Determining resource needs for specialty ambulatory clinics using classification/assignment methods

Matching resource supplies to demands for the same is a critical element in the management of an organization's operations. For health care providers, this is made even more so in this era of increased demands and government cutbacks. To best plan and provide appropriate service to patients, a health care provider has to predict resource requirements of future patients and match these requirements with available resources. In this paper, we discuss how neural networks perform in comparison with other assignment methods when used in conjunction with a patient resource classification system to predict future resource loads in a specialized ambulatory patient care setting. We compare and contrast the predictive performance of neural networks with those of decision trees, discriminant analysis and nearest neighbor approaches. In general, neural networks outperform the alternative approaches, however, their disparate performance across different field sites is intriguing.