Canadian researchers have developed and validated a decision tool to help physicians better discern which patients who visit an emergency department (ED) for heart failure (HF) should be hospitalized and which should be discharged. The evidence-based risk score was found to predict with high accuracy patients who were at risk of death within seven days of presentation.
The study was published June 5 in the Annals of Internal Medicine .
Douglas S. Lee, MD, PhD, of the Institute for Clinical Evaluative Sciences in Toronto, and colleagues designed the study to support the clinical decision making for patients who present with acute HF at EDs. Relying solely on clinical judgment may lead to some low-risk patients being hospitalized and some high-risk patients being discharged. Both scenarios contribute to higher healthcare costs, they wrote, and in the case of high-risk patients, possible death.
“An evidence-based clinical risk score that incorporates facets of the initial assessment rubric and facilitates prognostication would benefit clinicians and patients,” they wrote.
Previous research has focused on this problem, but excluded patients who were discharged. That prompted Lee and colleagues to try to develop a model that bridged both hospitalized and discharged patients.
Their study had two stages: developing a model for predicting acute HF mortality in an ED setting and then validating the model. To do so, they started with data on 12,591 patients who visited an ED in Ontario between April 2004 and March 2007 and who fulfilled Framingham criteria for HF. From that group, they randomly selected patients at randomly selected hospitals who were either hospitalized or discharged and abstracted data on them.
Those patients were further split between a derivation group (5,254 hospitalized and 2,179 discharged patients) and a validation group (3,560 hospitalized and 1,598 discharged patients). The primary outcome was seven-day mortality after initial presentation.
Data extracted included patient characteristics, clinical data, laboratory results, comorbid conditions, medication use and a Canadian Triage and Acuity Scale score assigned when the patients sought care in the ED.
The baseline characteristics between the two groups were found to be similar, but the validation group had a slightly higher creatinine concentration, had more patients who were transported via emergency medical services (EMS), a higher proportion of patients using diuretics and more patients with elevated troponin levels.
Based on their analyses, the researchers identified several predictors of seven-day mortality, including:
- Lower initial systolic blood pressure, oxygen saturation and hemoglobin concentrations;
- Higher leukocyte count and potassium, creatinine and nonnormal troponin levels;
- Being transported by EMS; and
- Being prescribed furosemide or metolazone before arrival.
They developed a weighted scoring system, the Emergency Heart Failure Mortality Risk Guide (EHMRG), based on a number of variables. When applied to both the derivation and validation groups, they found each 20-point increase in the score increased the odds of death within seven days, at 41 percent for the derivation group and 39 percent for the validation group. For each standard deviation, the odds of death increased 2.9-fold in both groups.
Results for mortality rates according to risk score quintile were similar in both groups. Those discharged had a 0.2 percent risk of death in the two lowest quintiles and a 21-fold risk in the highest quintile, while those hospitalized had a 0.4 percent risk and a 23-fold risk, respectively.
“Regardless of whether the HF subcohort was discharged from the ED or hospitalized, the EHMRG score was similarly effective in stratifying mortality risk,” the authors wrote.
Compared with a previously published hospital-based risk algorithm, their model showed superior discrimination overall, the authors wrote. They attributed that to the inclusion of a broad sample of patients.
“A major strength of EHMRG is that it encompasses all patients presenting to the ED, regardless of whether they were hospitalized or discharged,” they wrote. “If a model is intended to guide hospitalization-versus-discharge decisions based on acute prognosis, it is important to examine a patient sample whose inception is presentation to the ED and not only those who were hospitalized.”
The researchers emphasized that a decision tool does not replace