This study proposes a simple clinical prediction model that may help clinicians to decide whether a CTPA should be performed in hospitalized COVID-19 patients with suspected PE estimating the pretest probability. Age, common comorbidities, and classic risk factors for venous thromboembolism including obesity, recent immobilization and oestrogen therapy were different between PE and non-PE groups; however, there was no independent association with PE diagnosis. In our cohort, the independent predictors for PE were a high heart rate, high respiratory rate, increased serum CRP and D-dimer levels, and the extent of lung parenchymal damage on chest radiography at the time of suspected PE. These variables were used to construct a simple-to-use score that showed an excellent diagnostic accuracy and negative predictive value.
Several systematic reviews and meta-analyses have addressed the incidence of PE in COVID-19 patients with variable results (range from 2−79%) [19–22].In our study, the real incidence of PE is unknown because CTPA was performed based on clinical suspicion rather than systematic screening; therefore, the incidence might be underestimated, especially in oligosymptomatic patients with segmental or subsegmental PE. Consequently, to determinate the best cut-off point to rule out PE in CHEDDAR score, we used the reported prevalence in a recent systematic review that involved the largest series of COVID-19 patients [2].
PE diagnosis guidelines recommend a standardized protocol using models to determine the pretest probability of PE [23]. The most used models are the Geneva and the Wells scores, the PERC rule, and the YEARS algorithm. The combination of low or intermediate probability score and normal D-dimer level yield a NPV of 99%, and no further testing is required in these patients [24]. This recommendation is judged particularly relevant in COVID-19 due to the risk of hospital SARS-CoV-2 infection to other patients and healthcare professionals during the transport of the patient to the room radiology.
The most challenging aspect of diagnosing PE in COVID-19 patients is to establish when to perform a CTPA. Common symptoms of PE (fatigue, breathlessness and chest pain) show a wide overlap with COVID-19 pneumonia, and D-dimer levels are often elevated in the absence of thrombosis [25,26]. Clinical and pathological observations has highlighted the role of endotheliitis and the hyperinflammatory state in the fisiopathology of thrombosis in SARS-CoV-2 infection [27]. Because the pathophysiological mechanism of PE in COVID-19 patients seems to be a local thrombo-inflammatory response rather than an embolization of a deep vein thrombus, the scores that help to stratify the pretest probability of PE in non-COVID-19 patients have low accuracy in this clinical setting [9‒12].
We found that vital signs were significantly associated with PE. Heart rate was a risk marker of PE, in line with previous reports [28,29]. In a study conducted by Gil et al, tachypnea >22 bpm was a predictive factor of PE in patients with COVID-19 [30]. We found an association between the extension of pneumonia and the confirmation of PE. Previous studies showed that patients with severe lung damage (>50%) evaluated on CTPA had a higher PE incidence rate [31–36]. D-dimer level was higher in patients with PE than in those without PE, and this has been confirmed in nearly every publication on this topic [37–41]. D-dimer levels in COVID-19 patients in the absence of thrombosis may be elevated. Therefore, lower specificity as a predictor of thrombotic events may be expected. Consequently, a higher cut-off for diagnosis of PE in COVID-19 patients is proposed. In our study, as published previously [42, 43] a cut-off of D-dimer levels>3.000 ng/mL was associated with PE diagnosis. In addition, elevated CRP level was predictor of PE occurrence in line with previous reports [28, 29]. All the previously described factors are closely related to local excessive inflammatory response in COVID-19 patients.
We found a previous study that developed a model to predict PE in COVID-19 patients, the CHOD score [28]. In this work, CRP, Heart rate, Oxygen saturation, and D-dimer levels were associated with higher rates of PE during hospitalization. Compared to our model, the best cut-off points for heart rate, CRP and D-dimer were lower (≥90 bpm, ≥50 mg/L and ≥956 ng/mL, respectively). This score showed a high diagnostic accuracy (AUC 0.86; 95% CI: 0.8–0.93). CHOD score stratifies patients into three risk groups: low (0–2 points), moderate (3–5 points), and high risk (>5 points), with a PE rate of 4.5%, 36.8%, and 100%, respectively. Unlike to CHEDDAR score, the predictor’s variables for PE were collected at admission, and the model is applicable to patients with elevated D-dimer levels, because patients with normal D-dimer were excluded. This study was done during the first COVID 19 surge in Spain (from March 2020 to April 2020), whereas our study has been performed throughout different COVID-19 waves including different SARS-CoV-2 variants, including vaccinated population.
A novel finding of our study is the construction of a score to predict PE using five simple-to-measure variables (heart rate, respiratory rate, RALE score, CRP, and D-dimer) at the time of suspected PE.CHEDDAR score was constructed choosing the cut-off points with the best prognostic value for each variable. The diagnostic accuracy of the CHEDDAR score was remarkable (AUC: 0.87 [95% CI: 0.83−0.92]). A CHEDDAR score below 182 points showed excellent ability to rule out PE with a NPV of 92%.
The main limitation of our study is its retrospective nature. The RALE score is subject to interobserver variability, although the degree of agreement between RALE score points and lung involvement on CTPA is high [44].Finally, the model was constructed with COVID-19 patients and suspected PE who underwent a CTPA to obtain a certain diagnosis. Therefore, the model is applicable in this clinical scenario.