This was a prospective observational study reported according to the STARD guidelines and registered in Clinical Trials prior to conduction (identifier: NCT03905460). According to the Danish Act on Research Ethics Review of Health Research Projects, this quality improvement study did not require an approval from the research ethics committee system (Inquiry No. 208 / 2018). It was carried out in the Central Denmark Region, which is one of five political regions in Denmark, with a population of 1.3 million equivalent to 23% of the Danish population [14]. Ambulance services in the region are dispatched in two tiers. The first tier consists of ambulances staffed with emergency medical technicians and/or paramedics. The second tier consists of prehospital critical care teams staffed by physicians (anesthesiologists) dispatched either in one of ten regional rapid response vehicles and/or one of four nationally operated helicopter emergency medical services. Ultrasound competences among this group of physicians are heterogeneous, but the majority have extensive ultrasound experience from various courses and from in-hospital clinical use. POCUS competences were systematically implemented in prehospital critical care teams in our region from 2012 to 2018 by a combination of e-learning and through three hands-on courses held during continuous training courses. The use of POCUS was at physician discretion and in total, 40 of 136 physicians operating on one or more of the ten regional rapid response vehicles agreed to include patients for this study. These physicians routinely use POCUS for diagnostics in patients with dyspnea in the prehospital setting and agreed to report their clinical findings and the results of their ultrasound examinations.

Patients attended by one of these 40 critical care team physicians in the Central Denmark Region between May 21, 2019 and December 31, 2021 were included. Inclusion criteria comprised age ≥ 18 years old and dyspnea as the primary complaint and respiratory rate > 25 min-1 and/or saturation < 95% and/or need for oxygen therapy based on a clinical judgement. Exclusion criteria were trauma prior to dyspnea and prior enrollment in the study. Drop-out criteria were POCUS not completed or patient not admitted to hospital following examination, as this precluded adjudication.

Patients were screened on a tablet on-scene and, following inclusion, the physician reported the suspected diagnosis according to Fig. 1 based only on patient history and physical examination. To include the possibility of high degree of diagnostic uncertainty, possible answers to each suspected diagnosis was Yes/No/Unknown. Subsequently, the physician performed a POCUS examination of the patient’s heart and lungs, with views and sequence of own preference, and registered predefined ultrasonographic findings as listed in Fig. 1. After completing the POCUS examination, the physician registered findings of the examination and a revised suspected diagnosis including supplementary POCUS findings (index test). Finally, the physician registered if findings of the POCUS examination changed prehospital treatment and/or triage. All POCUS examinations were performed with a SonoSite (Bothell, Washington, USA) iViz portable ultrasound scanner.

Study design—overview of inclusion, exclusion and dropout criteria in addition to POCUS findings reported and diagnoses reported and adjudicated

For patients in whom the pre- and/or post-POCUS diagnosis was registered as “unknown” by the physician, the patients were described, but excluded from diagnostic accuracy analyses.

An endpoint committee, blinded to the result of the POCUS examination, adjudicated the final diagnosis (reference standard) according to Fig. 1 based on electronic patient records (MidtEPJ, Systematics, Aarhus, Denmark) including results from laboratory analyses, and descriptions of x-ray, computed tomography scans and echocardiography, when performed. The committee consisted of three consultants in emergency medicine, cardiology and pulmonary medicine, respectively. Again, possible answers to the diagnosis were Yes/No/Unknown. The final diagnosis was determined by a 2:1 majority (Yes, No or Unknown) in the committee and this diagnosis was used to determine the accuracy of the prehospital pre- and post-POCUS diagnosis. An endpoint committee meeting was completed for consensus agreement on those patients where adjudicators answered three different things. Lack of consensus led to the answer "Unknown". Cases with "unknown" reference standard were described but excluded from diagnostic accuracy analyses.

All data including inclusion forms and patients journal material was registered in REDCap© electronic data capture tools hosted at Aarhus University [15, 16]. Baseline characteristics in terms of preexisting cardiopulmonary conditions, prescribed drugs and smoking habits at the time of admission, and results of examinations during admission were collected from electronic patient records.

The primary outcome was the change in sensitivity for diagnosing acute heart failure by adding POCUS of the hearts and lungs compared to clinical examination alone.

Secondary outcomes were (1) other diagnostic accuracy measures for acute heart failure with/without POCUS (receiver operating curve (ROC) area under the curve (AUC), specificity, positive predictive value (PPV), negative predictive value (NPV), positive and negative likelihood ratio (LR + , LR-)) (2) diagnostic accuracy measures of other etiologies of dyspnea (AE-COPD or AE-asthma, pneumonia, pulmonary embolism, pneumothorax) with/without POCUS, (3) diagnostic accuracy of specific POCUS findings (4) change of patient treatment due to POCUS findings and (5) change of triage due to POCUS findings.

Based on previous findings, we assumed a sensitivity of prehospital clinical examination for the diagnosis of acute heart failure of 59% and a prevalence of acute heart failure of 20% [17]). We assumed a possibility of increasing the sensitivity to a clinically significant level of 80% with POCUS. In a two-sided analysis with a proportion of discordant pairs of 30%, 255 ultrasound examinations were needed to achieve a strength of 80% and α 0.05. On September 24, 2020, we calculated the prevalence of patients with acute heart failure in our population to be 32% based on endpoint-committee adjudications on the first 100 patients and revised the sample-size calculation to 182 ultrasound examinations.

A total of 214 patients were included in this study, even though our revised sample-size calculation estimated a need of 182 POCUS examinations. The number included was higher than the sample size to ensure sufficient patient data with definite diagnostics at the end of inclusion. This was based on the observation of differences in the endpoint committee’s judgement of diagnoses early in the inclusion process.

Change in sensitivity and specificity for the diagnosis acute heart failure with supplementary POCUS was examined using McNemars exact test. Change in ROC AUC for the diagnosis acute heart failure was examined in paired analysis as proposed by Delong ER et al. [18]. No other comparative statistics were predefined, and the remaining data are presented descriptively. All statistical analyses were done in STATA MP version 17.0 (LCC StataCorp, Texas, USA). P-values < 0.05 were considered statistically significant. Data is presented as numbers and proportions with the corresponding 95% confidence interval (CI) where appropriate.