Journal Club July 2017 - Risk Stratification of Acute Heart Failure

With nearly 1 million US emergency department (ED) visits attributed to acute heart failure (AHF) annually, heart failure is a serious illness frequently managed by emergency physicians. More than 80% of visits result in hospital admission, and readmission rates may range from 30-60% within 3 to 6 months of initial discharge. ED visits related to AHF are expected to continue to rise with the aging population and improved survival rates in patients with chronic heart failure and acute coronary syndromes. One critical issue facing emergency physicians caring for AHF patients is deciding upon disposition: admission (with or without monitoring) or discharge (with or without early follow-up). Consensus guidelines are available to support decision-making around diagnosis and treatment; however, no current guidelines provide an evidence-based approach to disposition. For this month’s journal club, we reviewed three papers examining decision aids developed to assist emergency physicians in determining risk in AHF patients.



1.      Stiell; et al. A risk scoring system to identify emergency department patients with heart failure at high risk for serious adverse events. Academic Emergency Medicine. 2013; Vol. 20, No. 1, 17-2.6

2.      Collins; et al. Identification of emergency department patients with acute heart failure at low risk for 30-day adverse events: The STRATIFY decision tool. JACC Heart Failure. 2015; Vol 3, No. 10, 737-747.

3.      Stiell et al. Prospective and explicit clinical validation of the Ottawa Heart Failure Risk Scale, with and without use of quantitative NT-proBNP. Academic Emergency Medicine. 2017; Vol 24, No. 3, 316-327.


1.  Collins & Storrow. Moving toward comprehensive acute heart failure risk assessment in the emergency department. JACC Heart Failure. 2013; Vol. 1, No. 4, 273-280.

2.  Heitz, et al. Hot off the press: Prospective and explicit clinical validation of the Ottawa Heart Failure Risk Scale, with and without use of quantitative NT-proBNP. Academic Emergency Medicine. 2017. Doi: 10.1111/acem.13192-17-215.





The goal of Steill and colleagues’ 2013 study was to develop a risk scoring system to guide the disposition decisions of physicians for emergency department patients with acute dyspnea secondary to heart failure. Their objective was to develop a highly sensitive scale for predicting the potential for development of significant adverse events in this patient population. Their hope was that the scale would improve and standardize disposition practices for patients with AHF while decreasing unnecessary admissions as well as unsafe discharges.

The study was conducted in six Canadian teaching hospital emergency departments with a combined volume of about 350 thousand visits annually. A convenience sample of adult ED patients ≥ 50 years old presenting for acute shortness of breath due to exacerbations of chronic heart failure or new-onset heart failure. Included patients had symptoms (dyspnea or fatigue), clinical signs of fluid retention (peripheral or pulmonary) in the presence of an underlying abnormality in cardiac function or structure. In unclear cases, a beneficial response to treatment (e.g. brisk diuresis and diminished shortness of breath) was also considered. In the EDs where it was available, B-type natriuretic peptide (BNP) values were reviewed.

Registered respiratory therapists or nurses conducted and recorded patient assessments with the target assessment period being 4 to 8 hours after initial ED treatment. Assessments were collected from history, general examination, laboratory tests, and the 3-minute walk test. The primary study outcome was occurrence of a significant adverse event including: death from any cause within 30 days of the index ED visit; admission to critical care or acute monitoring with 14 days; need for endotracheal intubation/non-invasive ventilation after hospital admission; MI; receipt of a major procedure including CAGB, PCI, new hemodialysis, or other cardiac surgery; relapse and additional hospital admission for those discharged.

Multivariate analyses were conducted using logistic regression for variables found to be significantly associated with serious adverse event on initial univariate analysis. These analyses facilitated development of the 10-item Ottawa Heart Failure Risk Scale (Figure 1). 

Steill and colleagues report that the risk of significant adverse events ranged from 2.8% to 89.0%, with good calibration between observed and expected probabilities. In their internal validation, the risk scores were very accurate across 1,000 bootstrapped replications. They determined that a threshold of 1, 2, or 3 total scores for admission would be associated with sensitivities of 95.2, 80.6, or 64.5%, respectively.

In sum, the authors felt that the use of the Ottawa Heart Failure Risk Scale could help emergency physicians to identify patients most at risk for adverse outcomes who are most in need of hospital admission or early follow-up. 





The aim of Collins and colleagues’ 2015 study was to develop an acute heart failure decision tool to identify ED patients at low risk of death or serious complications who could be considered for discharge from the ED and subsequent management as an outpatient. They conducted a prospective, observational cohort study (STRATIFY – Improving Heart Failure Risk Stratification in the ED) at 2 university-affiliated tertiary care EDs and 2 community EDs. ED patients diagnosed by the treating physician with acute heart failure (using modified Framingham criteria) were enrolled in the study. Blinded assessors abstracted data from medical records and research assistants collected data through direct questioning of participants and their treating physicians. The primary outcome was the most severe adverse event experienced within 30 days of the index ED visit (e.g. death, receipt of CPR, mechanical cardiac support, intubation/mechanical ventilation, emergent dialysis, PCI or CABG).

The STRATIFY decision tool was developed following assessment of a large number of candidate predictor variables. Potential predictors with a large degree of missingness or a lack of variation in response were removed. The tool was derived by fitting a proportional odds model to the 6-level ordinal outcome using 13 previously selected variables. The predictive accuracy of the resulting STRATIFY tool was assessed using the C (concordance) statistic. Calibration was calculated by constructing a smooth nonparametric calibration curve of predicted vs. observed outcome; the curve was optimized to minimize false negative results.

Ultimately, the C statistic was 0.68 (95% CI: 0.63-0.74). At risk thresholds of 1%, 3%, and 5%, this team observed 0%, 1.4%, and 13.0% patients were at low risk, with negative predictive values of 100%, 96%, and 93%, respectively. The STRATIFY nomogram is depicted in Figure 2.



In their recent study, Steill and colleagues sought to prospectively evaluate the accuracy, acceptability, and potential impact of the Ottawa Heart Failure Risk Scale (OHFRS) (Figure 3) that they reported on developing in the first article we reviewed. Their prospective, observational cohort study was again conducted in six Canadian emergency departments. Patients with acute heart failure were evaluated by their emergency physicians for the 10 Ottawa Heart Failure Risk Scale items and were then followed for 30 days. When feasible, quantitative NT-proBNP was evaluated and study analyses were completed with and without this laboratory data.

In this study, serious adverse events were defined as:

  • Death within 30 days
  • Admission to a monitored unit
  • Intubation
  • Noninvasive mechanical ventilation
  • Myocardial infarction
  • Heart failure relapse resulting in hospital admission within 14 days of the index ED visit.

1,100 adult (≥ 50 years) ED patients with acute heart failure were enrolled in the study. The primary outcome, a significant adverse event as defined above, occurred in approximately 16% of cases, 10% for discharged cases and 19% for admitted cases. As compared to actual clinical practice, using an OHFRS score for >1 as a threshold for admission would result in increased sensitivity (72% vs. 92%) but also in increased admissions (57% vs. 78%). For cases with NT-proBNP values available (n = 684), using an OHFRS score of >1 would have significantly increased sensitivity (70% vs. 96%) while also increasing admission rate (61% vs. 88%). Emergency physicians were generally comfortable using the OHFRS, with only 11% indicating discomfort with the tool. The authors concluded that the OHFRS is highly sensitivity for detecting significant adverse events in acute heart failure patients, with a resulting increase in admission rate. In addition, NT-proBNP values further improved observed sensitivity.

In sum, using the OHFRS may help to improve and standardize admission processes in acute heart failure, decreasing unsafe discharge dispositions for high-risk patients as well as decreasing unnecessary admissions for low-risk patients. 




Summary:  In sum, this literature provided an overview of the development and early evaluation of two heart-failure decision-making tools. In the case of the Ottawa Heart Failure Risk Scale (OHFRS), there is no data on the performance of the tool when applied in US ED's, which is a significant barrier to clinical use as US practice is very different than Canadian practice around heart failure (US admissions = 80% while Canadian admissions are 40-60%). At this time, the STRATIFY tool has not undergone external validation, which also limits clinical use. While there are promising features for both instruments, additional research is needed to further assess validity and determine how these might be used as a part of shared decision-making in acute heart failure.


download article summaries

STiell et al. 2013

STiell et al. 2013

collins et al.

collins et al.

Stiell et al 2017

Stiell et al 2017


Written by Tania D. Strout, PhD, RN, MS

Director of Research, Associate Professor of Emergency Medicine

Edited and Posted by Jeffrey A. Holmes, MD