Using Age-Specific Rates for Parametric Survival Function Estimation in Simulation Models

Area of Science:

• Epidemiology
• Biostatistics
• Health Economics

Background:

• Individual-level simulation models are crucial for health economic evaluations.
• These models often require individual patient data, which may not always be available.
• Age-specific event rates are frequently accessible, but direct individual data is lacking.

Purpose of the Study:

• To describe a procedure for incorporating parametric functions into individual-level simulation models.
• To sample time-to-event data when only age-specific rates are available.
• To facilitate simulation modeling in the absence of individual-level data.

Main Methods:

• Parametric survival distributions (Weibull, Gompertz, log-normal, log-logistic) were parametrized using age-specific event rates via regression analysis.
• The best-fitting distribution was selected using the R-squared statistic.
• The chosen parametric function was applied to assign random times to events in simulation models, using Spanish stroke rates as an example.

Main Results:

• The Gompertz, Weibull, and log-normal distributions demonstrated a good fit to the data up to 85 years of age.
• The Gompertz distribution was identified as the best-fitting distribution based on its goodness of fit.
• The procedure successfully incorporated parametric risk functions into simulation models.

Conclusions:

• A straightforward procedure is provided for integrating parametric risk functions into individual-level simulation models.
• This method enables the simulation of time-to-event data using readily available age-specific rates.
• The approach allows for the inclusion of parameter uncertainty in simulations, crucial for probabilistic sensitivity analysis.