Risk Scores May Not Reliably Predict Mortality Risk in ADHF

heart failure
heart failure, ecg. physician monitoring heart
Risk assessment tools may not accurately predict risk for mortality in patients with acute decompensated chronic heart failure.

Risk assessment tools may not accurately predict risk for mortality in patients with acute decompensated chronic heart failure (ADHF), according to study resulted published in Journal of Clinical Epidemiology.

Risk stratification is important for triaging of patients with ADHF, but the accuracy of risk assessment tools remains unclear. Investigators sought to evaluate the performance of 4 prognostic tools in predicting 180-day mortality after admission for ADHF.

Risk assessment for 1458 patients (27.8% women) were performed using the Acute Decompensated Heart Failure National Registry (ADHERE) predictive model, the Get With The Guidelines (GWTG) risk score, the updated ADHF/NT-proB-type natriuretic peptide risk score, and the Acute Study of Clinical Effectiveness of Nesirtride in Decompensated Heart Failure (ASCEND) risk score.

The study’s primary outcome of interest was mortality within 180 days of admission. Patients were followed for 180 days or until death, health transplantation, or implantation of a ventricular assist device.

Over the course of follow-up, 309 patients (21.2%) died, 66 (4.5%) underwent urgent heart transplantation or ventricular assist device implantation, and 30 (2%) had elective heart transplantation.

The ADHF/NT-proBNP risk score, the ADHERE model, and the ASCEND risk score were all well calibrated. For these models, actual mortality closely matched the average predicted risk.

A >30% probability of death was predicted for 25.0% of patients using the ADHF/NT-proBNP score, 19.3% using the ADHERE model, 20.6% using the GTWG risk score, and 25.6% using the ASCEND risk score. The ADHF/NT-proBNP risk score predicted a >30% probability of death in 55% of the patients who died, compared with 38.8%, 41.1%, and 50.8% for the ADHERE model, GWTG risk score, and ASCEND risk score, respectively.

At the >30% risk threshold, the positive predictive values ranged from 42.1% for the ASCEND risk score to 46.7% for the ADHF/NT-proBNP risk score. The negative predictive values at the same threshold ranged from 83.9% to 87.3%. The maximum sensitivity was 55.0% and specificity ranged from 42.1% to 46.7%.

The investigators noted that cardiovascular deaths were not distinguished from noncardiovascular deaths when calculating mortality rate.

“[A]lthough risk assessment tools work well for stratifying patients into risk groups, their use to estimate the risk [for] mortality for individuals is of limited clinical utility,” the study authors concluded. “Our data also suggest that the strategy of targeting high-risk groups for interventions using risk scores may leave a sizeable proportion of patients who will develop the event untreated, or could expose some patients to unnecessary overtreatment.”


Scrutinio D, Guida P, Ammirati E, Oliva F, Passantino A. Risk scores did not reliably predict individual risk of mortality for patients with decompensated heart failure [published online May 25, 2020]. J Clin Epidemiol. doi:10.1016/j.jclinepi.2020.05.020