Joseph C. Cappelleri, Pfizer Inc.

Christopher Schmid is professor in and chair of the department of biostatistics at the Brown University School of Public Health. He is a fellow of the American Statistical Association and served as chair of the Health Policy Statistics Section in 2013 and co-chair of the International Conference on Health Policy Statistics in 2008 with his longtime friend and colleague Joseph C. Cappelleri, who interviewed him here.

Family and Schooling

Where did you grow up? Tell me about your upbringing.

I grew up in Bethesda, Maryland, and attended college at Haverford, where I majored in mathematics. I was interested in statistics, so I went to work as an actuary for a couple of years at Prudential Reinsurance, but I did not like working in the insurance industry and decided to go to graduate school in statistics. I earned a PhD at Harvard and, while I was there, I worked at BBN Software as a trainer for its software. This was a great teaching experience, as I would visit clients who had purchased the software and give 2-4-day courses in statistics and experimental design.

When did you discover your passion for mathematics and statistics?

I became interested in statistics through sports. As a kid, I was a sports fanatic, and I knew all the numbers on baseball, basketball, and football. I learned my decimals through batting averages.

For my entrance essay for the PhD at Harvard, I chose to write about Bayesian shrinkage using the paper Carl Morris and Brad Efron wrote in Scientific American in 1977 about baseball batting averages. I thought the idea was very cool.

Please share your experiences at Haverford College and Harvard University.

At Haverford, I did a standard math degree and took a couple of courses in statistics. I took many courses in the humanities and social sciences, which is where I really learned how to write. It helps me write better grants and allows me to advise students and colleagues on their writing. At Harvard, I studied with many brilliant students from other countries, so I learned a lot about different cultures, but I also made a lot of wonderful friends and developed professional relationships with people who are leaders in the field. I learned quickly to know my place in the pecking order.

Tufts University School of Medicine/New England Medical Center

What led to your job there?

When I finished my degree, I was looking for a job in the Boston area because I was married and had two small children and we wanted to stay near my wife’s family. I was looking for a research job at the time. I was not interested in teaching. I took a job at Tufts-New England Medical Center and started working on developing predictive models for patients who had heart attacks and were candidates for thrombolytic therapy. We also had a small consulting service that eventually grew much bigger. Through that, I got involved in working with a lot of clinicians. Soon after I got to Tufts, fellow faculty member Joseph Lau received a substantial grant on meta-analysis and I started working with him and with you, Joe. That started a long and fruitful collaboration of many years and developed into many grants that led to the development of new methods and software.

What were some of your major studies and methodological advances?

The thrombolytic predictive instrument was an interesting study because we developed five logistic regression models to predict the risk of mortality at 30 days and one year, the probability of having a cardiac arrest, the probability of needing a blood transfusion, and the probability of having a stroke. These probabilities were printed at the top of an electrocardiogram (ECG) to help clinicians determine when and when not to provide thrombolytic therapy. The manufacturer, Hewlett-Packard, implemented this in their ECGs. Thrombolysis is effective in reducing mortality for some people, but there are also complications of stroke and major bleeding. Our models looked at treatment interactions, as well as factors that change the baseline probability, so we could advise clinicians properly on the risks involved.

I also did a lot of work with the nephrology department at Tufts. We did individual participant data meta-analysis looking at the effects of ACE inhibitors (which lower blood pressure) on risk in patients with nondiabetic renal disease to determine whether the benefit is due to blood pressure reduction or other causes. We also put together databases to construct better predictive models for estimating glomerular filtration rate (GFR), which is the amount of waste the kidney can process. It is the primary endpoint of interest to many clinicians in determining the extent of kidney disease in patients, but it is hard to measure exactly, so our predictive models used surrogates like creatinine, which could be more easily measured, along with demographics like age, sex, and race. These models are used all over the world now in laboratories when creatinine is measured to get a prediction of GFR. They are also used in all the renal disease research that looks at GFR.

In meta-analysis, we developed several important programs to speed up the process of doing systematic literature reviews and carried out several statistical methods assessments in the Tufts Evidence-Based Practice Center. Over the years, I developed relationships with scientists in ecology and other disciplines and eventually joined the Society for Research Synthesis Methodology (SRSM), which includes interdisciplinary experts in meta-analysis. We meet each year, and we also publish the journal Research Synthesis Methods, for which I am one of the two founding coeditors.

Brown University

What new opportunities did you find?

Along with my colleagues Joseph Lau and Tom Trikalinos, I started the Center for Evidence-Based Medicine (now Center for Evidence Synthesis in Health) at Brown in 2012. We were looking for a university environment, one that had students with whom we could do research. When I got to Brown, I also joined the department of biostatistics and was able to start teaching graduate students. Before then, I had been teaching medical doctors about statistics as part of the Tufts master’s program in clinical research that I helped start and worked with for almost 20 years. At Brown, I instead started to train statisticians. I directed the master’s program at Brown for my first four years.

What have been some of your major accomplishments?

Having seen the importance of teaching doctors how to do statistics, I realized how important it was to teach statisticians how to collaborate with doctors. Communication is so important. Statisticians need to be able to discuss statistical concepts and write clearly. The first course I taught was a consulting course in which I brought in both doctors and statisticians to talk about how to collaborate—including you, Joe. I also brought in speakers from the outside to talk to all our students, and we started the practice of bringing alumni back to talk about their experiences working in the “real world.” When I got to Brown, the program was small and fairly new, and we built it up to be a much bigger program that continues to grow. We now have about 40 students in the program and get more than 200 applications each year. I have worked with quite a few of the students as their master’s adviser and employed many in our center. One of them is now actually working at Tufts with the nephrology group I used to work with.

What is your mission as the new chair of the department of biostatistics?

As the new chair of biostatistics, I am taking over from two accomplished predecessors, Constantine Gatsonis and Joe Hogan. I would like to maintain the progress we’ve made as a new department and strive to be one of the top biostatistics departments. I’d like to continue to attract top students and faculty, give students more internship opportunities and other practical experiences, and increase our collaborations with other departments in public health and medicine at Brown. And I’d like to focus our curriculum on those areas that students need to be scientific leaders in this age of data science and big data. For our students to be leaders, they need to be able to not just compute and analyze and prove things, but they need to be able to have a vision of where their company or university needs to go and how to get there and be able to communicate that vision to their nonstatistical colleagues.

Service to the American Statistical Association, Especially the Health Policy Statistics Section (HPSS)

How did you get involved with HPSS?

I attended an HPSS mixer at the Joint Statistical Meetings (JSM) many years ago and, while talking with some of the members, was recruited to run for HPSS program chair for the next JSM conference. I served two years as program chair-elect and then chair. I then agreed to become the co-chair of the International Conference on Health Policy Statistics (ICHPS) in 2008 with you, Joe. We ran a successful conference in Philadelphia and attracted quite a bit of outside funding for it. After that, I became the HPSS representative to the ASA Council of Sections and, following that, I became chair of the section.

How would you describe your experience as an HPSS section member and officer, HPSS conference co-chair, and HPSS-sponsored invited speaker and short course instructor?

I have found the section to be a wonderful place to meet both junior and senior colleagues. It has provided me with a great network of friends and professional opportunities. The section also sponsored a JSM workshop in meta-analysis that I gave with Ingram Olkin for many years. I always look forward to the HPSS section mixer and attending and participating in sections sponsored by HPSS. ICHPS is one my favorite conferences because it is relatively small and focused on the kind of research I do.

How did HPSS contribute to your being elected fellow of the ASA?

My activities in HPSS were instrumental in getting me elected as an ASA Fellow not only because of the recognition it provided me, but also the work I was able to do professionally through my contacts there.

What colleagues were most influential to your career?

The most influential colleague I’ve had in my career has been Joseph Lau. I met Joseph in 1992 when he came to Tufts, and he was the one who got me into meta-analysis—which is where I’ve done most of my research and achieved most of my reputation. Through him, I was able to get involved in several organizations, including a working group in ecology and SRSM, and in evidence-based medicine and research that help guide health policy and medical recommendations.

What advice would you give to students and junior researchers, especially as it relates to a career in health policy?

Health policy sits at the intersection of so many fields. If you want to be successful in health policy research, you must know something about each of those fields. You need a wide skill set that includes statistics, computing, writing, health care knowledge, and the ability to work with people. Given the crucial position of health care in our economy and society, health policy is also one of the most rewarding statistical areas you can work in, with so many opportunities to make an impact.