Models for predicting short- to long-term mortality in older adults with hip fractures in clinical practice: a systematic review and meta-analysis.

BACKGROUND

Despite published summaries and appraisals of prognostic models for mortality after hip fracture, it remains unclear which models should be used with the evidence obtained to date. We systematically reviewed validated prognostic models for predicting mortality in older adults with hip fractures.

METHODS

A search of Medline, CINAHL, The Cochrane Library, CiNii, and Ichushi databases on April 2024. Model development with internal or external validation studies and external model validation studies of previously reported models were selected. Models predicting mortality outcomes at short- to long-term follow-up time points were included. We used the PROBAST instrument to assess the risk of bias and synthesized the studies' data to perform a meta-analysis. A structured five-step method was applied to determine which prognostic models were clinically valuable.

RESULTS

After screening 1,092 publications, we identified 24 studies (21 prognostic models) for this review. The model discrimination, measured by the area under the curve or C-statistic, ranged from 0.68 to 0.89 for the development models and 0.66-0.84 for the validation models. With the exception of the four validation models, all development and validation models were deemed to be at high risk of bias. Common concerns were participant and analysis domains. Our meta-analysis of the Nottingham Hip Fracture Score in 30-day mortality revealed a pooled C-statistic at 0.68 (95%CI: 0.66-0.71). A narrative synthesis identified the Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity and a mortality prediction score as predictors of in-hospital mortality and the Nottingham Hip Fracture Score as a predictor of 1-year mortality.

CONCLUSIONS

These results suggest that, based on the current evidence, recommended prognostic models for predicting mortality at each time point. However, given the high risk of bias in most studies, these models should be used with caution and only as adjunctive tools. Future studies should prioritize external validation and reducing the risk of bias.

TRIAL REGISTRATION: PROSPERO CRD42023462537.