Preference based scheduling in a healthcare provider network

Healthcare services are often provided by a large network of physicians and clinic facilities to patients with various levels of health conditions and preferences. Appointment scheduling is used to manage access to these services by matching patient demand with physician availability. This raises tremendous challenges for providers due to the heterogeneity in patient preference and physician availability. We propose a preference-based “nested” network model that consists of most practical operational constraints. Our model considers patients with varied priorities who can visit any clinic location and provider of their preference, and request the day and time of the appointment of their choice. The common challenges of patient no-show, cancellation, and uncertainty of physician availability are taken into account. We formulate this model as a Markov decision process and propose an approximate dynamic programming approach to provide robust scheduling policies. We also analyze the joint appointment scheduling and physician capacity planning problem as a mixed-integer nonlinear program. The proposed scheduling policies maximize revenue and minimize physician overtime and idle time while satisfying patient preferences. These policies are shown to perform within a close bound to the best achievable policy. They are also robust under patient demand uncertainties. We highlight the importance of considering patient heterogeneity and preference as well as systematic uncertainties to provide an optimal set of appointments.