A Prehospital Decision Support Tool for the Diagnosis of Acute Heart Failure: The RAPID-CHF Score

Diagnostic uncertainty is a major barrier to the timely treatment of heart failure (HF) in the prehospital setting. We aimed to develop and validate a decision support tool using readily available clinical variables to predict the probability of HF among dyspnoeic patients transported by emergency medical services (EMS).

A population-based cohort of all adults transported by EMS for dyspnoea in Victoria, Australia was chronologically split into derivation (2015-2017) and temporal validation (2018-2019) cohorts. Two models were developed: (1) a full multivariable logistic regression model using adaptive least absolute shrinkage and selection operator regression, and (2) a simplified points-based RAPID-CHF score derived from the nine most predictive variables. Among 271,204 patients with dyspnoea (176,269 derivation; 94,935 validation), 9.4% and 9.0% had HF, respectively. The full model included 19 variables and demonstrated excellent discrimination (AUC 0.861 derivation; 0.862 validation) and calibration. The RAPID-CHF score (range 0-13; comprising age, ECG rhythm, prior HF, conscious state, oxygen saturation, blood pressure, temperature, peripheral oedema, and crackles) retained strong performance (AUC 0.835 derivation; 0.836 validation) and calibration. HF prevalence increased across predefined risk categories: low (score 0-5; HF prevalence 1.7%), moderate (6-9; 13.6%) and high (10-13; 46.4%). Decision curve analysis demonstrated greater net benefit across clinically relevant thresholds than current EMS diagnosis or "treat all"/"treat none" strategies.

A risk score derived from routinely collected prehospital variables accurately estimates HF probability among EMS-transported patients with dyspnoea. The RAPID-CHF score may facilitate earlier diagnosis and timely initiation of HF therapy in EMS workflows.