What question did this study set out to answer?

To develop a Frailty Index based on Item Response Theory for better comparability across studies and cohorts.

January 23, 2026

Development and Evaluation of a Cross-Study Comparable Frailty Index based on Item Response Theory: Insights from Three Prospective Cohorts

SLShuai LiuSichuan University GRGuangquan RanSichuan University YWYan WangSichuan University

Key Points

To develop a Frailty Index based on Item Response Theory for better comparability across studies and cohorts.
Constructed a frailty index using data from three large cohorts: CHARLS, HRS, and SHARE.
Utilized a two-parameter logistic model to estimate frailty scores.
Employed Cox and logistic regression analyses to examine mortality associations.
Conducted tests for robustness regarding variable selection and population stability.
Frailty prevalence varied across cohorts: 16.3% (CHARLS), 28.5% (HRS), and 15.5% (SHARE).
FI-IRT demonstrated strong agreement with the traditional FI-CD method.
Higher FI-IRT scores significantly linked to increased risk of mortality across all cohorts (hazard ratios: 1.81-2.38).
Maintained stability even without certain domain-specific variables (intraclass correlation coefficient of 0.934).
Consistency across different population models shown with high Spearman correlation coefficients (0.971-0.999).

Abstract

Abstract Background The Frailty Index of Accumulative Deficits (FI-CD) is a key predictor of adverse health outcomes in older adults, but its utility for cross-study comparisons is limited due to assumptions of item homogeneity. Methods We constructed a Frailty Index based on Item Response Theory (FI-IRT) in frailty measures, including 51 health-related variables, in older adults from the China Health and Retirement Longitudinal Study (CHARLS) 2011-2020, the Health and Retirement Study (HRS) 2010-2020, and the Survey of Health, Aging, and Retirement in Europe (SHARE) 2010-2020. A two-parameter logistic (2PL) model estimated FI-IRT scores, and Cox/logistic regression assessed its associations with mortality. Several analyses tested its robustness to the variable set and population selection. Results The FI-IRT followed a slightly right-skewed, approximately normal distribution, with frailty prevalence of 16.3% (CHARLS), 28.5% (HRS), and 15.5% (SHARE). The FI-IRT showed a high level of agreement with the FI-CD, both on a continuous scale and a hierarchical scale. A higher FI-IRT was associated with an increased risk of all-cause mortality across all three cohorts (hazard ratios: 1.81-2.38, P 0.001). When domain-specific variables were excluded, the FI-IRT continued to offer relatively stable estimations (intraclass correlation coefficient: 0.934; 95% CI: 0.933-0.934; P 0.001). A high degree of consistency was observed between the FI-IRTs calculated from models constructed based on different subpopulations (Spearman’s rank correlation coefficients: 0.971-0.999, P 0.001; intraclass correlation coefficient: 0.945; 95%CI: 0.850-0.972; P 0.001). Conclusions The FI-IRT provides a useful, robust, and cross-study comparable measure of frailty.

AI에게 질문

Bookmark

View Full Paper