Lessons not learned from the generalist initiatives.

The grouping of the primary care specialties (general internal medicine, general pediatrics, and family medicine) for research purposes is at best limiting the value of the information that is found and, at worst, leading researchers to erroneous conclusions. For example, three large studies each showed differences in abilities to predict students' specialty choices in primary care (e.g., in one study, the investigators correctly predicted 3% of those choosing general internal medicine, 29% considering general pediatrics, and 51% considering family medicine). These and related findings suggest that medical students entering the three primary care specialties are not a homogeneous group. While there were some factors predictive for all primary care specialties, there were more factors that were unique to the individual specialties Grouping the specialties may not reveal factors that are significantly related to only one of the specialties. In addition, when a variable operates in different ways for different specialties, findings where the specialties are combined can show a reduced effect of that variable or even no effect, because the directions of effects cancel each other. Researchers can fruitfully examine the primary care specialties as a group but at the same time report their data for the individual specialties, which would greatly increase our knowledge both of primary care and also about the similarities and dissimilarities of its component specialties. However, the best models continue to be either research in which the sample size is large enough to compare specialty groups statistically or research with a focus on just one of the primary care specialties.