A nonproportional hazards Weibull accelerated failure time regression model.

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Cardiovascular disease remains a leading cause of mortality.
• Accurate risk factor assessment is crucial for predicting heart disease incidence.
• Longitudinal studies are essential for understanding disease progression and risk dynamics.

Purpose of the Study:

• To evaluate risk factors for heart disease incidence.
• To assess the adequacy of standard statistical models in analyzing longitudinal data.
• To propose an improved statistical model for predicting heart disease risk.

Main Methods:

• Utilized data from the Framingham Heart Study spanning 32 years.
• Employed survival analysis techniques, including accelerated failure time models.
• Compared standard Weibull and Cox proportional hazards models with a proposed model where dispersion depends on the location parameter.

Main Results:

• Standard Weibull and Cox proportional hazards models were found to be inadequate for the dataset.
• A novel model, where the dispersion parameter is dependent on the location parameter, provided a better fit.
• This improved model indicates that the cumulative hazard ratio between individuals converges over time.

Conclusions:

• The findings challenge the assumptions of standard survival models in long-term epidemiological studies.
• The proposed model offers a more accurate representation of risk factor dynamics in heart disease.
• This has significant implications for estimating differences in predicted outcome probabilities between individuals.