Predictive Models and Influencing Factors for the Work Relative Value Unit in Clinical Surgical Items Based on Generalized Linear Models and BP Neural Networks
- VernacularTitle:基于广义线性模型和BP神经网络的临床手术类项目人力资源消耗相对值影响因素及预测模型
- Author:
Haiyin WANG
1
;
Meifeng WANG
1
;
Liang FANG
1
;
Chunlin JIN
1
Author Information
1. 上海市卫生和健康发展研究中心 上海 201199
- Publication Type:Journal Article
- Keywords:
clinical surgical item;
work relative value unit;
influencing factor;
prediction model
- From:
Chinese Health Economics
2025;44(4):61-64
- CountryChina
- Language:Chinese
-
Abstract:
Objective:In order to establish benchmark values for new medical service items,it aims to investigate the predictive models and influencing elements that affect the work relative value unit in clinical surgical items in China.Methods:Generalized Linear Model(GLM)and BP neural network techniques were used to investigate influencing factors and build prediction models using the National Medical Service Project Technical Specification(2023 Edition)as the value database.Results:The average relative value of human resource usage was 41.9,with a total of 6 011 items across 16 systems and anesthesia.The GLM's mean prediction error was 4%and its linear correlation coefficient was 0.997.The top 5 predictor variables in terms of importance were technical complexity(0.45),risk level(0.30),physician time(0.08),number of physicians(0.06),and nurse time(0.03).With a mean prediction error of 1.5%,the neural network model obtained a correlation coefficient of 0.996.Technical difficulty(0.20),physician time(0.20),perfusionist time(0.19),risk level(0.15),and medical technician time(0.06)were the top five predictors.Conclusion:Both types of predictive models are well-fitted and valid,and future medical service items can provide relative values of human resource consumption,creating an integrated relative point system.
