A C-index for recurrent event data: Application to hospitalizations among dialysis patients

Area of Science:

• Biostatistics
• Epidemiology
• Health Services Research

Background:

• Assessing regression model discrimination for recurrent event risk is challenging.
• Existing measures are limited for complex event data, such as recurrent hospitalizations.

Purpose of the Study:

• To propose a novel C-index (index of concordance) for evaluating discrimination in regression models for recurrent event data.
• To address the need for robust measures in analyzing time-to-event data with multiple occurrences.

Main Methods:

• Derivation of theoretical properties of the proposed C-index under the proportional rates model.
• Development of computationally convenient inference procedures using perturbed influence functions.
• Application to hospitalization data from the Dialysis Outcomes and Practice Patterns Study (DOPPS).

Main Results:

• The proposed C-index demonstrates good performance in simulations with moderate sample sizes.
• Adding country indicators to a basic model significantly improved discrimination for recurrent hospitalizations.
• Further discrimination improvements were observed with the inclusion of comorbidity indicators.

Conclusions:

• The new C-index provides a valuable tool for assessing model discrimination in recurrent event data analysis.
• Country and comorbidity factors are important predictors of recurrent hospitalizations in end-stage renal disease patients.
• The methods offer practical improvements for risk prediction in clinical and epidemiological studies.