Elsevier

Annals of Emergency Medicine

Volume 50, Issue 2, August 2007, Pages 127-135.e2
Annals of Emergency Medicine

Cardiology/original research
Comparison of Four Clinical Prediction Rules for Estimating Risk in Heart Failure

https://doi.org/10.1016/j.annemergmed.2007.02.017Get rights and content

Study objective

We examine the performance of 4 clinical prediction rules prognostic of short-term fatal and hospital-based nonfatal outcomes in heart failure patients.

Methods

We used a retrospective cohort of 33,533 adult patients admitted to Pennsylvania hospitals in 1999 with a diagnosis of heart failure. We stratified patients into risk categories defined by each clinical prediction rule. We assessed prognostic accuracy according to sensitivity and specificity and compared discriminatory power according to area under the receiver operating characteristic (ROC) curves. The outcomes were inpatient death, 30-day mortality, and death or serious medical complications before hospital discharge.

Results

The 4 rules each created risk groups of various proportions and frequencies of outcomes. The proportion of patients assigned to the lowest risk group ranged from 13.3% to 73.0%. The rates of inpatient death or complications in the lowest risk group ranged from 6.7% to 9.2%, and 30-day death rates varied from 1.7% to 6.0%. Patients categorized at the highest risk of death or complication demonstrated similar variability. The area under the ROC curve for inpatient death and complications differed only slightly among rules (0.58 to 0.62). The area under the ROC curve for fatal outcomes tended to be higher and differed among rules (0.59 to 0.74)

Conclusion

Current acute heart failure prediction rules offer varying ability to predict short-term death or serious outcomes. Although each creates a risk gradient, differences in risk-group proportions and outcome frequencies should drive rule selection or use in clinical practice.

Introduction

Heart failure affects nearly 5 million Americans, with more than 550,000 new cases diagnosed in the United States each year1 and over 1 million admissions in 1999.2 The National Institute of Health estimates the health care cost of heart failure to be $33.2 billion a year in the United States, with hospitalization care ($17.8 billion) accounting for the greater part of the cost.1 The increasing burden is evidenced by the 19% increase in emergency department (ED) visits for the treatment of heart failure from 1992 to 2001.3 Recently, heart-failure-specific clinical prediction rules have been developed to help physicians estimate patient prognosis and discriminate between low- and higher-risk patients at hospital admission decisionmaking.4, 5, 6, 7

Use of these rules in the ED could safely reduce health care costs for the treatment of acute decompensated heart failure by aiding physicians in the identification of low-risk patients for whom less intensive medical management therapies might be appropriate. For example, patients identified by a clinical prediction rule as very low risk could be discharged for outpatient therapy or admitted to an observation unit. Conversely, those at the highest risk could be admitted for treatment in more intensive care settings. Additionally, these rules can help investigators to assemble patient cohorts or aid data analysis by risk adjustment.8, 9 Unfortunately, it is difficult to decide which prognostic tool should be used because direct comparisons of performance and use are lacking.

Our objective was to compare the prognostic accuracy and discriminatory power of several heart-failure-specific clinical prediction rules for estimating inpatient death, 30-day death, and death or serious medical complications before hospital discharge.

Section snippets

Study Design

We used a retrospective cohort study design of existing databases. The University of Pittsburgh institutional review board approved our study.

Setting

Our cohort consisted of patients discharged from all Pennsylvania general acute care hospitals in calendar year 1999.

Selection of Participants

We reviewed all patients with a diagnosis of heart failure, defined as an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)10 hospital primary discharge diagnosis code consistent with heart failure

Results

There were 36,370 patients who had an index hospitalization from the ED in 1999 and a discharge diagnosis of heart failure and were linked between the Pennsylvania Health Care Cost Containment Council and MediQual databases. Of these, a documented ED pulse, systolic blood pressure, and respiratory rate were available for 33,533 (92.2%) patients. We display demographic and clinical characteristics of this cohort in Table 1. The most common clinical characteristics of these patients were a

Limitations

Our study has several limitations. The MediQual system of truncating vital sign data may have reduced the accuracy of our findings; however, sensitivity analysis for the ADHERE Tree suggests that the impact of these truncated data was negligible. We also used surrogates for some variables; although this could influence the assessment, we believe our substitutions were clinically relevant. Additionally, the similarity between our findings and those reported previously for each rule suggests

Discussion

Clinical prediction rules are intended to assist clinicians in medical decisionmaking. Often, these rules try to assess the likelihood of a “bad outcome”—death, serious complications, prolonged care, or recovery or functional status—at a set interval. Our study focused on short-term outcomes because they can help guide the choice of initial site or intensity-of-care decisions.

Although no singular outcome is ideal, most prediction rules seek to identify risk of death. The 4 rules studied here

References (19)

There are more references available in the full text version of this article.

Cited by (47)

  • Current Emergency Department Disposition of Patients With Acute Heart Failure: An Opportunity for Improvement

    2022, Journal of Cardiac Failure
    Citation Excerpt :

    Consensus guidelines recommend risk stratification but provide few objective measures to assist with decision making about disposition of patients after the ED.29 Earlier studies focused on identifying high-risk markers that are associated with increased morbidity and likelihood of mortality.30 However, the challenge is that as much as 50% of patients in EDs lack these features,31 and the absence of high-risk features does not equate to lower risk.

  • Disposition of acute decompensated heart failure from the emergency department: An evidence-based review

    2021, American Journal of Emergency Medicine
    Citation Excerpt :

    Table 5 compares the currently ED validated AHF risk scores. Several other scores have sought to identify low-risk patients, including the acute decompensated HF national registry risk tree from the ADHERE Registry, the external testing of the enhanced feedback for effective cardiac treatment (EFFECT) score, and the Brigham and Women's Hospital (BWH) rule, but these were derived in hospitalized patients for AHF, with 16.3%–36.2% of these patients directly discharged from the ED without hospitalization (35,71-74). These scores also miss a major proportion of patients who may be low-risk and appropriate for discharge.

  • Moving toward comprehensive acute heart failure risk assessment in the emergency department: The importance of self-care and shared decision making

    2013, JACC: Heart Failure
    Citation Excerpt :

    Furthermore, as most ED patients with AHF are already admitted to the hospital, characterizing AHF clinical profiles that are safe for either immediate ED discharge, or after a brief period of treatment and observation, would be of much greater value. Prospective testing of 4 AHF prediction rules suggests they would not be useful in the ED (24). Previous studies have defined low-risk ED patients as having 30-day readmission rates of 15% to 20% and mortality rates of less than 1% (3,11,25–27).

View all citing articles on Scopus

Funding and support: By Annals policy, all authors are required to disclose any and all commercial, financial, and other relationships in any way related to the subject of this article, that might create any potential conflict of interest. See the Manuscript Submission Agreement in this issue for examples of specific conflicts covered by this statement. The authors received financial support from AHRQ R01 HS10888-01.

Reprints not available from the authors

Supervising editors: W. Brian Gibler, MD; Steven M. Green, MD

Author contributions: TEA, MH, JBM, and DMY conceived the study, designed the trial, and obtained research funding. TEA and MH organized the datasets; all authors participated in data analysis; TEA and MH performed all statistical testing. TEA and DMY drafted the article, and MH and JBM offered contributions to the article. DMY takes responsibility for the paper as a whole.

Available online April 20, 2007.

View full text