A Latent Survival Analysis Enabled Simulation Platform For Nursing Home Staffing Strategy Evaluation

TL;DR

This paper presents a data-driven simulation platform that models resident service demand and evaluates staffing strategies in nursing homes, aiming to reduce labor costs while maintaining care quality.

Contribution

It introduces a novel agent-based simulation framework combining survival analysis and domain knowledge to optimize staffing strategies under varying census scenarios.

Findings

Significant reduction in labor costs using the proposed staffing strategy.

Enhanced accuracy of length of stay predictions with multiple discharge considerations.

Framework effectively adapts staffing decisions across different census scenarios.

Abstract

Nursing homes are critical facilities for caring frail older adults with round-the-clock formal care and personal assistance. To ensure quality care for nursing home residents, adequate staffing level is of great importance. Current nursing home staffing practice is mainly based on experience and regulation. The objective of this paper is to investigate the viability of experience-based and regulation-based strategies, as well as alternative staffing strategies to minimize labor costs subject to heterogeneous service demand of nursing home residents under various scenarios of census. We propose a data-driven analysis framework to model heterogeneous service demand of nursing home residents and further identify appropriate staffing strategies by combing survival model and computer simulation techniques as well as domain knowledge. Specifically, in the analysis, we develop an agent-based simulation tool consisting of four main modules, namely individual length of stay predictor, individual daily staff time generator, facility level staffing strategy evaluator, and graphical user interface. We use real nursing home data to validate the proposed model, and demonstrate that the identified staffing strategy significantly reduces the total labor cost of certified nursing assistants compared to the benchmark strategies. Additionally, the proposed length of stay predictive model that considers multiple discharge dispositions exhibits superior accuracy and offers better staffing decisions than those without the consideration. Further, we construct different census scenarios of nursing home residents to demonstrate the capability of the proposed framework in helping adjust staffing decisions of nursing home administrators in various realistic settings.