Dr. Robert Rosati was the first director of the databank, beginning in 1976. In this interview, he speaks about growth and changes in the databank, as well as the databank under his leadership.
JESSICA ROSEBERRY: This is Jessica Roseberry. I'm here with Dr. Robert Rosati, who is Associate Professor Emeritus, Cardiology, in Medicine; and it's May 11, 2007; and we're here in his office in the home of the Rice Diet Program on Cole Mill Road [Durham, North Carolina]. And I want to thank you, Dr. Rosati, for agreeing to be interviewed today. I appreciate that very much.
ROBERT ROSATI: Thank you for coming.
ROSEBERRY: I wonder if I might start by asking you how you got into cardiology?
ROSATI: When I was a medical student, I worked with one of the professors in medicine, a cardiologist named Andy Wallace. I put together an elective and a couple of summer vacations so that I was able to work in his lab for about nine months. He was doing research in electrophysiology and from working in his lab, I got interested in cardiology, following him, really, that kind of thing. I guess that's how most of it happens. You find something that you're attracted to and then sort of head in that direction.
ROSEBERRY: Can you tell me a little bit about Dr. Wallace?
ROSATI: Oh, boy. Let's see. He, at that time, was on the faculty. I first saw him when he was the chief resident, and I was a medical student, and we had these conferences, where he'd present patients. And then he became a cardiologist, and that's when I worked with him; and then he became chief of Cardiology, and then he went up to Dartmouth to be whatever their equivalent of a dean of a medical school is. He had a very successful career as a physician, researcher and administrator. He's now retired and is living nearby.
ROSEBERRY: Can you tell me a little bit more about the work that you were doing with him?
ROSATI: He was doing research in electrophysiology of the heart. The work that I did involved looking at the effect of various drugs on some of the electrophysiologic properties of the canine heart.
ROSEBERRY: Do you remember any conclusions?
ROSATI: In one set of experiments, beta-blockers did not raise the fibrillation threshold unless they also had quinidine-like properties. In another, we showed a variety of effects of diphenylhydantoin in the denervated canine heart.
ROSEBERRY: Well, what was interesting to you about cardiology?
ROSATI: I think it was more being attracted to the people than anything else. I originally thought I was going to become a heart surgeon, then a neurologist. But, the experience with Dr. Wallace sold me on cardiology
ROSEBERRY: Did you work on the CCU [coronary care unit]?
ROSATI: Certainly, both as an intern and a resident. But, while I was still in medical school, I was attracted by two other faculty members: Harvey Estes and Tommy Thompson. They were both cardiologist but were also interested in computers and how they could be used in medicine. They and others from Duke were working with IBM people up in New York on a computer system for making medical diagnoses. And I asked Dr. Stead if I could go and work with them instead of taking senior medicine. He agreed and I went up there to the IBM place, in Yorktown Heights and worked with them on this computer diagnosis thing. It's how I first got interested in computers. So, by the time I became an intern, I had that computer background and the interest in cardiology.
ROSATI: I was an intern for one year and then a medical resident and then went on to cardiology fellowship. I wrote a grant with Wallace's help to spend a year with Ted Johnson, who was a physiologist'another person I'd met in physiology and medicine and been attracted to. He was into muscle mechanics and electrophysiology at a very basic cellular level. I was going to do that for a year and learn about muscle mechanics and electrophysiology at a very basic level, and then I was going to do clinical research with Vic Behar, who was one of the faculty in the cath [catheterization] lab and try to apply what I'd learned in the basic lab to people in the cath lab. We would be trying to measure Vmax, the intrinsic contractility of heart muscle, independent of other factors. I got the grant, and went to work with Ted Johnson. After about three months I said, (chuckles) 'This is not going to work, because this guy (Dr. Johnson) is so good that to compete with him in the future as a faculty member, I would have to spend all my time and more doing research' and I still wanted to be a physician. It didn't ever occur to me that I could do both well.
In the meantime I'd been reading a book called Clinical Epidemology by Alvan Feinstein. He'd written some articles first, but then he wrote the book, and basically he said you could take science and apply it to clinical medicine. He used lung cancer as one of the example, and rheumatic fever and rheumatic heart disease as another. He used clinical characteristics to predict outcomes. And so, I went to Dr. Stead, and I said, 'I've got this crazy idea that maybe what I really want to do is do clinical research in cardiology, doing the same kinds of things that Feinstein has been doing in cancer and other diseases.' And old Dr. Stead said, 'I've been waiting for you. I want you to go down to the MIRU lab,' which was the research unit that they'd developed. The myocardial infarction research unit was a project which Dr. Stead had gotten the NIH [National Institutes of Health], to fund. I think there were eight units at different universities to study myocardial infarction from hospital admission through day five. These were contracts, which didn't allow a lot of flexibility. Each unit was to get a computer. Typical Dr Stead: he wasn't sure what we'd do with them, but he figured we'd figure it out. And Duke got the best one. Everybody else had a Sigma Three, and we had a Sigma Five.
ROSEBERRY: What is that?
ROSATI: It's a computer. It fits in a room bigger then this room. With air conditioning specially made. And we had the best storage system, because we had five fixed-head, three-megabyte disks that meant you could really search things fast. In other words, we had fifteen megabytes in the thing that took fifteen or twenty feet of wall space and was six feet high.
ROSEBERRY: How did Duke get the best one?
ROSATI: Dr. Stead (laughs) and Dr. Wallace who was the chief investigator of that project. Now, contracts are different from grants. You're supposed to do what you say you're going to do and nothing else. So, Dr Stead sent me down to meet Frank Starmer. They were trying to figure out how to measure cardiac output continuously by measuring pulse pressures, looking at the wave forms and figuring out what the cardiac output was without having to recalibrate everything. It worked, but any time you did an intervention, you had to recalibrate by doing another cardiac output anyway. So along I came with my idea that we'd follow these patients, (laughs) but for more than five days. This was a problem. (laughs) So, Frank Starmer to set me up a little program that allowed me to start putting data in about the people on the coronary care unit with Galen Wagner's help. And we began to follow patients at six months and then one year and finally yearly thereafter. We'd call the patients and if they were alive, we'd ask them about symptoms, such as recurrent chest pain. And then we'd put the information in the computer, and that was the first databank. And then Frank Starmer wrote another program which was called 'What Is' that allowed you to form subgroups of patients, who had, for example, a diaphragmatic myocardial infarction as opposed to an anterior, who had heart failure or didn't have heart failure, who were men or women or whatever you wanted. And then the computer would come back and say, We've got ten of these or twenty and here's what happened, the survival at 6, 12, 24 months, number having symptoms at each interval, etc. Eventually, the MIRU contracts were replaced by SCOR grants (Specialized Centers of Research in ischemic heart disease). There were nine of these. They'd broaden the scope of it so that you could look at other kinds of ischemic heart disease, and you could go out as far as you wanted, not limited to five days. Andy Wallace was still the principal investigator. And then Al Bartel, who was a cardiology fellow after I was, came and said, 'Can we do this with the cath lab data? I'll collect the data and bring it to you; you put it in the computer.' And then I said, 'We'll start following the people, and then you can do the same kinds of things with the cath lab data that we were doing with the CCU data. The cath lab data had the advantage that patients were either treated medically or surgically, so we could examine the difference in outcome with different therapies. We developed a report, called the prognostigram. So a physician could find out what happened to patients similar to his patient, treated medically or surgically, and make a more informed decision. This is what a doctor tries to do with every patient, basing decisions on experience with similar patients. But, there was no way for a doctor to do this, because a doctor can never see back every patient that he ever takes care of, because they move or he gets too busy. If he saw a hundred patients the first year, and then the second year he saw a hundred patients and another hundred came back; that's two hundred. And then the next year he saw two hundred, another hundred, three hundred, and after a while you just can't do it anymore. So the idea was, the individual physician can't follow them, but we can. So we did the follow-up. We had to hire people to do the follow-up. Bernie McCants was the first and leader of this effort.
Then Frank Starmer, who had worked with Jim Grizzel, who was, I think, the chairman of biostatistics at UNC'went over and recruited Kerry Lee as the first statistician. Frank Starmer was a statistician, too, but I Kerry was the first specifically hired in this role. Then Frank Harrell came later. So we had an extremely good statistical unit.
Of course, collecting all these data was expensive and we wanted the databank to be part of clinical practice. Meanwhile, at the University of Vermont, Larry Weed had developed a medical information system on a computer, which allowed the house staff to do their workups on the computer using frame with specific choice. This in essence meant the reports looked like text but were really coded so that the information could be retrieved by a computer program. Paul Elliott and Sandy Simon created a similar system for us. And, it became possible for the doctor to generate a cath report, and at the same time the data were now in a form that we could take into the databank electronically. The folks in the cath lab had been filling out one form to collect the data and a separate form for the data bank. Now, they could fill out one form and they'd get a report and we'd get the data. So in the process of generating something that we were going to do clinically anyway, we got the data electronically. And then we could put it into our analysis system, and we were off and running. And mostly the databank wound up being centered around ischemic heart disease and medicine versus surgery.
Then, interested faculty, fellows and students could do research on the data, with the statistical help of Kerry Lee and Frank Harrell. We were surrounded by wonderful people. I mean, just the best people. Rob Califf. I don't know if you're interviewing him. Fred McNeer, Brant Mittler, Ted Fraker, Ron Yatteau, Marty Conley, Jim Margolis, Joe Kisslo, Gary Stiles and Ray Gibbons were some of the folks who did research using the information in the data bank. I guess what strikes me most looking back is all the wonderful people. Too much?
ROSEBERRY: I'm fine. Thank you. So who were some of the other key players in the databank?
ROSATI: Well, they were the surgeons, Bob Jones, Newland Oldham, Bob Anderson who went away and then came back as the chief of Surgery and of course Dr. David Sabiston who was the chief of Surgery at that time. And all the people in the cath lab like Vic Behar, Dave Kong, Jess Peter and Jim Morris. Glenn Young was one of the original surgeons and I think Dr. Will Sealy had retired by then. Then there was also Dave Clark who was another programmer, statistician kind of guy, and his wife who was our receptionist. And Dr. Stead and Andy Wallace were always present.
And certainly one of the most important contributors was Phil Harris. He was a fellow from Australia, and I think I introduced him (laughing) one time as, 'This is Dr. Harris, and I work in his lab.' He was a Rhodes Scholar. He was a very good scientist, and he wrote some of the really key.
Califf, of course. I think I already mentioned him. Dave Pryor, who took over as director of Clinical Epidemiology when I left, and Dan Mark and Mark Hlatky were other key people.
ROSEBERRY: So Surgery was involved right away?
ROSATI: Yeah. Well, pretty much right away. The surgeons wanted to know what was happening to their patients, just as the cardiologists did. And we started generating their op reports so we could get the data about what happened at the operation. The databank was the repository for all this information. It was a very collaborative effort. We were building the databank going forward. It wasn't like'I mean, I started doing this in '69, and that's when the first bypass was preformed. But, we didn't start collecting cath data until 1971. So, we went back retrospectively to the first bypass which was done in 1969. The MIRU had started in 1967 and here we started prospective collection in 1969 (maybe early 1970). So, we went back retrospectively to 1967, to the first MIRU patient. You could never do it now with HIPAA. You could never do it. (laughs). We went to the review board. What do you call those? IRB.
ROSEBERRY: IRB [internal review board].
ROSATI: The IRB, the IRB people to get permission to collect that data and they said, 'You're not doing research. You don't need permission.' (laughs)
ROSEBERRY: Excellent!
ROSATI: About 1982-83, I decided that bypass surgery was not prolonging life or preventing infarction in the vast majority of patients. So I got into cardiac rehab, and quickly realized that if you didn't change your diet, then you weren't going to get a maximal benefit. And most people were not changing their diet, and that's how I decided to go see Dr. [Walter] Kempner, again with Dr. Stead heading the way, and started working at the Rice Diet and was just absolutely amazed at the kinds of things that happened to people when they took the salt out of their food and ate less fat. And that's how I got on this career path, and I've collected information about the patients here, at the Rice Diet and followed a groups through six years, but I never published any of that information. It was just for my own interest. So the interest in finding out what happens to people, subgroups, is still there.
ROSEBERRY: You mentioned Dr. Stead. I wonder how much involvement he had later, after he had retired maybe.
ROSATI: Oh, well, he was retired, because he retired in '67, when I was an intern. He went up to Cornell to spend a sabbatical year and Dr. Wyngaarden took over. Dr. Stead got out of here to let him be the man. So there was no Dr. Stead that year unless you went to New York. But then he came back and had an office, and that's when I went to see him after he'd come back. It was at times a difficult endeavor. It wasn't thought of as real science, because real science was animals and test tubes and cells and things like this, and it wasn't clinical, cause clinical is taking care of patients. He was always encouraging. And if you had an intellectual problem, how do we do this, he was always there. I mean, he wasn't present on a daily basis or anywhere near a daily basis, but you could always walk down the hall, and there he was, and go see him, and he was'he was just an unbelievable person. He was always helpful. (laughs) This has nothing to do with the databank, but when I started working with Kempner, I was really on the lunatic fringe. Because Kempner and the real world didn't, they didn't go together. And I was running the cardiac rehab program, and that was the real world, and then I was working with Kempner: that was lunatic fringe. And so I decided I needed an assistant, and I was going to hire a physician's assistant to help me work these two places, and this had to be a person who could cross the vast gap between the real world and never-never land. And so I found this young lady, Carol Blessing-Feussner, who ran the hypertension clinic. She was a PA. And of course Dr. Stead had started the PA program way'(laughs) actually Wallace had. And so I went to Dr. Stead, and I said, 'I've been talking to this young lady, and I think I she could do the job; I wonder if you'd go down and talk to her?' And he said, 'Certainly.' He got up and walked right out of the office, and I went and talked to his secretary, Bess Cebe, who was another wonderful person. And talked to her about fifteen minutes or so, and I started to walk out, and met Dr. Stead's returning. He said, 'Hire her.' I said, 'Did you even see her?' He said, 'Yes'. I said, 'Did you talk to her?' He said, 'No'. I said, 'Well, how do you know I should hire her?' He said, 'She smiles a lot. You need that.' (laughs) And he was right. She was a gem. But I mean, he just had that'he could get down to the essence (laughs) of what was really important.
ROSEBERRY: Very good.
ROSATI: Great guy. I miss him.
ROSEBERRY: You had mentioned the prognostigrams as being kind of the'maybe your vision of what the databank could accomplish, and I wonder if you could tell me a little bit more about that and about them and whether that came to fruition at all? Whether they used'?
ROSATI: We did use them. All the physicians would order them. We never tested to see if they thought they were helpful, but they kept ordering them, so I guess they thought they were helpful. Later, we could match more completely using statistical models instead of subgrouping. But, somehow, I think we lost sight of the original idea. You see, most therapies only help a small percentage of patients. Since, we don't know who these patients are, we treat everybody. Somehow I always thought the databank could help us to identify the small percentage who really needed the therapy. Mostly, instead we are stuck with the average answer.
ROSEBERRY: Okay. Well, let's see, so tell me about the emergence of the clinical trials and'
ROSATI: That was after me. Then I went off into doing cardiac rehab and working with Dr. Kempner, and I never looked back.
ROSEBERRY: Okay. But, what was fun about the databank?
ROSATI: It was the people and looking at the problems and trying to find answers to important clinical questions.
ROSEBERRY: How would you describe your own leadership style?
ROSATI: Oh, It was pretty much laissez fare. You find some good people who are going in your general direction and you go with them. You influence them, they influence you.
ROSEBERRY: Well, what was the direction, kind of the stated path that you were hoping to take from the beginning?
ROSATI: You mean me, personally?
ROSEBERRY: Sure. Were the'when everyone was moving in the same direction, what was'?
ROSATI: That was it. I wanted to figure out a way to collect the information in the practice of medicine without spending extra money doing it, which we did, I think. And then to figure out some way of finding out what happened to people, which was going to take extra money, be cause you have to hire people to follow people up and make it available to doctors to know what happened to the patients they took care of. And I think we did that. But I did have fun doing it, so'and I never worried too much about what I planned to be. (laughs) Whoever would have thought I'd wind up here?
ROSEBERRY: Well, Dr. Rosati, what have I not asked you today?
ROSATI: I think you've gotten more then you wanted. (laughs)
ROSEBERRY: It was great.
ROSATI: Thank you. (laughs)
ROSEBERRY: Thank you very much. I really appreciate it.
ROSATI: I hope I didn't ramble too much.
ROSEBERRY: Not at all.
(end of interview)
|
Questions? Comments (919) 919-383-2653 dumc.archives@mc.duke.edu Last modified: September 27, 2007 |
|