ObjectivesThe day-to-day management of diabetes is often the responsibility of resident physicians in teaching hospitals, yet studies have consistently shown that learners lack confidence and knowledge of inpatient insulin titration. We sought to elucidate factors that influence inpatient management of diabetes by medical learners. We predicted that there would be conscious and unconscious variables at play, and we sought to determine the process by which learners arrived at insulin titration decisions. MethodsWe administered a survey using a hypothetical case based on script concordance testing in 88 medical students and residents at the University of Calgary. We used multilevel, mixed-effects linear regression and random-effects meta-analysis to compare clerks’ and residents’ insulin prescribing, confidence in prescribing decisions and concordance with endocrinologists. ResultsResidents had higher confidence than clerks, yet were less concordant with the insulin decisions of staff endocrinologists. Both residents and clerks prescribed more insulin when the reported blood glucose was higher, or the patient had positive ketones and prescribed less insulin when the patient had hypoglycemic unawareness. Residents prescribed a higher dose of insulin when the charge nurse reported the result compared with a student nurse, but this variable did not significantly impact prescribing by clerks. ConclusionsWe propose a model in which learner factors (knowledge, confidence, experience), patient variables and perceived expectations influence insulin titration decisions via analytical and heuristic processes. We propose that educational interventions designed to improve inpatient glycemic control must, therefore, address both knowledge and systems factors.