# Measuring frailty: a comparison of the cumulative deficit model of frailty in survey and routine data

## TL;DR

This study compares how well survey and routine data can measure frailty using the cumulative deficit model, finding that both methods give similar results despite differences in the specific health issues measured.

## Contribution

The study demonstrates that the cumulative deficit model of frailty is robust to differences in deficit selection across diverse data sources.

## Key findings

- Frailty index scores were similar in survey and routine data despite differences in deficit types.
- Differences in deficit capture were offset, supporting the model's robustness.
- Sex and age influenced frailty similarly in both datasets.

## Abstract

We assess whether similar frailty prevalence is observed in survey and routine data despite differences in the number, types and prevalence of the individual deficits measured in frailty indices.

Frailty index calculations using established principles showed strong comparability between survey and routine data. Including a sufficient range of deficits does not significantly alter population-level frailty measures.

The cumulative deficit model’s robustness to deficit selection supports flexible approaches to population frailty assessment across diverse data sources, such as survey and routine data.

The online version contains supplementary material available at 10.1007/s41999-025-01251-7.

Frailty, a state of increased vulnerability to adverse health outcomes, impacts individuals and healthcare systems. The cumulative deficit model provides a flexible frailty measure but its application across diverse data remains underexplored. This study compares frailty indices derived from survey and routine data.

Frailty indices in the Clinical Practice Research Datalink (CPRD) Aurum (N = 1,625,677) and the English Longitudinal Study of Ageing (ELSA) (N = 5190) were compared for adults aged 65 + in England. Deficits were categorised as “one-to-one”, “one-to-many”, and “one-to-none”. Age-sex-standardised deficit prevalence, frailty distribution and associations with demographics were analysed using summary statistics and regression.

Mean frailty index scores were similar (CPRD: 0.13 ± 0.10; ELSA: 0.13 ± 0.12) but differences were observed in the capture of specific deficits. The majority of deficits had a “one-to-none” or “one-to-many” mapping. Among 14 comparable deficits, visual impairment, fractures and heart failure were more common in CPRD, while falls, sleep disturbance and arthritis were more frequent in ELSA. Severe frailty and greater fitness were more prevalent in ELSA than CPRD. Sex and age influenced frailty similarly in both datasets, with frailty index scores increasing more rapidly with age in CPRD.

Differences in the number and types of deficits measured offset each other overall, supporting the cumulative deficit model’s premise that including a sufficient range of deficits does not significantly alter population-level frailty measures. This interchangeability may alleviate concerns about deficit selection, supporting more flexible approaches to population frailty assessment across both survey and routine data.

The online version contains supplementary material available at 10.1007/s41999-025-01251-7.