1. No significant differences in targeted interventions (medication prescription, implantable cardioverter-defibrillator, palliative care referral) occurred between patients with heart failure managed by clinicians with additional prognostic information compared to those without.
2. There was no significant difference in 30-day readmission rates and 1-year mortality between the intervention and control groups.
Level of Evidence Rating: 1 (Excellent)
Study Rundown: Heart failure is a complicated clinical diagnosis which may vary in presentation and course from patient to patient. Identification of high-risk patients and timely intervention is a critical skill in the management of heart failure. Recently developed risk prediction tools can help to simplify clinical decision-making by using patient characteristics to assess their risk of morbidity and mortality, thus advising clinicians on what therapy may be appropriate. Many of these tools have not yet been evaluated in a clinical context. The purpose of this study was to determine whether having information from a risk prediction tool available to clinicians influences their management of heart failure, and subsequent patient outcomes.
3124 patients were randomized in this trial: 50.9% were assigned to the control group and 49.1% to the alert group. The demographic makeup and baseline characteristics of the two groups were comparable at baseline. 48.7% of patients were estimated to have a low risk of 1-year mortality (5-15%), 20.0% were at medium risk (15-30% 1-year mortality) and 25.1% were in the very low risk group (<5% 1-year mortality). The median number of alerts viewed by each clinician was 5 over the course of the 1-year study period. The primary endpoint of 30-day readmission to hospital was 38.9% in the alert group and 39.3% in the control group. The 1-year mortality rates were also similar between the two groups: 27.1% in the alert group and 26.1% in the control group. There was no difference in the discharge management of patients in the alert versus control groups (i.e., medication prescriptions, palliative care referrals) which would suggest that the alerts influenced clinician decision-making.
This randomized controlled trial by Ahmad et al demonstrated that additional prognostication information presented to clinicians while managing patients hospitalized with heart failure did not influence management decisions and outcomes as predicted. This is an interesting finding, as developing risk prediction tools can be a resource and time intensive process but may not be clinically useful in a real-world setting. A strength of this study was the pragmatic design as well as the large size of the trial. However, a primary drawback was the oversimplification of clinical decision making in this study (i.e., based on interactions with an electronic medical record alert); to overcome this aspect, qualitative data about clinicians’ experience with the alert information system may be helpful.
In Depth [randomized controlled trial]: A pragmatic, multi-center, randomized controlled trial was performed. This trial is known as the Risk Evaluation and its Impact on Clinical Decision-Making and Outcomes in Heart Failure (REVEAL-HF) study. Patients aged 18 or older who had an elevated N-terminal pro-brain natriuretic peptide (NT-pro BNP) greater than 500pg/mL and received intravenous loop diuretics within 24 hours of admission were eligible. These patients were automatically identified through the electronic medical record (EMR) at participating hospitals. Patients were randomized to the alert group (treating clinicians would receive an “alert” on the patient’s EMR displaying the predicted 1-year mortality rate and other prognosticators, the risk score being previously developed in this population) or the control group (care as usual, no additional information provided). Physicians treating the alert group patients were able to interact with the risk estimate to determine whether they thought it was accurate or not.
Baseline characteristics were well-balanced between the two groups. The most common comorbid conditions were atrial fibrillation, diabetes, chronic kidney disease, chronic obstructive pulmonary disease and depression. The risk prediction model achieved a mean area under the curve of 0.74 (standard deviation 0.02) which was similar to the cutoff point described in the original validation study. There were no significant differences in outcomes between the intervention and control groups: the p-value for 1-year mortality was 0.89 and for 30-day readmission was 0.39. The median length of hospital stay (interquartile range) was 4.4 days (2.2-9.4) in the alert group and 4.3 (2.1-9.0) in the control group, p-value 0.28.
In the analysis of clinician interaction with the alert tool, 63.6% of clinicians were “not sure” about their agreement with the alert’s risk assessment, 29.3% felt that it “seemed appropriate”, 5.3% thought it was “too low” and 1.7% felt that it was “too high”. Clinician estimates of risk correlated well with the actual mortality of patients, particularly those in higher risk groups (p=0.05).
©2022 2 Minute Medicine, Inc. All rights reserved. No works may be reproduced without expressed written consent from 2 Minute Medicine, Inc. Inquire about licensing here. No article should be construed as medical advice and is not intended as such by the authors or by 2 Minute Medicine, Inc.