Validation of the Health Assessment Tool (HAT) based on four aging cohorts from the Swedish National study on Aging and Care

Abstract

Abstract Background As global aging accelerates, routinely assessing the functional status and morbidity burden of older patients becomes paramount. The aim of this study is to assess the validity of the comprehensive clinical and functional Health Assessment Tool (HAT) based on four cohorts of older adults (60 + years) from the Swedish National study on Aging and Care (SNAC) spanning urban, suburban, and rural areas. Methods The HAT integrates five health indicators (gait speed, global cognition, number of chronic diseases, and basic and instrumental activities of daily living), providing an individual-level score between 0 and 10. The tool was constructed using nominal response models, first separately for each cohort and then in a harmonized dataset. Outcomes included all-cause mortality over a maximum follow-up of 16 years and unplanned hospital admissions over a maximum of 3 years of follow-up. The predictive capacity was assessed through the area under the curve (AUC) using logistic regressions. For time to death, Cox regressions were performed, and Harrell’s C-indices were reported. Results from the four cohorts were pooled using individual participant data meta-analysis and compared with those from the harmonized dataset. Results The HAT demonstrated high predictive capacity across all cohorts as well as in the harmonized dataset. In the harmonized dataset, the AUC was 0.84 (95% CI 0.81–0.87) for 1-year mortality, 0.81 (95% CI 0.80–0.83) for 3-year mortality, 0.80 (95% CI 0.79–0.82) for 5-year mortality, 0.69 (95% CI 0.67–0.70) for 1-year unplanned admissions, and 0.69 (95% CI 0.68–0.70) for 3-year unplanned admissions. The Harrell’s C for time-to-death throughout 16 years of follow-up was 0.75 (95% CI 0.74–0.75). Conclusions The HAT is a highly predictive, clinically intuitive, and externally valid instrument with potential for better addressing older adults’ health needs and optimizing risk stratification at the population level.