An update of the statistical methods underlying the FELS method of skeletal maturity assessment.

Background: Evaluation of skeletal maturity provides clinicians and researchers a window into the developmental progress of the skeleton. The FELS method for maturity assessment provides a point estimate and standard error based on 98 skeletal indicators. Aim: This paper outlines the statistical methodology used by the original FELS method and evaluates improvements that address the following: serial correlation in the calibration sample is now considered, a Bayesian estimation method is now employed to improve estimation near ages 0 and 18 years and uncertainty in the calibration due to sampling is now accounted for when computing confidence limits. Subjects and methods: The original FELS method was calibrated using 677 Fels Longitudinal Study participants. In the improved method, serial correlation is accounted for using GEE, a Bayesian analysis with a prior centred on chronological age is used and the bootstrap is used to account for all sources of variation. Results: Accounting for serial correlation resulted in larger slopes for ordinal indicators. The Bayesian paradigm led to narrower confidence limits and a natural interpretation of skeletal age. Sampling variability in the calibration parameters was negligible. Conclusion: Improvements to the statistical basis of the FELS method provide a more effective method of estimating skeletal maturity.