In this study, we highlight the importance of adding POCUS evaluation to patients diagnosed with AHF in the ED. The proportion of misclassified patients is consistent with available literature [9] and presents a particularly concerning number, mainly because initiating specific AHF treatment can be detrimental for conditions detected as alternatives. Notably, the high number of significant pericardial effusions detected draws attention. This phenomenon was previously described by Blaivas [10], where misclassification as “classic AHF” and initiation of intensive diuretic treatment could potentially lead to hemodynamic instability.

This study also introduces two concepts not previously described in the literature: the asymmetry index and the temporal progression of B-lines adjusted for evolution time. The asymmetry index quantifies differences between interstitial pneumonia caused by SARS-CoV-2 and AHF. Since the distinguishing factor between them is the presence of a symmetric B-pattern [11], we sought to quantify symmetry, leading to the asymmetry index which, with significant implications, opens the potential for application in other patient groups. The index could be valuable in the differential diagnosis between asymmetric interstitial pathologies (interstitial pneumonia, ARDS) and symmetric interstitial or alveolar pathologies (pulmonary fibrosis or AHF), given that both circumstances exhibit a bilateral B-pattern, making B-lines indistinguishable. The second concept, the density of B-lines, offers relevant information, particularly when considering evolution time and the probable location of B-lines. Additionally, it indicates lung reaeration patterns with a centrifugal tendency, as observed in the graphs.

Regarding the evaluation of the inferior vena cava, the standalone precision values are consistent with other studies in the available literature, but we know that the precision is reinforced with multi-window evaluation, which forms the basis of the decision tree (Fig. 4).

Final decision tree for alternative diagnosis (calculated with a random 80% of sample and tested on remaining 20%)

With all these aspects considered, the decision tree suggested by our statistical model exhibits excellent classification ability. It’s worth noting that this model does not consider the presence of significant pericardial effusion initially. Clinically, the initial cardiac assessment appears necessary to exclude both pericardial effusion and acute cor pulmonale patterns, which if absent, would lead to subsequent pulmonary and inferior vena cava evaluations, as suggested. However, we are currently developing a “random forest” model that includes all variables, where both the asymmetry index and the presence of pericardial effusion are incorporated. Initial testing indicates perfect classification ability (100% sensitivity and specificity), though further studies are needed to confirm this.

Conceptually, the study highlights that the competencies required to find an alternative diagnosis don’t demand high expertise and should be part of basic ultrasound training for Emergency Medicine specialists [12]. This is where the study has significance in the day-to-day operations of an Emergency Department. On the other hand, in cases where an alternative diagnosis isn’t found, detecting relevant information for AHF management, such as valve pathology or non-preserved LVEF, requires quantitative measurements (Doppler quantification, biplane Simpson method, etc.) and, thus, more extensive training [13].

Limitations of this study include its single-center nature, conducted in a Level 3 hospital with a high-demand Emergency Department, potentially affecting external validity when applied to centers of different complexity or lower demand. Additionally, patients were already selected since they had been diagnosed with AHF by at least one treating team, rendering the protocol’s application invalid for patients with isolated clinical suspicion of AHF upon ED arrival. Nonetheless, the information derived from the early-stage B-line density map along with the asymmetry index could be applicable in these cases. Moreover, all diagnostic precision values need contextualization due to the highly selected patient population and the prevalence environment in which they are found. While the initial sample size calculation suggested around 87 patients for the main variable (detection of an alternative diagnosis), a larger sample size might be advisable to strengthen the description of the other variables or to include a larger cohort to test the protocol’s application. Finally, given the baseline characteristics of the patients included in this study always considering the main diagnosis, the possibility of the temporal coincidence of a respiratory infection and heart failure is not reflected, something that, according to some studies, is relatively common [14] an can be consider as a main limitation.