Using multilevel models for assessing the variability of multinational resource use and cost data

Area of Science:

• Health Economics
• Statistical Modeling
• International Healthcare Research

Background:

• Multinational economic evaluations often pool cost data, assuming uniformity across countries.
• Assessing cost variation requires identifying factors influencing resource use and expenditure.
• Ordinary least-squares (OLS) regression has been traditionally used but may not suit hierarchical data.

Purpose of the Study:

• To compare the appropriateness of OLS and multilevel models (MLMs) for analyzing international cost variations.
• To identify patient- and center-level factors associated with length of hospital stay (LOS) and total cost in stroke admissions across European countries.
• To evaluate the validity of pooling international cost data.

Main Methods:

• Utilized a multinational dataset of 1300 stroke admissions from 13 centers in 11 European countries.
• Employed both OLS and MLMs to estimate the effects of patient and center covariates on LOS and total cost.
• Compared MLMs with normal and gamma distributions for within-center data.

Main Results:

• OLS models indicated both patient and center factors influenced LOS and total cost.
• MLMs revealed no center-level characteristics affected LOS.
• Health spending level was the primary center-level factor associated with total cost in MLM analyses.

Conclusions:

• OLS regression can lead to incorrect inferences when assessing international cost variation.
• MLMs are more suitable for accurately analyzing variations in resource use and healthcare costs across different centers and countries.
• Findings highlight the importance of appropriate statistical methods in multinational health economic evaluations.