Cost Control of Treatment for Cerebrovascular Patients Using a Machine Learning Model in Western China

Area of Science:

• Health Economics
• Public Health
• Medical Informatics

Background:

• Cerebrovascular disease is a leading cause of death in China, necessitating urgent control of associated medical expenses.
• Diagnosis-related groups (DRG) are globally adopted for healthcare cost reduction, but regional variations in classification variables exist.
• The inclusion of length of stay (LOS) as a DRG classification variable remains a point of debate.

Purpose of the Study:

• To identify key factors influencing inpatient medical expenditure for cerebrovascular disease patients.
• To compare the performance of two DRG classification rule sets, with and without LOS.
• To evaluate the impact of controlling unreasonable medical treatment on cost savings.

Main Methods:

• Analysis of data from 45,575 inpatients in western China.
• Kruskal-Wallis H tests and multiple linear stepwise regression to identify cost influencers.
• Development of a chi-squared automatic interaction detector (CHAID) decision tree model for DRG grouping.

Main Results:

• Factors such as gender, age, insurance type, hospital level, LOS, surgery, outcomes, comorbidities, and hypertension significantly impacted inpatient costs (P < 0.05).
• DRG grouping with LOS resulted in seven groups, while without LOS yielded eight groups, requiring more clinical variables for comparable results.
• The DRG classification including LOS demonstrated a smaller coefficient of variation and lower upper cost limits, projecting annual savings of $3.35 million.

Conclusions:

• Incorporating LOS into DRG classification for cerebrovascular disease patients is effective for cost control.
• This approach optimizes healthcare resource allocation and reduces financial burden.
• The findings support the inclusion of LOS in DRG systems to enhance economic efficiency in managing cerebrovascular diseases.