Reliability of reconstructed population survivorship from transition analysis age estimation

Abstract

Skeletal age markers are often used to estimate age-at-death to reconstruct population survivorship in the past, which is essential to draw conclusions about population health, fertility, and mortality. Transition Analysis age estimation was developed to address common issues facing traditional age estimation methods such as age mimicry and broad ranges for middle and older aged adults. However, some studies have shown Transition Analysis' limitations in overcoming these challenges, with regards to the method's accuracy and bias when applied at the individual-level. It was previously found that the method performed poorly in a sample from the Hamann-Todd Osteological Collection due to a series of factors. This study re-examines the application of Transition Analysis (ADBOU) age estimation in a sample from the Hamann-Todd Collection for reconstructing the survivorship profile of this skeletal sample. Although Transition Analysis is argued to be better suited for reconstructing trends in population mortality compared to traditional age estimation methods, results of Kaplan–Meier and log rank tests show that Transition Analysis produced significantly different survival curves compared to the known survival curves for White males (χ² = 6.828, p = 0.009) and females (χ² = 11.270, p = 0.001), but not for Black males (χ² = 1.363, p = 0.243) and females (χ² = 0.071, p = 0.790). Age-at-death distributions between the estimated maximum likelihoods and known ages also differed significantly (t = 4.249, p < 0.001). Therefore, caution should be employed when drawing conclusions from trends in survivorship based on skeletal age estimates, even when using more recently developed methods such as Transition Analysis, as the informative priors in ADBOU may not be accurate for all populations.