Exploring the optimal allostatic load scoring method in women of reproductive age

Aims The aim of this study was to determine the optimal allostatic load scoring method. Design This is a secondary analysis of data on women of reproductive age from the 2001-2006 National Health and Nutrition Examination Survey. Methods We created allostatic load summary scores using five scoring methods including the count-based, Z-Score, logistic regression, factor analysis and grade of membership methods. Then, we examined the predictive performance of each allostatic load summary measure in relation to three outcomes: general health status, diabetes and hypertension. Results We found that the allostatic load summary measure by the logistic regression method had the highest predictive validity with respect to the three outcomes. The logistic regression method performed significantly better than the count-based and grade of membership methods for predicting diabetes as well as performed significantly better for predicting hypertension than all of the other methods. But the five scoring methods performed similarly for predicting poor health status. Conclusion We recommended the logistic regression method when the outcome information is available, otherwise the frequently used simpler count-based method may be a good alternative. Impact The study compared different scoring methods and made recommendations for the optimal scoring approach. We found that allostatic load summary measure by the logistic regression method had the strongest predictive validity with respect to general health status, diabetes and hypertension. The study may provide empirical evidence for future research to use the recommended scoring approach to score allostatic load. The allostatic load index may serve as an 'early warning' indicator for health risk.