Rapid global aging has led to an increasing demand for long-term care services for the elderly; however, current long-term care systems are underdeveloped and under-resourced. It is essential to develop an effective classification system to guide resource allocation that is tailored to the needs of the elderly.

This study aimed to construct a classification system for long-term care service needs among the elderly based on cluster analysis and machine learning.

Multi-center, cross-sectional study.

Community and nursing homes in Changsha, Hunan Province, China.

1270 elderly aged ≥ 65 years, who were randomly divided into the training set (70%) and test set (30%).

Cluster analysis was conducted based on service time from caregivers, nurses, and doctors. Machine learning approaches were used to determine classification criteria based on sociodemographic information and 17 secondary indicators of long-term care service needs. The best model was selected based on its accuracy and area under the curve in the test set and was then interpreted using Shapley Additive exPlanations.

Five clusters of needs for long-term care services were identified, with the number and proportion of participants in each cluster as follows: 612 (48.2%), 299 (23.5%), 172 (13.5%), 150 (11.8%), and 37 (2.9%), respectively. The five clusters varied significantly in sociodemographic characteristics and long-term care service needs. The random forest model demonstrated the best predictive performance among the five models tested. The Shapley Additive exPlanations method identified the 10 most influential features that impacted the random forest model predictions.

The classification system for long-term care service needs can accurately distinguish among the elderly with varying levels of long-term care needs, guiding optimal service allocation and promoting the sustainable development of the long-term care service system.

The trial is registered at the Ethics Committee of Xiangya Hospital, Central South University (No. 202105083). Registration date 05/2021. First recruitment 06/2021.