When survival analyses are utilized in decision analytic models the tails of survival curves are often checked against mortality from general population life tables to ensure hazards from the disease-specific survival curves are not incongruent (less than) than the general population. This approach uses the average age of the modeled population and corresponding age-specific mortality from life tables with updates to mortality as the model cycles.

This approach works when the model population is homogeneous in terms of age. However, in a heterogenous population the “wall of human mortality” can cause the average age approach to overestimate the hazards of model population sans disease. An alternate methodology that recreates the model population using survival of multiple age groups rather than a single can provide more accurate hazard adjustments. The impact of the traditional method and the new age-matched life table (AMLT) method is compared in an illustrative example.

In the AMLT method, the age distribution of the modeled population is used to distribute patients into reasonable age groups. Next, lifetime survival for the same population sans disease within each age group is calculated from life tables. Finally, the age groups are rolled up based on the population surviving in each year and group to create survival/hazard estimates for an age-matched cohort.

A survival analysis utilizing mortality data following kidney transplant from the United Kingdom Transplant Registry was conducted. The mean age of patients in the cohort was 45.4 years and age distribution was 0-17 =6.4%, 18-34=12.8%, 35-49=38.6%, 50-59=22.8%, 60-69=16.3%, and 70+=3.2%. Utilizing the average method resulted in tail-corrected mean survival of 24.70 years. Utilizing AMLT resulted in tail-corrected mean survival of 26.82 years, a difference of 2.12 years.

This methodology will provide more accurate survival tail corrections in models when the population is heterogeneous in terms of age.