EHR Alerts Have No Impact on Inpatient HF Treatment or Outcomes

Ted Bosworth

November 15, 2021

When provided at the bedside of patients hospitalized for heart failure, an electronic health record (EHR) alert with prognostic information did not improve outcomes or appear to have any impact on what treatments were offered.

There was no signal that the EHR prognostic alerts, which identified the 12-month risk of mortality, had any impact on any of a variety of clinical-decision-making metrics or on any of the primary or secondary outcomes, according to Tariq Ahmad, MD, who reported results of the randomized REVEAL-HF trial, presented at the American Heart Association scientific sessions.

"These results call into question the hypothesis that accurate prognostic information alone will lead to better clinical decision-making," said Ahmad, medical director of the Heart Transplant and Mechanical Circulatory Support Program at Yale University, New Haven, Conn., and principal investigator of REVEAL-HF.

He acknowledged that the possibility that many clinicians pay little or no attention to EHR alert might have played a role in the negative results.

At four participating Yale-affiliated clinical centers, all patients hospitalized for acute heart failure were randomized as long as they were over the age of 18, had an N-terminal pro-brain natriuretic peptide (NT-proBNP) level above 500 pg/mL, and had been placed on IV diuretics within 24 hours of admission. In the experimental arm, the provider at the time of entering orders received an EHR alert with an estimate of the risk of all-cause mortality at 12 months. There was no such alert for patients managed in the control arm.

Twelve-Month Mortality Estimates Calculated

The all-cause mortality risk was calculated on a sizeable list of variables that included laboratory results, such as cell counts, and patient characteristics, such as weight and age. The risk estimate was displayed along with a five-category color-coded bar to provide context for the risk in the spectrum of very low, low, medium, high, and very high likelihood of death within 12 months.

The 1,590 patients randomized to the experimental arm and the 1,534 patients randomized to the usual care arm did not differ significantly in any baseline characteristics. The median age was about 77 years, the mean left ventricular ejection fraction (LVEF) was 55%. About 29% had an LVEF below 40%, about 40% had chronic kidney disease, and about 30% had chronic obstructive pulmonary disease (COPD).

The composite primary outcome of all-cause mortality or rehospitalization within 12 months was reached by 38.9% and 39.3% (P = 0.82) of the intervention and control arms, respectively. The components of the primary outcome were also nearly identical, as was inpatient mortality (8.4% vs. 8.8%; P = 0.72).

There were no significant differences in any of the secondary outcomes, which included rates of 30-day rehospitalizations, discharge on guideline-recommended heart failure therapies, implantation of a cardioverter defibrillator, use of a left ventricular assist device, or heart transplant.

The proportion of patients referred for palliative care was almost identical in the very low, low, and medium risk groups. In the high (23.4 vs. 15.6; P = 0.19) and very high (50% vs. 40%; P = 0.92) groups, there were numerically more referrals in the group randomized to usual care, but these rates did not reach significance.

No Differences Seen in Discharge Meds

There was essentially no difference between groups in the rates at which patients were discharged on beta-blockers, renin-angiotensin system inhibitors, sodium-glucose co-transporter type 2 (SGLT2) inhibitors, or mineralocorticoid antagonists.

When prespecified subgroups, such as those older than age 75 years relative to those younger, males relative to females, Black versus White participants, patients with reduced ejection fraction (HFrEF) relative to preserved ejection fraction (HFpEF), and intensive care unit versus non-ICU patients, were compared, there were no indications that the EHR alert improved outcomes.

Invited discussant Harriette G. C. Van Spall, MD, director of digital health and virtual care and associate professor of cardiology at McMaster University, Hamilton, Ont., did not dispute the conclusions, but she pointed out several potential explanations for the neutral result.

Not least, nearly 75% of those enrolled had low risk or very low risk for adverse outcomes within 1 year, so the opportunity to show a reduction in events, including all-cause mortality, was limited.

"This was largely a HFpEF population, for which there are no treatments for which a risk score would change therapy," she said.

EHR Alert Efficacy Questioned

There is considerable evidence that risk prediction tools "are common but underutilized in HF," Van Spall added. She noted that many clinicians find alerts in the EHR more annoying than informative, and it remains unknown what proportion of clinicians pay attention to them, particularly in the absence of evidence that they lead to meaningful improvements in care over their own clinical judgment.

Ahmad agreed.

"I think that we need to study these alerts in a clinical trial format," he said. Acknowledging that alerts have been poorly received by many clinicians, Ahmad said that trials to validate the impact of any specific alert are needed to improve their credibility. If a positive impact cannot be shown, he said the alert should be eliminated, leaving only the alerts with proven clinical value.

Ahmad reported financial relationships with Amgen, AstraZeneca, Boehringer Ingelheim, Cytokinetics, Novartis, and Relypsa. Van Spall reports no potential conflicts of interest.

This article originally appeared on, part of the Medscape Professional Network.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.