Although most hospitals in the United States provide medical services in English, a significant percentage of the population uses languages other than English. In most hospitals, an interpreting department is responsible for securing interpreters for patients who cannot speak English or who have limited English proficiency (LEP). This includes inpatient, outpatient, and emergency patients. Hospitals usually have full-time interpreters but when the demand exceed the capacity of full-time interpreters, the department has to find part-time interpreters, or what are called “on-demand interpreters,” to cover the shortage.
There are two main challenges facing interpreter schedulers. Firstly, there are many interpreting agencies in the market from which part-time interpreters can be chosen from. Therefore, selecting part-time interpreters with the best service quality and lowest hourly rate makes the scheduling processes more challenging. In addition, the arrival of LEP emergency patients is stochastic, but the system must be able to predict that the right interpreter is available and to avoid any service delay.
To overcome these challenges, this research proposes a framework for scheduling full-time and part-time interpreters for medical centers. The framework is composed of two phases: firstly, the researchers develop a prediction model to forecast LEP patient demand at the emergency department. The demand of LEP emergency patients along with the demand of LEP inpatients and outpatients are then fed into a mixed integer linear programming model to assign interpreters to all patients. The goal is to minimize the total interpreting cost and maximize the quality of the interpreting service. The model considers the availability of both full-time and part-time interpreters, and language compatibility. The framework is implemented at a hospital in Minneapolis, Minnesota, in which five languages are considered: Hmong, Spanish, Vietnamese, Somali, and Russian. Various experiments are conducted to show the robustness and practicality of the proposed framework. The goal is to show that framework is practical and can be generalized for any interpreter-scheduling problem at any hospital.