Abstract 14711: Prediction of Persistent High-Need High-Cost Cardiology Patients With and Without Renal Failure

Abstract

Introduction: Heart failure (HF) and coronary artery disease (CAD) are increasingly complex disease entities, in particular in older patients with multiple long-term conditions. High-Need High-Cost (HNHC) patients are at risk of receiving inefficient care that does not contribute to better outcomes. Prediction models may help to identify patients at risk of becoming HNHC with potential benefit for care coordination interventions. Aim: To predict persistent HNHC status at outpatient presentation for cardiovascular populations with HF, CAD, with and without renal failure (HF&RF; CAD&RF). Methods: We performed a retrospective cohort study in the claims database of a large healthcare insurer in the Netherlands. We included patients with consecutive data between 2015 - 2019, at their first outpatient appointment in 2016. Potential predictors were based on the previously published Persistent High Utilizer (PHU) model and collected in the year prior to presentation. HF, CAD and RF were defined by Diagnosis Treatment Combination insurance codes. The outcome of interest was PHU status, defined as belonging to the top 10% of total annualized healthcare expenditures for three consecutive years. To account for inflation, all costs were scaled towards 2015. Predictor effects were quantified through logistic multivariable regression analysis as odds ratios. Models were validated internally by bootstrapping procedure. Model performance was assessed as discrimination, quantified by the C-statistic. Results: 219.567 patients were included, of whom 2.0% ( n =4286) became PHU. The most important predictors were chronic kidney disease diagnosis (OR 3.1 95%CI 2.4-4.1), gastroenterology involvement (OR 3.0 95%CI 2.7-3.3) and COPD diagnosis (OR 2.1 95%CI 1.8-2.5). Model performance in the overall population was good (C-statistic 0.77). Model performance varied in the more homogeneous populations with HF ( n =2261; n =157 PHU), CAD ( n =7221; n =292 PHU), HF&RF ( n =163; n =40 PHU) and CAD&RF ( n = 196; n =34 PHU) resulting in C-statistics of 0.74, 0.78, 0.60 and 0.67 respectively. Conclusions: Persistent HNHC status can be predicted in the outpatient setting for patients with HF and CAD. These models can be used to select cardiology patients for HNHC care coordination.