Assessing differences among persistent, episodic, and non- high-need high-cost hospitalized children in China after categorization by an unsupervised learning algorithm

Peng Zhang; Bifan Zhu; Linan Wang

doi:10.1186/s13690-024-01442-x

Assessing differences among persistent, episodic, and non- high-need high-cost hospitalized children in China after categorization by an unsupervised learning algorithm

Arch Public Health. 2024 Nov 20;82(1):218. doi: 10.1186/s13690-024-01442-x.

Authors

Peng Zhang^#¹, Bifan Zhu^#², Linan Wang³

Affiliations

¹ School of Humanities, Shanghai Institute of Technology, Haiquan Road 100, Fengxian District, Shanghai, China.
² Shanghai Health Development Research Center, Shanghai, 20040, China.
³ Shanghai Health Development Research Center, Shanghai, 20040, China. wanglinan@shdrc.org.

^# Contributed equally.

PMID: 39568040
DOI: 10.1186/s13690-024-01442-x

Abstract

Background: High-need, high-cost (HNHC) patients are a major focus of international healthcare reform. However, research on HNHC children in China remains limited. This study aims to classify HNHC pediatric patients, analyze the differences among groups, and explore the factors influencing HNHC status.

Methods: Data were obtained from a retrospective observational cohort of hospitalized children in Shanghai, China from 2017 to 2023. K-means clustering, one of the unsupervised learning algorithms, was employed to classify patients according to their HNHC status. Descriptive statistical analysis and the Kruskal-Wallis H test were used to describe and test the differences among different groups, with the logit regression models to analyze the predictors.

Results: 688,131 hospitalized children were classified into three groups: 1,871 persistent HNHC, 32,539 episodic HNHC, and 653,721 non-HNHC. Significant differences were observed among these groups. Persistent HNHC patients have significantly higher costs and longer HNHC durations compared to episodic and non-HNHC patients, who were more likely to be aged 30 days to 1 year or 13-18 years, female with only one type of health insurance, and leukemia was the most prevalent and costly disease. They exhibited distinct healthcare utilization patterns, including emergency admissions, higher surgery rates, longer hospital stays, more frequent hospitalizations, and a preference for tertiary and specialized hospitals in city centers. Multiple influencing factors of persistent HNHC versus episodic HNHC and non-HNHC were identified.

Conclusion: This study provides valuable insights into the classification, characteristics, and influencing factors of persistent, episodic, and non-HNHC hospitalized children in China. Persistent HNHC patients warrant targeted interventions to improve health outcomes and reduce healthcare costs. Enhanced medical coverage for key diseases, high-quality healthcare services tailored to their needs, and early interventions are crucial.

Keywords: China; Healthcare utilization; Hospitalized children; K-means clustering; Persistent high-need, high-cost.

Grants and funding

No. 23692109000/Science and Technology Commission of Shanghai Municipality "Research on Optimization Strategy of Children's Multi-Level Health Security System in Shanghai"