By Paul Jolly, PhD, senior director, Special Studies, Association of American Medical Colleges, Washington, DC.
There is a serious concern that the pipeline for the production of the physician workforce is inadequate to meet future needs, especially for primary care. Using the database on residency training sponsored by the Association of American Medical Colleges (AAMC) and American Medical Association (AMA), which contains annual data on individual residents, it is possible to monitor the size and shape of this pipeline. Because we can follow individuals longitudinally, we can determine who does and does not enter subspecialty training, implying that they will not practice primary care. Of course, some physicians trained in a subspecialty nevertheless practice primary care, and some who have not been trained in a subspecialty become hospitalists or limit their practice to a subset of patients in a subspecialty of their discipline.
AAMC has been measuring primary care output in this way for some time. The blue line in the graphic below shows how this works. We start with the number of new categorical residents in internal medicine or another primary care discipline, and we follow them through their graduate medical education, using the GMETrack database. For those beginning internal medicine in 2000, a small number are reported in a fellowship in 2002; most of the residents who will subspecialize have their first opportunity to enter a fellowship in 2003. As the blue line on the graph indicates, however, more than one third of those who will subspecialize are not reported in a fellowship until 2004, 2005, or even later. In order to determine the eventual number who will subspecialize, we need at least eight years of data, which is the problem with this longitudinal method. Although we can determine what fraction of a cohort will subspecialize and thus how many may be assumed to be planning a primary care practice, we can’t tell what is happening in recent years.
For our recent report in Academic Medicine, we studied these longitudinal measurements, which have shown that the number who will subspecialize in each successive year is changing only slowly. We can use this fact to develop a simple estimation method that can be applied to recent years, even to the immediate past year. We simply assume that the number from the cohort who are beginning internal medicine or another primary care specialty in a given year and will eventually subspecialize is equal to the number entering subspecialties that same year.
The red line on the graph shows the effect of applying this method to the cohort beginning internal medicine in 2000. As you can see, the estimate is very close to the result of the longitudinal analysis. The longitudinal result is slightly higher, as we would expect when the subspecialization fraction is gradually increasing. But the beauty of this method is that we can apply it to recent years.
Our report employed this method to estimate the percentage of new program year one residents who would practice primary care, and we found that the percentage has stabilized in recent years. We will need to continue to follow these trends, but the resulting projections indicate that the long decline in interest in primary care may be ending.