Deriving injury risk curves using survival analysis from biomechanical experiments

Area of Science:

• Biomechanics and Automotive Safety Engineering
• Statistical Modeling for Injury Prediction
• Occupant Protection Systems

Background:

• Injury risk curves are crucial for automotive crashworthiness and occupant safety.
• Current methods using binary regression and survival analysis have limitations.
• International Standards Organization guidelines suggest survival analysis but require refinement.

Purpose of the Study:

• To present an improved, robust, and generalizable methodology for injury risk curve analysis.
• To address deficiencies in existing survival analysis approaches for injury prediction.
• To enhance the accuracy and applicability of biomechanical data analysis in automotive safety.

Main Methods:

• Statistical identification of optimal independent variables using Area Under the Receiver Operator Curve (AUC).
• Quantitative determination of the best probability distribution via Akaike Information Criterion (AIC).
• Objective comparison of distributions, identification of influential observations, and appropriate confidence interval estimation.

Main Results:

• A refined methodology for developing injury risk curves was developed and validated.
• The approach demonstrated feasibility using post-mortem human subject data and 24 thoracic/abdominal injury metrics.
• The method is applicable to various loading scenarios, including underbody blast events.

Conclusions:

• The proposed methodology offers a robust and generalizable framework for injury risk assessment.
• Implementation with commercial or open-source packages is feasible for retrospective and prospective studies.
• This advancement can significantly contribute to improved automotive occupant safety and experimental design.