Nonparametric analysis of competing risks data with event category missing at random.

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Competing risks data often have missing event categories (e.g., cause of death).
• Treating missing categories as censored data underestimates hazard functions.
• Accurate estimation is crucial for understanding disease progression and outcomes.

Purpose of the Study:

• To develop nonparametric estimators for cumulative cause-specific hazards and incidence functions.
• To address the challenge of missing event categories in competing risks analysis.
• To provide a robust methodology for analyzing complex survival data.

Main Methods:

• Utilized the Nadaraya-Watson estimator to incorporate events with missing categories.
• Developed and derived properties of novel nonparametric estimators.
• Determined optimal bandwidth to minimize mean integrated squared errors.

Main Results:

• Proposed estimators provide more accurate estimates of cumulative cause-specific hazards.
• The methodology effectively handles missing event category information.
• Demonstrated improved performance compared to traditional methods that ignore missing data.

Conclusions:

• The developed nonparametric estimators are effective for competing risks data with missing event categories.
• This approach mitigates underestimation of hazard functions.
• The methodology is applicable to real-world health data, as shown in the Cystic Fibrosis Registry example.