Clinical Prediction Models for Nonmortality Outcomes in Older Adults With Hip Fractures: A Systematic Review

Several hip fracture clinical prediction models have been developed. We conducted this study to (i) map outcomes used in clinical prediction models for hip fracture, (ii) identify the domains and instruments of predictors, and (iii) assess the risk of bias.

We performed systematic searches of studies published from June 2002 to June 2023 in the PubMed, Cochrane Library, CINAHL, CiNii, and Ichushi databases. After the relevant articles were identified, we performed the data extraction and bias risk assessment. We used the Prediction Study Risk Of Bias Assessment Tool (PROBAST) to assess each study's risk of bias. Outcome mapping was performed for the core outcome set of hip fractures. Qualitative synthesis and the PROBAST evaluation were performed on other-than-mortality core outcomes, which are difficult to target in rehabilitation.

We screened 3 206 studies for eligibility; 45 studies were included in the outcome mapping, and 10 studies were included in the qualitative synthesis. Outcomes included mortality (n = 35), mobility (n = 8), and activities of daily living (n = 2). No clinical prediction models had pain or health-related quality of life as an outcome. Predictors were reported in 8 domains and 38 measures. The PROBAST evaluation showed a high risk of bias in all 10 studies that were eligible for a qualitative synthesis.

The clinical prediction models had only mortality, mobility, and activities of daily living as outcomes. The development of clinical prediction models with pain and health-related quality of life as outcomes is necessary. Clinical prediction models overcoming the risk of bias identified in this study are also needed.