Addressing the issues that arise in analysing multicentre cost data, with application to a multinational study.

Area of Science:

• Health Economics
• Biostatistics
• Multilevel Modeling

Background:

• International cost data analysis is complicated by variations in mean, spread, and skewness across countries.
• Patient and national characteristics influence healthcare costs, necessitating sophisticated analytical approaches.
• Hierarchical data structures are common in multinational health economic studies.

Purpose of the Study:

• To develop and apply generalized linear multilevel models for analyzing multinational cost data.
• To investigate the impact of patient and national characteristics on healthcare costs.
• To address the challenges posed by heterogeneity in cost data across different centers and countries.

Main Methods:

• Development of generalized linear multilevel models.
• Application of gamma distributions and multiplicative effects for patient characteristics.
• Comparison with models assuming normal distributions or additive effects.
• Utilization of a multilevel gamma model to account for center-specific heterogeneity.

Main Results:

• Gamma distributions with multiplicative effects provided a better fit for cost data compared to normal distributions or additive models.
• The developed multilevel gamma model effectively handled heterogeneity in patient case-mix effects across centers.
• The models successfully estimated the influence of patient and national factors on costs.

Conclusions:

• Multinational cost data analysis must account for differences in data distribution (mean, spread, skewness) across centers.
• Generalized linear multilevel models, particularly with gamma distributions, are suitable for analyzing complex, hierarchical cost data.
• Recognizing data heterogeneity and hierarchical structure is crucial for accurate interpretation of international health economic studies.