Monitoring time to event in registry data using CUSUMs based on relative survival models

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Disease surveillance requires monitoring changes in survival time distributions.
• Health registries provide data for real-time or retrospective monitoring.
• Uncertain or missing cause-of-death data presents a challenge in survival analysis.

Purpose of the Study:

• To propose a novel cumulative sum (CUSUM) procedure for monitoring changes in survival time distributions.
• To adapt existing methods for time-to-event data to the excess hazard setting.
• To incorporate relative survival methods for handling uncertain cause-of-death information.

Main Methods:

• Development of a CUSUM chart based on a survival log-likelihood ratio.
• Modeling the total hazard as the sum of population hazard and excess hazard.
• Accounting for changes in population risk and excess hazard explained by covariates.

Main Results:

• The proposed CUSUM procedure effectively monitors changes in survival time distributions.
• The method is applicable even when cause-of-death information is missing or uncertain.
• Demonstrated application using cancer registry data.

Conclusions:

• The CUSUM procedure offers a robust method for disease surveillance.
• It enhances the analysis of survival data with excess hazard models.
• This approach improves the quantification of disease burden from registry data.