199 Previous Visits as a Predictor of Revisits: A Retrospective Cohort Analysis

Abstract

The proportion of patients discharged from emergency departments (EDs) with short-term return visits has been scrutinized as both an internal and external quality metric, as well as a target for interventions. The ability to accurately predict which patients are more likely to revisit the ED could allow emergency departments and health systems to develop more focused interventions, but efforts to reduce revisits have not yet found success. A separate but related thread of literature has focused on patients with a high number of ED visits (frequent visitors) but the relationship between these two phenomena remains unexplored. Using a unique dataset with encounter-level data spanning several states we set out to identify the predictors of short-term return visits, with a particular focus on an often-overlooked characteristic: previous ED visits. We conducted a retrospective analysis of patients discharged from EDs from 80 hospitals in seven states from January 2014 - July 2016. Encounter data were obtained from CEP America, a multistate physician partnership that contracts with hospitals to provide ED provider staffing. Using a multivariable logistic regression we regressed short-term return visit on patient demographics and comorbidities, visit characteristics, physician, hospital characteristics, community characteristics, and an indicator for frequent visitor. We defined frequent visitor as a patient having two or more visits in the six months preceding the index visit; we conducted sensitivity analysis using several thresholds for number of previous visits to define frequent visitor, and also for alternate time horizons for return visit. Over the study period, there were 11,871,943 total visits; after excluding non-eligible visits the sample size was 6,699,717. The overall risk of 14-day revisit was 12.6%. Patients defined as frequent visitors were significantly more likely to have a 14-day return visit than non-frequent visitors (odds ratio [OR] 3.52, 95% confidence interval [CI] 3.50 - 3.53); they accounted for 18.7% of all visits and 40.2% of all 14-day revisits. In the full model, frequent visits was associated with the highest odds of a revisit: OR 3.06 (95% CI 3.04 - 3.07). Other predictors of 14-day revisits were skin and subcutaneous tissue infections (OR 2.13 (95% CI 2.11 - 2.16), alcohol-related disorders (OR 1.58, 95% CI 1.55 - 1.61), congestive heart failure (OR 1.58, 95% CI 1.52 - 1.63), and public insurance: Medicaid (OR 1.51 95% CI 1.5 - 1.3) and Medicare (OR 1.60, 95% CI 1.58 - 1.62). All other variables had odds ratios below 1.5 or greater than 0.8. Sensitivity analyses did not find any meaningful differences using a cutoff of 1, 3, or 4 or more previous visits to define frequent visitor, 3-, 7-, or 30-day return visit time horizons, or a sample restricted to adults. We found that recent ED visits was associated with a return visit. This finding was clinically significant, held true when controlling for patient, visit, hospital, and community characteristics, and was the strongest predictor of a return visit in the full model. We found that one previous visit was associated with a more-than-twofold risk of return, and that higher thresholds defining frequent visitors were associated with an even higher risk of return visit. Clinicians and policymakers should consider previous ED use when identifying or designing policies targeting patients at risk of short-term return visits.