To understand the predictive value of medical student application characteristics on rural practice intent. The authors constructed a linked database of 2012-2017 medical school matriculants from American Medical College Application Service applications and Association of American Medical Colleges Matriculating Student Questionnaire (MSQ, 2012-2017) and Graduation Questionnaire (GQ, 2016-2018). Using logistic regression, they compared application variables (birth, high school, childhood county, and self-declared geographical origin) to students' MSQ and GQ intent to practice rurally. Rural practice intent from matriculation to graduation was compared using the McNemar test for paired nominal data. The number of students meeting inclusion criteria was 115,027. More students self-declared rural origin (18,662; 16.4%) than were identified using geographically coded variables (6,097-8,784; 6.1%-8.1%). Geographically coded rural variables were all strongly and similarly associated with rural practice intent, with rural high school being the most predictive on both MSQ (odds ratio [OR], 6.51; CI, 6.1-7.0) and GQ (OR, 5.4; CI, 4.9-6.0). Self-declared geographical origin was associated with a similar rural practice intent on both MSQ (OR, 6.93; CI, 6.5-7.3) and GQ (OR, 5.69; CI, 5.2-6.2). Rural practice intent declined for all groups from matriculation to graduation. Considering students who self-declare as rural identifies a larger group of rural medical school applicants than more "objective" geographic variables, without negatively impacting students' predicted interest in eventual rural practice. Further research should track actual practice location and explore strategies to mitigate declining rural career interest.