Abstract 168: The ADHERE Risk Model for In-hospital Worsening Heart Failure

Abstract

Background: A subset of patients hospitalized with acute heart failure (HF) experience in-hospital worsening heart failure (WHF), defined as persistent or worsening signs or symptoms requiring an escalation of therapy. These patients are at increased risk for mortality and readmission resulting in increased costs, yet predictors for the development of WHF remain poorly characterized and an in-hospital WHF risk model has not been derived and validated. Methods: We analyzed data from the Acute Decompensated Heart Failure National Registry (ADHERE) linked to Medicare claims to develop and validate a risk model for in-hospital WHF. Based on previous work, in-hospital WHF was defined by any of the following: use of inotropes or vasodilators >12 hours after admission; initiation of mechanical circulatory support, hemodialysis, or ventilation after the first inpatient day; or transfer to an intensive care unit. We considered candidate risk prediction variables routinely assessed on admission including age, medical history, biomarkers and renal function. Logistic regression with robust standard errors was used to generate a risk model for in-hospital WHF in a 66% random derivation sample; the model was validated in the remaining 34% sample. We evaluated the calibration and discrimination of the model in both samples. Results: We evaluated 23,696 patients hospitalized with acute HF. Baseline characteristics were well-matched in the derivation and validation samples. The mean age was 81 years (standard deviation [SD] 7.7), mean systolic BP was 145 mm Hg (SD 29.8), and mean creatinine was 1.5 mg/dL (SD 0.9). The occurrence of in-hospital WHF was similar in the derivation, 15.4% (2403 of 15,640), and validation samples, 15.6% (1257 of 8056). After multivariable adjustment, the strongest predictors of in-hospital WHF were positive admission troponin and admission creatinine (Table). The model was well-calibrated. The risk model had strong discrimination in the derivation (c statistic=0.74) and validation samples (c statistic=0.72). Conclusion: The ADHERE WHF risk model is a validated clinical tool with good discrimination for use in hospitalized acute HF patients to identify patients at risk for in-hospital WHF. This tool may be useful to target treatment strategies for patients at high risk for in-hospital WHF.