Stochastic Modeling of Patient Arrival Offset Times in Scheduled Visits

A new model for patient offset times (i.e., patient deviation from scheduled appointment time) is developed. In previous studies, offset times was mostly assumed to be sampled from a normal distribution. Alexopoulos et al.[1] offered Johnson SU as the most suitable fit. A thorough analysis of patient offset times, obtained from workflow observations in a broad sampling of ambulatory care sites, revealed these assumptions are often not valid. Although Johnson SU is still largely acceptable, it is not the most stable fitted distribution of the observed data. Our study suggests that three distributions (Generalized Logistic, Johnson SU and Log-Logistic) are more suited to modeling patient offset times with Hosking[2] Generalized Logistic (GL) distribution the most stable in its estimated parameters. We will also consider uncertainty associated with computing parameters of a Generalized Logistic distribution fitted to observed data. This model is central in devising efficient scheduling strategies to reduce patient waiting time and improve patient throughput and satisfaction.