesota and Blue Cross and Blue Shield of Minnesota contain smoking information?

A. No, they do not.

Q. When did you begin collecting claims records from Blue Cross Blue Shield of Minnesota?

A. Again, about the beginning of 1980.

Q. Did you have access to Blue Cross's computers?

A. Yes. In the same way as with the state, their programmers did work for us, but also myself and some of my programmers accessed their computers on site for many occasions over this time period.

*16 Q. Can you tell us what is in the claims records of Blue Cross Blue Shield of Minnesota that you reviewed.

A. Again, these claims records contain disease information, they contain expenditure information, they contain type of services, hospital, doctor visit, and they contain information about the persons such as age and gender.

Q. Do they also contain identification information?

A. Well the -- of course the original information in the computer does that -- that Blue Cross uses in -- in running its business, but as with the state, we wanted to be extremely careful about confidentiality, and before we saw any records we worked out with their programmers again some rules for either eliminating completely identifying information, or when it came down to just a code like the subscriber number, that was replaced with a nonsense code so we could track a person, but we had no way of inadvertently seeing an actual subscriber number or other I.D. for any of these people in the -- in the Blue Cross records.

Q. What if any role did you or your colleagues have in encrypting the claims record identification information for the state of Minnesota and Blue Cross?

A. None whatsoever. I was very specific that I didn't want to know how it was done. I laid out what we wanted to accomplish and left them to it.

Q. Now in addition to any usual audits, did you conduct your own accuracy checks of Blue Cross Blue Shield of Minnesota records?

A. Yes. We carried out a lot of checks on -- on these claims records and billing records that we were assembling. Blue Cross did not have a set of tapes. As a business, they don't have to report to the Health Care Financing Administration in the way that Medicaid does. And so we did some additional checking there, kind of replacing the sorts of checks that the Health Care Finance Administration typically runs on the state of Minnesota tapes.

MR. BIERSTEKER: Your Honor, Your Honor, I object to this line of inquiry. This was not something that was disclosed in the expert report. The work product that was generated by these kinds of checks was not produced to us. This is impermissible, I believe.

MR. HAMLIN: Your Honor, if I may. Those tapes from the -- that were prepared for HCFA were in fact produced to the defendants. Those are the tapes we gave them.

MR. BIERSTEKER: I'm sorry, this is Blue Cross Blue Shield, not HCFA.

MR. HAMLIN: He's not --

THE COURT: All right.

MR. HAMLIN: He's not suggesting, Your Honor, that HCFA reviewed the Blue Cross tapes.

THE COURT: Why don't you re-ask the question so we clarify it.

MR. HAMLIN: Okay.

BY MR. HAMLIN:

Q. What if any accuracy checks did you yourself conduct of the Blue Cross Blue Shield claims records?

MR. BIERSTEKER: Objection, Your Honor, same basis.

THE COURT: No, you may answer that.

A. There were a variety of checks. One thing we would do is have meetings and go over the programming code with the programmers to make sure we were on the same page with what we were trying to accomplish. We did detail-level checks where we would again write programs that would create complete claims histories for individual subscribers randomly selected. Blue Cross has its own programs that do that. For example, if a claims rep wants to look at a claims history, someone has a question about -- about a past bill, and we had that, Blue Cross people go back and prepare the results of our printouts of claims histories with what was coming out of their existing program, and so we had two independent computer programs and made sure they were producing the same results.

*17 We also, at a -- not at a detailed level but at a more aggregated level, compared the totals we were getting with various quarterly and annual reports from Blue Cross that addressed these rated plans we were working with, and made sure that we were replicating the total dollars and the total health costs that they were showing in their own reports.

In addition we did a number of other checks. One example is -- is looking at -- at any trends over time from 1978 to 1996 in -- in various aspects of the claims data we had, and any time there was a -- a jump in a trend, we would make sure that we had an explanation of it from the Blue Cross people and that it wasn't some error in gathering the computerized data.

Q. Did you have occasion to review the Blue Cross Blue Shield of Minnesota claims records with Drs. Samet, Zeger and Miller?

A. Yes. We went over quite a few of these claims records after we printed them out for randomly selected people in this claims database.

Q. Now with respect to the state and Blue Cross, did you have computerized billing information for all of the years; that is, 1978 through 1996?

A. No, we didn't.

Q. What years were you missing for the state?

A. Well for Medicaid we had computerized billing records from 1982 through 1995, although 1995 was not entirely complete. For GAMC we had billing records from 1987 through 1995, although again not complete in 1995. And for Blue Cross we had computerized records from 1980 through 1995.

Q. Now what other information did you look at for the state for the years where there was no computerized data?

A. For the state we assembled a number of annual or quarterly reports that the state programs produced, again in their normal course of business, back in those years when we didn't have computerized data, and from those reports we got the total health costs paid for by those programs in the years prior to or after the years in which we had the actual computerized records.

Q. What other information did you look at for Blue Cross?

A. The same thing. Again, annual reports and quarterly reports in some instances that gave us total dollars expended in rated group plans.

Q. And do you have an attachment to the magnetic board which indicates the use of the annual reports?

A. Yes.

Q. For the record, I'm now placing that attachment on the mag board. Can you tell us what that says, Dr. Wyant?

A. "State of Minnesota and Blue Cross Blue Shield Annual Reports."

Q. And those are the annual reports to which you were referring.

A. That's correct.

Q. Now based on your experience with large data sets, how does the quality of this data set compare with those that you have used in the past?

A. It's a very comprehensive data set, seems to be of very high quality, and has a really unusual amount of reliable information in there.

Q. Now you mentioned earlier in your testimony ICD-9 codes. Is there an ICD-9 code manual?

A. Yes, there is.

*18 Q. And did you make reference to that in your work in this case?

A. Yes.

Q. Could you turn to Trial Exhibit 14968, which should be to your right.

A. Yes.

Q. Okay. And can you identify that.

A. This -- this is the ICD-9 manual, International Classification of Diseases, 9th Revision.

MR. HAMLIN: And for the record, Your Honor, the trial exhibit has been previously admitted.

Q. And did you make use of this manual in your discussions with your colleagues, Dr. Samet, Dr. Miller and Dr. Zeger?

A. Yes, I did.

Q. What if any use did you make of the manual?

A. Well the manual basically explains these ICD-9 codes. It's a mapping. It says when you see this code, what disease does it mean, and it just lists all those in order so that you could -- you can go back and forth from -- from the real description of the disease to the code or from the code to the description.

Q. Did Dr. Samet provide you with a list of ICD-9 codes for diseases caused by smoking?

A. Yes, he did.

Q. If we could have on the Elmo Trial Exhibit 30153, which has been previously admitted.

Is that the list provided to you by Dr. Samet --

A. Yes, it is.

Q. -- of the diseases caused by smoking and their ICD-9 code numbers?

A. Yes.

Q. And can you identify those for the record, please.

A. These are the major smoking-attributable diseases, and after each disease or set of diseases are ICD-9 codes.
So it starts with atherosclerosis, the ICD-9 codes 440 or 441, and 444; bladder cancer is code 188; cerebrovascular disease or stroke, 342 and 430 to 438; chronic obstructive pulmonary disease or COPD is 491 to 492 and 496; coronary heart disease, 410 to 414, 425, 427 to 428; esophageal cancer, 150; kidney cancer, 189; laryngeal cancer, 161; lung cancer, 162; oral cancer, 140 to 141 and 143 to 149; pancreatic cancer, 157; peptic ulcer disease, 531 to 533; and diminished health status/respiratory morbidity and mortality, including respiratory infections, 460, 464 to 466 and 480 to 486.

Q. Now are those code numbers found in the manual that you just identified?

A. Yes.

Q. And in fact that's where those numbers were taken from; is that right?

A. That's correct.

Q. Dr. Wyant, could you give us an example of what a claims record looks like for an individual on GAMC. And with the court's permission, come down and use the flip chart.

A. Well this is just an example of the -- the kind of information you see in a claims record and the kinds of things we look at when we do printouts to check and make sure we have a good understanding of this information.

I'll give an example from the GAMC program, a man 51 years old. Again, this is the kind of information we get off the system. We've taken out any I.D. information, but we do know what their gender is and how old they are, although we did take off the actual birth date.

The data in these claims records, they do come in the form of these ICD-9 codes, and one you may see in these claims is like a 786.5. The codes you see up on the Elmo are all to three digits. The classification system can really go finer than that, and so when you see the actual codes on the computer, they'll actually have another digit attached to them. And this is a code for chest pain. And this is not a real person I'm writing up here, but these are generally the kinds of dollars and diseases that -- that we would typically see in our work.

*19 You know this is chest pain by either looking it up in that manual, if you're doing it by hand, or you use versions -- what we did was use versions of that manual that are on the computer, so when we saw 786.5, that was translated into chest pain for us so we could see what was happening.

Then the next thing you might see is 162.9 in the kinds of claims we're looking at, which is a lung cancer. Again, this is one of the 162 codes that you see up in the Elmo. Lung cancer is 162, and point one, point two, point three they use for different sites of lung cancer in the lung. And this cost, we see substantial dollars associated with that code.

And another thing we see in these records is what's happening. In this case the person went to the hospital. And to give you -- as I go through these with Dr. Samet, I try to get, and any biostatistician tries to get, kind of a general, common-sense understanding of what's going on here, and the kind of understanding that I developed, what typically goes on in these records is that here's a person that comes in with chest pain, and then it turns out the diagnosis, what was causing that chest pain, is lung cancer. As the person gets out of the hospital, you're likely to see a different kind of code, one starting with a V, a V72.5, and that would be a code for a routine x-ray exam. Might only cost fifty dollars. Then something you see not uncommonly in these records might be a 998.5, postoperative complication, which again might cause the person to have to go back into the hospital for treatment. Could easily cost a couple thousand dollars.

And another thing you see a lot in the records is prescription drugs, and particularly you see a lot of prescription drugs where there's serious illnesses like this, and that might be five hundred dollars. Not be unusual for several months for a person like this. Now there is no ICD-9 code for prescription drugs. The reason for that is generally people go to their pharmacist to get prescription drugs and the pharmacist doesn't fill in ICD-9 codes for diagnoses, so you don't typically see any codes for prescription drugs in the system.

And then another code you might typically see in a sequence like this for a person is something like a V58.0, that's for radiation therapy, or a V58.1, which is chemotherapy. Generally those sessions are a little more expensive. And you might see a V70.8, which is a general medical exam for a hundred dollars. All of these things, again, just speaking of my general understanding from -- from talking with Dr. Samet and getting a common-sense picture of what goes on in these records and how to deal with them in a statistical model.

Of course this is very common-sensical. You have an operation, may have complications, you got prescription drugs, couple of other problems, and of course the cancer radiation and chemotherapy are standard courses of treatment. And then as time goes on, might see something like this, 844.9, this is just a leg and knee sprain, maybe a doctor's visit for a hundred bucks, and then you might see something like this, 724.5 for a backache, and then you might see 198.5, secondary bone cancer. See a lot of these with the lung cancer claims just going through the data. It costs a lot to treat.

*20 And again this could be -- just as when a person comes in, they have chest pain and then it's diagnosed as lung cancer -- some of these orthopedic problems like backache could be the first signs of the secondary bone cancer.

And then after this you can sometimes see a code like this, 733.1. that's a code for pathological fracture, and that's where the cancer invades the bone, the bones become weak and the bones start breaking. And then you see another thing that's contained in claims data: person typically dies, for a patient like this.

And there are a couple aspects of this that are important for understanding the statistical analysis. This one is oversimplified, because there are dates on these and -- and usually you wouldn't see just this many services. You can kind of think that every time I wrote a line here, you're going to see several lines in the computer record with different dates. For example, you wouldn't typically have just one chemotherapy session. You see in these records a person like this would be going in for visits and hospital stays just over and over and over again over many weeks. But I've just given one line apiece here.

Another aspect of this that's -- that's important for statistical modeling is that only one of these codes appears on the chart up there, just that one.

Q. Which one is that, doctor?

A. That is 162.9, lung cancer. Now as you look through all the things that are happening to this person, and if you sit down as I did with Dr. Samet, it's pretty clear that if you walked into this just as a statistician and say, well, I know how to get all the lung cancers out of the Minnesota claims data, I'll just look for codes 162, and if you did that in this person, for example, well you'd miss the chest pain, which may be the first sign of the lung cancer, you'd miss the follow-up x-ray exam because that doesn't have a 162, you'd miss the post-op complications and you'd miss any of the prescription drugs that are related to the lung cancer and what's going on, because over in this ICD-9 code you're not seeing a 162 code. And then you'd miss the radio- therapy, which is the V58, you'd miss the chemotherapy, and you'd miss the general follow-up medical exam.

Now the knee and leg sprain we'll talk about a little later when we talk about the third reduction. Maybe that one's okay in this.

Backache, well no. It could be related to the secondary bone cancer, but we missed that. We certainly missed the secondary bone cancer, that's a 198.5 and that's not a 162. And we'd miss the pathological fracture.

So part of doing the statistical modeling and using this data in a -- in a common-sensical way is to make sure that we don't go into it and miss all these things when we're trying to look, for example, at smoking-attributable expenditures due to lung cancer.

The other aspect of this that's fairly typical for a GAMC claim is that we're basically not seeing the person until they have lung cancer, and the majority of the GAMC claims for lung cancer are like that. And again, my understanding talking from administrators in the program as -- as well as looking at the claims data, that a typical situation is there will be someone who doesn't have much financial resources but is getting along okay, and then all of a sudden they're hit with all this stuff. And if they didn't -- if their resources were low enough to begin with, once they start having these bills, it's likely that they can qualify for General Assistance Medicare -- Medical Care in order to pay for this sequence of treatments.

*21 Q. What would happen to the final smoking-attributable expenditure in the model if you included just the ICD-9 code for lung cancer and the expenditure connected to lung cancer?

A. Well I haven't added this list up, but a common-sense view of this, again from my discussions with the medical people, is that most of these expenditures are related to the lung cancer, and if you didn't -- if you only took this five thousand dollars out, you'd be missing fifty dollars, you'd be missing 2,000, you'd be missing 500, you'd be missing 100, you'd be missing 500, and on down the list.

Q. Now did you do any other accuracy checks on the claims records in addition to those you described?

A. Yes.

Q. And do you have a trial exhibit setting out those accuracy checks?

Let me show you Trial Exhibit 30174.

A. Yes.

Q. And did you prepare this exhibit?

A. Yes.

MR. HAMLIN: Your Honor, we offer Trial Exhibit 30174 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: The court will receive 30174 for illustrative purposes.

MR. HAMLIN: Can we have those on the Elmo.

BY MR. HAMLIN:

Q. Could you tell us, doctor, the title of this exhibit.

A. This is called "Validation screens for identifying major smoking- attributable diseases in the computerized billing records."

Q. Could you take us through the exhibit and describe what - - what is on there.

A. This exhibit basically talks about how we identify people in the claims records who have one of the major smoking- attributable diseases; that is, where did we come up with the number 90,000. And there's a little more to it than just going through and finding every person that has, say, a 162 code.

The first screen is an age screen, and if a person shows up with one of the codes for the major diseases that were on the list that we had up on the board a few minutes ago, but their age is under 35 or under 40 for COPD, we don't count that as a major smoking-attributable disease. These diseases on that list occur very unusually at that young an age, and so the possibility is is there is a misdiagnosis or it was some unusual set of circumstances that we don't feel comfortable, at least in consultation with Dr. Samet, of classifying those as just one of the major smoking- attributable diseases when the person getting that disease is younger than those ages you see there.

So, for example, if this person here, instead of being 51, had come in and we saw them at -- at 34 years old in the claims record, we wouldn't count that.

Even though this 162 is here, we wouldn't count them as one of our 90,000 people.

Q. Can you tell us what the next screen is.

A. The next screen is the lab and x-ray screen. "Disregard major disease identification from lab and x-ray alone."

I can do that on the flip chart and give a sense of what's going on there. But it's similar to this person, what I'll write down, in that someone will show up for chest pain, except after a test it turns out they don't have lung cancer, they have something else. So let's say in this example it's a Medicaid woman 50 years old who comes in -- I'll put some labels on here -- ICD-9, disease. Say this person comes in and we see her first with a 786.5, chest pain, but then the next record, instead of being a hospitalization, looks like this. Has the 162.9, the 162 being the code for lung cancer, but instead of a 5,000-dollar hospitalization, we see a hundred dollars, and instead of an indication of hospital out here after that first doctor visit, we see an indication that this person went to like a imaging center to get an x-ray. And then maybe we don't see anything for a long while, and then we see them again with, who knows, some infection or some other condition.

*22 Now what went on here, typically, at least in talking with Dr. Samet and based on my own experience in dealing with ICD-9 codes and billing records, is that a doctor will send the person to the imaging center and say need to check out whether this person has lung cancer. They might say -- rule out lung cancer, they might say possible lung cancer, and so the code will get on the bill from the imaging center. And they say lung cancer because there's no place in here to say check for lung cancer. And what you'll see here is the person, since there was no further treatment and they didn't come in for quite a while and then just some normal medical visit, it's very likely that that 162.9 means that the person was checked for lung cancer but probably didn't have it. So we don't want to count this person as one of our 90,000 if we see someone like this. That would probably be overcounting the number of people with major smoking-attributable diseases in the state and in Blue Cross.

And it's these kinds of things that you need to screen out with the lab and x-ray screen, which is the second screen on the Elmo.

Q. What is the third screen on the Elmo.

A. That's called the independent confirmation screen, and that's after applying the other screens, we also want to see at least two identifications within a two-year period of whatever major smoking-attributable disease we're seeing here. It is what it says, an independent confirmation check.

And again, we can look at a similar example of a person coming in, let's say to Blue Cross. This person, again, let's say, comes in with a 786.5 or chest pain for a hundred dollars for an office visit, and the next thing we see is, again, a 162.9, another office visit, and then again we don't see the person for a while and they have some routine condition. Well this looks exactly like the previous person, except that this one occasion where we see a 162, it doesn't say x-ray, it says office visit. And one way you can get that happening is exactly the same way it happened on the previous claim. Went to a doctor, doctor said got to check you out for lung cancer. "I don't have the equipment, but my buddy down the hall does. Go see him. He's not a radiologist, but he can do this check." So the person goes and checks -- you know, a second doctor checks for lung cancer, does an x-ray, comes out clean, and so all you see again is --

One way this can happen is this one record that says 162.9, which probably meant check for lung cancer, but you don't see it again, and almost always if you have a condition like this you're going to see it appearing more than once in the records. So if it just pops up once like this, we don't count the person as one of our 90,000 Minnesotans who have these major smoking-attributable diseases.

Q. You've used the word "screens" in this context. What do you mean by the term "screen" now that you've explained how they work?

A. Again, that's just really a different way of saying -- saying what I -- I just said, that -- that a person can come in and have one of the ICD-9 codes on the master list of major smoking- attributable diseases, but we go through people that are first identified like that and we screen out anyone who gets caught up in one of these three checks here, and then they're not counted as having one of the major smoking-attributable diseases.

*23 Q. How are these screens developed?

A. They were developed in consultation with Dr. Samet. And again, this goes back to what I was talking about earlier, that as we got into these claims databases and started printing out claims for randomly selected people, claims histories, you would sit down and review them with Dr. Samet so that we could get an understanding of what they meant, how we should use them in the model, and how we should apply and -- and put together any validity checks or screens like the ones up on the board.

Q. And -- and what is the essential purpose of these three screens, Dr. Wyant?

A. Well we don't want to overcount or over-identify people with major smoking-attributable diseases; that is, we want to make sure we're making a reasonable effort to screen out anyone that might have a mistaken diagnosis of one of these diseases.

Q. And did you apply these screens to the claims records of the state of Minnesota and Blue Cross Blue Shield of Minnesota?

A. Yes.

Q. In the model?

A. Yes.

Q. Now can these screens exclude or eliminate people who actually have a smoking-attributable disease?

A. Oh, sure.

Q. How does that work?

A. Well there are a variety of ways, but -- but again, during my discussions with Dr. Samet, there -- there's a way it could happen. Let's say Blue Cross -- we're looking at 50-year-old woman, and say this person is going along over time and having occasional medical expenditures, nothing serious, but all of a sudden has a major heart attack -- I don't know the ICD-9 off the top of my head for that -- has a major heart attack and gets picked up by the ambulance, goes right to heart attack ward at the hospital and stays there for some time, but nonetheless dies. That can happen with heart attacks sometimes. So this could generate a lot of expenditures here for care in that intensive ward, but the only time we see heart attack would be the hospital admission, and then we wouldn't see the person again. So this is an instance where we go ahead and apply that third screen. We're going to be missing someone, we're going to be mistakenly excluding someone from that count of 90,000 even though they had a major smoking-attributable disease.

Q. Then why do you apply these screens?

A. Well we want to be careful about looking at these data, and we want to be sure that, if anything, we've got a slight undercount and that we can reliably trust these counts as reflecting at least what the disease burden was in this population over this time period.

Q. Thank you. You can return to the stand now.

Now we've -- we've talked about the claims records from the state of Minnesota and Blue Cross Blue Shield of Minnesota. Are there any other data sources that included Minnesota information that you used in the statistical model?

A. Yes.

Q. What is that?

A. We made use of data from the Minnesota Behavioral Risk Factor Surveillance System.

Q. And do you have a attachment to the magnetic board that signifies the Minnesota Behavioral Risk Factor Surveillance System?

*24 A. Yes.

Q. Now I'm going to place the attachment on the board. Could you identify that, Dr. Wyant.

A. Yes. That reads "Minnesota Behavioral Risk Factor Surveillance System (BRFSS)."

Q. And can you tell us what that data set or source of information is.

A. There is a cooperative program between the Centers for Disease Control in Atlanta, CDC, and the states to run these telephone surveys to collect information on behavioral risk factors, such as smoking. Those are conducted every year in almost every state. Minnesota has been the leading state in the nation for many years in terms of the number of people surveyed.

Q. Now which agency coordinates this survey?

A. It's the Minnesota Department of Health.

Q. And do they work with any federal agency?

A. Yes, the Centers for Disease -- excuse me, the Center for Disease Control in Atlanta, or I guess the actual name they say is the Centers for Disease Control and Prevention.

Q. Which agency actually conducts the interview?

A. The Minnesota Department of Health.

Q. And what kind of interview is it?

A. This is a telephone survey.

Q. How often is the interview or the survey taken?

A. Well actually it's kind of a rolling survey, and they generally are calling people every month of the year. And then they compile the results up into annual surveys for each year that they do the survey.

Q. When did Minnesota begin conducting this survey?

A. I believe the first year they conducted it in conjunction with CDC was 1984, although I think Minnesota had started its own similar survey a couple years before that.

Q. Can you tell us whether it's been conducted every year since then?

A. Yes.

Q. And approximately how many interviews have been compiled?

A. About 34,000.

Q. And how many is that per year?

A. Oh, that's roughly about 3,000 per year.

Q. Now is the interview that's conducted by the Department of Health conducted using a standard data collection questionnaire for assessing behavioral risk?

A. Yes.

Q. And is a standard protocol used for interviewing procedures and data collection?

A. Yes.

Q. Is the survey designed to allow participating states like Minnesota to add questions?

A. Yes.

Q. And these questions are state-specific?

A. Yes.

Q. And has Minnesota added questions regarding assessment of general health status?

A. Yes.

Q. Has Minnesota added questions regarding use of health care resources?

A. Yes.

Q. Now -- now what does that mean?

A. Hospital stays, doctor visits.

Q. Has Minnesota added questions regarding health insurance coverage?

A. Yes.

Q. And are public aid persons included in the survey?

A. Yes.

Q. Is one of the risk factors surveyed smoking?

A. Yes.

Q. And you said that we use data from this survey. For what years?

A. 1984 to 1994.

Q. And that works out to how many interviews that we used in the statistical model?

A. Thirty-four thousand.

*25 Q. Can you tell us how you obtained this data from the Minnesota Department of Health.

A. Well we went to them and asked them to put these data on -- on computer disks for us along with file layouts and other descriptive information, and they did so.

Q. And what did you do with the data?

A. Well we wrote computer programs that accessed it and summarized, for example, the percentage of smokers in Minnesota for different years and various other things from those surveys, and then we sent our results back to the Department of Health and had them write programs or use existing programs to produce the same numbers to make sure that we were matching up exactly with what we should in terms of summarizing these computer disks that -- on which were the results of each telephone interview.

Q. And what was the purpose of the matching up?

A. Well again, we wanted to be sure that we understood exactly what was in the data file and understood how to -- how to access it correctly and that our computer programs were correct.

Q. Who makes use of the Behavioral Risk Factor Surveillance System data?

A. Researchers interested in looking at the prevalence of risk factors or trends in risk factors or the relationship of risk factors to other health care payments.

Q. For what purpose?

A. Again, for example, it's one source of data to track the extent to which people have quit smoking in any state or in the U.S.

Q. And have you compiled a bibliography of articles that have made use of the Behavioral Risk Factor Surveillance System?

A. Yes.

Q. Can you turn to Trial Exhibit 26054, which should be in your testimony notebook. Do you have that?

A. Yes, I do.

Q. And can you tell us what that is.

A. This is just a bibliography of scientific references to the Behavioral Risk Factor Surveillance System.

Q. Who prepared it?

A. I did.

Q. How many articles are listed?

A. One hundred twelve.

Q. And how long is that bibliography?

A. Fifty-one pages.

Q. Is it a complete bibliography?

A. No, it's just an example.

Q. Dr. Wyant, what are the sources from Minnesota of the data that you used in the statistical model?

A. Well the three sources that you see up there on the board, the billing records, the Minnesota Behavioral Risk Factor Surveillance System, and the state of Minnesota and Blue Cross Blue Shield annual and quarterly reports.

Q. Do all of the estimates of smoking-attributable expenditures in plaintiffs' statistical model rely on these data sources?

A. Yes.

Q. Are there any that do not?

A. No.

Q. Now I want -- I want you to assume that counsel for the defense told the jury the following in its opening statement: "Plaintiffs' experts are going to come in here and tell you that you should find that the state has incurred 1.3 billion dollars in increased health-care costs, and Blue Cross 460 something million dollars as the result of smoking, and there's a very fundamental flaw in how they came up with that number. How they came up with the number, there is a fundamental flaw, and that flaw is they didn't look at the right people. They didn't look at -- when they calculated this number, they didn't look at Medicaid people, GAMC people, Blue Cross people in Minnesota. What they based their statistical estimate on was a National Medical Expenditure Survey. In fact that's what it's called, the National Medical Expenditure, not Minnesota, not Medicaid, not GAMC and not Blue Cross."

*26 Dr. Wyant, is that an accurate statement?

A. No, it's not an accurate statement.

Q. Why?

A. Because every one of our estimates uses information from these three sources, including, for example, the people such as the example person that I -- I put up here earlier in the claims records from GAMC.

Q. Dr. Wyant, how much time did you spend gathering and analyzing the Minnesota data shown on the magnetic board?

A. Over two years.

Q. Did you also use records from the National Medical Expenditure Survey?

A. Yes.

MR. HAMLIN: Your Honor, I'm not sure what Your Honor has in mind about lunch. We can break now or I can continue until 12:30. It's up to the court. We are moving into another area.

THE COURT: Okay. Well did you want to know what I was going to have for lunch?

(Laughter.)

MR. HAMLIN: That's okay. I think that would fall into the category of too much information.

THE COURT: All right. Why don't we recess, we'll reconvene at 1:30.

THE CLERK: Court stands in recess.

(Recess taken.)
 
 
 
 

*1 TITLE: STATE OF MINNESOTA AND BLUE CROSS AND BLUE SHIELD OF MINNESOTA, PLAINTIFFS, V. PHILIP MORRIS, INC., ET. AL., DEFENDANTS.

TOPIC: TRIAL TRANSCRIPT

TRANSCRIPT OF PROCEEDINGS

DOCKET-NUMBER: C1-94-8565

VENUE: Minnesota District Court, Second Judicial District, Ramsey County.

YEAR: February 26, 1998

P.M. Session
 

JUDGE: Hon. Judge Kenneth J. Fitzpatrick, Chief Judge
 

TEXT:
 

AFTERNOON SESSION.
 

THE CLERK: All rise. Court is again in session.

(Jury enters the courtroom.)

THE CLERK: Please be seated.

THE COURT: The record should show that Exhibit -- Trial Exhibit 11116 is received into evidence.

Counsel, go ahead.

MR. HAMLIN: Thank you, Your Honor.

BY MR. HAMLIN:

Q. Dr. Wyant, did you also use information from the National Medical Expenditure Survey in plaintiffs' refined statistical model?

A. Yes, we did. Yes, we did.

Q. I'm placing on the magnetic board another attachment and placing it in the data -- the last data box. Can you identify that attachment, please.

A. Yes. It says "National Medical Expenditure Survey" or NMES, N-M-E-S.

Q. What is the National Medical Expenditure Survey?

A. It's a survey, a national survey undertaken every 10 years by the Agency for Health Care Policy and Research, which was part of the U.S. Department of Health and Human Services. The particular survey that we used was the 1987 survey -- it's done every 10 years -- and that was a survey of approximately 35,000 people.

Q. Now is it a sample of the civilian non-institutionalized population --

A. Yes.

Q. -- of the United state?

A. Yes, it is.

Q. Now can you tell me what you understand the term "sample" to mean.

A. Well in this study they had people keep a diary and keep records of their medical expenditures, and every time they went to the doctor through the year -- and that would be quite an undertaking to ask everyone in the United States to do that for a year -- so what they do in this survey is take a representative sample of 35,000 people and have them do this for a year, and they're re-interviewed several times, and it's this representative sample that we're talking about when we say a survey.

Q. How many samples were taken from NMES in 1987?

A. Just the one of 35,000 for the civilian non- institutionalized population.

Q. Was the sample designed to provide a larger representation of population groups of special interest to the federal government?

A. Yes.

Q. And was one of those groups poor and low-income families?

A. That's correct.

Q. Was each family in the survey interviewed five times over a period of 16 months?

A. That's correct.

Q. Can you tell me whether, in your experience, this is a common practice?

*2 A. No, this is an unusually comprehensive survey. Most surveys, I would say the typical kind of survey that looks at this kind of information has usually just one interview.

Q. And did the interview obtain information about the family's health and health care during 1997?

A. Yes, it did.

Q. Did the survey obtain information on house -- household composition?

A. Yes.

Q. Employment?

A. Yes.

Q. Health insurance?

A. Yes.

Q. And was that information updated at each interview?

A. Yes, it was.

Q. Did the survey obtain information on illnesses, use of health services and health expenditures for each family member?

A. Yes.

Q. Was a supplement sent out between the first and second interviews on cigarette smoking?

A. Yes.

Q. Have you reviewed the format of any of these questionnaires?

A. Yes, I have.

Q. Now to verify and supplement the information provided by these households, were additional surveys done?

A. Yes, they were.

Q. And was one called a Medical Provider Survey?

A. Yes.

Q. And did that survey obtain information from physicians, hospitals, outpatient clinics, emergency rooms and home health agencies?

A. Yes.

Q. Was there also a Health Insurance Plan Survey?

A. Yes.

Q. Did that survey obtain information on the private insurance of persons in the household sample?

A. Yes.

Q. Did NMES obtain information on smoking?

A. Yes, it did.

Q. Did NMES obtain information on illness?

A. Yes.

Q. Did NMES obtain information on health-care services and health-care expenditures?

A. Yes.

Q. How much did it cost to do the NMES survey and the preparation of the results for distribution?

A. Tens of millions of dollars.

Q. Now you mentioned that the NMES survey is done every 10 years?

A. Yes.

Q. Have you been able to obtain information from the -- well let me ask you this: Is there a current NMES survey being done now?

A. Yes.

Q. Have you been able to obtain any information from that survey?

A. We've obtained the results of the first round of interviews from that survey.

Q. Were you able to make use of that information in the model?

A. No. It only became available within the last couple of months.

Q. Are the NMES results publicly available?

A. Yes.

Q. In what form?

A. You can get them on computer tapes or on CD-ROMs to put into your computer.

Q. And approximately how many records are on those CD-ROMs?

A. For the 1987 NMES, there were approximately a million records.

Q. And did you obtain those records?

A. Yes.

Q. Who uses the National Medical Expenditure Survey?

A. Researchers interested in looking at health insurance coverage, health costs, and various factors that might affect those kinds of things.

Q. And have you compiled a bibliography of articles that have made use of the NMES survey?

A. Yes, I have.

Q. Could you turn to Trial Exhibit 26053. Do you have that in front of you?

*3 A. Yes, I do.

Q. And is that the bibliography that you prepared?

A. Yes.

Q. How many articles are listed?

A. Eighty-four.

Q. And how many pages is the bibliography?

A. I'm not sure, but looks to be about 60.

Q. Is it a complete bibliography?

A. No.

Q. Why did you make use of the National Medical Expenditure Survey information in plaintiffs' statistical model?

A. Well it's a reliable source of data that has all of the key factors in one place; that is, it has information on smoking, it has information on disease, and it has information on health- care costs.

Q. And with that information, what can you do?

A. Well what we use the National Medical Expenditure Survey for is to calculate what Professor Zeger called the three reductions.

Q. In your prior work when you looked at one population, have you used information from another population?

A. Yes.

Q. Is that a common practice in biostatistics?

A. Yes.

Q. Dr. Wyant, could you list the steps taken to obtain a smoking- attributable expenditure in the refined model, and with the court's permission, would you list those steps on the flip chart.

THE COURT: Go ahead.

A. One step is to identify people with major smoking- attributable diseases, and for that we use Minnesota claims data. We identify their age, and again, that's the Minnesota claims data; their gender, again that comes from Minnesota claims data; we get the total health-care expenditures, again from Minnesota claims data; and then we can calculate the three reductions. For that we use a combination of National Medical Expenditure Survey and the Minnesota Behavioral Risk Factor Surveillance System.

Q. You recall in Dr. Zeger's description of the core model, he testified that the claims data was used in the third reduction.

A. That's correct.

Q. Could you explain why you don't use the Minnesota claims data for the third reduction in the refined model.

A. In the refined model, we want to take more factors into account, we want to do a better job of stratifying, as Dr. Zeger put it, to more -- do -- to more carefully compare smokers to non- smokers in similar groups, comparing likes to likes, and the best way to do that is to use information from the National Medical Expenditure Survey in the third reduction rather than the claims data because information on the additional factors is in the National Medical Expenditure Survey. But because we can also compute this third reduction using the claims data in Professor Zeger's core model, we can have a lot of confidence in the results given that we can do it one way or do it the other and make sure that, just choosing one source of the data, there will be other -- it's not having any unusual or worrisome effect on the results.

Q. Thank you.

Dr. Wyant, did you make any comparison of the Minnesota population and the population in the United States?

A. Yes, I did.

Q. And is one of those comparisons contained in Trial Exhibit 26056?

*4 A. Yes, it is.

Q. And could you identify that.

A. That's called "Table 3. Minnesota smoking rates compared to U.S., 1987."

Q. And this was prepared by you?

A. That's correct.

MR. HAMLIN: Your Honor, we offer Trial Exhibit 26056 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 26056 for illustrative purposes.

BY MR. HAMLIN:

Q. We've now put the exhibit on the overhead.

First of all, can you tell us again what the title is? And maybe we could focus in on the title.

A. That's -- it says "Table 3. Minnesota smoking rates compared to U.S., 1987."

Q. And can you describe for us what's in the exhibit.

A. Yeah. These -- this exhibit shows the percentage of people who are either current smokers or who were either current smokers or used to smoke, and compares Minnesota with the U.S. for the year 1987. This was prepared from the Behavioral Risk Factor Surveillance System data for Minnesota and the National Medical Expenditures Survey data for -- for the U.S.

And the first column is persons with health insurance through their employer or union, and those are people basically like those in the Blue Cross Blue Shield rated groups for whom we've looked at the claims data. And as we go down the rows, there are two groups of rows. The first rows are for current smokers, and then the next for current or former smokers. And in the current smokers group, for example, we see the first one, 24.8 percent, that's the percentage of adults in Minnesota in 1987 who reported that they were current smokers in the Behavioral Risk Factor Survey that the Department of Health conducted. And then below that is 26.5 percent, a percentage point or so higher, for the U.S. in 1987, and those are people who reported in the National -- not people, but adults who reported in the National Medical Expenditure Survey that they were current smokers. And that again is 26.5 percent.

And then if we move down to the next set of rows, we see the same kind of comparison, except this time we're looking at the percentage of people in Minnesota who, when they're called up by the Department of Health for the Behavioral Risk Factor Survey, say I'm either a current smoker or I used to smoke, and that was 51.9 percent. And the percentage of adults in the National Medical Expenditure Survey who said that they smoke or used to smoke in 1987 was 51.6 percent, almost identical.

If we go to the next column, persons covered by Medicaid, and again the source is the Behavioral Risk Factor Survey or the National Medical Expenditure Survey, the first set of rows we see 49.5 percent of persons in Minnesota covered by Medicaid saying that they were current smokers in 1987, and 41.1 percent for the U.S. And then again asking people if they smoke now or used to smoke, we go down -- in Minnesota when we were talking to people, again, the Department of Health was talking to people on public aid, 65.6 percent say yes, I either smoke or I used to, and in the U.S. in the National Medical Expenditure Survey, 57.5 percent of people said yes, either I used to smoke or I smoke now.

*5 The final column is for all adults 19 and older, and the similar figures there, 25.1 for Minnesota compared to 28 for the U.S., and 52.3 for Minnesota compared to 52.9 for the U.S.

Q. And what conclusions have you drawn from the data in this exhibit?

A. Well these smoking rates are very similar between --

Minnesota looks very much like the U.S. in terms of smoking rates.

Q. Did you also do a comparison of mortality rates for major diseases caused by smoking?

A. Yes.

Q. And what was the source or sources of your data?

A. The source of data for that comparison was the American Cancer Society's Cancer Prevention II study, CPS-II.

Q. And let me show you in your testimony notebook Trial Exhibit 15980. Do you have that?

A. Sorry, I'm --

Q. Yeah. I think it's the first one.

A. Yes.

Q. Can you identify that paper?

A. It's a paper called "Excess Mortality Among Cigarette Smokers: Changes in a 20-Year Interval."

MR. HAMLIN: And that -- for the record, Your Honor, Trial Exhibit 15980 is in evidence.

Q. Can you tell us the subject matter of this paper.

A. This talks about excess mortality among cigarette smokers and changes from the first American Cancer Society study, CPS-I, to the second one, CPS-II.

Q. And did you base your comparison of mortality rates on the data described in this article?

A. Yes, I did.

Q. And are the CPS-II study results also reported in the 1989 Surgeon General's report?

A. Yes.

Q. What other data sources did you use?

A. I used the U. S. Census data on the proportions by sex in the U.S.

Q. And can you tell me why you used that data?

A. Because in making this comparison we wanted to standardize to a similar age/sex distribution to make the comparison more relevant.

Q. What did you find?

MR. BIERSTEKER: Your Honor, I -- Your Honor, I object to any further questions along this line. This was the subject of a motion we filed. The analysis of mortality here in Minnesota was something that the witness did not disclose in his reports. We've not received the work product for it.

MR. HAMLIN: Your Honor, yes. In fact we've submitted an affidavit; that is, the witness submitted an affidavit January 10th in response to a motion. The expert reports by the defendants were then filed on January 15th. Following the expert reports the defendants had an opportunity to depose Dr. Wyant. They asked him questions about that affidavit. Moreover, their own expert is well- versed in CPS-II data and has testified about it extensively in other trials.

We have submitted papers on this. We believe it's well within the scope. There's absolutely no surprise and no prejudice whatsoever.

MR. BIERSTEKER: Your Honor, we have not had the underlying calculations. As you can tell from this witness's testimony so far today, these are not -- there's a lot of devil in details, and we're entitled to have those calculations to have a meaningful opportunity to cross-examine the witness. We did not get those. He filed a very cursory summary affidavit. We did not get the work product that Mr. Hamlin says underlies it. It would be unfair.

*6 THE COURT: You may answer the question.

THE WITNESS: I'm sorry? I'm sorry, Your Honor?

MR. HAMLIN: I didn't -- I'm sorry, I didn't hear that.

THE COURT: I said he may answer the question.

MR. HAMLIN: He may answer the question.

THE COURT: I suggest you reread it, however, it's been about five minutes.

MR. HAMLIN: Thank you, Your Honor.

BY MR. HAMLIN:

Q. Yeah. When you compared the mortality rates for smoking- attributable diseases in Minnesota to the mortality rates for smoking-attributable diseases in the United States, what did you find?

A. They were very similar. In Minnesota the chances of dying from one of the major smoking-attributable diseases was about twice as high for a smoker as a non-smoker, actually about 2.1 times as high. In the United States, the chances of dying from one of the major smoking-attributable diseases was again just about twice as high for smokers compared to non-smokers in the United States, it was actually 2.2 times as high as compared to 2.1 in Minnesota.

Q. Did you do any other comparisons of the Minnesota and United States populations?

A. Yes, I did.

Q. Let me direct your attention to Trial Exhibit 30173. I believe that's in your demonstrative notebook.

Do you have that?

A. Yes.

Q. Can you identify that exhibit, please.

A. An exhibit called "Comparison of plaintiff expenditures with national averages from NMES."

Q. And did you prepare this exhibit?

A. Yes, I did.

MR. HAMLIN: Your Honor, plaintiffs offer Trial Exhibit 30173 for illustrative purposes.

MR. BIERSTEKER: On that basis, no objection, Your Honor.

THE COURT: Court will receive 30173 for illustrative purposes.

MR. HAMLIN: Have that up, please.

BY MR. HAMLIN:

Q. Dr. Wyant, could you again tell us the title of this.

A. This is "Comparison of plaintiff expenditures with national averages from NMES."

Q. Could you take us through the exhibit, beginning with the bar charts at the top first. And perhaps we could focus on those.

A. The bar charts at the top compare information from Blue Cross Blue Shield of Minnesota from the claims records that we collected with information for people in the National Medical Expenditure Survey who said that I'm covered through my employer for my insurance; in other words, people who were like the people in Blue Cross rated groups. And the overall message from this chart is the yellow bars, which look at the U.S. based on NMES, and the blue bars, which come from Blue Cross Blue Shield Minnesota claims records, follow basically the same pattern as we go from left to right; in other words, where the yellow bars are high, the blue bars are also, and wherever the yellow bars are low, the blue bars are also.

And what this is looking at is the pattern of expenditures as reported in the National Medical Expenditure Survey and also from the claims records, and each set of bars is for one category of expenditures. The two bars at the left show prescription drugs, and then the next set of bars show ambulatory, which means doctors' visits, clinics visits, anything you walk in to basically. And then the next small set of bars there are for home health care, particularly for home health attendants helping someone at home, for example, who's unable to completely care for themselves. Inpatient hospital, it's basically staying overnight in the hospital. And then everything else is in the final pair of bars.

*7 So going back over to the left again, you see in the blue bars that when we go and spin through the claims records, the computerized billing data from Blue Cross for 1987, you'll see eight percent of the total expenditures going for prescription medications, and the comparable figure for the U.S. as a whole on the yellow bar is the seven points -- excuse me, 7.2 percent.

Then moving on to the outpatient visits, the ambulatory category, we see 40.9 from the Blue Cross claims records compared to 36.2 in the National Medical Expenditure Survey.

And moving on, for both surveys home health moves down close to the line.

And then inpatient at the highest category at 49.3 percent or 53.8 percent, that is most of the expenditures -- or just about half of the expenditures either in Blue Cross or in the U.S. for this group of people are for inpatient hospital stays. And then there's a little bit left over over on the right.

The chart just below that makes a similar comparison, except this time we're looking in the blue bars at claims records from Minnesota Medicaid, and on the yellow bars from the National Medical Expenditure Survey. So again we're comparing Minnesota to the U.S., but this time just the Medicaid programs.

Again, when you kind of take the bird's-eye view here and really -- the message in this chart is the blue bars follow a very similar pattern to the yellow bars, so if the blues are high, the yellows are high, and vice versa.

And again, we have same categories going left to right. I know that's hard to read there, but prescription medications on the left, and, for example, 7.7 percent of the expenditures of Minnesota Medicaid in that year were for prescription drugs, and 7.1 percent in the national sample. And moving across, the next bars are both at around -- between 25 and 30, and the next one's between five and ten, and then again inpatient being the biggest source of cost of a little over half of the expenditures going for inpatient care where they were talking about Minnesota Medicaid or the U.S. as a whole based on the National Medical Expenditure Survey.

Q. What conclusions have you drawn from this expenditure pattern comparing Blue Cross and the state of Minnesota to the National Medical Expenditure Survey?

A. Well again, as for the previous comparisons, Blue Cross is very similar to the nation as a whole when we're looking at a comparable population of people.

Q. And what about the state of Minnesota; that is, Medicaid?

A. The exact same thing, Minnesota Medicaid looks very much like U.S. Medicaid.

Q. Can you use this chart to compare expenditures for Medicaid and private insurance?

A. Yes. You can see comparing the chart on top and the chart on the bottom, they basically follow the same pattern, inpatient being number one, and then the outpatient visits and ambulatory being number two. The one difference that's noticeable is the home health in the bottom chart, that's the middle set of bars, those are higher for Medicaid than they are for Blue Cross, and that reflects the fact that there are -- there is a fairly sizable disabled population on Medicaid, and the Medicaid program, those try to care for a lot of those people at home rather than having them be in an institution of some kind.

*8 Q. Now in plaintiffs' statistical model, what if anything have you done to take into account any differences between the national population and the population in Minnesota?

A. Well in the statistical model we use the Behavioral Risk Factor Survey data from Minnesota. Although Minnesota, as we can see, is very similar to the U.S., it's not identical, and the small differences might make some difference in the smoking-attributable expenditure rates in Minnesota. So to would ensure that we've taken that into account as far as possible, we've used the Behavioral Risk Factor Survey data from Minnesota in our -- in our smoking- attributable expenditure models.

Q. Once again I'm placing on the easel Trial Exhibit 30182, which is the magnetic board of the statistical model. We've now discussed the data sources, but let me direct your attention to the schematic that is listed as "Statistical Analysis." What does that signify?

A. Well that is where our statistical model is, and that's where we apply what Professor Zeger called the three reductions to the total dollars from the claims data that are coming in to that circle from the box on the upper right.

Q. Can we have Trial Exhibit 30197 on the overhead.

MR. HAMLIN: That has already been introduced into evidence, Your Honor.

Q. Dr. Wyant, can you identify the exhibit that is currently on the overhead.

A. Yeah. An exhibit called "The Core Statistical Model, Three Reductions."

Q. Now are the reductions listed in this exhibit also used in the refined model?

A. Yes. They're at the heart or the core of the refined model.

Q. Could you go over the first reduction, please. Tell us what that means.

A. The first reduction is what percentage are smokers; that is, when we get the claims records from Minnesota, we don't have smoking information, and so one of the first reductions we apply is to adjust for that fact. We take account of the fact that only a certain percentage of those claims records are for smokers, even if it's for a disease like lung cancer.

Q. Could you go through the second reduction listed on the exhibit.

A. Again, and taking lung cancer as an example, when we bring in in the refined model claims records from among the 90,000 people in Minnesota with major diseases who have lung cancer, we know that some of those people would likely have gotten lung cancer anyway even if they hadn't smoked, and so again we reduce the dollars being spent on people with lung cancer to account for that fact, that there's a certain percentage that would have gotten it anyway, and those don't go into the smoking-attributable dollars.

Q. Could you take us through the third and final reduction.

A. Well the third reduction, again, is addressing the same thing in the refined model as it did in the core model, and that's what dollar percentage is attributable to disease caused by smoking. And that's addressing the fact that people with lung cancer, even if they're smokers and even if their lung cancer was caused by smoking, they would have had a certain amount of expenditures in all likelihood anyway, they would have been spending some money on other diseases and conditions, and so we want to subtract that baseline level of -- of health-care expenditures out from whatever goes into the smoking- attributable dollars for lung cancer.

*9 Q. I want to ask you a few questions about this -- about the third reduction, and what we have now on the flip chart is your GMAC claims record that you prepared this morning. Do you see that?

A. Yes.

Q. Now I want to direct your attention to the entry for leg and knee sprain. Do you see that?

A. Yes.

Q. The expenditure there is a hundred dollars?

A. Yes.

Q. Does the statistical model count all of those dollars as smoking- attributable expenditures?

A. No. The whole purpose of the third reduction is to adjust for exactly those kinds of dollars. The only way that any of those dollars go into smoking- attributable expenditures is if smokers spend more for that kind of injury than non-smokers, than similar non-smokers, once we've stratified and done the other statistical methods that Dr. Zeger has talked about.

Q. Did the refined model consider factors not considered in the core model?

A. Yes.

Q. And do you have an exhibit that compares the factors considered in the core model and in the refined model?

A. Yes.

Q. Is that Trial Exhibit 30188?

A. Yes, it is.

Q. Was that exhibit prepared by you?

A. Yes.

MR. HAMLIN: Your Honor, plaintiffs offer Trial Exhibit 30188 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 30188 for illustrative purposes.

BY MR. HAMLIN:

Q. Dr. Wyant, I've placed Trial Exhibit 30188 on the easel. Could you first of all tell us what this exhibit is.

A. This is a comparison of the core model to the refined model, and in particular it's a comparison with regard to the factors considered in the two models.

Q. Could you go through the column marked "Core" and tell us what that refers to.

A. Well the core model that Professor Zeger talked about takes into account smoking, it takes into account gender, age group, 35 to 64 or 65 plus, and disease, those were the major smoking-attributable diseases, lung cancer, COPD or CHD/stroke, and public or private health coverage.

Q. Would you go to the column marked "Refined" and tell us what that is.

A. The refined model, the first factors are also identical, smoking and gender. Age is a little different because in the refined model you take into account actual age in addition to age group; that is, on top of looking at people aged 35 to 64, we also look at people aged 37 or aged 32 and take the exact age into account in the model.

Disease is also a factor in the refined model, as is public or private health coverage. That would relate to Blue Cross Blue Shield, for example, versus the state programs.

Type of medical service, now we're into factors that are considered by the refined model but not the core model. "Type of medical service" relates to the different types that we saw in the bar chart a minute ago, prescription drugs, hospitals, outpatient visits, and home health care.

Income group, categories of people's income, those are put in categories by the federal government in terms of household size and level of income. Those groups are taken into account in the refined model.

*10 Marital status, single, married, widowed, separated, divorced.

Education level.

Seatbelt use, which is in the refined model, it's a standard way in these kinds of studies to look at risky behavior.

Overweight is a ratio of weight to height.

And region of a country.

Q. Why did you take into account these additional factors in the refined model?

A. Well as both Professor Zeger and Dr. Samet have said, one goal of these kinds of studies and statistical models is to compare like to like, to stratify so that we're isolating the effect of smoking comparing smokers to similar non- smokers, and using these additional factors, we can do a much finer stratification and a much better job of comparing like to like.

Q. How do you decide --

How did you decide what variables to stratify at all?

A. Dr. Zeger, myself, Dr. Miller, met together and met with Dr. Samet, got extensive medical input from Dr. Samet on the general medical issues. He reviewed from all of our perspectives what would typically be in a study like that, what would be useful, and in particular taking into account the quality and extent of the information that was in the National Medical Expenditure Survey, and as a collaborative -- collaborative effort, came to a decision on exactly what should be there.

Q. Did that collaboration include a discussion of prior scientific knowledge about smoking and disease?

A. Oh, yes.

Q. Now is it possible to stratify too much?

A. Oh, yes.

Q. Now what does that mean, stratifying too much?

A. Well it's not the case that you can make a statistical model better simply by adding more and more factors. At some point, basically, you hit diminishing returns. And in fact it's worse than that, overstratifying can produce I believe what Dr. Samet called overadjustment where in fact you may be masking some of the things you are trying to identify or isolate or measure from the model rather than improving your comparisons like to like groups in the stratification.

Q. So you had discussions with Drs. Samet, Zeger and Miller about the factors to be included in plaintiffs' statistical model?

A. Yes. We had general background medical discussions with Dr. Samet, and specific discussions between Dr. Miller, Dr. Zeger and myself on the specific factors to be included in this refined model.

Q. Did you discuss whether alcohol should be included?

A. Yes.

Q. Did you discuss whether exercise should be included?

A. Yes.

Q. Did you do your own investigation with respect to these two factors?

A. Yes. We did an investigation involving Minnesota Behavioral Risk Factor Survey data, which had information on exercise, alcohol, smoking, and use of hospital services or outpatient visits, and we found that information on exercise or alcohol had no material effect on the relationship of smoking to use of these medical services, which was entirely consistent with what we understood to be the bulk of the information in published studies as well.

*11 Q. Dr. Wyant, what other factors need to be taken into account in a model such as this one?

A. I don't think any other factors need to be taken into account in this model. Part of that comes from discussion with Dr. Samet. At the conclusion of creating this model, he told us that he didn't believe that adding any other factors would make any material difference in the smoking-attributable expenditure estimates, and that was entirely consistent with numerous other studies that we had looked at as well as the statements by the Surgeon General of the United States.

Q. Let me direct your attention now to Trial Exhibit 3821, which is a portion of the 1989 Surgeon General's report. It's in your demonstrative notebook, the one you just had, the small book. That's Trial Exhibit 3821. It's the first one.

Do you have that exhibit?

A. Yes, I do.

Q. Okay. And is that an excerpt from the 1989 Surgeon General's report?

A. Yes.

Q. And specifically an excerpt from page 129. Do you see that?

A. Yes.

MR. HAMLIN: For the record, Your Honor, that exhibit has been admitted into evidence.

Q. And can you read for us the sentence that is highlighted on the overhead.

A. "As discussed in Chapter 2, numerous attempts to control statistically for confounding and stratifying variables have not materially altered the estimated relative risks for cigarette- related diseases."

Q. This was the conclusion of the Surgeon General in 1989?

A. That's correct.

Q. Now what does that sentence mean in regards to developing the factors to be included in plaintiffs' statistical model?

A. Again, a part of these three reductions, a significant part of the three reductions and that related to disease has to do with relative risk of smokers versus non-smokers, so this sentence is consistent with the notion as expressed by Dr. Samet that controlling for additional factors in our model would not be likely to materially affect our estimate of smoking-attributable expenditures.

Q. Now this particular sentence dealt with mortality; correct?

A. I believe that was the thrust of most of the studies to which it was referring; that is, the studies that were summarized and assembled for the Surgeon General's report.

Q. What if anything can those mortality studies tell you about the cost of disease?

A. Well one would expect, as is stated in a number of places in the scientific literature, that although the costs of diseases are not going to perfectly track mortality, there's likely to be some correlation. For example, major smoking-attributable diseases, the typical course is that there will be a lot of expenditures after the disease is diagnosed, and if the person dies from those diseases, significantly elevated expenditures just before death, so that the expenditures in a rough way track mortality.

Q. As part of your review of the literature that you just discussed, did you also review the nurses health study?

A. Yes.

*12 Q. Can you turn to Trial Exhibit 16039, which is in your testimony notebook. I believe it's the second one.

A. Yes.

Q. Do you see that?

MR. BIERSTEKER: Objection, Your Honor, there is no foundation at this juncture.

MR. HAMLIN: We can take that off a minute.

BY MR. HAMLIN:

Q. Can you identify that?

A. This is an article on relative and absolute excess risk of coronary heart disease among women who smoke cigarettes.

Q. And does it make any reference to the nurses health study?

A. Yes, it does.

Q. All right. And is it a report on the nurses health study?

A. Yes.

Q. And are you also familiar with a board prepared by Dr. Samet regarding the results of the nurses health study?

A. Yes.

Q. Is that Trial Exhibit 30166?

A. Yes.

Q. I'm now placing on the easel Trial Exhibit 30166, which has been previously admitted into evidence, and can you tell us the title of that exhibit.

A. The title of that exhibit is "Nurses Health Study."

Q. And can you describe for us, first, the left-hand column marked "Design." Tell us what the nurses health study is about.

A. That left-hand column describes the nurses health study. There were 120,000 nurses enrolled in 1976 into the study and were followed for six years until 1982. They were aged 30 to 55 at the time of their enrollment in 1976. The study was looking at coronary heart disease and cancer. There were biennial questionnaires and medical record reviews, and this study was looking at both the incidence of disease and mortality.

Q. What do you mean by "the incidence of disease?"

A. Whether or not a person was sick from coronary heart disease or cancer.

Q. Can you now turn to the second column marked "Results" and describe that for us.

A. On the right-hand side, the results addressing fatal coronary -- excuse me, fatal coronary heart disease and non-fatal heart attacks, myocardial infarctions, it's looking at the current smokers versus never smokers.

May I step up to the board?

Q. With the court's permission, sir.

A. In these columns we're looking at risk estimates, in particular the relative risks of either dying from one of these coronary diseases or getting a non-fatal heart attack.

Q. Dr. Wyant, maybe we could --

A. Excuse me.

Q. Yeah. Thank you.

A. In this column we see cigarettes smoked per day, so we have a baseline here for the non-smokers, no cigarettes per day, and the risk estimate is one. For one to 14 cigarettes a day it's more than twice the risk of either getting one of these fatal heart attacks or non-fatal. And as we go up to 15 to 24 cigarettes, there's almost four times the risk. And at 25 plus it's about six times. That is, smokers, say, who smoke one to 14 cigarettes a day are about twice as likely to have one of these conditions, either dying of coronary heart disease or having a non-fatal myocardial infarction or heart attack.

Now there's two columns of risks here and they're very similar, both about two here, both about four here and both about six here. And the second column -- this is another statistical model. The second column takes -- this is from a model that takes into account not only age, these are all adjusted for age differences, but also time period, Quetelet's index, which is overweight, menopausal status, hormone replacement therapy, family history of heart attack, personal history of diabetes, personal history of hypertension and personal history of high cholesterol, and adjusting for all of these factors, including all of those in the statistical model, it really didn't change these relative risks appreciably, and in fact including them in the model actually increased slightly the estimated relative risk due to smoking as we go from this column to this column.

*13 Q. Now what does this tell you about developing plaintiffs' statistical model with respect to factors?

A. Again, this is one of many studies that are consistent with the Surgeon General's statement and Dr. Samet's statements to us that including additional factors beyond those we already included would not be likely to have a material fact on our smoking-attributable expenditure estimates.

Q. In fact in the nurses health study when those additional factors were included, the relative risk increased; right?

A. That's correct.

Q. I've once again --

I've once again placed on the easel Trial Exhibit 30188, which is entitled "Comparison of Core Model to Refined Model, Factors Considered."

Dr. Wyant, were you in this court on Tuesday when Dr. Zeger was testifying?

A. Yes, I was.

Q. And do you recall that defendants' counsel asked him whether he had taken certain factors into account in the simple or core model?

A. Yes, I do.

Q. Let me go through some of them and ask you whether they were included in the refined model.

Was marital status included in the refined model?

A. Yes.

Q. Was race included in the refined model?

A. Yes.

Q. Was education included in the refined model?

A. Yes.

Q. Was income included in the refined model?

A. Yes.

Q. Was other risky behaviors included in the refined model?

A. Yes.

Q. Does the epidemiology provided by Dr. Samet and your own review of the Surgeon General's reports and other epidemiology indicate to you that you needed to take into account in plaintiffs' model any of the other factors that defendants' counsel listed, such as place of birth and family history?

MR. BIERSTEKER: Objection, Your Honor, leading.

THE COURT: It is leading.

MR. HAMLIN: I'm sorry?

THE COURT: It's leading.

MR. HAMLIN: All right.

BY MR. HAMLIN:

Q. Dr. Wyant, what did the epidemiology provided by Dr. Samet and your own review of the Surgeon General's report and the epidemiology indicate to you --

A. With respect to --

MR. BIERSTEKER: Objection.

A. I'm sorry, were you finished?

Q. -- with respect to factors to be included in the statistical model?

MR. BIERSTEKER: Objection, Your Honor, I think it's beyond the scope of the witness's expertise, and he's not a medical doctor, he's not an epidemiologist.

THE COURT: He may rely on Dr. Samet, however.

You may answer.

THE WITNESS: I'm sorry, could I have the question read back?

(Record read by the court reporter.)

THE COURT: Excuse me. I think that is an objectionable question with respect to "other epidemiology."

MR. HAMLIN: Okay.

THE COURT: You'll have to rephrase the question.

MR. HAMLIN: I will, Your Honor.

BY MR. HAMLIN:

Q. Dr. Wyant, what did the epidemiology provided by Dr. Samet indicate to you regarding what you needed to take into account in plaintiffs' statistical model with respect to factors?

A. It indicated to me that beyond the factors listed there, that we did not need to include other factors in the model as there was not likely to be any material effect on the smoking- attributable expenditure estimates.

*14 Q. Why not?

A. Well although some of these factors, I'm sure, are useful for medical doctors and individual patients, that doesn't mean that they are useful in population studies that are focusing on the relationship of smoking to disease or smoking to the use of health- care services.

Q. Dr. Wyant, with the permission of the court, could you come down and place on the flip chart the structure of the disease model.

A. The basis of the model was laid out by Dr. Samet of smoking causing disease and resulting in health-care costs. We looked at two general categories of health-care costs, medical services, that would be hospitals, doctor visits, prescription drugs, home health care and nursing home fees, and those are the monthly fees to maintain someone in a nursing home, board and room and daily care.

Under medical services we first look at major smoking- attributable disease, and then we look at diminished health. Within major smoking- attributable disease we have the group that Dr. Zeger talked about, lung cancer/COPD, and we had a second group of the other major diseases that we called CHD/stroke.

The diminished health, we also have two groups, we have diminished health that's captured by reported health status, and we have diminished health that's not captured by that called other diminished health effects.

Q. Now Dr. Wyant, has the Surgeon General of the United States made comments regarding conditions described by Dr. Samet as falling under the diminished health category?

A. Yes.

Q. We have on the overhead Trial Exhibit 20179, which has been previously admitted into evidence, specifically page 42.

Now can you see the title of that exhibit?

A. It says "Table 2. Statements from U.S. Surgeon General's reports and the International Agency for Research on Cancer concerning smoking as a cause of the listed diseases."

Q. Is it your understanding this table was prepared by Dr. Samet?

A. That's correct.

Q. Let me direct your attention to the portion of the table marked 1967, specifically the second entry. Do you see that?

A. Yes.

Q. Now on the left-hand side we see "Disease;" right?

A. Yes.

Q. Now what -- what is under "Disease?"

A. That is "Diminished health status/respiratory morbidity."

Q. And then the model column is marked "Statement"?

A. Yes.

Q. And on the right-hand side is "Surgeon General's report," see that?

A. That's correct.

Q. And what's your understanding of what's in that column?

A. Those are dates of the different Surgeon General's reports.

Q. Okay. Let me direct your attention now to the second entry for 1967. Could you read that for us.

A. Sure. "Cigarette smokers have higher rates of disability than non- smokers, whether measured by days lost from work among the employed population, by days spent ill in bed, or by the most general measure - days of 'restricted activity' due to illness or injury."

Q. Let me direct your attention now to page 43, the next page. Again I want to direct your attention to the entry marked "Diminished health status/respiratory morbidity." Do you see that?

*15 A. Yes.

Q. The first entry there is from the 1975 Surgeon General's report; is that right?

A. Yes.

Q. Can you read that for us.

A. "In addition to an increased risk of COPD, cigarette smokers are more frequently subject to and require longer convalescence from other respiratory infections than non-smokers. Also, if they require surgery, they are more likely to develop possible operative respiratory complications."

Q. Let me direct your attention now to the entry from the 1979 Surgeon General's report. It's the second entry that begins "A wide variety...." Do you see that?

A. Yes.

Q. Can you read that.

A. "A wide variety of alterations in the immune system have been observed due to cigarette smoking."

Q. Now if we turn back to page 42, now if you look at this page, Dr. Wyant, that was prepared by Dr. Samet, could you identify for us the years in which the Surgeon General made comments regarding the conditions identified by Dr. Samet as falling in the diminished health status category.

A. Right here we see the years 1964, '67, '71 and '72, '75.

Q. Then if we could turn to the next page.

A. '79.

Q. And then finally the third page.

A. '80, '84, '90.

Q. And have you reviewed the diminished health status summary table prepared by Dr. Samet?

A. Yes, I have.

Q. And in fact that's Trial Exhibit 30109. I've placed your demonstrative notebook right in front of you.

A. That's correct.

Q. Okay. How many studies are listed in this table?

A. I believe there are 99 studies.

Q. And how many pages is the table?

A. Twenty-eight pages.

Q. Now let's go back to the structure of the disease model. Can you direct your attention to the diminished health portion and tell me what you mean by reported health status.

A. In the National Medical Expenditure Survey at one time in a year -- in fact it was the same survey in which we asked for whether or not a person smoked, whether they were a current or former smoker, and they also asked to assess -- asked people to assess their health as excellent, good, fair or poor, and these self-assessed health status questions are widely used and a generally reliable method of looking at general health status and are frequently used in the scientific literature for that purpose.

Q. Could you describe for us the other category under diminished health which you've identified as other diminished health effects.

A. The people in the National Medical Expenditure Survey were followed for a year, and they were only asked this question at one point during the year, so it is likely that not all of the diminished health associated with smoking is captured by this particular question. In fact it's almost certain that it is not. In addition, there are other aspects of diminished health, such as complications of other conditions due to a background smoking- related disease or condition that may not be reported as poor or fair health, and all the remaining health effects of diminished health for smokers compared to non- smokers that aren't captured by this reported health status are included in this second category under diminished health.

*16 Q. Thank you. You can return to the stand. And you should probably take your demonstrative notebook with you.

Dr. Wyant, I want to refer you to the chart of the structure of the model which you have just placed on the easel, and I want to refer you to the major smoking-attributable portion. See that?

A. Yes.

Q. Who is in that category?

A. When we do the refined model, we first go to the claims records, and if a person, after --

Well the first thing we look for in the claims records is one of the ICD-9 codes indicating one of the major smoking- attributable diseases, and then we apply the validity screens that we talked about before lunch. If after that application of validity screens the person is still considered as having a smoking- attributable disease, a major smoking-attributable disease, then the dollars for that person are put into either the lung cancer/COPD category or the CHD/stroke category, and that person for that year of being treated is a part of major smoking- attributable disease. And that's true for all of the 90,000 people that we've identified as being treated for such diseases.

Q. Let me direct your attention to the CHD/stroke category. Does that category refer to more diseases than just CHD; that is, coronary heart disease and stroke?

A. Yes, it does.

Q. Could we have Trial Exhibit 30153 on the overhead, please.

Could you identify, first of all, what that exhibit is.

A. Those are the diseases identified by Dr. Samet as caused by smoking. The final line, diminished health status, is just some examples of what goes into the pervasive effects of diminished health which, as Dr. Samet has said, cannot really be captured by some set of ICD-9 codes.

The rest of the codes are what we call the major diseases. Going down fourth from the top, there's chronic obstructive pulmonary disease, and down a little below that is lung cancer, and those two categories combined make up the lung cancer/COPD part of major smoking-attributable disease. All the other diseases on that list are what we call CHD/stroke.

Q. So they would include the remainder of the diseases, with the exception of diminished health status?

A. That's correct.

Q. All right. Let me direct your attention again to the chart of the structure of the disease model and to diminished health. Who is in that category?

A. If a person is not in a major smoking-attributable disease category, then we address their expenditures in the diminished health category. So it's basically everyone else is addressed in diminished health, and the three reductions appropriate to that group were applied to them.

Q. Are all the conditions and diseases of diminished health listed in the ICD-9 codes on Trial Exhibit 30153, which is on the overhead?

A. No, those are just examples of some of the ICD-9 codes that might be considered part of diminished health.

Q. And who are you relying on for that opinion?

A. Dr. Samet.

Q. What are the data sources for each of the categories of the disease model that are listed on the chart? If you could begin with major smoking- attributable disease.

*17 A. Well the major smoking-attributable disease, we've used the Minnesota claims data, we've used the Behavioral Risk Factor Survey, we used the National Medical Expenditure Survey.

Q. And is that --

You used that information for both lung cancer/COPD and the remainder of the smoking-attributable diseases?

A. That's correct.

Q. What data sources are used for the diminished health category?

A. Diminished health, again I used Minnesota claims data, Minnesota Behavioral Risk Factor Survey, and the National Medical Expenditure Survey.

Q. Now there's an additional category of nursing home fees. What kinds of expenditures are included in that category?

A. Those are what we call maintenance fees. They're not costs for going to the doctor or going to the hospital, they're the monthly charges when a person is in a nursing home that provide for their -- their room and board and their daily care.

Q. What are the sources of data for that portion of the model?

A. Again the Minnesota claims data, the Minnesota Behavioral Risk Factor Survey. But in nursing homes we do not use the National Medical Expenditure Survey, we use another federal survey called the National Health and Nutrition Evaluation Survey, or NHANES.

Q. And can you tell us about NHANES?

A. NHANES was a study done by the National Center for Health Statistics in which baseline health characteristics were collected on some 40,000 people in the early 1970s, early to mid-'70s, and then a subset of those people were followed up and recontacted in 1982 and in, I think, three other times in the 1980s, with the last time being in 1992.

Q. And does NHANES include smoking information?

A. Yes, it does.

Q. What kind of information?

A. It's the standard question that appears on the Behavioral Risk Survey and the National Medical Expenditure Survey: Have you smoked more than a hundred cigarettes in your life?"

Q. Does NHANES include disease information?

A. Yes, it does.

Q. What kind?

A. It's coded with ICD-9 codes, and it relates to diseases that are recorded either at hospital admissions or discharges or on admission to facilities such as nursing homes.

Q. Have you reviewed the NHANES data?

A. Yes.

Q. And what if anything did you find about whether people entering nursing homes had diseases caused by smoking?

A. They did have diseases caused by smoking.

Q. I'm going to place the magnetic board once again on the easel. I'll take down the National Medical Expenditure Survey attachment and put up a new attachment. Can you tell us what that is?

A. That attachmentment says "National Health And Nutrition Examination Survey, NHANES, N-H-A-N-E-S.

Q. And that's the survey that we've just been discussing; right?

A. That's correct.

Q. Who uses that information?

A. Researchers who are interested in studying kinds of health issues, which data were collected at the baseline survey or follow-ups, as to what the results are for different people in that survey as they're tracked from 1970s through 1992.

*18 Q. Can you direct --

Well, have you prepared a bibliography of some of the articles and papers that have used NHANES?

A. Yes, I have.

Q. Could you direct your attention to Trial Exhibit 26055. Do you have that?

A. Yes, I do.

Q. Could you identify it, please.

A. That's called "Attachment G, Scientific References to the NHANES Epidemiologic Followup Study."

Q. And is that bibliography prepared by you?

A. That's correct.

Q. And approximately how many articles does it contain?

A. Forty-nine.

Q. And how many pages are in the bibliography?

A. Thirty-six pages.

Q. Is that a complete bibliography of all of the articles that have used NHANES?

A. No. It's only a sample.

Q. How does the smoking-attributable expenditure arise in the nursing home setting of the model?

A. Well I believe, as Dr. Samet testified, just as an example, smoking causes stroke, it's a cause of stroke, and if someone gets a stroke because of smoking and the stroke is severe enough that the person has to be cared for in a nursing home, that's a way that smoking-attributable expenditures can arise in the form of nursing home fees.

Q. What if any differences are there in the methodology regarding the nursing home portion of the model and the disease portion of the model?

A. In the nursing home portion of the model there are only two reductions, not three reductions. In the medical portion of the model, the third reduction was to take account of normal baseline medical expenditures, expenditures that would have occurred anyway. In a nursing home situation you don't assume that there's a normal background level of nursing home expenditures, that someone is typically in a nursing home for a week or so a year up till they're admitted as a permanent resident. And so if a person goes into the nursing home -- and we measure this in our model -- as a result of smoking-attributable conditions, then all of those nursing home fees due to that admission are considered part of these smoking- attributable expenditures.

Q. So you don't need a third reduction?

A. No.

Q. Do you know of any scientific papers that address smoking and nursing home entry?

A. Yes.

Q. Let me direct your attention to Trial Exhibit 18943. I believe that's in your testimony book. Do you have that in front of you, Dr. Wyant?

A. Yes, I do.

Q. Can you identify that paper, please.

A. That's a paper called "High-risk Profiles for Nursing Home Admission" by Alan Jette, Ph.D., and others, from Gerontologist from 1992.

Q. And have you relied, in part, on this paper for your opinions in this case?

A. Yes.

Q. Do you consider this paper to be a reliable authority in the published scientific literature?

A. Yes.

MR. HAMLIN: Your Honor, we offer Trial Exhibit 18943 as a learned treatise.

MR. BIERSTEKER: On that basis, no objection, Your Honor.

THE COURT: Court will receive 18943.

BY MR. HAMLIN:

Q. Dr. Wyant, could you summarize for us the study in the Exhibit that we've just identified regarding high-risk profiles for nursing home admission.

*19 A. A very major factor for nursing home admission is neurological conditions, including stroke, but in addition to stroke there's an elevated risk found for nursing home admission for smokers, particularly male smokers, having approximately two to three times the probability of a nursing home admission as a non- smoker.

Q. And are those the findings and results of this paper?

A. Yes.

Q. Now have you examined the National Medical Expenditures Survey with respect to nursing home entry rates?

A. Yes.

Q. And what have you found?

A. The National Medical Expenditure Survey was a survey of people outside of nursing homes, but it does track where people go to if they leave a survey mid-year, and some of those people leave to go into nursing homes. And if you look at the number of people who go into nursing homes, although it's a small sample, you find that smokers in nursing homes are greater than non-smokers.

MR. BIERSTEKER: Objection, Your Honor. Again this is work product that was not disclosed to us. It's not mentioned in his report. It's really way beyond the scope of what we anticipated this witness was going to talk about.

MR. HAMLIN: Your Honor, this is well within the scope. Defendants have deposed Dr. Wyant three times; they've had plenty of opportunities and have asked about his conclusions regarding nursing homes.

THE COURT: You may answer the question. Did you finish your answer?

THE WITNESS: Yes, I did.

THE COURT: Thank you. The answer will stand.

BY MR. HAMLIN:

Q. Dr. Wyant, what if any statistical methods did you use to estimate the health-care costs paid by the state of Minnesota and Blue Cross to treat diseases and conditions caused by smoking?

A. In our model we used the method of stratification that Professor Zeger talked about, we used methods of maximum likelihood, which are standard statistical methods, and among those methods are regression analysis and probit regression analysis, and we used a method of attributable risk, just a standard method in biostatistics and epidemiology.

Q. We've talked about stratification. Could you explain briefly what you mean by "maximum likelihood."

A. Maximum likelihood is a statistical method for finding the best formulas to summarize a set of data. Our model focuses on the three reductions, and as you may remember, part of Professor Zeger's testimony had to do with expressing statistical models as a formula to bring in more factors such as the additional factors in the refined model. And maximum likelihood is a standard method for finding the best formulas for fitting a particular set of data, such as the data we've collected here from those various sources.

Q. You also mentioned regression analysis. Can you tell us what that is?

A. That's in most instances a form of maximum likelihood estimation, and it's a particular form that's used in finding a formula that best explains or measures a quantity such as dollars, number of dollars of health-care costs for example.

*20 Q. And I believe you mentioned that there were several kinds of regression analyses, including probit analysis?

A. Probit analysis.

Q. What is that?

A. That's a -- basically a form of regression analysis where instead of developing a formula for dollars like health-care costs, what you're developing a formula for is a probability, such as the probability that someone will go into the hospital at any time during the year. And that probability may be related to a number of factors, including smoking, and a probit analysis is a standard statistical method for calculating such formulas.

Q. You also mentioned attributable risk. Can you explain what that is.

A. Attributable risk is really what's captured in the first two reductions. In the refined model those reductions have additional adjustments based on other factors, but what attributable risk is is just the standard means in epidemiology of looking at excess cases of disease due to some factor such as smoking.

Q. And is that also a commonly accepted statistical method?

A. Yes.

Q. So the methods that you have identified -- stratification, maximum likelihood, regression analysis, and attributable risk -- are all commonly accepted statistical methods?

A. That's correct.

Q. Dr. Wyant, with the court's permission, could you put one of the equations from the refined model on the flip chart?

THE COURT: Maybe we should take a short recess.

MR. HAMLIN: Okay.

THE CLERK: Court stands in recess.

(Recess taken.)

THE CLERK: All rise. Court is again in session.

(Jury enters the courtroom.)

THE CLERK: Please be seated.

THE COURT: Counsel.

MR. HAMLIN: Thank you, Your Honor.

BY MR. HAMLIN:

Q. Dr. Wyant, I've asked you to place on the flip chart an example of one of the formulas used in plaintiffs' statistical model, and have you done that?

A. Yes, I have.

Q. You did that during the break?

A. Yes.

Q. Could you briefly describe for us what that formula means.

A. This formula is basically a top-level view of the application of the three reductions in the model for major smoking- attributable diseases.

Q. Now were there other formulas used in the statistical model?

A. Yes, a great number of them.

Q. Court will be pleased to know we're not going to go over any other formulas.

Did you use computers in this project?

A. Oh, yes.

Q. And can you tell me about that.

A. Well to assemble the state billing records, what we would typically do is on the state's computer, it's their big computer, takes up a room over at Intertech and runs -- basically runs the state's systems during the day, and we would run at night. And, for example, I think we ran for 250 nights in 1996 to sort and compile the billing information we're talking about here. I think at any given time we had approximately 2,000 data tapes on the tape robots there; each one of those tapes is equivalent to about 600 of those diskettes that most of you have probably seen, computer diskettes.

*21 Q. Can you speak up?

A. Sorry. Each one of those tapes is equivalent to about 600 computer diskettes. There were hundreds of thousands of lines of computer programs involved in doing this. And that was -- basically a parallel kind of effort went on over at Blue Cross as well.

Q. Now when you say "lines of computer programs," what do you mean?

A. Computer programs are written, as some of you may know, on a terminal or a computer screen, and they're lines of instructions for the computer to do different steps in sequence to accomplish sorting and moving claims records around to match them up where they need to be matched up, and so each one of those lines, when we count them up over all the programs that are being used to process these years of data, adds up to well over a hundred thousand lines of code.

Q. Dr. Wyant, I've placed before you three computer disks marked as Trial Exhibit 14936. Do you have those?

A. Yes.

Q. And could you identify those.

A. These were disks prepared by me under -- under my direction.

Q. And do they contain the estimates of smoking-attributable expenditures for the state and Blue Cross Blue Shield of Minnesota for the years 1978 through 1996?

A. Yes, they do.

Q. And were these disks prepared under your direction?

A. Yes.

Q. Do the disks contain computer programs and formulas derived from the data sources which we have described; specifically, the Minnesota Medicaid and GAMC claims records, the Blue Cross Blue Shield of Minnesota claims records, the National Medical Expenditure Survey, the Behavioral Risk Factor Surveillance System, and the National Health and Nutrition Examination Survey?

A. Yes.

Q. Do the disks contain results obtained by applying these formulas to the state and Blue Cross Blue Shield of Minnesota claims records?

A. Yes.

MR. HAMLIN: Your Honor, we offer Trial Exhibit 14936 into evidence.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 14936.

BY MR. HAMLIN:

Q. Now Dr. Wyant, based on your education, your training and your expertise in biostatistics, do you have an opinion as to whether the damages model constructed from the Minnesota and national data sources you have described provides a reasonable basis on which to estimate the smoking-attributable expenditures incurred by the state and Blue Cross?

A. Yes.

Q. What is that opinion?

A. It does provide a reasonable basis.

Q. Based on your education, your training and your expertise in biostatistics, do you have an opinion as to whether the application of this model to the claims data from the state of Minnesota and Blue Cross Blue Shield of Minnesota produces the most probable estimate of the smoking-attributable expenditures they incurred from 1978 to 1996?

A. Yes.

Q. Now Dr. Wyant, I've placed on the easel the magnetic board. With the court's permission, if you could come down.

What is the estimate for the smoking-attributable expenditure for the state of Minnesota during the period 1978 through 1996?

*22 A. That estimate is 1.31 billion dollars, or as it's stated here, 1,310 million dollars.

Q. What is the estimate of the smoking-attributable expenditure for Blue Cross Blue Shield of Minnesota from the period 1978 through 1996?

A. That estimate is 460 million dollars.

Q. Now if you could move to the flip chart, what is the total estimate of the smoking-attributable expenditures for the state of Minnesota and Blue Cross Blue Shield of Minnesota for the period 1978 through 1996? If you could write this on the flip chart.

A. That estimate is 1.77 billion dollars.

Q. And how did you arrive at that figure?

A. By the application of the statistical model I've just described that applies the three reductions using the factors from the refined model to the data sources that are described in this exhibit, including the Minnesota claims data, the national surveys, and the Minnesota Behavioral Risk Factor System.

Q. And how did you reach the total of 1.77 billion?

A. That is the sum of 1.31 billion for the state and 460 million, or .46 billion, for Blue Cross Blue Shield of Minnesota.

Q. Okay. You can return to the stand.

Now have you broken down the smoking-attributable expenditures for the state of Minnesota into disease categories?

A. Yes.

Q. And do you have an exhibit that breaks down the smoking- attributable expenditures for the state into disease categories?

A. Yes.

Q. Now let me direct your attention to Trial Exhibit 30203. Is that the exhibit to which you've just referred?

A. Yes.

Q. Was that prepared by you or under your direction?

A. Yes.

MR. HAMLIN: Your Honor, plaintiffs offer Trial Exhibit 30203 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 30203 for illustrative purposes.

BY MR. HAMLIN:

Q. Dr. Wyant, I've placed on the easel Trial Exhibit 30203. Could you identify this exhibit by title first.

A. This is an exhibit titled "State Expenditures Attributable to Smoking by Disease Category, Full Model 1978 to 1996."

Q. And starting with the total, could you tell us what the exhibit shows.

A. May I --

Q. With the court's permission, certainly, come down.

A. The total is 1.31 billion, or to make it easy to add these numbers up, all of which are in millions, we've written it here as 1,310 million. Those are the estimates of the expenditures attributable to smoking for Medicaid and GAMC. Out of all the expenditures during this time period for adults in those programs, that 1.31 billion represents seven percent, that's seven percent of the total expenditures in these groups were Medicaid and GAMC from 1978 to 1996.

Q. How much is the total state expenditures for Medicaid and GAMC addressed here in this exhibit?

A. I believe it's approximately 19 billion dollars.

Q. And how did you get that total?

A. By assembling all of the computerized claims data and also the additional annual reports that we talked about earlier.

*23 Q. What Medicaid and GAM expenditures are we not counting as part of that total of 19 billion?

A. Okay. These expenditures don't count expenditures for children; that is, up through age 18, so this seven percent is seven percent only of the expenditures for adults 19 and older.

Q. Would you go through each of the categories of disease and explain to us what is written on the exhibit.

A. The lung cancer/COPD, 140 -- excuse me, 154 million dollars attributable to smoking, and that represents 45 percent of the total expenditures in these programs to treat people who have lung cancer/COPD. We identify those people by going through the claims records using the ICD-9 codes and applying the three validity screens that we saw this morning. So when we do that, we build up to a set of dollars which is not shown here. We take after applying the three reductions, the three reductions result in a 45 percent figure, and when we take 45 percent of that total amount, which is not shown here, for lung cancer/COPD people, we get 154 million dollars.

Q. Could you take us through the CHD/stroke category.

A. Again the same thing. We identify people with CHD/stroke using the ICD-9 codes, the claims data, the validity screens, and find all the expenditures on such people during this time period. Applying the statistical model gives us three reductions. Those reductions result in us taking 16 percent of those expenditures for those people being treated for CHD/stroke, and that 16 percent results in a smoking-attributable estimate of 184 million dollars.

Q. Now the CHD/stroke category there, is that an indication of other diseases as well?

A. Yes. Those are all the other major diseases on Dr. Samet's ICD-9 list.

Q. Go to the next category and tell us about that.

A. The next two categories are the two parts of what Dr. Samet calls diminished health status. The first part is that part related to reported poor health as estimated from the National Medical Expenditure Survey, again, out of all the dollars being expended for medical services, hospitals, outpatient visits and so on for adults who were not in one of the major disease groups, and we apply the reduction formulas, we get five percent of those dollars as related to reported poor health, and that five percent comes up to 371 million dollars.

Q. Can we go through the next category.

A. The other effects of diminished health, those not captured by reported poor health, again when we apply the statistical models and the three reductions to the total dollars being spent on medical care outside of the major smoking- attributable disease groups, we calculate an additional four percent, or 341 million dollars.

Finally --

Q. And how do you get to the 341 million dollars? Use the three reductions?

A. Yes. All of these percentages are defined using the three reductions. In the diminished health mixed effects there are only two reductions, as with the nursing homes. These two last categories are the ones where there are two reductions. The reasons there are two reductions here are that you've used disease information that's available in the previous categories, and here we're looking at the remaining reductions, percentage of smokers, and the third reduction, dollars attributable to smoking.

*24 Q. And what does the four percent signify?

A. Again, once we've taken out all the dollars for people being treated for lung cancer/COPD and CHD/stroke, we have dollars for all the other medical services. Some of those dollars are attributable to diminished health among the smokers. We divide that into two segments, two parts of the model, two formulas, then apply the reductions, and end up with five percent smoking- attributable that's related to reported poor health and four percent that's related to other effects. So in other words, nine percent, five plus four, of the remaining dollars for health-care services are being estimated by the model as attributable to smoking.

Q. Can we talk about the last category, nursing home fees.

A. Nursing home fees, again we take all the maintenance fees for the state claims records and apply the statistical model, apply the reductions, we end up taking three percent, estimating three percent of those total nursing home maintenance fees, approximately nine billion dollars over this period as smoking-attributable, and that results in 260 million smoking-attributable dollars.

Q. And what is the total -- the percentage of -- of dollars that we're asking for?

A. Well the percentage of total dollars spent by the state during this time period that is smoking-attributable as estimated by the model is seven percent out of the total dollars spent by these two programs for adults -- I'm sorry, GAMC and Medicaid.

Q. So it's seven percent of the total spent on these categories.

A. That's correct.

Q. That's the smoking-attributable expenditure?

A. Yes.

Q. Now what if anything is significant about the pattern of percentages that we see on this exhibit?

A. Well these percentages follow the epidemiology. Smoking is the major or the predominant cause of lung cancer/COPD, and when we apply the models and the formulas we get 45 percent of expenditures for people with those diseases attributable to smoking.

For CHD/stroke, from the Surgeon General's report and based on our understanding from Dr. Samet, smoking is a major cause, and here we see an intermediate percentage. Not as high as for lung cancer, but higher than for diminished health.

For diminished health smoking is a cause, and here we see the lowest percentages in the medical care group.

Q. And what about nursing home fees?

A. Smoking is a cause of nursing home fees following Dr. Samet's discussion of stroke. And again, as a cause we're only seeing, and consistent with that the calculations are only showing, three percent smoking-attributable expenditures out of all the nursing home maintenance fees.

Q. Have you also broken down the smoking-attributable expenditures for Blue Cross Blue Shield of Minnesota into categories?

A. Yes.

Q. Let me direct your attention now to Trial Exhibit 30205 in your book here, and is that the exhibit?

A. Yes, it is.

Q. And that was prepared by you?

A. Yes, it was.

*25 MR. HAMLIN: Your Honor, plaintiffs offer Trial Exhibit 30205 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 30205 for illustrative purposes.

BY MR. HAMLIN:

Q. I'll place Trial Exhibit 30205 on the easel. Can you identify this exhibit.

A. These are Blue Cross Blue Shield expenditures attributable to smoking by disease category -- (clearing throat) excuse me -- full model, 1978 to 1996.

Q. Starting with the total, could you explain what the exhibit means.

A. Total dollars for the smoking-attributable dollars from the model for Blue Cross Blue Shield Minnesota, if you add up all the dollars being spent on adults in the Blue Cross rated plans from 1978 to 1996 and take 15 percent of those dollars, that yields 460; in other words, for the smoking-attributable percentage of these dollars, it's 15 percent. That's of all dollars being spent in these programs for adults.

Q. For adults in the categories identified?

A. Correct.

Q. Can you take us now through the first category, lung cancer/COPD.

A. Again we go through the claims records as before using ICD-9 codes, the validity screens, identify Blue Cross expenditures for people who have lung cancer or COPD. When the models calculate the three reductions, they tell us to take 61 percent of those expenditures as smoking-attributable, and that 61 percent results in 52 million dollars of smoking-attributable expenditures associated with lung cancer and COPD.

Q. Can you take us through the CHD/stroke category.

A. Again, with the claims data we use ICD-9 codes, find the dollars expended to treat people with these diseases, apply the three reductions, we take 34 percent of those total expenditures as attributable to smoking. Applying 34 percent to those total expenditures yields 168 million dollars of smoking-attributable expenditures related to CHD/stroke in Blue Cross Blue Shield of Minnesota.

Q. Again, what does the category CHD/stroke represent in terms of diseases?

A. Those are all of the ICD-9 codes identified by Dr. Samet as major diseases, except for lung cancer and COPD.

Q. Now turn to the reported poor health category and tell us how you arrived at the smoking-attributable expenditure for that category.

A. Reported poor health, we look at the remaining expenditures for people who are not being treated for one of these major diseases, and using our data sets and our statistical model we estimate that after applying the reductions, four percent of the remaining expenditures are related to reported poor health as one part of diminished health effects.

Q. And can you take us through the last category.

A. The final category is other effects of diminished health, and again we take all the dollars being spent on medical care for adults who don't fall into one of the two major disease groups, application of the model, calculation of the three reductions, there's an additional five percent related to other diminished health effects, or another 136 million dollars that is smoking- attributable as part of Dr. Samet's general diminished health category.

*26 Q. Can you give us an example of the kinds of expenditure that you will see in that category.

A. One example would be complications of conditions due to having a smoking-related condition that makes treatment more costly or more difficult for some condition that might not be reported and often aren't reported as poor health.

Q. Are there any nursing home expenditures for Blue -- for Blue Cross?

A. No, there is no line here for nursing home expenditures. Blue Cross spent some money on nursing home expenditures during 1978 to 1996, but it was a very small amount of their business, so we did not try to calculate a smoking- attributable expenditure percentage from those nursing home expenditures. That's why they're not on this chart.

Q. Can you tell us what if anything is significant about the pattern of percentages that we see in the column marked "Percentage of health-care expenditures" for Blue Cross.

A. Again, the pattern of percentages follows the epidemiology that Dr. Samet laid out for us. If smoking is the predominant cause of lung cancer and COPD, then we would expect the formulas, if they're making sense, to produce the highest smoking- attributable percentages for that group, and that's exactly what happens with 61 percent. After application of the three reductions to remove people with lung cancer who don't smoke and the other elements that we talked about, after removing those, 61 percent of the lung cancer/COPD dollars are smoking-attributable.

The epidemiology says that smoking is a major cause of CHD and stroke. It's not the predominant cause, a major cause. And what we see, the percentage is moving as you'd expect from 61 percent of the dollars being smoking- attributable down to 34 percent.

And the remaining categories, smoking is a cause, not a major cause, and as a cause we're seeing the smallest percentages being calculated as smoking- attributable.

Q. Now the percentage for Blue Cross is 15 percent. The percentage of total expenditures is 15 percent; is that right?

A. That's correct.

Q. And the percentage for the state of Minnesota is seven percent; is that right?

A. That's correct.

Q. What happens if you remove the nursing home fees, what happens to the percentages?

A. If we removed this line here for nursing home fees where we're estimating only three percent of the dollars are smoking- attributable and look only at the rest so that we can get a fair comparison between Blue Cross and Medicaid; that is, looking at all the medical expenditures outside of nursing homes, hospitals, doctors' visits and so on, if we calculated the total percentage only going down as far as here, diminished health and the major diseases, then we would get not seven percent but about 11 and a half percent down here. So that's the comparison between Blue Cross and state if we only look at the medical expenditures and not at nursing home maintenance fees, 15 percent versus about 11 and a half percent.

*27 Q. Okay. Thank you.

What methods do you as a biostatistician use to gain confidence in the reliability of the smoking-attributable expenditures for the state of Minnesota and for Blue Cross Blue Shield of Minnesota?

A. Well one thing that I do as a biostatistician is to look at other studies and other calculations that address the effects of smoking.

Q. Have you done that in this case?

A. Yes, I have.

Q. Let me show you what has been marked as Trial Exhibit 30200, and I want to ask you whether or not that's one of the comparisons that you made. That should be in your demonstrative book.

Have you found that?

A. Yes, I have.

Q. And this is one of the exhibits that you have prepared; is that right?

A. Yes.

Q. And can you identify it for me, please.

A. This exhibit is entitled "Death from Smoking in the United States."

MR. HAMLIN: Your Honor, we offer Trial Exhibit 30200 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 30200 for illustrative purposes.

BY MR. HAMLIN:

Q. I'm placing Trial Exhibit 30200 on the easel. Can you identify this and tell us what it is.

A. The first part is an excerpt from an article in the Journal of the American Medical Association in 1993. Two of the conclusions of that article is 19 percent of all deaths in the United States are caused by smoking, and that amounts to about 400,000 deaths per year, or at the time of the article.

Q. Now why does this give you confidence in the reliability of the smoking- attributable expenditures that you've calculated for the state of Minnesota and Blue Cross Blue Shield of Minnesota?

A. Although one would not expect that the percentage of deaths is going to mirror the percentage of expenditures, one would expect that there's some relationship, particularly since many of the major diseases result in highly increased expenditures right before death or in the year preceding death. And the fact that other researchers have found that 19 percent of all deaths are caused by smoking and that we find that 15 percent for Blue Cross down to seven percent of expenditures for Medicaid, or 12 percent - - 11 and a half percent if we look at only the medicals, those are very reassuring numbers to me as being in a reasonable ballpark given the percentage of deaths that are attributed to smoking in the U.S. every year.

Q. Have you made other comparisons?

A. Yes, I have.

Q. Let me direct your attention to Trial Exhibit 30201. Do you have that exhibit?

A. Yes, I do.

Q. Can you identify that.

A. This is exhibit called "Death from Smoking by Major Diseases."

Q. And was this exhibit prepared at your direction?

A. Yes.

MR. HAMLIN: We offer Trial Exhibit 30201 for illustrative purposes.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 30201.

MR. HAMLIN: Put that on the overhead, please.

BY MR. HAMLIN:

Q. Dr. Wyant, can you identify this exhibit, tell us what it is.

*28 A. These are percentages calculated and recorded by a U.S. agency that assists -- is used to assist Congress in analyzing various factors relating to the economy, and they're estimates of the percentage of lung cancer deaths attributable to smoking, it's 83 percent, for COPD 80 percent, for CHD 20 percent, and for stroke 16 percent. Those are the percentages of deaths in those disease categories estimated to be attributable to smoking.

Q. Now why do these percentages give you confidence in your smoking- attributable expenditures for the state of Minnesota and Blue Cross Blue Shield of Minnesota?

A. Again, when you track the rough ballparks of the percentages that we calculate as attributable to smoking in terms of expenditures, you see a pattern that's very similar to this, the expenditures up around 50 to 60 percent in the lung cancer/COPD category, and then dropping from our intermediate figure not that far removed from the 16 and 20 percent in the CHD/stroke category.

Q. Now have you looked at studies of smoking and health-care costs in addition to the studies of mortality that we have just examined?

A. Yes, I have.

Q. Let me direct your attention to Trial Exhibit 26047. That's in your testimony notebook. Do you have that?

A. Yes, I do.

Q. Can you identify that study.

A. This is a study called "Health Risks and Their Impact on Medical Costs," a study by Milliman & Robertson, Incorporated.

Q. And is this a study that you have reviewed?

A. Yes, it is.

Q. Does this study form part of the basis of your opinions in this case?

A. Yes.

Q. Is this a reliable authority in published literature on health-care costs?

A. Yes.

MR. HAMLIN: Your Honor, we offer Trial Exhibit 26047.

MR. BIERSTEKER: No objection, Your Honor.

THE COURT: Court will receive 26047.

MR. BIERSTEKER: I assume you -- Your Honor, I assume it was admitted as a learned treatise?

MR. HAMLIN: That is correct.

THE COURT: As a learned treatise, yes.

BY MR. HAMLIN:

Q. Let me direct your attention now to Trial Exhibit 18945. Can you identify that study.

A. This is a study called "Risk for Cardiovascular Disease in the Elderly and Associated Medicare Costs: The Framingham Study."

Q. Is this one of the studies that you have reviewed in this case?

A. Yes, it is.

Q. This is a study on the health-care costs?

A. Yes.

Q. And does this study form part of the basis of your opinions in this case?

A. Yes.

Q. Is this a reliable authority on health-care costs?

A. Yes.

Q. Is it a reliable authority in the published literature on health-care costs?

A. Yes.

MR. HAMLIN: Your Honor, we offer Trial Exhibit 18945.

MR. BIERSTEKER: Your Honor, I would have no objection to the article coming into evidence as a learned treatise but for the fact that there is marginalia on the first page. If we substitute a clean copy of the first page, I would have no objection.

THE COURT: Do you have a clean copy, counsel?

*29 MR. HAMLIN: We do, Your Honor.

THE COURT: All right. Use the clean copy.

MR. HAMLIN: Yes.

THE COURT: We'll receive 18945 as a learned treatise.

BY MR. HAMLIN:

Q. Dr. Wyant, with the court's permission, could you come down to the flip chart and list on the flip chart the two studies that we have just identified.

Could we begin with the Chrysler study.

A. Chrysler study was a study about 2500 employees of the Chrysler Corporation that were covered for their health care by Blue Cross Blue Shield of Michigan. The study has a date of about 1995. And what they did was look at health-care costs for smokers compared to non-smokers as well as health-care costs for other health behaviors. This study was done by Milliman & Robertson. Milliman and Robertson are professional actuaries. The study was done in their Milwaukee office. And these are the kinds of people that corporations like Chrysler want to do a study of smoking costs among their employees or other factors, these are the kinds of professionals that are usually retained to do that. And what they found was that increased costs due to smoking was 31 percent in this group.

The other study, Framingham study, Framingham study I believe Dr. Samet talked about, that's a famous study in Massachusetts in which a large number of people were followed for many years, in particular looking for factors that influence coronary heart disease. This particular study of that data dealt with health-care costs, so it's published in the scientific literature, I believe about 1993, and it looked at another government program, not Medicaid, but Medicare.

Q. Can you tell us what Medicare is?

A. Medicare is a federal government program that largely covers people 65 and older. And the people in this study, about 1,000 persons age 65 plus. And in this study an increase in expenditures due to smoking of 16 percent was observed.

Q. Approximately when was that study done, or the report on that study done?

A. I'm going from memory here, but I believe it was about 1993.

Q. Now can you compare the Chrysler study to the results that you obtained with respect to smoking-attributable expenditures?

A. When we compare the Chrysler study with Blue Cross Blue Shield of Minnesota --

Q. And why do you compare it with Blue Cross Blue Shield of Minnesota?

A. Again, these were people covered through their employer and in fact covered through another Blue Cross plan, this time in the state of Michigan.

Q. And what do you find upon making that comparison?

A. Well here we're talking about hundreds of thousands of people in the claim records, an additional tens of thousands from our other data sources, and when we use our models and process all that data, we come up with a figure of 33 percent for Blue Cross Blue Shield of Minnesota.

Q. Can you explain how you came up with 33 percent? Based on the earlier chart, the percentage appeared to be 15 percent.

A. The earlier chart had 15 percent here, and that's because those charts calculated smoking-attributable expenditures as a percentage of everyone's expenditures. In these studies they calculate smoking-attributable expenditures as a percentage of just the smoker expenditures, so in order to compare them, we have to calculate in Minnesota smoking-attributable expenditures as a percentage of smoker expenditures, and that's why the 33 percent is higher than the 15. It's really based on the same figure, we're just dividing by smoker expenditures rather than everyone's expenditures according to all the estimates from our refined model. And the only reason for doing that is so that we can compare the results of our study by doing the arithmetic in the same way that they did present the results of the Chrysler study.

*30 Q. Okay. Upon making that comparison, what have you concluded?

A. Well these are obviously very close. And we didn't have this Chrysler study until recently, until well after we had calculated all of our results here.

Q. What -- why is that significant?

A. Well this is a totally independent study and a totally independent group of employees. This was not something we were shooting at in any way, shape or form or -- just another corroboration coming in of more recent published work as showing percentages very similar to what we're calculating.

Q. And can you now turn to the Framingham study and can you compare that to smoking-attributable expenditures in this case?

A. We can look at the state program, Minnesota Medicaid, and to make a comparison reasonable we'll look only at persons age 65 and older so that we're comparing smoking-attributable percentages that we're finding in our study for persons age 65 and older with those found in the Framingham study for persons aged 65 and older, and when we do that calculation, our figure is 19 percent compared to the 16 percent found in the Framingham study.

Q. And how do those percentages compare?

A. Again you can see how these percentages are matching up quite closely.

Q. And what do you conclude from the comparison of the Framingham study of people on Medicare to the Minnesota Medicaid portion of people who were 65 and over?

A. Well again, results like this are very reassuring. You see that researchers using different data and different persons and slightly different methods -- some of the methods are identical. On all of these studies, smokers are compared to non-smokers adjusting for age in an attempt to make a comparison of similar -- of smokers to similar never smokers. But some of the factors are different, and they're done in slightly different ways. But nonetheless, with all those differences, in all of our spinning of the computers late at night to go through all those numbers, these two paths converged, 16 percent and 19 percent here. And to a statistician, those kinds of approximate results, results that are close to each other, are very reassuring that -- that these are reasonable and reliable estimates that we're calculating with our models.

Q. What if any confidence do you draw in your estimates of the smoking- attributable expenditures for the state of Minnesota and Blue Cross Blue Shield of Minnesota from your comparison with the Chrysler study and with the Framingham study?

A. Briefly --

MR. BIERSTEKER: Objection, Your Honor, I think this question has been asked and answered.

MR. HAMLIN: I don't believe that that question was asked and answered, Your Honor.

THE COURT: I don't think so, counsel.

MR. BIERSTEKER: Fine. Overruled.

THE COURT: Just -- just in case, we'll let him answer it again, but I don't think it was.

Go ahead.

A. It increases my confidence in the reliability of our results.

Q. Thank you. You can return to the witness stand.

*31 Dr. Wyant, what was the number of people in the state and Blue Cross Blue Shield of Minnesota claims records that -- that you reviewed, approximately?

A. Oh, it was well over a million.

Q. And what -- what was the number of people surveyed in the National Medical Expenditure Survey in 1987?

A. Approximately 35,000 people.

Q. And you used the NMES survey in the -- in plaintiffs' statistical model; right?

A. Yes.

Q. What was the number of people surveyed in the Behavioral Risk Factor Survey during the 11-year period that you testified about?

A. About 34,000.

Q. And what was the number of people surveyed in the NHANES survey which you used as a basis for the nursing home portion of the model?

A. Well there were some 40,000 people in the original survey, and I think about 14,000 in the follow-up that we looked at.

Q. Now defendants claim that the plaintiffs' model is unreliable because of estimates arrived at for two 94-year old women in nursing homes. Let me ask you this: Was the model designed as a tool for estimating smoking-attributable expenditures for small groups such as women in nursing homes who are 94 years old?

A. No, it wasn't.

Q. Can you give us an example of why the model is not designed to look at small groups, using Dr. Zeger's skate boarder example.

A. Well --

Q. I think you may have to come down to the flip chart.

THE WITNESS: If I might.

A. I have a cheat sheet for this.

Q. There it is.

A. Excuse me. This will take just a minute to put up here.

Q. Well first of all, perhaps you could review with us Dr. Zeger's skate board example.

A. Sure. Dr. Zeger's skate board example had to do with a need to control for age in studies of medical costs and other medical conditions, that if you went out to the lakes and you were looking at what the average costs of skate boarding were for skate boarding injuries and just picked some skate borders and then picked some non-skate boarders and compared average costs, what you would likely see is a lot of young skate boarders being compared to a lot of people who are older who don't skate board, and the older people would tend on average to have higher medical costs just due to being older. So unless you stratify by age and compare 10- to 20-year-old skate boarders with 10- to 20- year-old non-skate borders, you have a lot of trouble isolating the risk factors you're looking to, which is the extent to which using skate boards increases your medical costs.

So if we look at skate boarders in the age range of 10 to 20 -- again this is a hypothetical experiment -- we'll look at average costs for skate boarders and average costs for non-skate boarders, and we'll look at ages 10 to 20. And we can put these up like this, so that skate boarders, showing around a thousand dollars in annual medical expenditures -- this amount is increasing on average with age so that by the time we get to age 20, we're up to around 1500 dollars. And this shows why we might want to still stratify by age even though here in the age 10 to 20 group there are still differences, although much smaller than, say, compared to 65-year- olds, there's still differences in average expenditures across these age groups.

*32 Now let's put up the comparison expenditures for the non- skate boarders. And these expenditures also tend to go up from around nine hundred to a thousand dollars in the 10 to 12 age range, up to around 1500 dollar range when we're 18 to 20.

And now we can look at the differences. A hundred dollars more for the skate boarders and 150 more, a hundred more, then even, 100 more, I have 150, 100, 400, 100, minus 200, and 100. And I believe I've correctly put this down, the average for the total population is one hundred dollars. But what always happens in one of these studies, one of these statistical studies when we're comparing costs and stratifying by age, is that there's statistical fluctuation. Lots of times you see a hundred dollars more, or hundred dollars more, a hundred dollars more, but there's no guarantee of that, and in some groups you're clearly going to see 150 and some at zero, but by and large the pattern when you look at the whole population is for an increase of about a hundred dollars a year, skate boarders compared to non-skate boarders when we compare across age groups.

Q. And what if any statistical principle is at work here?

A. Well this statistical principle is the law of averages. This is what Dr. Zeger talked about. And in particular with the coin-flipping example, if you went in with a sequence of coins where you were expecting about a 50 percent measure and picked out a sequence of four heads and said, huh, maybe that means something, it would be like going down this list and picking out, well gee, 400 dollars --

Q. Is this for 17-year-olds?

A. That's for 17-year-olds.

Q. So the difference between skate boarders and non-skate boarders is 400 dollars.

A. That's correct.

Q. Can you tell us about that group.

A. Well, you can see from the fluctuation going on here that sometimes you're going to get a fluctuation that's higher than the rest, and you can always go back through and pick out the one that's highest just like you can go through a sequence of coin flips and say, "Gee, here's four heads, that looks unusual, let's take that out. We might want to throw that away."

If you took that out and threw it, you would be losing the law of averages on your side, because what you would end up with is a total percentage of heads that's always going to be too low because you've gone in there and systematically, every time you saw four heads in a row, said "I'm not going to count that."

And similarly here, another nature of fluctuations when you see these studies that compare like to like in terms of age is that among the 19- year- olds we even see skate boarders in that group, whoever we found out at the lakes, costing less than the non-skate boarders. Now do you look at a study like this and say, well, now I'm ready to believe that skate boarding causes injuries and it costs money, but -- obviously you can fall off skate boards, but I believe it will cost you money at age 10 and it will cost you money at age 11 and it will cost you money at age 15, but if you luck out and you're in that one year, age 19, you're not going to fall off your skate board, or if you do, it's not going to hurt you and you're not going to spend any money. Well that's not a reasonable conclusion because there's normal fluctuation in any study. And that's what you expect to see, hundred here, 150 here, and occasionally a 400 or minus 200.

*33 To get the law of averages to work for you, get a total estimate for 10- to 20-year-olds, you need to include all of these differences for each age group and come up with a hundred dollars in this hypothetical example as a typical amount by which it costs skate boarders in terms of treatment of skate boarding injuries compared to similar non-skate boarders.

Q. If you restricted your information to the group of 17- year-olds, what would that tell you about the average for the entire population?

A. Well it really wouldn't tell you much of anything, especially if you'd gone through and picked out 17-year-olds simply by virtue of the fact that they're showing in this little study the highest difference between skate boarders and non-skate boarders.

Q. What if anything would it tell you; that is, selecting out that group of 17-year-olds, about the validity of the study for the entire population?

A. Nothing.

Q. Why is that?

A. Well you expect fluctuation in statistical studies like this, and unless you allow that fluctuation to work in both directions, just as you count all the heads and all the tails when you're flipping a coin, you're not going to arrive at a sensible and reliable final estimate for the group as a whole.

Q. So suppose you get a large estimate for two 94-year-olds in the smoking- attributable expenditures in the plaintiffs' model. Do you leave it in?

A. Sure.

Q. Why?

A. Because otherwise you're going to make the same mistake as a person who flips the coin 20 times and then looks through to see, ah, here's four heads, take those heads out. It's a violation of basic statistical principles to go through and arbitrarily pick out groups at one end or the other and simply discard them.

Q. Now in the nursing home model, did you get any negative fluctuations?

A. Absolutely.

Q. And can you give us an example.

A. Sure. There were four men in the 57- to 59-year-old age group that contributed more than a hundred million dollars in negatives, and we were adding up nursing home expenditures in a fashion similar to this to arrive at an overall smoking- attributable nursing-home-expenditure estimate.

Q. And did you leave them in?

A. We left all of those in. All of those negative dollars are in the summary.

Q. And -- and why did you leave them in?

A. For exactly the same reasons, that just simply going in and picking out small groups on some arbitrary basis is destroying your ability to take advantage of the law of averages to arrive at a reasonable and reliable estimate.

Q. Thank you.

Did you do a further analysis of the nursing home data to help the jury understand how the nursing home model works?

A. Yes.

Q. And what did you do?

MR. BIERSTEKER: Your Honor, I don't know whether this is leading, but we've only had one analysis of the nursing home model.

MR. HAMLIN: Your Honor, this is --

Q. Well let me ask you this, Dr. Wyant: Did you do a further sensitivity test?

*34 MR. HAMLIN: This is not a new model.

A. Yes.

THE COURT: Okay. All right. You may answer that.

MR. BIERSTEKER: Well Your Honor, I -- just for record, I'd like to interpose an objection to the extent that he is being permitted to testify about calculations, no matter what they are called, that have not been provided to us.

THE COURT: Okay. This is not a different model.

MR. HAMLIN: It is absolutely not a different model, Your Honor.

THE COURT: All right. Go ahead.

MR. HAMLIN: Under no circumstances.

BY MR. HAMLIN:

Q. Can you tell us about the sensitivity --

Well first of all, tell us what a sensitivity test is.

A. Well a sensitivity test is a reworking of the calculations where if there's some factor of interest that may be contributing something to the model that you want to isolate, you run it with and without the factor to see just how much difference it makes.

Q. And what was the sensitivity test that you performed with respect to the nursing home model?

MR. BIERSTEKER: Your Honor, can I have a continuing objection to testimony about reworked calculations that have not been provided to us?

MR. HAMLIN: Well, Your Honor, what we're attempting to do is explain the sensitivity test in response to statements made by counsel in opening statements.  

THE COURT: Okay. Are these different calculations?

MR. HAMLIN: They're not different. No, they're not different calculations.

MR. BIERSTEKER: Your Honor, Your Honor, the witness just testified that the sensitivity test is a reworking of the calculations.

THE COURT: Well the question is are they different calculations.

MR. HAMLIN: They use exactly the same data, and it's just a different look at the same data. That's -- that's all it is, Your Honor.

THE COURT: All right. Go ahead.

MR. HAMLIN: That's what a sensitivity test is.

THE COURT: You may answer.

MR. BIERSTEKER: Your Honor, may I have a continuing objection, please?

THE COURT: Well the record will show your objection to this question.

Q. Can you tell us what you found.

A. Based on the NHANES data, which is what we used in calculating the reductions of the nursing home model, close to 95 percent of the data from people in nursing homes is for the ages from 60 to 90. That means there's very little information in the age groups in the 50s and in age groups over 90 to get a look at a sensitivity test of how important the fluctuations are outside those ranges, because there will be more fluctuations where you have fewer people in the nursing home, that occurs both at the high end and at the low end, and one simple way to address that in a balanced way is to simply re-add up the attributable contributions from the different age groups just from age 60 to 90 instead of from age 55 to 95. In that way you're taking a look at the possible effect of fluctuations where you have fewer people in the nursing home, and you're doing it in a balanced way that simply has a rule that says, look, here's a data cutoff from 50 to 90, let's just look in that range where just about 95 percent of our data are. And if you do that, if you simply add up the contributions from 50 to 90 instead of the nursing home estimate of smoking-attributable expenditures of 260 million dollars, you get an estimate of smoking-attributable nursing home expenditures of 262 million dollars, so virtually identical. And what you see happening here is a -- is a very good example of the law of averages in practice.

*35 Yes, there are more fluctuations above 95 -- or excuse me, above 90, there were fluctuations below 60, but those fluctuations are essentially cancelling each other out and you get basically the same estimate regardless of which age range you want to look at.

Q. Did you leave any of the 94-year-old women in the sensitivity test that you just described?

A. No.

Q. But you still got numbers close to the original estimate.

MR. BIERSTEKER: Objection, no basis for that, Your Honor.

THE COURT: You may answer that.

A. Only two million dollars' difference compared to the original estimate of 260 million.

Q. Now defendants also claim that our model is wrong because they say we're asking for 700 million dollars for 19- to 30-year- old men in the diminished health portion of the model. Now was this 700 million dollars subject to offsets just like in the nursing home model?

A. Yes, absolutely.

Q. And can you give us an example of an offset.

A. Well males 35 to 64 contributed just such an offset. It's a situation very similar to that in the skate boarder chart where we see the 400 at age 17 being offset by a minus 200 at age 19.

Q. Now are these males 35 to 64, are they smokers or non- smokers?

A. Well they're both.

Q. And what effect do they have on the estimate? Do they drive it up or do they drive it down?

A. The contribution from that group drives the estimate down.

Q. Was the model designed to estimate smoking-attributable expenditures for small groups like males 19 to 34 years old?

MR. BIERSTEKER: Objection, Your Honor, he's leading.

THE COURT: You are leading, counsel. Sustained.

BY MR. HAMLIN:

Q. What if any use can be made of our -- of the plaintiffs' statistical model to estimate smoking-attributable expenditures for small groups like males 19 to 34 years old?

A. I would not use it for such a purpose.

Q. And why is that?

A. The model was designed to take advantage of the law of averages and to take advantage of all the data available to us to come up with the most reliable estimate for the population of people covered by these programs during 1978 to 1996.

Q. Now would it be proper statistical practice to remove this group of males 19 to 34 from the model?

MR. BIERSTEKER: Objection, leading.

THE COURT: No, you may answer that. It is somewhat leading though, counsel.

A. I don't know of any basis for doing that.

Q. And can you tell us whether or not there are statistical practices that would allow that?

A. I don't know of any statistical justification for that.

Q. Even though it isn't appropriate, if you took out those males 19 to 34 years old, would the plaintiffs' estimate go down by 700 million dollars?

A. No.

Q. Dr. Wyant, is there variability in the statistical -- in statistical estimates generally?

A. Yes.

Q. And can you tell us why.

A. Well one reason relates to what we've talked about earlier, that part of our smoking-attributable expenditure estimates in the formulas are based on surveys, national surveys like the National Medical Expenditure Survey and like NHANES, and those surveys are only of 30,000 people. Those are very big in terms of the study, but not very big in terms of the total population of the United States. And if those surveys were redone in the same way with just pulling in a different sample of 30,000 people instead of the sample that was actually taken, then there would be some variation in the smoking-attributable expenditure estimate that we came up with using our models, but using this hypothetical re-draw of the survey.

*36 Q. What do you mean by "hypothetical re-draw of the survey?" Could you be a little clearer about that?

A. Well the survey was clearly only done once and we got the data from that survey. We know that if they had done it several times, each time you would have gotten a different set of people and there would have been some statistical variability in the estimates, because with each group of people, you would have gotten a somewhat different estimate of smoking-attributable expenditures.

Q. Now what were our estimates based on in terms of samples?

A. Well, our estimates were based on the samples that were drawn, the actual samples drawn in the National Medical Expenditure Survey and the National Health and Nutrition Evaluation Study.

Q. Now is there a statistical-measure variability that is most appropriate here?

A. I believe so.

MR. HAMLIN: Your Honor, we're going to get into a brand-new area here. Perhaps this might be a logical breaking point.

THE COURT: All right. We'll recess and then reconvene tomorrow morning at 9:30.

THE CLERK: Court stands in recess to reconvene tomorrow morning at 9:30.

(Recess taken.) Return to tobacco index

Return to The Putnam Pit