Predictive models for identifying risk of readmission after index hospitalization for hip arthroplasty: A systematic review.

Background: An aging United States population profoundly impacts healthcare from both a medical and financial standpoint, especially with an increase in related procedures such as Total Hip Arthroplasty (THA). The Hospital Readmission Reduction Program and Comprehensive Care for Joint Replacement Program incentivize hospitals to decrease post-operative readmissions by correlating reimbursements with smoother care transitions, thereby decreasing hospital burden and improving quantifiable patient outcomes. Many studies have proposed predictive models built upon risk factors for predicting 30-day THA readmissions.

Questions: (1) Are there validated statistical models that predict 30-day readmissions for THA patients when appraised with a standards-based, reliable assessment tool?. (2) Which evidence-based factors are significant and have support across models for predicting risk of 30-day readmissions post-THA?

Methods: Five major electronic databases were searched to identify studies that examined correlations between post-THA readmission and risk factors using multivariate models. We rigorously applied the PRISMA methodology and TRIPOD criteria for assessment of the prognostic studies.

Results: We found 26 studies that offered predictive models, of which two presented models tested with validation cohorts. In addition to the many factors grouped into demographic, administrative, and clinical categories, bleeding disorder, higher ASA status, discharge disposition, and functional status appeared to have broad and significant support across the studies.

Conclusions: Reporting of recent predictive models establishing risk factors for 30-day THA readmissions against the current standard could be improved. Aside from building better performing models, more work is needed to follow the thorough process of undergoing calibration, external validation, and integration with existing EHR systems for pursuing their use in clinical settings. There are several risk factors that are significant in multiple models; these factors should be closely examined clinically and leveraged in future risk modeling efforts.