Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)
57 peer-reviewed publications in journals including Nature Communications, PNAS, JAMA, and Nature Machine Intelligence.
1 publication matching filters
Appropriate nurse staffing is a critical challenge for healthcare delivery institutions, that can affect the delivery of quality patient care. Reliable forecasting of nursing demand can optimize resource utilization and support preemptive hiring strategies; however, the use of advanced machine learning approaches to address this challenge has been limited thus far. We propose a probabilistic forecasting approach for predicting nursing workforce demand across multiple hospital units using DeepAR, an autoregressive recurrent neural network algorithm. We analyzed 5-year historical workforce data from a large US-based health system in New York, encompassing both full-time and temporary nurses across multiple specialties and hospital units. Our implementation leverages DeepAR's ability to create a single global model for heterogeneous time-series, incorporating the impact of the COVID-19 pandemic, through binary feature encoding. The model was trained on 38 business unit-specialty combinations, each containing over 100 employees, with a 12-month context and prediction length. The model successfully captured complex temporal patterns, with most ground truth values falling within the 90% prediction intervals and accurately predicting both gradual trends as well as sudden magnitude changes. The model achieved consistent performance across multiple test sets through rolling window analysis, with only 9 out of 38 combinations showing significant temporal sensitivity. Furthermore, validation on 91 smaller business unit-specialty combinations (with 15-100 employees) showed no statistically significant performance degradation, indicating strong generalizability. This approach provides healthcare administrators with a reliable tool for probabilistic workforce demand forecasting, enabling more informed staffing decisions across diverse hospital settings.