Modifiable predictive factors and all-cause mortality in the non-hospitalized elderly population: An umbrella review of meta-analyses.

Abstract

This umbrella review aimed to summarize the association between modifiable predictive factors and all-cause mortality in the non-hospitalized elderly population, and estimated the credibility and strength of the current evidence. PubMed, Embase, Web of science, and EBSCOhost were searched up to February 28, 2022. Random-effect summary effect sizes and 95% confidence intervals (CIs), heterogeneity, small-study effect, excess significance bias, as well as 95% prediction intervals (PIs) were calculated. Methodological quality was assessed with the Assessment of Multiple Systematic Reviews 2 (AMSTAR-2) tool. The credibility of the included meta-analyses was graded from convincing to weak using established criteria. This umbrella review was registered with PROSPERO, CRD 42021282183. In total, 32 predictive factors involving 49 associations extracted from 35 meta-analyses were analyzed. Forty-three of the 49 (87.8%) associations presented nominal significant effects by the random-effect model (P<0.05), of which 34 had harmful associations and nine had beneficial associations with all-cause mortality. Frailty (FRAIL scale), low short physical performance battery (SPPB) score, and fewer daily steps carried a more than three-fold risk for all-cause mortality. Convincing evidence showed that weight fluctuation, prefrailty and frailty status, sarcopenia, low SPPB score, fewer daily steps, and fatigue increased the risk of all-cause mortality, while daily moderate-to-vigorous physical activity (MVPA) duration and total physical activity participation reduced the risk of death. There were twenty, nine, five, and six associations that yielded highly suggestive, suggestive, weak, and non-significant grades of evidence. Thirty-four (69.4%) of the associations exhibited significant heterogeneity. Twenty-two associations presented 95% PIs excluding the null value, two indicated small-study effects, and three had evidence for excess significance bias, respectively. The methodological quality of most meta-analyses was rated as low (37.1%) or critically low (42.9%). A summary of the currently available meta-analyses suggests that a broad range of modifiable predictive factors are significantly associated with all-cause mortality risk in the non-hospitalized elderly population. The most credible evidence indicates that physical function represented by frailty and sarcopenia, as well as physical activity, are significant predictors for all-cause mortality. This umbrella review may provide prognostic information to direct appropriate diagnostic evaluation and treatment goals in the future. More solid evidence is still needed coming from moderate-to-high quality meta-analyses.