Prevalence of frailty in 62 countries across the world: a systematic review and meta-analysis of population-level studies

Abstract

Introduction

The prevalence of frailty at population level is unclear. We examined this in population-based studies, investigating sources of heterogeneity.

Methods

PubMed, Embase, CINAHL and Cochrane Library databases were searched for observational population-level studies published between 1 January 1998 and 1 April 2020, including individuals aged ≥50 years, identified using any frailty measure. Prevalence estimates were extracted independently, assessed for bias and analysed using a random-effects model.

Results

In total, 240 studies reporting 265 prevalence proportions from 62 countries and territories, representing 1,755,497 participants, were included. Pooled prevalence in studies using physical frailty measures was 12% (95% CI = 11–13%; n = 178), compared with 24% (95% CI = 22–26%; n = 71) for the deficit accumulation model (those using a frailty index, FI). For pre-frailty, this was 46% (95% CI = 45–48%; n = 147) and 49% (95% CI = 46–52%; n = 29), respectively. For physical frailty, the prevalence was higher among females, 15% (95% CI = 14–17%; n = 142), than males, 11% (95% CI = 10–12%; n = 144). For studies using a FI, the prevalence was also higher in females, 29% (95% CI = 24–35%; n = 34) versus 20% (95% CI = 16–24%; n = 34), for males. These values were similar for pre-frailty. Prevalence increased according to the minimum age at study inclusion. Analysing only data from nationally representative studies gave a frailty prevalence of 7% (95% CI = 5–9%; n = 46) for physical frailty and 24% (95% CI = 22–26%; n = 44) for FIs.

Conclusions

Population-level frailty prevalence varied by classification and sex. Data were heterogenous and limited, particularly from nationally representative studies making the interpretation of differences by geographic region challenging. Common methodological approaches to gathering data are required to improve the accuracy of population-level prevalence estimates.

Protocol registration

PROSPERO-CRD42018105431.