A data-driven approach for estimating the travel times of operators for the assignment problem of home health care services
Date & Time
22 September 2014 - 15:00
Abstract: Home Health Care service is an alternative to conventional hospitalization and consists of delivering medical, paramedical and social services to patients in their homes. This is one of the recent service systems where human resource planning has a great importance. The assignment problem in the served territory is an important issue that the service provider must address and it consists of deciding which operator will provide services to which patients. Because the routing optimization is not considered and exact travel times between patients are unavailable when the assignment problem is solved, an estimation of the travel time necessary to reach each patient is required to solve the assignment problem. Traditionally, these travel times are considered based on only the geographical criterion with Euclidean traveling times. However, service provider should also take into account several characteristics of patients and operators while considering the travel times. Examples of such characteristics can stem from features related to patient care requirements (i.e., their care profile) or the geographical aspects of
the territory. Modeling and integrating these features would not be computationally tractable since one would have to formulate each feature as a new decision variable or a new constraint and integrate it in the formulation of the model. Alternatively, such features can be captured by the available historical data that would give information regarding the choices made in previous routes accomplished by a given operator. Then, such data-driven approach would enable to estimate (future) travel times of operators based on their past behaviors. Hence, in this work we propose data-driven method to estimate travel times of health care operators in the assignment problem. Numerical results based on realistic problem instances are performed as well to show the performance of the proposed method.