Brian D Steinhart1, Phillip Levy2, Hilde Vandenberghe3, Gordon Moe4, Andrew T Yan4, Ashley Cohen5, Kevin E Thorpe6, Melissa McGowan7, C David Mazer8. 1. Department of Emergency Medicine, St Michael's Hospital, Toronto, Ontario, Canada; Division of Emergency Medicine, Department of Medicine, University of Toronto, Toronto, Ontario, Canada. Electronic address: steinhartb@smh.ca. 2. Department of Emergency Medicine and Cardiovascular Research Institute, Wayne State University, Detroit, Michigan. 3. Department of Laboratory Medicine, St Michael's Hospital, Toronto, Ontario, Canada; Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada. 4. Division of Cardiology, St Michael's Hospital, Division of Cardiology, Toronto, Ontario, Canada; Department of Medicine, University of Toronto, Toronto, Ontario, Canada. 5. Applied Health Research Centre, St Michael's Hospital, Toronto, Ontario, Canada. 6. Applied Health Research Centre, St Michael's Hospital, Toronto, Ontario, Canada; Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada. 7. Department of Emergency Medicine, St Michael's Hospital, Toronto, Ontario, Canada. 8. Departments of Anaesthesia and Critical Care, Keenan Research Centre for Biomedical Science and Li Ka Shing Knowledge Institute of St Michael's Hospital, Toronto, Ontario, Canada; Department of Anesthesia, University of Toronto, Toronto, Ontario, Canada.
Abstract
BACKGROUND:Diagnosing acute heart failure (AHF) in undifferentiated dyspneic emergency department (ED) patients can be challenging. We prospectively studied a validated diagnostic prediction model for AHF that uses patient age, clinician pretest probability for AHF, and N-terminal pro-B-type natriuretic peptide (NT-proBNP) as a continuous value to determine its utility and performance. METHODS AND RESULTS: This was a multicenter randomized controlled trial of undifferentiated dyspneic patients with an indeterminate pretest probability of AHF as assessed by the treating emergency physician (EP). After recording its components, the calculated model results with validated treatment threshold guidelines were provided to EPs for patients randomized to the intervention arm. Final diagnoses with the use of 60-day follow-up information were adjudicated by 2 independent cardiologists. The primary outcomes were accuracy of the model and of physician diagnosis comparing intervention and standard care arms. A total of 197 patients were randomized and had outcome data recorded; 41% were determined to have had heart failure. Final EP diagnostic accuracy was 76% (sensitivity 68.2%, specificity 83.9%) with no significant difference between exposed versus blinded arms (accuracy 77% vs 74%; P = .77). Area under the model receiver operating characteristic curve was 0.93. Using the model treatment thresholds would have redirected 48% of patients with 95% accuracy. CONCLUSIONS: This study prospectively validated the diagnostic accuracy of our AHF model in a significant proportion of indeterminate dyspneic ED patients, but provision of this information did not improveEP diagnostic accuracy. Future studies should determine how such a clinical prediction tool could be effectively integrated into routine practice and improve early management of suspected AHF patients in the ED.
RCT Entities:
BACKGROUND: Diagnosing acute heart failure (AHF) in undifferentiated dyspneic emergency department (ED) patients can be challenging. We prospectively studied a validated diagnostic prediction model for AHF that uses patient age, clinician pretest probability for AHF, and N-terminal pro-B-type natriuretic peptide (NT-proBNP) as a continuous value to determine its utility and performance. METHODS AND RESULTS: This was a multicenter randomized controlled trial of undifferentiated dyspneicpatients with an indeterminate pretest probability of AHF as assessed by the treating emergency physician (EP). After recording its components, the calculated model results with validated treatment threshold guidelines were provided to EPs for patients randomized to the intervention arm. Final diagnoses with the use of 60-day follow-up information were adjudicated by 2 independent cardiologists. The primary outcomes were accuracy of the model and of physician diagnosis comparing intervention and standard care arms. A total of 197 patients were randomized and had outcome data recorded; 41% were determined to have had heart failure. Final EP diagnostic accuracy was 76% (sensitivity 68.2%, specificity 83.9%) with no significant difference between exposed versus blinded arms (accuracy 77% vs 74%; P = .77). Area under the model receiver operating characteristic curve was 0.93. Using the model treatment thresholds would have redirected 48% of patients with 95% accuracy. CONCLUSIONS: This study prospectively validated the diagnostic accuracy of our AHF model in a significant proportion of indeterminate dyspneic ED patients, but provision of this information did not improveEP diagnostic accuracy. Future studies should determine how such a clinical prediction tool could be effectively integrated into routine practice and improve early management of suspected AHF patients in the ED.