Risk Prediction in Patients With Heart Failure

Abstract

There are several clinically useful and well-validated death prediction models in patients with heart failure. Although the studies differed in many respects, the models largely included a few common markers of risk.

Regional Validation and Recalibration of Clinical Predictive Models for Patients With Acute Heart Failure

“…For this analysis, “compatible CPMs” were defined by the following characteristics: (1) the index condition in the derivation cohort was similar to the index condition in the validation cohort (here AHF), (2) CPM predicts an outcome captured in the validation cohort (here mortality), (3) all variables in the CPM were captured in the validation data sets and can be assigned a value, and (4) CPMs were derived in patient samples from a single world region (here, North America). We identified compatible models by reviewing a recently published systematic review of CPMs for HF 6. For this analysis, we present a sample of the compatible CPMs developed in North America that predict mortality at 3 different time points (in‐hospital, 60 day, and 1 year) following hospitalization for HF.…”

“…In the case of heart failure, CPMs have been proposed to inform decisions for advanced therapies and palliative care4 and also the common and costly admission decision for patients with acute heart failure (AHF) in the emergency department 5. While many different CPMs exist for predicting mortality for HF,6 CPM performance is often significantly better for the population on which the model was derived compared with similar yet distinct “validation” populations 7…”

Prognostic factors in heart failure — are they all equally important?

“…Furthermore, although their performance seems acceptable at the population level, they do not reliably predict one-year outcome of an individual patient [26]. The discriminatory ability of the models for prediction of HF hospitalisation appears to be even lower than that for estimation of the risk of death [20,21]. Finally, no evidence from randomised clinical trials exists to support the superiority of any of these models in qualification for HTx or any other specific therapy.…”

“…Understandably, the abundance of those variables: 1) derives from complex pathophysiological pathways of HF development as well as from the fact that advanced HF affects function of other critical organs, and 2) denotes the need for more comprehensive means of assessment of prognosis in these patients because no single parameter is sufficient on its own. Thus, different risk scores, encompassing various numbers of predictive variables, have been proposed for risk stratification in HF [4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22]. A recent meta-analysis reported as many as 117 models, using 249 different variables [20].…”

“…Other strong predictors of death in HF, consistently reported across different models, include diabetes, BNP/NT-proBNP concentration, weight or body mass index, and exercise capacity [4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22]. Of all prognostic variables, LVEF and NYHA class belong to the most useful and powerful predictors of long-term HF outcomes.…”