Constructing block schedules for internal medicine residents

Before medical school graduates can practice independently, they must spend a minimum of three years in a residency program, where each year is divided into different clinical assignments called blocks. In this study, the goal is to generate annual block schedules that hierarchically (i) minimize the imbalance in the number of clinic sessions each resident must attend over the year (each rotation has a given clinic requirement), (ii) limit the number of times residents are given a night float block immediately before or after an intensive care unit block, and (iii) minimize the maximum deviation between the number of patients that can be seen in clinic during any month over the year from the average. The problem was formulated as a mixed-integer program (MIP) and solved with a heuristic that yielded locally optimal alternative solutions. Using these solutions as input, a second MIP was formulated with the objective of obtaining a diversity of solutions with minimum similarity. These provide the chief resident with a range of options when assigning blocks and trying to accommodate individual preferences at the same time. The models are illustrated with data provided by University of Texas Health Science Center in San Antonio.