Modeling absolute differences in life expectancy with a censored skew-normal regression approach

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Traditional survival regression models do not directly quantify life expectancy differences in years.
• Gaussian linear models are unsuitable for life expectancy due to skewed time-to-death distributions.
• Existing methods lack appropriate handling of censored and left-truncated data in life expectancy modeling.

Purpose of the Study:

• To present a skew-normal regression approach as a viable alternative for modeling life expectancy.
• To demonstrate how parameter estimates from skew-normal models can be interpreted as survival time differences.
• To adapt skew-normal regression for censored and left-truncated observations.

Main Methods:

• Utilizing a skew-normal distribution to model time-to-death data, accommodating negative skewness.
• Implementing regression techniques to account for censored and left-truncated survival data.
• Applying the method to Swiss National Cohort Study data and official life expectancy estimates.

Main Results:

• The skew-normal regression model provides interpretable parameter estimates reflecting survival time differences.
• The approach successfully models differences in life expectancy across various covariates.
• Comparison with standard survival regression models shows the advantages of the skew-normal method.

Conclusions:

• A censored skew-normal survival regression approach is effective for modeling life expectancy differences.
• This method accurately accounts for left-truncated observations in survival data.
• The proposed approach enhances the analysis of life expectancy variations in epidemiological studies.