Development and validation of an Index of Social Attributes Predictive of Frailty Among Older Adults
Abstract
Background: Frailty captures accumulating age-related deficits in the normal functioning of an adult. The existing frailty models include physical and behavioral attributes; however, they ignore the complex social environments that impact them. The varying environment and estimation processes further make cross-country comparisons of frailty difficult. The Commonwealth Fund’s International Health Policy (IHP) survey data on older adults allows one to compare social predictors of frailty across high-income countries.
Methods: Our objective was to develop and validate a social risk measure to predict frailty across the countries using comparable social characteristics based on Social Production Function Theory. We employed the least absolute shrinkage and selection operator (LASSO) regression to identify a small subset of predictors for our logistic model using IHP 2017 data and including country-level fixed effects. A 10-fold cross-validation was performed to identify the lambda parameter within one SE of the minimum. The model was validated using IHP 2021 data to determine discrimination and calibration. Outcomes considered for analysis were inability to perform instrumental activities, hospitalization, and number of emergency room (ER) visits.
Results: In preliminary results, 24 attributable risks, including demographic characteristics, were identified, of which 16 appeared in the final model. The validation results showed the excellent discrimination capacity of the model (AUC ranging between 0.75 and 0.80). The confusion matrix showed the accuracy of the predictions ranged between 0.70 and 0.86.
Conclusion: A novel social frailty index can predict frailty using cross-country data. The first of such efforts allows for predicting frailty and associated adverse health outcomes for older adults using comparable social indicators across 11 high-income countries.
Keywords
Frailty, Older Adults, Commonwealth Fund, Social Production Function Theory, LASSO Regression
Development and validation of an Index of Social Attributes Predictive of Frailty Among Older Adults
Background: Frailty captures accumulating age-related deficits in the normal functioning of an adult. The existing frailty models include physical and behavioral attributes; however, they ignore the complex social environments that impact them. The varying environment and estimation processes further make cross-country comparisons of frailty difficult. The Commonwealth Fund’s International Health Policy (IHP) survey data on older adults allows one to compare social predictors of frailty across high-income countries.
Methods: Our objective was to develop and validate a social risk measure to predict frailty across the countries using comparable social characteristics based on Social Production Function Theory. We employed the least absolute shrinkage and selection operator (LASSO) regression to identify a small subset of predictors for our logistic model using IHP 2017 data and including country-level fixed effects. A 10-fold cross-validation was performed to identify the lambda parameter within one SE of the minimum. The model was validated using IHP 2021 data to determine discrimination and calibration. Outcomes considered for analysis were inability to perform instrumental activities, hospitalization, and number of emergency room (ER) visits.
Results: In preliminary results, 24 attributable risks, including demographic characteristics, were identified, of which 16 appeared in the final model. The validation results showed the excellent discrimination capacity of the model (AUC ranging between 0.75 and 0.80). The confusion matrix showed the accuracy of the predictions ranged between 0.70 and 0.86.
Conclusion: A novel social frailty index can predict frailty using cross-country data. The first of such efforts allows for predicting frailty and associated adverse health outcomes for older adults using comparable social indicators across 11 high-income countries.
