Purpose: Primary care (PC) career choice has been declining among U.S. medical graduates and residents since the 1990s. 1,2 By 2034, an estimated shortage of 17,800–48,000 PC physicians is expected. 2 Some medical schools have developed programs to increase students’ interest in PC specialties and practice in underserved areas. 3,4 To our knowledge, the durability of early PC-specialty preference for long-term PC practice has not been reported. Method: The Association of American Medical Colleges provided linked, deidentified data for a national cohort of 48,533 U.S. MD-granting medical school matriculants from 1998–1999 through 2000–2001. Data from the Student Records System (sex, race/ethnicity, and graduation year), Matriculating Student Questionnaire (MSQ; specialty board certification plans and attitudinal variables: importance of prestige and social responsibility in choosing a medical career), Graduation Questionnaire (GQ; total debt, plans to practice in underserved areas, and student ratings [excellent vs not excellent] of PC and non-PC clerkship experiences), and American Medical Association (AMA) Physician Masterfile practice data through May 2020 were used. The outcome was office-based practice in a specialty from the AMA Physician Masterfile, categorized as follows: (1) general-PC (i.e., internal medicine, pediatrics, family practice, general practice, internal medicine/pediatrics, or internal medicine/family practice as primary specialty, with another general-PC or unspecified secondary specialty), (2) PC-subspecialty (i.e., general-PC primary and PC-subspecialty secondary specialty, or PC-subspecialty as primary specialty), and (3) non-PC specialties (i.e., non-PC primary specialty). Multivariable logistic regression models identified variables independently associated with general-PC vs each of PC-subspecialty and non-PC-specialty practice. Results: Of 46,698 graduates in 2002–2013, 30,435 (65.2%) had complete data for analysis (7,700 [25.3%] general-PC physicians, 5,327 [17.5%] PC subspecialists, and 17,408 [57.2%] non-PC specialists). Female (vs male) students, and students with plans to practice in underserved areas (vs no plans) were more likely to report general-PC (vs PC-subspecialty) practice. Underrepresented racial/ethnic minorities in medicine (URiM), Asian/Pacific Islander (Asian/PI) students (each vs White), and students planning to pursue non-PC specialties or no board certification (each vs PC board certification plans) on the MSQ were each less likely to report general-PC (vs PC-subspecialty) practice. In the second model, female students and students reporting plans to practice in underserved areas (vs no plans), endorsing higher importance of social responsibility, or rating only PC clerkships excellent (vs rating both PC and non-PC clerkships excellent) were each more likely to report general-PC (vs non-PC-specialty) practice. URiM, Asian/PI students, students planning to pursue non-PC or no board certification, and students endorsing higher importance of prestige were each less likely to report general-PC (vs non-PC-specialty) practice. Discussion: Several factors were associated with future PC practice. The greater likelihood of women and lower likelihood of URiM graduating students to choose PC specialties on the GQ has been reported. 1 We extend this work, finding lower likelihood of URiM and A/PI physicians to be in general-PC practice. Importantly, early preference for PC specialties was positively associated with general-PC practice in both models. This may reassure medical schools considering early specialty preferences for admissions or entry into early, PC-pathway programs. Although clerkship experiences might influence students to change specialty interests, 5 to our knowledge, associations between clerkship experiences and PC practice have not been examined. We observed that students reporting excellent PC-clerkship experiences were more likely to report general-PC (vs non-PC) practice. Greater importance of social responsibility in choosing medicine also was positively associated with general-PC (vs non-PC) practice. Significance: Medical school experiences that are amenable to intervention were associated with greater odds of general-PC practice. Admitting students with early interest in PC specialties or who highly value social responsibility, fostering students’ plans to practice in underserved areas, 3,4 and improving PC-clerkship experiences may help mitigate the PC-physician shortage. Acknowledgments: The authors thank Hershel Alexander, PhD, Marie Caulfield, PhD (retired), David Matthew, PhD, and Lindsay Roskovensky, at the Association of American Medical Colleges (AAMC), for their efforts to obtain and prepare the deidentified datasets from various sources and their assistance with data coding; the National Board of Medical Examiners (NBME) Office for Research Strategy for permission to use Step 1 licensing examination data; Kristy Vanderplow at MMS, Inc., for provision of the American Medical Association (AMA) Physician Masterfile data; and Maria Pérez, MA, at Washington University School of Medicine, for data management and preparation of the data for analysis.