A relative survival model for clustered responses.

Area of Science:

• Epidemiology
• Biostatistics
• Medical Care Research

Background:

• Relative survival analysis estimates disease impact by comparing patient survival to expected survival in similar populations.
• Existing regression models for relative survival do not adequately handle clustered data, limiting their application in certain epidemiological studies.
• Accurate estimation of disease-specific mortality is crucial for public health and healthcare planning.

Purpose of the Study:

• To extend existing regression models for relative survival analysis to accommodate clustered responses.
• To apply these novel methods to an epidemiological dataset investigating medical care provision for lung cancer patients.
• To enhance the estimation of disease effects in registries where cause of death may be uncertain.

Main Methods:

• Development of Generalized linear mixed models (GLMM) for relative survival analysis.
• Embedding relative survival regression within the GLMM framework to account for data clustering.
• Application and demonstration using data from the HALLUCA study on lung cancer care.

Main Results:

• The proposed GLMM-based approach effectively models relative survival with clustered data.
• The methodology provides a robust framework for analyzing epidemiological data with complex structures.
• Demonstration on the HALLUCA study data highlights the practical utility of the extended models.

Conclusions:

• Generalized linear mixed models offer a powerful extension for relative survival regression, particularly for clustered data.
• This approach enhances the accuracy of disease impact estimation in epidemiological studies.
• The developed methods are valuable for analyzing healthcare provision and outcomes in specific patient cohorts, such as lung cancer.