Predictors of Post-Discharge Outcomes from Information Acquired Shortly After Admission for Acute Heart Failure: A Report from the PROTECT Study.
Circ Heart Fail. 2013 Nov 26;
Authors: Cleland JG, Chiswell K, Teerlink JR, Stevens S, Fiuzat M, Givertz MM, Davison BA, Mansoor GA, Ponikowski P, Voors AA, Cotter G, Metra M, Massie BM, O'Connor CM
BACKGROUND: -Acute heart failure (AHF) is a common reason for admission and outcome is often poor. Improved prognostic risk stratification may assist in the design of future trials and in patient management. Using data from a large randomized trial, we explored the prognostic value of clinical variables, measured at hospital admission for AHF, to determine whether a few selected variables were inferior to an extended data-set.
METHODS AND RESULTS: -The prognostic model included 37 clinical characteristics collected at baseline in PROTECT, a study comparing rolofylline and placebo in 2,033 patients admitted with AHF. Pre-specified outcomes at 30 days were death or re-hospitalization for any reason, death or re-hospitalization for cardiovascular or renal reasons and, at both 30 and 180 days, all-cause mortality. No variable had a c-index >0.70 and few had values >0.60; c-indices were lower for composite outcomes than for mortality. Blood urea was generally the strongest single predictor. Eighteen variables contributed independent prognostic information but a reduced model using only eight items (age, prior heart failure hospitalization, peripheral edema, systolic blood pressure, serum sodium, urea, creatinine and albumin) performed similarly. For all-cause mortality at 180-days, the model C-index using all variables was 0.72 and for the simplified model, also 0.72.
CONCLUSIONS: -A few simple clinical variables measured on admission in patients with AHF predict a variety of adverse outcomes with similar accuracy to more complex models. However, predictive models were of only moderate accuracy, especially for outcomes that included non-fatal events. Better methods of risk-stratification are required. Clinical Trial Registration-URL: http://www.clinicaltrials.gov. Unique identifiers: NCT00328692 and NCT00354458.
PMID: 24281134 [PubMed - as supplied by publisher]