Dynamic Ranking of Physicians in Online Healthcare Platforms with Multiple Service Types

Online healthcare services are developing rapidly in recent years. They provide a convenient way for patients to consult with physicians across the country. However, they also bring new challenges to online healthcare platforms. Physicians’ resources are usually limited. What’s more, patients are heterogeneous, and vary in their severity of illness, choice behaviors and preferences for different service types. In this work, we consider a multiple service type setting, where we characterize the choice behaviors of patients, and formulate the problem of the platforms using dynamic programming. With the huge number of choices usually displayed on online healthcare platforms and great uncertainty in patient behaviors, our goal is to design personalized physician ranking methods to maximize the total pairing reward. We propose two types of methods based on improving the inventory balancing methods and the value function approximation methods to better fit our setting. Through numerical experiments, we show that our methods generally outperform the existing methods. Our findings can help online healthcare platforms adopt the most suitable ranking method in different situations to improve service quality.