| Introduction | Overview | Timeline | Credits & Copyright | Archives | heart beat

Dr. Galen Wagner's Interview


Dr. Galen Wagner, MD. Associate professor of medicine, cardiology, DUMC.

Dr. Galen Wagner was director of the Duke Cardiac Care Unit (CCU). He was involved in the early uses and development of the databank, and he directed fellows who did databank research. He also organized the Duke Cooperative Cardiology Fellows Program (DUCCS) for fellows to use databank information after leaving Duke. In these two interviews, he speaks about the history and philosophy of the databank concept from his perspective.

Click here to go back to timeline.

INTERVIEWEE: Dr. Galen Wagner

INTERVIEWER: Jessica Roseberry

Interview 1 of 2

DATE: February 14, 2007

PLACE: Dr. Galen's office in the North Pavilion

JESSICA ROSEBERRY: This is Jessica Roseberry. I'm here with Dr. Galen Wagner. He's associate professor of medicine in cardiology. Today is February 14, 2007, and we're here in his office in the North Pavilion, and we're speaking in particular about his recollections of the Duke Cardiovascular Databank. Thank you very much, Dr. Wagner, for agreeing to be interviewed today. I appreciate that.

DR. GALEN WAGNER: You're welcome.

ROSEBERRY: I thought I might ask you just about some initial recollections of coming to Duke, if that's all right, if we could start that far back.


ROSEBERRY: Sure. If you don't mind telling me just kind of what your impressions were when you first came and kind of what you were assigned to do.

WAGNER: Well, I came to Duke as a college student in September of 1957, and I had come on a visit during high school, just to view the campus and talk to some people. And my interest was in finding a place that was about one-day drive from where I grew up in Pennsylvania, near Pittsburgh, and a place that had a strong premed program. So.

ROSEBERRY: And were you looking at cardiology at that time?

WAGNER: Oh, no. I was just interested'I was an undergraduate. I was seventeen years old, and I was just looking to find someplace where I could get a good general college education and a place that had what was considered a premed program that was pretty comprehensive, had a good reputation.

ROSEBERRY: What was the medical school's reputation then?

WAGNER: I didn't know anything about it, just that Duke had one. At that point, as a student, you're not really looking for any particular thing about a med school; it's just that there is one there, or that they have quite a good record of getting people who go through their undergraduate program then moving along into medical school.

ROSEBERRY: And did you attend Duke's medical school as well?


ROSEBERRY: And maybe you can tell me now about how cardiology came about for you.

WAGNER: I suppose that during medical school I had some particular interest in cardiology, although I don't remember how strong that was. I do remember that during medical school I worked in the summertime. Those days, we still had summers off, and so I worked in the summertime in a hospital near my home, and the internists'I was interested in internal medicine. The internists I worked with were particularly interested in cardiology, and I think of the things in cardiology I was most interested in, the electrocardiogram was most foremost. And I had some education during those summers in interpreting electrocardiograms.

ROSEBERRY: What was the field of cardiology like at the time?

WAGNER: It was very much focused on heart sounds, the auditory kinds of things, picking up heart murmurs, being able to have, quote, 'good ears' so that you could hear heart murmurs. Most of the heart disease that we were concerned with was valve disease in adult situation, and there was a lot of predominantly rheumatic heart disease present, which of course meant that there were a lot of abnormal valves, and you had to be able to listen and hear sounds that told you which valve was most involved. The only real diagnostic test would be the chest X ray, and the ECG. Cardiac catheterization was there, but very few patients got catheterized, mostly those who might be considered candidates for surgery, because heart surgery was in its early stages, particularly open heart surgery, and therefore patients with these various severe valve disease, anybody that had congenital disease that lived to be an adult would potentially get surgery, so catheterization was done usually as a way of evaluating for possible surgery. Coronary artery disease was something that didn't make any noises that you could hear with a stethoscope, and had very little that you could do for the patient except medical management, and so there was no surgery for patients with coronary disease.

ROSEBERRY: Who were some of the people at Duke who were important in cardiology during the time that you were beginning your interests?

WAGNER: There was no such thing as a division or section of cardiology. There were cardiologists who did various things. There was the group of private cardiologists who saw private patients, and the foremost of those was Edward Orgain. Orgain had come here during the 1930s and was the leading clinical cardiologist. He'd spend his days in the clinic seeing patients, and then he'd round on the patients in the hospital. He had a big inpatient service. Then there was the cardiac cath lab, and Henry McIntosh was the chief of the cath lab, and these were people who really spent their days in the cath lab doing these very complex'not so many caths but complex caths, and these guys also had their experimental work with animals or whatever, so they divided their time between patients and experimental animal physiology. And then there were a couple of basic science laboratories in cardiology. Joe Greenfield had a basic science lab. There were a couple of others.

ROSEBERRY: Was there overlap between these sort of'?

WAGNER: Not much, because any training was done, you were an 'Orgain fellow' or a 'McIntosh fellow' or a 'Greenfield fellow,' so you signed up with an individual to be trained after you finished your internal medicine. There was no such thing as being a 'Cardiology fellow.'

ROSEBERRY: And what was your training?

WAGNER: I had never done any research at all during medical school except one project with a neurologist, with a family with a particular kind of neurologic problem that I followed up on, wrote up. And I was really interested in clinical medicine and interested in being a clinical physician, internal medicine, possibly thinking about doing something special in an area like cardiology. But I ran into a surprise in that in those days, the interns in medicine were the ones, the only ones that could actually write orders on patients and have them done. Even residents or attendings'any orders they wrote had to be countersigned by an intern, so it was very much the interns' service. It was the job of the residents to teach. And so as an intern, I was busy trying to learn how to be a general doctor, and when I became a junior resident then the next year, then of course I was to step back and let the intern be the person that did the 'doctoring'; and my job was to be the teacher for the intern, the students and the nurses and everybody else on the patient care team. That meant I was the one to evaluate the patient, go over the patient with the others and then go off and try to find information that might guide diagnosis and therapy. I found that when'the information that could be used to guide diagnosis was pretty definite, and usually I could come back with some pretty definite things to do, which the intern then would say, 'Oh, fine. I'll go do them.' But when it came to how to treat the patient, then it was very unusual that you'd find a very definite treatment. And so you'd have to come back saying, 'Well, this article says this; this article says that. This attending says this; this attending says that.' And it might vary as widely as 'have the patient into vigorous exercise versus keep the patient in bed for a month.' 'Fully anticoagulate the patient or give blood transfusions,' But you couldn't find very good evidence, and so consequently what I ended up doing usually was saying to the interns, 'Well, you could do this or that, and let's kind of follow the patient and see, but I really don't know what's best.' And it was that uncertainty that was kind of in my mind, although I wasn't conscious of it, on the day that I received in the mail a piece of paper that said I should'it was my contract for being a senior resident the next year. And most people did three years of internal medicine: intern, junior resident, and senior resident. And so here I was in November, December, being faced with: What are you going to do next July? And I found I didn't want to sign that contract. So I went to talk to my chief resident in medicine, who sent me to the chairman of the department, who was Eugene Stead, and I remember when I told him of my dilemma, he made the interesting statement that there were, of course, two ways to learn: one was by sitting at the feet of the master and seeing what the master had to teach, and the second way was to try to do something yourself; and, 'Maybe you are tired of sitting at the feet of the master; maybe it is time for you to do something yourself.' And I''Well, okay. It sounds okay to me.' But I wasn't quite sure what he meant until he mentioned the word 'research' and said, 'Well, why don't you see if you can find somebody around the Department of Medicine who is doing research and has some money and is willing to pay you as a research trainee next year? If you can't find such a person, why don't you come on back, and I'll give you a month to make up your mind, but by then, you have to either decide you're going to take this senior residency or not. I can't hold the place for you beyond a month.' So I talked to a variety of people, including a senior faculty guy who was a cardiologist named Andy Wallace, and I was very fortunate because Andy had, about a year before that, actually begun the Duke Coronary Care Unit [CCU], which was a ward in the hospital, which was the first time that patients with a single problem were kept at a single place. And the reason was that by that time, they had figured out that many people who had myocardial infarcts would have sudden death. And so they better monitor the patient's rhythm to see if there was something that was a warning or whether the person just all of a sudden stopped. The technology of monitoring was such that they had these beds where the patients had continuous ECG monitoring. Previously you would take a 12 lead ECG once a day But now you would also take a single lead ECG continuously. The patient would be on the CCU monitor, so that you could see some warning arrhythmia before it became serious. So because of monitoring, our CCU was developed, so that all patients with a certain problem could be at a certain place, and the nurses had better be particularly trained to watch the monitors and be able to respond. The other thing was that by that time defibrillators had been invented, so that meant that if a patient did have a cardiac arrest and you did find that they were actually fibrillating, you could then shock him and often get him back. Also pacemakers had been developed, so now you could put wires into hearts and you could pace slow rhythms. So the availability of the diagnostic stuff that was there via monitoring and the therapeutic stuff that was there by defibrillators and pacing made it sort of imperative that a, quote, 'first-rate' or tertiary care or whatever one wants to call it medical center should have a coronary care unit. Andy Wallace had established a unit like that with ten beds in Duke Hospital, and in the first year, Andy had taken care of everything himself, but then the next year, really the year I was a junior resident, he had two people who were cardiology fellows, and they alternated weeks on the CCU, while one of them was going to be doing to be doing something else the following year, so Andy said, 'Well, how'd you like to be every other week the CCU fellow?' Sounded good to me. The alternate week, I'd be in a dog lab because Andy also had an animal physiology lab, so on the alternate weeks I'd be the fellow in the dog lab, doing various kinds of experimental stuff with whole animal models that Andy was doing in his lab.

ROSEBERRY: Now, am I right in thinking that there were private and public patients on the CCU?

WAGNER: Yes. If it was private patient, then Ed Orgain or another private attending would continue to be the physician of record, and Andy Wallace's only responsibility would be to make sure the nurses, the interns, the residents, et cetera were taught and that special procedures were performed. There were ten beds; these were all private rooms, and we made no differentiation between private or public except it was the house staff with Andy's supervision were responsible for the care of the 'public patients.'

ROSEBERRY: You had mentioned that the nurses had a very specific, important role there. Do you mind elaborating on that as well?

WAGNER: Nurses tended in the hospital to be in passive roles; they could suggest to a physician what orders might be written, but unless the physician wrote the orders, it wouldn't be done. But in a situation like the coronary care unit [CCU], there really wasn't time for this process, so the nurses had to be empowered, based on their education, to actually do things under, quote, 'standing orders.' For example, 'If a patient has a cardiac arrest, and the monitor it shows ventricular fibrillation, you shock the patient.' And so one of Andy's jobs was to find nurses around the hospital who could be CCU nurses and put them through an education program.

ROSEBERRY: Can you tell me a little bit more about Dr. Wallace?

WAGNER: Andy had come to Duke as an undergraduate from Columbus, Ohio, probably four or five years before me. Andy accelerated and actually went to Duke Med School after just two years. He knew he wanted to be a cardiologist, and did a lot of cardiology research with mainly Harvey Estes while he was in medical school. Andy also was really a top-of-his-class kind of student, Dr. Stead recognized he had a very capable person, and so, as he would in those days, he accelerated Andy through the house staff training, so that when Dr. Stead decided to have a coronary care unit, he asked Andy, then, to start it. When Dr. Stead decided he was interested to have a physician's assistant program, he asked Andy to set up the education program Dr. Stead had been part of an NIH [National Institutes of Health] group to decide if they should fund research in myocardial infarction. Andy had then had two years in the Public Health Service at the NIH in Stan Sarnoff's Cardiac Physiology Lab. He returned as Chief Medical Resident when I was in medical school. The recommendation was to fund myocardial infarction research units, or MIRUs. Dr. Stead's contribution was, 'Make sure that every application you consider says they're going to do something with a computer.' And then Dr. Stead turned to Andy and said, 'You put together the application for the Duke MIRU. And of course, you've got to include something you want to do with a computer, because that has to be in the application.' And Duke, with Andy as the leader then was one of the five institutions that got awarded a MIRU contract. The application might have been in 1966, because when I was a CCU resident in May 1967, I heard that the computer was being installed. This process was under the leadership of Andy Wallace.

ROSEBERRY: Do you know why Dr. Stead felt strongly about computers?

WAGNER: His feeling was that one of the great weaknesses in clinical medicine was the ability of physicians to really remember the characteristics of all of their patients, and, to the extent that experience helped a physician, a lot of it was because of being able to recall, 'I've seen someone like that before, and I did thus and so and had so-and-so outcome.' Dr. Stead was aware that such a recollection was based on the physician's limited memory, so maybe the physician actually had seen forty patients with that problem but was using the experience with the he remembered and ignoring the experience with the thirty-nine others that he just didn't remember. Dr. Stead was very practical, and he understood that the value of the computer was to have memory that was perfect and therefore far, far better than any human memory. His concept was that doctors would do best if they really had access to vast experience with patients who were like their new patient. The doctors then could use their judgment to ask, 'What's the characteristics of my patient that are most important?' and then search in a computer memory to ask, 'Who have I seen before that was like that patient in these important ways?' So Stead's concept was that in coronary care, since we were caring for only one clinical problem'that is, myocardial infarction'we should determine the twenty or thirty most important variables about patients who had that problem: 'how old they were, had they ever had a previous infarct, did they have diabetes, did they have hypertension, et cetera.' We should make certain for every one of the patients we saw, we enter those variables into a 'databank' The databank, in his mind, was useless really unless you knew what happened to the patient's outcome over time,. Once you entered CCU data on the patient in the databank, you're responsible every year for follow-up information: 'Were they alive? Did they have heart failure? Had they had another infarct, et cetera.' Andy Wallace, when he started the CCU, developed a form to be completed on every patient. I became a CCU fellow in July 1967, and then when it was clear there was going to be a computer, my job was to take and expand that form based on the experience Andy and I had and the others had gained. We would collect a certain amount on admission, and then each day, we would collect a certain amount of information as to what happened: 'What happened day one? What happened day two? What happened day three?' And then when the patient went home: 'What are some summary information about the patient?' And then each year, 'What's happened?' Dr. Stead called this a 'computerized textbook of medicine.' He said, 'There are a bunch of textbooks, but this is a textbook that is based on experience, not just based on some wise person's opinion.' So if you go back to Dr. Stead's advice to me of the two ways to learn: most textbooks you sit at the feet of the master, and you're learning what the master wrote in the textbook. Now, I, as a young cardiologist, was producing a kind of a textbook which would be more useful over time, because it was based on ongoing experience. So in a way, the responsibility that I had to work with Andy Wallace and the others to develop that databank really solved my problem as a young physician of, 'How do you now get another way of finding information about how to treat the patient?' When the question is, 'Who gets a pacemaker?', rather than going to some textbook or asking Dr. Orgain, I can consult the databank: 'Who have we ever put a pacemaker in? How many patients really had a slow rhythm? How many got pacemakers? How many didn't? And what happened to them? How many have had sudden death? How many of them needed a pacemaker later? How many of them had complications with their pacer?'

ROSEBERRY: Now, was it difficult for some people to support, since perhaps you were no longer sitting at the feet of the master? It sound like it's kind of a new way of looking at'

WAGNER: This system makes a lot of sense for people who do not believe they know what to do for every patient. The private attendings, Orgain, et cetera, obviously had a tremendous experience themself, and for years they were the experts. And so for most clinical problems, they really believed they had the answer and were much less likely to consult the computer. The private attendings would sometimes come to me and ask, 'What does your computer say about this?' Usually those weren't the common problems. For example, coronary bypass surgery came along in 1970. There were many patients who had a lot of chest pain, that couldn't be managed medically, so when the cardiologists decided who needs coronary bypass surgery, they weren't about to look it up in a computer. They said, 'We've got the surgery here now; this patient's got chest pain, (makes clicking noise with tongue), let's go do it.' It wasn't that the senior cardiologists didn't use the system, but the younger doctor without experience was of course more likely to consult it on a regular basis.

ROSEBERRY: How difficult was it to make a query?

WAGNER: Oh, it was very easy. Because our computer file was a fixed-length file, like a spreadsheet. In other words, we collected exactly the same pieces of information for every patient we saw, so if the patient was there for two hours and died, we collected the same amount of information as if the patient was there for two weeks. Now, because of that, the computer was organized in such a way so that for patient number one, let's say a variable such as, 'Does the patient have heart failure?' would always be in column number fourteen, so for patient number two, that variable was also in column number fourteen. So if I asked, 'Let me see all the patients who have heart failure,' the computer could very easily (makes clicking noise with tongue) just give me the information in column fourteen. It was lined up in such a way. So on rounds, it was very easy. I had the computer terminal, and if I saw a patient, let's say, that was a young woman with her first myocardial infarction and she didn't have diabetes or hypertension and she had an anterior infarct location, I could ask, 'How many women under the age of so-and-so have we seen with their first myocardial infarction, no hypertension, no diabetes?' and bang, there was the answer. And then I asked, 'How did they do over time?' Well, to the extent they were followed'of course, in those days we didn't have very long follow-up, but at least for however long there was, you could find out what happened to them: how many of them were alive in a year, how many of them had this, how many of them had that. I could do my searches, my queries, right there on the screen, so it was really very easy. Let's suppose a patient had an unusual situation. For example, I remember one time there was a patient who had done very well for the first few days and then suddenly, just as we were about to transfer to the general ward, the patient had a cardiac arrest. The monitor showed no arrhythmia at all, just normal rhythm, but the heart stopped. That meant, even though there was a rhythm, there was no contraction, and we couldn't resuscitate the patient. So they said to me, Have we ever seen this before? And I said, 'I don't remember anybody like that. Let me look it up.' Well, there were ten patients we had seen like that before. Seven of them had autopsies, and we could ask, 'What did the autopsies show?' But that was a way that we were able to use the computer as a complete record of our experience.

ROSEBERRY: Now, you were directing the CCU at this time.

WAGNER: It was just a lucky thing for me in that I began as a fellow in the CCU in July of 1967. I was two years out of medical school. On alternate weeks, I was in the dog lab. By the end of 1967, Andy Wallace said, 'I'm taking sabbatical. I'm going to go do more basic science work. I'm not going to be on the CCU,' and he asked me if I would take over as the director of the CCU. So here I was, just about to have my twenty-eighth birthday, and I was now the CCU director. My colleagues were senior residents. I mean, they weren't even in cardiology yet. I, six months into cardiology, was the CCU director. So by that chance situation'I was at the right place at the right time'I became the Duke CCU director January 1 of 1968, which'by the way, was the day that the Duke cardiology division was begun. Henry McIntosh moved from being director of the cath lab to become the first chief of cardiology.

ROSEBERRY: Were you putting energy and effort into creating this databank, or was that Dr. Rosati's'?

WAGNER: Bob Rosati graduated from medical school two years after me in June of '67, so he would have been a medical intern when the databank began. It had been my job and the other CCU fellow, Doug Zipes, to work with Andy to develop the variables that would go in the databank. I think I was more interested than Doug, so I probably did most of it. And then it was my job to collect the data on the patients and put them into the computer: I would either enter what I could on my daily CCU rounds, and then I'd sit with one of the data techs once a week to get all of the data into the computer. I had a follow-up clinic, so I would see the patients back in follow-up, and I'd make sure and put the outcome data in. As CCU director, it got easy because I had two fellows working under me, so now they could, of course, be doing a lot of what I had primarily done before.

ROSEBERRY: How did the biostatisticians come into the picture?

WAGNER: There were no biostatisticians back then. The person who worked in some ways as a biostatistician was Frank Starmer. Frank was an engineer, who had been somehow found by Dr. Stead, and he had begun to bring Frank as the first nonphysician onto the faculty. Frank had a big role working with Andy Wallace in deciding what computer to buy and how to set up the facility. Around the corner from the CCU, there was a patio in the middle of the current Duke South, which was Duke Hospital at that time. The hospital had its outside windows, and then rather than being a solid structure, it had these open areas where there would be windows from a patient room into an internal patio. It was a big square area about, oh, maybe thirty by thirty feet. And of course, it went up to the sky. So the decision was made to build a 'tin hut.' It was some kind of a metal building (chuckles) with a roof on it. It was one-story high with a ramp going up to it. The floor was elevated by about two feet. You had to have an elevated floor because you had to have fans underneath to keep the computer cool. And the computer was a 'Sigma 5' computer. It took up I'd say all the central 75 percent of the space. And around the computer was a little place you could walk, and various people had their offices around there. Frank had a certain number of people that were helpers, and I suppose they were computer programmers. Even after Andy turned over the CCU to me, he remained the MIRU director. One of the things that Dr. Stead believed in was that if somebody came into a medical center who was not a physician, that they should be trained, to a limited extent, as a physician. So Dr. Stead asked me to do was to have Frank Starmer take physical diagnosis in my group, along with the medical students. So I had four or five medical students, and I was teaching them how to do history and physical, and Frank was right there with them. And so Frank went through that process so that Frank learned, 'How do you go and talk to a patient? How do you examine a patient? How do you put the information in the record?' Subsequently, Dr. Stead decided that we really needed real statisticians, and so he made a deal with the guy that headed up the School of Public Health at Chapel Hill, that some graduate student in statistics would, on a part-time basis, come and be the first real professional statistician at Duke and work with the MIRU. And so he got Kerry Lee, and Kerry can tell you the story of what happened with him. And Kerry shared an office with Brant Mittler, I think, when he first came. Kerry also became a student in my physical diagnosis course, and so also, learned how to do this process of, 'How do you get data from patients?'

ROSEBERRY: Tell me about the cath lab at that time as well.

WAGNER: The cath lab, up until 1970, was a place for studying patients with congenital and acquired valve disease. The only reason you'd ever catheterize a patient with coronary artery disease was to make the diagnosis. If you weren't certain whether they had coronary disease or not, then you'd do an angiogram of their coronary arteries to look. But when, in 1969, coronary bypass surgery was developed, then David Sabiston, who was chairman of Surgery then, said, 'Okay, that's something we need to be doing.' And as Duke surgery began to do bypass surgery, Duke cardiology began doing the catheterizations in people with known coronary disease, to find out where was the disease and then to consider with the surgeons if bypass could be done and how to do it. The databank, which had begun on the CCU as a fixed-length file, then was also adapted for the cath lab, so from the very first patient who had coronary angiography with the purpose of evaluation for bypass surgery would have their data put into databank. But a difference was that every patient who was catheterized'since it was a procedure, there had to be a report generated, and somebody had to type it up. So they had the idea, 'Well, why type up a cath report? Why not put the data into the computer and have the computer generate the cath report?' When you do that, you can't have a fixed amount of data for every patient. The data have to be customized according to what was done in the cath lab. So that means if you had a very simple procedure, you have a little bit of data; if you had a major procedure, mere data, the data that went into computer was put in a variable-length file. So that means that a particular variable wasn't lined up in the same column. To do research, you had to write an 'extractor' program to organize the data from this variable-length file into fixed tables. On the CCU, we just entered data into a fixed-length file. I remember one of the programmers who worked with Frank named Sandy Simon, and Sandy's responsibility with Frank was to build an extractor program. Our computer file that we used for the CCU was called the MIDAS [Mycoardial Infarction Data Analysis System] File. The file for the cath lab was called CIRCE. I'm sure that meant something.

ROSEBERRY: The work you were doing'did it more lend itself to chronic'or did the databank, I guess, lend itself to more chronic diseases or study of more chronic diseases?

WAGNER: In those days, you would not take a patient in the acute phase of coronary disease, so these populations didn't overlap very much. A patient who was being considered in the chronic phase for possible bypass surgery'their data would have been entered when they came to the cath lab. In the acute phase, the patients were intensely monitored because changes occurred rapidly in both rhythms and hemodynamics. Treatment decisions were primarily guided by serial physiologic data. Bedside interventions were done and results observed. As you have suggested, the databank was only valuable to support the CCU decisions with longer term outcomes such as whether to insert a permanent pacemaker and when to discharge from the CCU.

ROSEBERRY: Do you know about the funding, the continued funding for the computer?

WAGNER: A lot of the funding for the CCU component, of course, came from the MIRU contract, which went on for many years. In the cath lab, I'm sure some came from the MIRU, but that would only have been a start. They got some money from the National Library of Medicine and some insurance companies. There was a lot of piecing together of funding, but they realized they had to attach the databank to patient care dollars as with the cath report.

ROSEBERRY: Can you tell me what a prognostigram is?

WAGNER: Go back to this concept that Dr. Stead had, that you could use a computer as a textbook. Let's suppose that you arrive with with chest pain and get a catheterization, and I as your physician want to be able to give you advice as to whether you need coronary bypass surgery. A 'prognostigram' would be developed for an individual patient. The physician would identify the key baseline variables of that individual, put them in the computer, find out who we had seen like that before, and then identify which patients had been treated with bypass surgery and which ones with only medicine. Then let's suppose you were the patient. Your cardiologist would say, 'Miss Roseberry, we've seen fifty patients that are like you in these ways that we think are important. Twenty-five received surgery; twenty-five did not. The outcome of the twenty-five with surgery is quite good, whereas for the people without surgery, many who have had infarcts or sudden death. In fact, you can look at the way the outcomes diverge. Therefore your prognosis for you'the prognosis for you is better with surgery than without. Therefore I recommend bypass surgery for you.' And so giving advice to the patient based upon their prognosis as determined by this computerized databank, that was a prognostigram. We actually had forms, so we could do prognostigrams on individuals.

ROSEBERRY: Were those used often?

WAGNER: I don't know the answer because I don't think that any such record was kept. I would say in general not often, because of a variety of reasons, which usually centered upon the physician believing they knew what was best therapy without such a formal consultation with the computer.

ROSEBERRY: How often were you using these tools?

WAGNER: Very seldom. I was never responsible for the patients' chronic care. I was in a CCU teaching role, and so I was using the computer primarily to do clinical research projects. Many of these clinical research projects were based on the principle of the prognostigram. For example, I had recognized that in general patients were staying in the hospital a long time with their infarcts; and the typical thing was to stay for about a week on the CCU, even if they were uncomplicated, and then about a week on a 'step-down unit.' So I worked with Fred McNeer and others to devise a study to see how could we use the prognostigram principle to tell us which patients, based on their early course, would be predicted to have such a low incidence of late complications and so could be sent home early. The specific question was: 'How long do you have to watch a patient be stable before you can predict they're so unlikely to have a serious complication, you can safely send them home?' So on the CCU, the intervention wasn't bypass surgery, the intervention was hospital discharge. In the mid-1970s we did a study showing that if a patient with acute MI remained free of serious complications for three days, their prognosis was so good that you could then prepare them for early discharge, to go home at the end of the first week. We showed those who were sent home did just as well as those that stayed in the hospital. That was a major breakthrough because then you didn't have to keep patients around for an excess period of time, just wondering whether they're going to get a complication. So that was a way in which the prognostigram principle was used to change care for people with acute infarcts.

ROSEBERRY: Do you feel that the databank was used more often for those kinds of research questions than for patient care?

WAGNER: Yes. One of the things I did was to make sure that every fellow that came through the CCU had a research project, and many of those projects were projects that related to using the computer for its prognostic capability. For example, we saw patients who had a block in rhythm between their atrial and ventricular chambers, so-called AV block [atrioventricular block] during their infarct. We had to treat them with temporary pacemakers, but almost always after three or four days, their rhythm was back to normal. So the question was: 'Which ones of those need a permanent pacemaker?' So some we'd put permanent pacemakers, some we wouldn't. There were other patients who, in acute phase of their infarct, didn't have any slow rhythm at all but had conduction problems on their ECG: left or right bundle branch block. Although every beat was conducted from atria to ventricles, perhaps they needed a permanent pacemaker? But since their bundle branches were infarcted, maybe after they went home, they would have syncopy or even sudden death. Well, we did a study to ask: 'Okay, how do we actually predict, based upon watching the patient for a few days, which ones are going to need permanent pacemakers after they leave the hospital?'

ROSEBERRY: Can you talk a little bit more about the fellows and their relationship, their research relationship to the databank? Perhaps you've covered that.

WAGNER: No. I made teaching rounds, as the CCU director, every day from ten to twelve. My job was to go over all the new patients and review the care plan, and review the plans for the other patients and make sure the data got in the computer. A fellow would be assigned to the CCU for a three-month period of time. I would work then with each fellow to make sure I identified one research project that could be begun during the time the fellow was there, and then I'd work with the fellow to complete the project even after they left the CCU. In those days, if a fellow was in the cath lab, the people in the cath lab made sure they did research, even on the Orgain private service. It was part of the faculty's responsibility to do a research project. It was always a matter of trying to find a project that related to the interest of that individual fellow and something that was also practical for the care of the CCU patients.

ROSEBERRY: What would some of those projects look like?

WAGNER: I've mentioned two of them relating to early discharge or pacing. Another project might be relating to heart failure. We used a lot of very potent diuretics if a patient had even noises (rales) in the lung or shortness of breath. But we found that if we gave too much diuretic, then we dropped their blood pressure. There was a only certain amount of 'dryness' that you wanted to have in a patient. One of the things that was developed in another one of the MIRUs, was the ' Swan-Ganz catheter.' We were afraid to use stiff catheters in these patients with heart attacks, but this catheter was not a stiff catheter. It was a flexible catheter that you inserted into an arm vein and then inflated its balloon, then with venous flow it just carried right on into the heart and through the right side of the heart into the pulmonary artery; so it was behind the left side of the heart. You could monitor the pressures in both the right and left sides of the heart and calculate the cardiac output of the heart. Once we started being able to use the Swan-Ganz catheter, we added fields to the databank, so that we collected data on the cardiac output, pulmonary artery pressure, oxygen saturations, et cetera. Then we could say: 'If patients have signs of heart failure, let's then randomly have half of them get, say, a diuretic; half of them not. And then let's look to see what happened to these, quote, 'hemodynamic variables.'' In other words, it wasn't just a matter of measuring the amount of urine or measuring the blood pressure or measuring how the person felt, we could actually measure, on a several-times-an-hour basis'what happened to the right and left pressure, and what happened to the cardiac output? We were able to do a lot at the bedside that was really experimental: we could measure patient data, do an intervention, and then in the minutes to hours afterwards, see what the intervention did. So before the patient got into clinical trouble from the intervention, you could observe that the data aren't going in the right direction. For example, one of the big problems was patients who would have so much infarction that they would go into shock, meaning they wouldn't be able to perfuse their body, so they would drop their blood pressure, become disoriented, become cold, they stop making urine et cetera. The standard treatment was to give them a drug that increased the resistance of their arteries and increased their blood pressure. But when you monitored that patient's hemodynamics, you found out that the cardiac output had actually dropped. So we asked, 'What if instead of constricting the patient's blood vessels, suppose we give a vaso-dilator? Maybe the blood pressure might decrease a little more, but the cardiac output might go up, and maybe even the weakened heart would perfuse their head, their kidneys, their vital organs. And so we totally changed from treating shock with vasoconstrictor drugs to treating shock with vasodilator drugs. We used that new strategy a lot, and of course the data went into the databank.

ROSEBERRY: Were there any things like that that you were able to use the computer as part of the study itself?

WAGNER: Yes. There was no need to use prognostigrams because each patient serves as their 'own control,' but searches of the databank produced the results that generated the fellows' clinical research projects.

ROSEBERRY: And were the fellows able to carry on some of the tools that they gathered during that time and take that further, beyond the fellowship in cardiology?

WAGNER: Sure, because obviously a fellow who went through this experience got education in ways that a fellow that would not have access to a computer databank didn't get. And depending on the fellow's interest, they might carry it on in various ways. Some fellows would be interested in the computer side, so wherever they went, they'd want to have a computerized databank. Other fellows would not be so much into the computer side, but they might care for their patients differently because of what they learned from the databank. One of the things all our fellows had to do was collect data, and so they came through a training system where data collection was required; and they learned things about their patients they couldn't otherwise learn.

ROSEBERRY: Do you know of any examples of fellows who maybe went on to have a fairly direct connection with what they learned?

WAGNER: There were fellows like Brant Mittler and Fred McNeer, who decided they wanted to stay on the faculty for some period of time and continue to be attached to the databank and do research using the databank. There would be other fellows, like Charlie Bethea and Alan Bartell, who would leave Duke and would then develop a databank where they went. The former group like McNeer and Mittler were mainly interested in using the computer to do research. These other group like Bartell and Bethea would want to use the computer to guide care of patients. They were interested in using the computer to help build systems allow they could use those data for decisions about patients they cared for in their practice.

ROSEBERRY: I know that you mentioned that you were not as much on the patient care side, but I'm going to ask you if you feel that maybe the databank was living up to Dr. Stead's original vision for it.

WAGNER: I would say that since his vision was a patient-care vision, the answer is no, because it wasn't his primary interest that the research be done. I would suppose that if you had asked Dr. Stead at that time, he would have predicted that every patient coming through Duke Cardiology would have major treatment decisions based upon these computer searches. And I think there were several reasons that did not occur. One was there was never really a computer-oriented Duke investigator who became very interested in testing the prognostigram in patient care. There was nobody who said, 'The thing I really want to do is to do research in the use of the databank for clinical care.' Nobody ever really emerged with that as their interest. I think if they had, they would have done the kind of research that would have tested the value of the prognostigram, the value of using the computer for care of individual patients. For a variety of reasons, there were very few of the cardiology fellows who had primary interest in doing research using the databank. One reason was that it was challenging for them to be able to fund a research career because they wouldn't be able to get money like a basic scientist does to do research. Another was that to the extent they were interested in clinical care, their time would be consumed by the clinical care. There were a few fellows like David Pryor, who came here because of the computer'he really developed a career in what was then called informatics. Dr. Stead's son, Bill, was interested in nephrology, not cardiology, and he built an outstanding career in informatics. Others, like Mark Hlatky, Dan Mark, Rob Califf'cardiology fellows who built academic careers in what I would call, in general, informatics. I think that what Rob Califf did is an example of an individual who used the databank to go in a different direction. When Rob came out of training, coronary care was changing because there now became the possibility of doing reperfusion therapy and not just allowing the clot that caused the infarct to remain but very aggressively going after the clot. And so a major decision that had to be made was, 'Which patients should best have their clot attacked, with a drug or a catheter?' That question could not have been answered by the prognostigram approach, since surgery was not required. Cardiologists would not have been willing to arbitrarily withhold reperfusion therapy while follow-up was being accumulated in the databank. Another way was the way that Rob actually adopted, the randomized multicenter clinical trial. You enroll patients who meet criteria for having reperfusion treatment, and then they get randomized, with half, then, getting the treatment and half not. And so the randomized trial, then, took the place of the prognostigram as a way of determining the best therapy. As soon as you randomized enough patients so that the statistician would be able to say, 'Now you have statistically significant results that the patients who got the treatment do better than the patients without, so stop the study.' Or if you collect enough patients so the statistician says, 'The outcomes are so much similar, you could collect thousands of patients but are not going to show a difference, so stop the study.' Rob became the individual at Duke who, starting with the so-called TAMI [Thrombolysis and Angioplasty in Myocardial Infarction] trials, built the doing of clinical trials into the culture at Duke and into making decisions about what's the best way to take care of patients with acute infarcts.

ROSEBERRY: I think you've touched on this some, but can you tell me more about some of the differences between the ways that now the databank is being used as to before randomized clinical trials, when they were being used in a certain way, and now they're being used by Dr. Califf?

WAGNER: One of the weaknesses of the randomized trial is the only patients who get included are the ones that fit criteria for the trial; so by definition, this is not everybody. It might be patients under the age of whatever, or patients who do not have this comorbidity, et cetera. The research databases used for the trial were really what I call 'numerator databases,' because they a were a portion of the entire population (denominator). Another weakness was that the database was specific for that trial. When the next trial was developed, it required whole new database. Now, those databases had the advantage of the data being very clean; you had 'data cleaning' so that you made sure there were no missing data, every parameter was as correct as it could be. To make sure that when you did your statistics, the data were as clean as possible, and so there were 'locks.' In other words, you had a time when the trial when the trial was over. (claps once) No more patients were enrolled. Then you had a time when you had cleaned the data. (claps once) The data are clean; they can't be changed after that. And then the data analysis was done. And so you had these time locks that were very defined. Every trial had a 'data and safety monitoring committee,' and none of these collecting the data could on it. They looked intermittently at the results, and if they saw that now there was a major difference between outcomes, they'd say, 'Stop the study, because it's unethical to continue on. We calculated that were going to need 1,000 patients, but now with 550 patients enrolled, we've got the statistically significant answer.' They were also looking at complications. Maybe they would find out that the patients who were being treated had significantly more bleeding than the others, 'Let's stop the trial.' After Bob Rosati stopped as director of the databank, Joe Greenfield, who was chairman of Medicine at that time then made Rob and David Pryor co-directors of the databank, David was mostly interested in the clinical databank, but after a couple of years he left Duke to take a hospital administration job in Boston. I recently sat with Rob and David, sort of reflecting, 'Well, suppose it went the other way.' Suppose Rob had left and David had stayed, because when David left, there was really nobody who was going to carry on the 'clinical database' with the same kind of fervor Rob then carried on the 'research database.' For a variety of reasons, Rob now is coming back to developing clinical databases actually beyond just the hospital. One of his responsibilities now is to develop the database that is Durham County wide, so data from all the citizens of Durham County don't even have to get sick enough to come into the CCU or cath lab. So I think that Rob now, in this stage of his career, is coming back toward building the clinical databases, which of course he was not building during the time when he was developing the randomized trials.

ROSEBERRY: Now, those clinical databases'just so that I've gotten it straight'are collecting all kinds of information on all kinds of patients, so you come in and you have a wide spectrum of maybe twenty-five, more, however many variables about you that are just entered in, and because you're a part of this, because you're in a certain program, at the CCU or whatever, everything is being entered, and so there are no restrictions as to what is'

WAGNER: There are always restrictions as to the amount of data elements entered; otherwise, it's impossible.


WAGNER: We created a denominator database on the CCU, and we were limited because our computers were so clunky they could only hold fifty pieces of data on everybody. In fact, I brought here today the first 'computer' I ever used. This computer is an eight and a half by eleven piece of cardboard, (Roseberry chuckles) and it has a bunch of holes all around the perimeter of the paper, and each hole has a number that was assigned a question with a yes or no answer. And so based upon whether the answer was yes or no, you'd either clip the slot or punch a small hole in it. There are a total of 146 holes around here, so we could only ask 146 questions with yes or no answers. But if question number one was 'male or female' and we decided that males would have the hole punched out and the female would have the slot removed, then if we wanted to find all the males, we'd stack the cards up, we'd put a knitting needle through the hole in slot number one. All the cards that stayed on the knitting needle would be males and all the ones that fell on the table would be females. If we wanted to find out which of the males were alive in a year and 'alive in a year' was question number 144, then we'd just the needle through number 144 and the ones that stayed on were all the males who were still alive in a year. So that was our searching method; we were limited by the number of data pieces, the number of holes that could go around the cardboard. When computers came along, they were also limited because there was only a certain amount of data that they would hold, so we had to only collect that number of pieces of data. But on every patient, we had to collect that number. Today, with computers being able to hold millions of pieces of data, it's not physically possible to have milliosn pieces of clean data on every human being: so now it's the human who has to set the limits. If Rob Califf wants to go collect information on everybody in Durham County, he's got to figure out, 'What are the important things to collect?' because you can't collect every darn thing about every darn person. You can collect data on every darn person, but you can't collect every darn thing. So you got to say: 'Here are the pieces that in my judgment ought to be collected on everybody.'

ROSEBERRY: Can you tell me more about what you mean by clean data?

WAGNER: Clean data means that in every data slot you've got the most correct answer. For example, here, on the cardboard, there are 146 slots. There better not be any 'don't know's.' It either ought to be 'yes' or 'no'; you can't say, 'Male or female? Well, I'm not sure.' Or, 'Age? We didn't collect that.' Or, 'Did they have a prior infarct? Well, we didn't ask them.' You had better have answers. And any time you have missing data, that's dirty because you don't have it. Also wrong data. Suppose you have, 'Does the patient have heart failure?' And let's suppose that you say, 'Yes, heart failure,' and you count it for heart failure, and somebody asks, 'What's your definition of heart failure?' And you say, 'It means that if they ever got short of breath.' Well, wait a minute. You can be short of breath for a lot of reasons. How do you know 'short of breath' was 'heart failure'? In other words, you had better have a definition of heart failure that'll stand up as a definition of heart failure, so if you haven't bothered to write a firm definition, the data element really ought to be 'short of breath'. Same thing with diabetes. The question is, 'Diabetes? Yes or no.' What's the definition of it? If the definition is, 'Were they ever told they had diabetes?' Hey, wait a minute. Is that clean? No, it's not clean. In a research database that Rob collected for the trials, they had very definite criteria, and if it said 'heart failure,' well, you can could be certain they had heart failure.

ROSEBERRY: So when you were first coming up with the variables, how were you determining what those variables might be?

WAGNER: We knew we were limited in the number of variables we came up with, so we had to make sure, and we'd sit around and ask, 'Do we really want to know if the patient had diabetes?' Yes. Well, then, 'How are we going to define it?' If it's so difficult to define, we had better kick that one out, even though it would be important to know. Hypertension. 'Has the patient had hypertension?' What do you mean by hypertension? 'Did they ever have a blood pressure over so-and-so?' You better have a pretty good definition, not just somebody said, 'Hey, you had high blood pressure.'

ROSEBERRY: And who all was coming up with these definitions and variables? That was you and Dr. Rosati?

WAGNER: Remember that the process began with Andy Wallace when we started the CCU in 1965, and accelerated when we got the MIRU computer in 1967. Bob had primary responsibility in 1969 for starting the databank in the cath lab. We all were responsible for having a definition for everything we wrote down, so if we said 'heart failure,' we had to have written down our definition of heart failure. The New York Heart Association classification said 'Grade zero is they don't have any heart failure.' Grade 1 is 'They get shortness of breath if they try to do something'; Grade 2 is 'They get shortness of breath at rest'; Grade 3 is, 'They can't even get out of bed.' The Killip Classification, for acute heart failure, Grade 1 was 'absolutely no evidence of heart failure'; Grade 2 was 'When you listened to their chest, you hear crackling rales'; Grade 3 would be, 'They're actually short of breath at rest'; Grade 4 would be, 'They're in shock'.

ROSEBERRY: Okay. How big was that initial computer? What did it look like?

WAGNER: I would say it was probably twenty by twenty feet, although it wasn't a solid block of twenty by twenty; there were, I suppose, modules that might be two feet by two feet, and they would be next to each other so that you'd have space to walk around or whatever, as a maze, through there. And each module was, at least as high as people were, maybe six feet. And they got hot, so they had to be kept cool by fans running underneath the floor. But it was still in hot in there because those things gave out a lot of heat.

ROSEBERRY: How many different computers were there?

WAGNER: There was one computer called a Sigma 5. Now, you had various components; like, you had tape drives and you had places for the cards. I never bothered to understand any of this stuff, but the people that did, each piece was a part of the computer. Today each piece would probably take up a little chip or something. But at that time, it took up a lot of space, and you had to kind of walk around it, and it was warm. To me, the computer was a tool I used for answering the questions I wanted to answer, and I was asking a lot of them. Frank Starmer first, then Bob Rosati were responsible for the computers. They were hiring the people to take care of it and fixing it if it was broke.

ROSEBERRY: So was it that one computer throughout the time that you were'?

WAGNER: As far as I remember. I'm sure they replaced some things, but as far as I knew, it was the Sigma Five computer. Now, there might not have been a single piece of it that was still the same at the end of the time as it was at the beginning.

ROSEBERRY: So tell me about the time when Dr. Califf came and began to be in the CCU and kind of take over the operation there.

WAGNER: During the time when I was directing the CCU, we had the ability to give thrombolytic drugs to break up clots, but we were only using these drugs for things like pulmonary emboli. We didn't use them to break up clots in the coronary arteries, for a number of reasons. We believed that any clot that was there was on top of a very tight plaque, and so if we removed the clot, we might take it from 100 percent blockage to, say, 90 percent. And so we believed that we would leave the patient with their tight obstruction; and since we were afraid to do a cardiac catheterization on such an unstable patient, we never would have taken the patient to the cath lab to find out. There was no such thing as angioplasty, so nobody would have thought of putting a catheter in and trying to break up the plaque. So we always thought it was best to just make sure the patient was as safe as possible during the time on the CCU. Let them have whatever damage they were going to have, but try to minimize it by putting them in bed rest, keeping their heart rate low, but don't try to do anything about the place in the artery that's causing the problem. Now, by the end of the 1970s, this was proven wrong, then the whole process of reperfusing patients now came in. I remember it as being a guy name Peter Rentrop over in Germany who began to do cardiac catheterizations in acute patients and gave drugs that could break up clots. He found out that sometimes what was left behind was really not a very tight obstruction, and sometimes a plaque that wasn't really a very impressive plaque had ruptured, or something, and then a clot formed. So when you got rid of the clot, instead of having a 90 percent obstruction, you had a 25 or 40 percent obstruction. Wow! That was brand-new information. So now giving these thrombolytic drugs'streptokinase, urokinase or whatever'became very important to do. We had been afraid if we did limit the size of the infarct during the acute phase but sent the patient out with that tight lesion, they might then close up again next week or next month and drop dead. At least we had gotten them through the acute MI process. They weren't going to die of it. So when it became known that you could reperfuse people with either putting a catheter in or putting in these drugs intravenously, then the job on the CCU became very different. The chairman of Medicine made the decision that no longer could the private cardiologist keep taking care of these patients on the CCU; the CCU director had to be the patient's physician. And so there were conversations about it, and very clearly I wasn't going to be that person. It wasn't my interest. That wasn't what I wanted to do. And so then the decision was made between me and the chairman of Medicine'Joe Greenfield'that I then would go totally into research, and somebody else would take the job of CCU director. And the first person was a guy named Eric Conn, who had come through the fellowship. Eric did it for a year and then he left, and then Rob Califf was given the opportunity, and Rob took over to do it and did it then from then on. So Rob then, as a co-director of the databank, was also the director of the CCU. That was very appropriate because in those days, patients with chest pain would come to their community hospital, where if an infarct was diagnosed, they would perhaps start a drug like streptokinase, put a patient in a helicopter or ambulance, send them to Duke. So patients were arriving at Duke on these drugs. And of course, in Duke they would do an acute catheterization, find out what was there, see if the blockage was still there, do something about if it was. So CCU became a very, very different kind of place, with very aggressively going after the primary problem, trying to open up that artery and trying to keep it open. So it was not so much watching, waiting, trying to keep the patient out of trouble; it was more aggressive: Let's get rid of the problem that brought him here in the first place. So the best place for me was somewhere else, where I could study the process but let somebody else, then, be the one that was there. And of course, Rob was that one, and he did iy wellt v, and he did it for a long time.

ROSEBERRY: Now, were those clotbusters part of the clinical trials that he was doing?

WAGNER: Yes, right, they were, and Rob was at that point working with mainly people that he had been in residency with at the University of California, San Francisco, to then set up what was called the TAMI studies, Thrombolysis and Angioplasty in Myocardial Infarction, because thrombolysis is the drug; angioplasty is the catheter going in and breaking up the plaque, and so a lot of question was, 'What should you do? Should you go in acutely and break up the plaque or should you just give the lytic drug and then perhaps take care of the plaque later?' So a lot of trials were set up, and the TAMI trials were set up, patients randomized to receiving this or that, TAMI-1, TAMI-2, TAMI-whatever, and so Rob really built up the clinical trials operation here based upon the Duke CCU being one of the sites. Now, he added other sites as well, so there were a number of TAMI sites.

ROSEBERRY: Tell me about the work that you were doing during this time.

WAGNER: I was doing a whole bunch of different kind of work, but you mean sort of the different pieces of what I was doing?


WAGNER: One of the problems I faced from even the early days was it was clear I was not going to develop academically in either of the ways that people typically developed academically. I wasn't going to have a big patient practice of my own like Orgain. I wasn't going to be in the cath lab. I wasn't going to be doing basic science and getting grants. So how was I going to be sustained as a faculty person over time? And so Dr. Stead gave me the challenge of, 'Well, how do you plan to support yourself to stay here over time?' And since at that time I wasn't even worried about that, he said I should get worried about it, and one of the things I might do is consider working in a community round-about that really wanted to make a commitment to having a collaboration with Duke for clinical research. And he had identified in 1972 a community near Charlotte, in Cabarrus County, where there was a lot of money coming from the local industry foundation, (Cannon Clinical Trust) and they were interested in having somebody who would go into that community and be a Duke faculty person based there and help them develop. Mr. Cannon had been unwilling to fund Duke education in Durham, but he was very willing to fund Duke education in his own Cabarrus County. And so Dr. Stead said I ought to consider doing that, because that would give me a firm source of funding over time. As it was, I actually worked out a way so I could support the development of the Cabarrus education program and still be based at Duke rather than moving to the community. So for many years I'd go spend one day a week in Cabarrus County, working with the program, and the rest of my time at Duke directing the CCY and doing research. And then about that time, in the seventies, a thing got going called AHECs, Area Health Education Centers, and federal and state monies supported getting medical education out into communities. And so North Carolina had about seven or eight AHECs. A bunch of them were run by UNC and one by Wake Forest and one by Duke. There was money then for communities without a local charitable source having educational activities develop, with guys from Duke, UNC or Wake going out there and delivering that education. So I started to use AHEC money to build up these sort of outreach programs in many, many communities around about. And so by the time I stopped directing the CCU, I had pretty major responsibilities for keeping these so-called Duke outreach programs going in many communities. I also moved into the dean's offices as an associate dean of education to try to help stabilize the Duke unique curriculum where one year of medical school was a research year. It had started in the sixties and now into the seventies'there was a question of should it continue. I had a job in the dean's office to try to develop systems to stabilize and further develop this research experience for all medical students in medical school, From my experience in the CCU, I had seen medical students like Brant Mittler, Fred McNeer, and Rob Caloiff get involved and actually do research. I got involved in trying to help find places for funding for medical students to have scholarships to do research, and so I started working with people to do that. The first was Stan Sarnoff who had made money from a company (Survival Technology) that he began while mentoring investigators like Andy Wallace while at the NIH. I had also begun, in the seventies, to work with the ambulance service because it was clear that you should do more in the community to get the ambulance guys doing as much as they could for the patients to stabilize them and keep them in the best possible shape as they came into the hospital. I started working with the ambulance service to strengthen that and to do studies to see how could the people in the ambulance service do a better job, whether was taking care of people who were arresting in the community or people having infarcts in the community. So I started doing a bunch of studies with the ambulance people, and that was sort of an outreach also but into the out-of-hospital care of people. Another thing I did was to work with the cadiac fellowship alumni. I had worked with these fellows in the CCU, and as they went out, I became very interested in developing an alumni group, so we would keep track'we'd have a database, and we'd keep track of these people, find out where they went, keep up with them, organize a sort society of fellows trained at Duke. And then in the eighties, as Rob Califf then needed other sites for his TAMI studies, as David Pryor wanted to find sites for some of his kind of studies, I was able then to go to these people who had trained at Duke and help them to organize the ability to do research in their practices, wherever they were. Now, in many cases it was their experience with the databank that made them interested in that, because of course they had done research while they were a fellow. They were interested in perhaps doing research later, and they knew that to do research you had to collect data. Well, why not collect data not just at one place but at many places? So being able to, in their practice, right where they were, do some research as a site became attractive to some of them. So I helped to set up a thing called the DUCCS, Duke University Cooperative Cardiovascular Studies, so that wherever people were, if they were alumni of Duke, they could be in DUCCS and they could do research together. So I did a lot of that development on into the eighties. So I guess those are some of the major kind of things that I was doing as I went into the eighties and as I moved out of the CCU into these more research and administrative kind of aspects of what I put myself into. Another thing was that as the clinical trials were done, then one of the pieces of data would be from an EKG, and so somebody had to read those EKGs, and the concept came of setting up a core lab; so that from whatever site there were, EKGs would flow in and I would have an EKG core lab. So I would read those EKGs, not knowing anything else about the patients, but EKG that data into the research databank. So every research databank, like TAMI-1, TAMI-2, would have an EKG file. And so I would produce'the data from the EKG variables would go into those databanks, and so I had a core lab, then, in the research institute that later developed into the DCRI [Duke Clinical Research Institute]. So there were a variety of those things that kind of together made up the fabric or the pieces of what went into to my own research career, and I was able to use funding from the core lab'in the clinical trials. Because I went into communities and helped the community to be able to get education, money would also come in. I really had sort of a small business. And so as money came in, I could use that money to pay my salary, to pay for students that might work with me or assistants or so forth. And so I guess from the time I stopped the CCU onward, I guess for the last twenty-five or so years, I've had that kind of a small business and that kind of a responsibility to make sure I was able to pay for myself and my activities so that I didn't have to ever be somebody who said, Okay, Duke, you gotta pay my salary or else I'm outta here, or whatever, like that. I was able to do the things I wanted to do as long as I could pay for it.

ROSEBERRY: I wonder if I might ask what you feel are maybe the promises or the potential realized from the databank.

WAGNER: Let's turn it off for just a minute.

(Pause in recording)

ROSEBERRY: This is Jessica. Dr. Wagner has been called away. Our session will conclude for today, but we'll finish up at another time, and there will be another recording that will conclude this interview.

(end of interview 1)

Interview 2 of 2

DATE: February 16, 2007

PLACE: Dr. Galen's office in the North Pavilion

ROSEBERRY: This is February 16, 2007, continuing a conversation with Dr. Galen Wagner about the Duke databank. This is Jessica Roseberry, and we are here in his office in the North Pavilion. Dr. Wagner, I had asked you at the end of our last interview just kind of about the promises and the potential of the databank and wanted to kind of see what was fulfilled in that.

What was the databank able to accomplish?

WAGNER: Their accomplishments I think could be looked at now, in hindsight, in several areas. I think we could consider that the direction that Rob Califf took with the research, in his role as a director of the databank, moving into the research databanks for clinical trials'that really formed the basis for the Duke Clinical Research Institute [DCRI], so whatever the Duke Clinical Research Institute has become is certainly built on the foundation of the databank. And now in 2006 [2007], as Rob turns over responsibility for the DCRI to Bob Harrington and takes on a broader responsibility that includes DCRI and this translational medicine community database and so forth, whatever that activity is'and you can get the spread of that, the breadth of that from Rob'again, it's built on this. I think another part of it is the incorporation of the databank into practices, both clinical and academic, of all the various people who trained here, because you can imagine over the years that the record-keeping system here was different because the databank. So everybody who rotated through the cath lab here had the responsibility of getting the data on each patient into the databank, had the opportunity to do research using data, and that influenced their training, so obviously when they went out wherever they went, then the principles of the databank were more or less that in them from that experience. There's a lot of examples of how this occurred, but one individual, Brent Muhlestein had been from Utah, went back on the staff at Latter Day Saints Hospital in Salt Lake City, with the expectation that he bring the Duke databank with him. And what he really brought back was an adaptation that he built together with colleagues at the Intermountain Health System there, that is a manifestation of the Duke databank in cardiology at LDS Hospital. There was a time in'oh, I forget what year, but it must have been ten or so years ago, when Brent Muhlestein asked me then to work with him to put together a documentation of the function of the Duke databank in LDS Hospital. He identified the University of Utah medical student, Gregg Taylor to develop the study. We worked together and wrote a manuscript, which is in the American Heart Journal, the about the function of the databank in many different ways at LDS Hospital. That experience subsequently was that once they'd built the databank into cardiology practice at LDS, they then spread it through cardiology practices in a bunch of the other hospitals in that Intermountain Health System in that part of the country. Further south, in Oklahoma City, Charlie Bethea did a similar thing, where he took the Duke databank. He incorporated it into his practice at Baptist Medical Center, collaborated with the people at Duke and took the software that at that time was called DISC, D-I-S-C software. I guess he started off with David Pryor and the software at that time, and then as Don Fortin took over from David and developed DISC software, Charlie incorporated that into his practice in Oklahoma City. Now that's part of the INTEGRIS Health System, which goes through a lot of the breadth of Oklahoma. So if you look at those two parts of the country, you'll find a lot of those kinds of manifestations of the databank functioning in the care of patients. And a lot of the principles, the tools of the databank, such as the prognostigram are being used, for example, by Brent Muhlestein at LDS Hospital. So I think if you looked at, oh, any number of practices of people who trained here in cardiology, you'd some manifestation, something different they're doing in their record-keeping system, care of patients, et cetera because of their work with the databank.

ROSEBERRY: Now, is there something specific about cardiology that makes it a nice fit with the databank, or is it just because it happened to begin here in cardiology?

WAGNER: I'd say it's just because it happened to begin here, because the use of the databank has minimal value in a very acute illness, because there, the outcome is so quick that the clinician can pretty much see it right then and can therefore learn the association between treatment and outcome. But the databank is much more valuable in a chronic illness, where you have variables at one point in time that may well predict something that may be one year, five years, ten years down the road. So if you think of any chronic problem, the databank has equal value as it would have in cardiology. There was one other thing about cardiology, and that is that early on'when I say 'early on' I mean in the sixties'we could already quantify a lot of clinical data. Not just quantify 'What's the blood pressure?' or, 'What's the pulse?' but, 'What are the particular quantity of arrhythmias?, What's the size of obstructions in coronary arteries?, What's the quantity of exercise performance on a stress test?, What's the quantity of hemodynamic abnormality in somebody who has a low heart function?' So these very quantifiable things made cardiology a ripe place for a databank back then, not that qualitative aspects aren't valuable in a databank, but any time you can measure'and you can imagine the cardiovascular system'it lends itself to measurements, say, better than, say, the neurosystem or some other parts of the body. The answer to your question I think is that it really could have been valuable back then in lots of other areas outside cardiology where databanks are alive and doing very well now. I think one of the things that happened in Duke cardiology really facilitated the spread of the databank, and that was the development of first the Duke Cardiology Fellow Society (DCFS), which was our alumni group, and then DUCCS are the people in the alumni group who are interested in collaborating in Duke clinical research wherever they might be. And so because they were adding research components to their practice, it was much more likely that DUCCS members would be using the databank or principles of the databank in some way. And so I think there was dissemination of the databank out to these people, not just through the time they were here in fellowship but then later on, as they interacted with us through DUCCS clinical trials. If you are collecting data for clinical research, then this spreads off also into clinical practice. David Pryor, who was co-director of the databank with Rob, came to me at one point because he had been approached by a clinical research organization out in the Research Triangle, to get together a bunch of sites to test the very first statin drug. And this group had organized sites to randomize the statin versus a placebo, and they needed some more sites, so David and I worked to develop a bunch of DUCCS sites. They weren't even called DUCCS at that time, just DCFS. There were about twenty-five of these people who trained here, were out somewhere, and then collaborated to collect data, randomizing patients to no statin or statin. Very first statin trial. That was really the very first time that we had a bunch of DUCCS groups that collaborated to bring data into a research database, because it was the database put together for that particular statin trial.

ROSEBERRY: What have been maybe some of the limitations of the databank?

WAGNER: A major limitation is just how busy cardiologists are. I would suppose that cardiology has changed the way it's practiced more than just about any other area of medicine. One reason it's changed so much is because of the databank, because when you can look at your past experience, you say, 'Well, maybe we're keeping people in the hospital too long; maybe we're putting pacemakers in people that we don't need to, or not enough people; maybe we're not doing interventions in as many people as we should.' The databank and clinical trials led to such advances in cardiology that the cardiologists became very busy, and that very busyness with clinical care made it more difficult to actually carefully collect, so that the actual use of the databank, no matter how much one might want to use it for clinical care or for research, often is limited by just how busy we are. For example, back in the days when I was on the CCU, when a patient an acute infarct'you put them to bed, tried to make sure this and that was safe. Now even in the middle of the night the cardiologist has to get up, get the patient to the cath lab quickly, open up the artery. That takes time. That guy's tired. Next day, well, how much time is there to make sure you collect good data? So the clinical urgencies produced by all the things that cardiologists became able to do have been one of the main things that caused limitation in the use of the databank. Another thing is that at Duke the use of databanks for research has gone further than the use of the databank for clinical care. There are lots of reasons for this, but what is left, I think, then is a underutilization of the databank for just clinical use apart from research. I think that with Rob Califf's responsibilities now, with his broader venture at Duke, I think this is going to change, and I think there are going to be a lot more movement now toward'and maybe it took time to get that, because obviously you don't want to use dirty data'a databank can only be useful if the data are clean. How do you take responsibility for having clean data on the whole population? That's a daunting thing. It's daunting to even have clean data on a clinical trial population. But I think it's taken a lot of time to learn about how do you collect data, how do you clean data, how do you use computers to hold it, how do you use computers to enter the data. Maybe it's just taken a lot of that experience with the stronger research use of the databank to make it so here we are, at the early part of this century, now struggling to ask, 'Can we use it better for clinical medicine?'

ROSEBERRY: Why is it used less for clinical than for research?

WAGNER: Because you can do clinical medicine without it. You can't really do clinical research without it, so that we could argue that the cardiologists will do a better job with their patient if they make certain to enter data not just on the general medical record but into a place where it can be retrieved, not only for the care of that patient but for comparison with other patients. But now it's really, I would say, at the level of a luxury in clinical medicine versus a necessity in clinical research. You can't do clinical research without strong, clean database. You can do clinical medicine without any databases at all. I think in the future, as there are a whole bunch of other realities that will be brought to bear on clinical medicine to showing its outcomes, using banks of clean data will going to stop being a luxury and become a necessity. So my guess is there won't be any such thing as cardiology clinical care, or perhaps any kind of clinical care in the future, without computerized databases. I just think the paper record business is eventually going to go away. The question is: When is eventually? And a lot of the answer to that is sort of: When we're ready, when we've learned enough. But that means that we've got to recruit a lot of doctors, nurses, et cetera into informatics, help them gain the skills of, What does it take to build informatics into the daily work? How do you make it so that a doctor, when they're very busy and don't have much time, can still collect data? Well, one way is if in their dictation which they do, if there's word recognition, voice recognition. So just as you dictate, you're populating a database, so you don't have to do anything else; you just dictate. If you ever heard a doctor dictate, you realize, my gosh, there's going to be a lot of errors if you try to pick up those words. So the technology of how do you take the very active, excellent, time-challenged physician, cardiologist, and make it so that it doesn't take additional time to put the key pieces of data someplace it can be retrieved. That's the piece we really haven't figured out well enough.

ROSEBERRY: Even with additional personnel, it sounds like it's still'

WAGNER: You're right; this can't be really solved with only personnel. It's got to be'if you're the physician and you're getting the data from the patient and you're responsible for the patient record, you can't have both a clinical record and a research record. That means as you document your clinical record, in the very act of doing so, has got to populate the research database. Now, obviously not every bit of information you put in the record needs to go in the database, but if you go back to what we did on the CCU, select the key pieces on every patient. You can't sort of sometimes collect this and sometimes not collect. It's got to be collected every time. And we were able to build in discipline because we were dealing with one problem, myocardial infarction, and we had time. So now, with less time, we need more concentrated improvements in the technology. You can imagine a clinician could say, 'The patient has an acute infarct?' Yes. Could say, 'Old infarction?' No. Acute infarction, anterior, dah-dah-dah-dum, dum-dum. Just particular key words. And in much of our daily life, voice recognition is used, for telephone and so forth. The technology will get there. And when it does, then it'll just be routine that our entire medical record will have two components. One will be the text component that just tells in general about us, and the other will be specific stuff that then goes into the databank. Now, to the extent that the data in the databank can be numbers, that's pretty easy now: blood pressure, heart rate, whatever. To the extent that the data in the databank are based on words, it's more difficult: 'How much heart failure do you have? It is a little bit? Is it more? Is it more? What happens to that heart failure if you get treated with this, that or whatever?' It's much more challenging to take any clinical variable that has to do with qualitative information, with words instead of numbers, and get it in the databank. Now, as you or I are seen by our physician, do we want data in the databank? We sure do. Do we want only the numbers? We sure don't, because we don't want to be known as people, as patients only by numbers. We'd like to also be known by these key words. And that's much more challenge. How do you get those key words in there? And key words that are well defined, not just sort of how every individual person might want to say them. So that's some of the short'I'm answering as kind of shortcomings that I think will get solved but aren't solved really at this point so that the principles of the databank can be realized in just everyday clinical care of everybody who has in this case cardiac disease, but certainly on into the future has any kind of health care problem. Dr. Stead predicted that the databank can only be really useful when it contains information about the individual's brain as well as their heart, lungs, et cetera.

ROSEBERRY: Is there a relationship between this and the electronic medical record? I think if Dr. Hammond, Ed Hammond developed some'is there a relationship there?

WAGNER: Sure, it is, because that's what we're talking about. We're talking about the data. We're talking about the electronic medical record being used to populate the elements in a databank'

ROSEBERRY: I see. Okay.

WAGNER: 'so that as you're creating electronic medical record, how do you then extract? Think of it in the situation I mentioned about the cath lab. The fact that they were generating a cath report way back in 1970'that's an electronic medical record of a procedure. Well, let's suppose that for every catheterization there's fifty pieces of discrete data that you want to have populating a spreadsheet that is very searchable. Frank Starmer and Sandy Simon had to build an extractor program to take a piece of data out, stick it in the place where an electronic search could be done. That's still the same process that is needed now. The process isn't any different. With all the strength of computers, you can imagine that it would easier now. 'How much real estate on a computer screen do you need as you walk around to see a patient?' Okay, you can imagine you've got a little screen; it's got to be a certain size; you're putting certain pieces in. Those pieces then are in the databank. All you've got to do is touch a screen, or all you've got do is talk, and that goes in the databank. Those systems will be developed. And they'll be able to be used in two ways. One is to make the text chart, so your text record then will come from it. But it will also make something that is not text at all, that is just populated information in data fields that can be easily searched and retrieved for comparison with other patients or for generating research projects or for whatever the job might be. It's like doing one thing and having it branch into two things, so you say, (claps once) I've done it. It didn't take any extra time. Here's the electronic medical record. It can be for the patient, the referring physician, whatever, and here are the elements in the databank. And they're there for all these: computerized textbook and so forth. So it's getting sort of a two-for-one, having your cake and eating it, whatever, with the thing that has to happen before the entire potential of the databank is realized.

ROSEBERRY: What have I not asked you that I should have asked you?

WAGNER: Mmm, I'd say that you've covered almost everything I can think of. Maybe at some point in time, as you talk to other people, it might be useful to ask some additional things, but I think you've pretty much covered everything I think might be useful.

ROSEBERRY: Good. Well, thank you, Dr. Wagner.

WAGNER: You're very welcome.

ROSEBERRY: I appreciate it.

(end of interview)