key: cord-0959071-ckauc92d
authors: Rindi, Lorenzo Vittorio; Moghazi, Samir Al; Donno, Davide Roberto; Cataldo, Maria Adriana; Petrosillo, Nicola
title: Predictive scores for the diagnosis of Pulmonary Embolism in COVID-19: A systematic review
date: 2021-11-27
journal: Int J Infect Dis
DOI: 10.1016/j.ijid.2021.11.038
sha: 3b994ca7e5f226df2f846b653c7b33768175a4f6
doc_id: 959071
cord_uid: ckauc92d

Objectives During the COVID-19 pandemic, several studies described an increased chance of developing pulmonary embolism (PE). Several scores have been used to predict the occurrence of PE. In this systematic review, we aim at summarizing the literature on predicting rules for PE in hospitalized COVID-19 patients (HCPs). Methods PUBMED and EMBASE databases were searched to identify articles (1 January 2020-28 April 2021) presenting data pertaining the use of a prediction rule to assess the risk for PE in adult HCPs. The investigated outcome was the diagnosis of PE. Studies presenting data on the use of a single laboratory assay for PE prediction were excluded. Included studies were appraised for methodological quality using the Newcastle - Ottawa Quality Assessment Scale for Cohort Studies (NOS). Results We obtained a refined pool of 12 studies, for a total of 5 scoring systems (Wells score, Geneva score, CHADS2/CHA2DS2VASc/M-CHA2DS2VASc, CHOD score, Padua Prediction Score), and 4526 patients. Only one score was specifically designed for HCPs. Three and nine included studies were prospective and retrospective cohort studies, respectively. Among the examined scores, CHOD score seems promising in terms of predictive ability. Conclusion New prediction rules, specifically developed and validated for estimating the risk of PE in HCP, differentiating ICU from non-ICU patients, and taking into account anticoagulation prophylaxis, comorbidities and the time from COVID-19 diagnosis are needed.

One and a half years after the beginning of the coronavirus disease 2019 pandemic, the high morbidity and mortality across the world is still a concern. To date, more than 255 million confirmed cases have been reported, including more than 5.1 million deaths (WHO, 2021 ). An extenuating effort in research and clinical efforts led to improvements in diagnostic strategies and therapeutics, however data gathered by the World Health Organization (WHO) about the case fatality rate across the countries are threatening (Ioannidis, 2021) .

Among factors contributing to a worse prognosis in COVID-19 patients, an important role is played by the increased chance of developing pulmonary embolism (PE) (Klok et al., 2020) . As already well known, immobility, inflammatory state, and altered coagulation are factors associated with increased chances of developing deep vein thrombosis (DVT) and PE (Elias et al., 2016) . All these factors are common in symptomatic COVID-19 patients, especially during severe disease, and since the beginning of the pandemic, several studies and case series described the occurrence of PE (Danzi et al., 2020; Suh et al., 2020) . Indeed, according to the latest meta-analysis by Suh and colleagues, PE and DVT occurred in 16.5% and 14.8% of patients hospitalized for COVID-19, respectively (Liao et al., 2020; Suh et al., 2020) . Notably, more than half of patients with PE lacked DVT (Suh et al., 2020) .

In COVID-19, two distinct pathophysiological mechanisms are believed to independently and simultaneously cause PE: immobility and local immune thrombosis (Van Dam et al., 2020) . The first pathological mechanism, as known, is characterized by blood stasis, the main risk factor for thromboembolic genesis. The second is to be ascribed to pulmonary microvascular endothelial damage, associated with systemic and local proinflammatory factors, in turn leading to coagulation cascade. Evidence points that some patients respond to the infection by an immune overactivation, leading to the so called "cytokine storm" and to activation of the coagulation system, in turn increasing the risk for Acute Respiratory Distress Syndrome (ARDS), Disseminated Intravascular Coagulation (DIC) and PE (Liu et al., 2020) . Of note, COVID-19 related PE most commonly involves the basal lung lobes, precisely in areas of ground glass opacities (GGO) (Van Dam et al., 2020; Mueller-Peltzer et al., 2020) .

Thrombotic lesions found in COVID-19-related PE more frequently involve distal, peripheral arteries of the lungs, when compared to PE found in non-COVID-19 patients (Van Dam et al., 2020) . These elements, together with a typically decreased total clot burden, as expressed through the Qanadli score (Qanadli et al., 2001) , brought the researchers to hypothesize a different PE phenotype in COVID-19 patients (Van Dam et al., 2020) . PE in COVID-19 patients is seen at various phases during the illness and can occur despite thromboembolic prophylaxis with low molecular weight heparin (LMWH) (Helms et al., 2020) . The insidious onset of PE has led to the need for clinicians of frequent monitoring of D-dimer, inflammatory markers, and clinical symptoms, in order to early identify signs of PE, to promptly perform imaging diagnostics, and eventually to start anti-thrombotic treatment.

Before COVID-19 pandemic, the most used scores to predict PE in the general population were the Geneva and the Wells scores, used either alone or in combination with D-Dimer (Guo et al., 2015) . After the beginning of the COVID-19 pandemic, due to the increasing occurrence of PE in hospitalized COVID-19 patients (HCP), several scores have been used or adapted. However, several flaws hamper the extensive use of predictive scores, especially in the context of COVID-19: i) the uneven predictive ability of available scores, in terms of sensitivity and specificity; ii) some of these scores contain variables that are not screened outside a few limited settings, such as in the case of interleukins; iii) predictive scores are often developed in the contest of research projects and are seldom validated in clinical settings.

Despite these limitations, the use of scores aiming at identifying patients at risk for this complication could represent an added value to the clinical management of COVID-19.

Since a wide variety of predictive scores are available, with different sensitivities and specificities, tested in several settings and in patients with heterogeneous clinical characteristics, we performed a systematic review of published data on prognostic model to predict the risk for PE in COVID-19. We aimed at providing a comprehensive picture of predictive values, including pros and cons of each score, and to give suggestions on their use in the clinical practice.

Studies concerning the use of at least one prediction rule to assess the risk for PE in HCPs were identified through computerized literature searches using free text searching, MEDLINE (National Library of Medicine Bethesda MD) and EMBASE and by reviewing the references of the retrieved articles.

Index search terms included the Medical Subject Heading "Covid-19" and "pulmonary embolism" and "score". The search term lines are available in supplementary material.

The search was restricted to the English language articles. The literature search period ranged from 1 January 2020 to 28 April 2021. No attempt was made to obtain information about unpublished studies. Reviewed articles were recorded in a master log and any reason for exclusion from analysis was documented in the rejected log.

The systematic review was reported according to the 2020 Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) guidelines (Page et al., 2021) .

Studies were considered eligible if they presented data pertaining the use of a prediction rule to assess the risk for PE in adult HCPs with a laboratory-confirmed SARS-CoV2 infection. The investigated outcome was the diagnosis of PE. Studies presenting data on the use of a single laboratory assay for PE prediction, such as ddimer, C-reactive protein (CRP), ferritin etc. and of imaging diagnostics alone, such as Lung Ultrasound (LUS) were excluded.

Observational studies were considered eligible, if randomized controlled trials (RCTs) were not available. Reviews, letters, editorials, abstracts, and case reports were excluded. Studies gathering data for less than 10 patients were excluded as well.

Extraction of data was performed independently by two investigators (S.A.M. and L.V.R). Each investigator was blinded to the other investigator's data extraction. In the case of disagreement between the two reviewers, a third reviewer was consulted (M.A.C.). Data from each study was entered into a standardized form, verified for consistency and accuracy, and entered a computerized database.

Abstracted information included: country and period in which patients' enrollment took place; number and setting of enrolled patients; patient characteristics, including relevant demographic variables; criteria for selecting patients to be assessed with imaging studies; imaging technique used to assess the presence of PE; number of patients with PE confirmed at imaging; number of patients completing follow up; type of scores used for the prediction, including the reported threshold, statistical information about predictive ability of the score used with sensitivity/specificity and/or AUC of the Receiver Operating Characteristic (ROC) when available.

No automatic tool was used during any phase of the present study.

Included studies were appraised for methodological quality independently by two authors (S.A.M. and L.V.R.) without blinding to journal or study authorship.

Discrepancies were resolved by discussion or involvement of a third review author if required.

The quality of observational studies was assessed using the Newcastle -Ottawa Quality Assessment Scale for Cohort Studies (NOS). Detailed assessment of the risk of bias via NOS table is available in the supplemental material in S1.

Our search retrieved 158 articles of which 12 were eventually included (Whyte et al., 2020; Baccellieri et al., 2020; Monfardini et al., 2020; Zotzmann et al., 2020; Kirsch et al., 2020; Melazzini et al., 2020; García-Ortega et al., 2021; Fang et al., 2020; Scardapane et al., 2021; Polo Friz et al., 2021; Kampouri et al., 2020; Caro-Codón et al., 2021) , comprising 4526 HCPs. Figure 1 shows the selection process of studies included in the systematic review.

A summary description of the included studies is reported in Table 1 and Table 2. Among the twelve included studies, three and nine were prospective and retrospective cohort studies, respectively. All but one studies assessed the presence of PE by means of CTPA, one used perfusion (Q)-single-photon emission computed tomography (Q/SPECT) instead. Table S1 shows the quality appraisal of the included studies.

In our search, five prediction rules were identified. Only one was specifically designed for HCPs; the remaining ones were already in use before COVID-19. Table   3 reports the prediction rules and the variables included in these scoring systems.

Wells score was the most frequently used, found in 8 out of 12 studies.

Kirsch et al. validated the utility of Wells score in predicting PE in a retrospective cohort of 64 HCPs (median age 54.9 years, 54.7% males), 12 of whom developed PE. In this study, a Wells score above 4 was significantly associated with PE development (p= 0.04), even though 4 out of 12 patients with PE had a score of 0 (Kirsch et al., 2020) . The AUC-ROC curve for the prediction of PE in HCPs, calculated for an optimal value of Wells score between 1 and 2, was 0.54 (Kirsch et al., 2020) .

In another study, no significant correlation between Wells score and PE was found in a cohort of 43 HCPs (median age 65 years, 51.16% males) (Scardapane et al., 2021) . Similar findings were confirmed by Whyte and colleagues, in a cohort of 214 HCPs (median age 61 years, 60.28% males). Authors did not find a significant association between Wells score and the probability of having PE. Of the main components of the Wells score, only the presence of DVT signs and symptoms were found to be significantly associated to PE (p = 0.0017) (Whyte et al., 2020) .

In a retrospective cohort study of 34 HCPs (median age 61 years, 77% males) with moderate-to-high pre-test probability of PE, as suggested by Wells score > 4, 76% of the subjects showed signs of PE at computerized tomography pulmonary angiogram (CTPA) (n = 26) (Monfardini et al., 2020) .

In a retrospective study on 443 HCPs (median age 68.68 years, 57.7% males), Kampouri et al. found that a Wells score > 2 in combination with a d-dimer value > 3000 ng/L provided a very specific predictive rule with a sensititity of 57.1%, and a specificity of 91.6%. A Wells score > 2 in combination with a d-dimer value > 1000 ng/L provided a more sensitive prediction rule, with a sensitivity of 92.9%, and a specificity of 46.9%. Furthermore, PE was less likely upon admission in cases where Wells score was ≤ 2 and D-dimer value was ≤ 1000 ng/ml (Kampouri et al., 2020 (Melgaard et al., 2015) . No score showed a significant correlation with PE in COVID-19 patients. All the three above mentioned scores showed poor predictive value for PE (AUC 0.497, 0.490, and 0.541, respectively) (Caro-Codón et al., 2021).

Our search identified only one study designed for the purpose of predicting PE in HCPs, i.e. the CHOD score, acronym of CRP concentration + Heart rate + Oxygen saturation + D-Dimer levels. Patients with an elevation in d-dimer were randomly selected from a cohort of 372 HCPs; 73 patients were included (median age 65.4 years, 71% males) and underwent CTPA assessment. PE was diagnosed in 35.6% of them. A multivariate analysis showed that heart rate [Hazard ratio (HR), 1.04], oxygen saturation in room-air (spO2) (HR, 0.87), d-dimer (HR, 1.02), and CRP levels (HR, 1.01) at the time of admission, were independent predictors of PE in HCPs. The AUC-ROC method was used to determine the diagnostic value of each selected quantitative variable and dichotomized variables were included in another multivariable logistic regression in order to construct the CHOD score. This score showed a high predictive value (AUC-ROC of 0.86; 95% CI: 0.8 to 0.93) (García-Ortega et al., 2021) . Furthermore, CHOD score was able to stratify patients in 3 risk groups, low (0-2 points), moderate (3-5 points) and high risk (more than 5 points), with a PE rate of 4.5%, 36.8% and 100%, respectively (García-Ortega et al., 2021) .

The importance of assessing a predictive score for PE stems from the need to promptly diagnose the acute thrombotic complications in HCPs, thus reducing an adverse outcome. Indeed, routinely performing CTPA for all HCPs would be costly, time consuming, poorly feasible, and risky for both patients and operators. Candidate selection for contrast imaging is therefore a clinical decision based on experience, observation, and laboratory findings. Ad hoc prediction tools could help in selecting patients who would benefit from CTPA in a more efficacious way.

Our systematic review found 12 studies assessing the role of 5 clinical scores in predicting PE in HCPs.

The most frequently used scores in studies included in our systematic review was the Wells score. It has been extensively used to predict PE for over twenty years and is still used nowadays for stratifying the general population into three groups. i.e., low (1.3% prevalence), moderate (16.2% prevalence), and high risk (37.5% prevalence), according to their pre-test chance of developing PE (Wells et al., 1997) . The score had an AUC of the ROC calculated for predicting PE in the general population of 0.632 (CI 0.574-0.691) (Coelho et al., 2020) .

In the studies included in our review, heterogeneous results on Wells score were obtained. Five studies did not report a significant association between this score and the risk of PE in HCPs. The only study reporting the AUC of the ROC for Wells score predicting ability for PE in HCPs was calculated to be equal to 0.54, when used alone (Kirsch et al., 2020) .

It has been hypothesized that the low prediction ability of Wells score in HCPs might be correlated with the PE pathophysiological mechanism in HCPs. Indeed, PE might result from a direct endothelial cell injury by viral action, or from an inflammatory reaction secondary to the alveolar damage (Scardapane et al., 2021) . Therefore, scores designed for investigating PE as a main diagnosis, rather than a complication of another pathology (i.e. COVID-19), may not be the best option in these patients (Fang et al., 2020) . This hampers the predictive ability of Wells score, mostly because it stems from the assumption that PE results from a DVT. COVID-19-related PE is most frequently a pulmonary local phenomenon, and not a result of immobilization, or DTV. In fact, 85% of PE cases were not associated with DVT at US of the lower limbs, as described by Monfardini and colleagues (Monfardini et al., 2020) .

Regardeless the use in HCPs, the main limitation of the Wells score pertains the inclusion of a subjective opinion of the physician among variables, i.e. "PE is most likely diagnosis", as already described by Klok and colleagues (Klok et al., 2008) .

Especially in COVID-19 management, physicians will most often suspect PE if the patients present with hypoxemia and tachycardia, de facto limiting the utility of this score in predicting PE (Kirsch et al., 2020) .

A non-accurate history taking could also lead to a miscalculation of the Wells score, as highlighted by Whyte and colleagues who found a poor predictive ability in HCPs (Whyte et al., 2020) .

Of note, it has been reported that the concomitant use of Wells score and d-Dimer would enhance the sensitivity and specificity of the test (Zhang et al., 2020; Girardi et al., 2020; Touhami et al., 2018; Kampouri et al., 2020) .

Regarding other scores utilized in studies included in our review, the usefulness of PPS in predicting PE in HCPs was evaluated in two studies (Baccellieri et al., 2020; Melazzini et al., 2020) . The PPS has been validated before COVID-19 to identify the need for anticoagulation in hospitalized patients based on their risk of VTE (Barbar et al., 2010) . Results appear to confirm the utility of PPS in HCPs only when the score is >4. Nevertheless, a larger prospective study would be necessary in order to clarify the predictive ability of PPS in predicting PE in HCPs.

The revised Geneva score includes risk factors, such as age, previous PE/DVT, surgery in the month prior to the admission, active malignant condition, and symptoms, including hemoptysis and unilateral lower-limb pain. In the general population, a score below 4 suggests a low clinical probability of PE (<10%); a score between 4 and 10 defines an intermediate risk group (10-60%), and a score ≥ 11 a high-risk group (>60%) (Le Gal et al., 2006; Wicki et al., 2001) . In HCPs, the performance of the revised Geneva score appears to be fairly satisfactory, with an AUC of the ROC of 0.727; however, the variable with the highest predictivity in this study was the mean d-dimer value (Scardapane et al., 2021) . CHA2DS2-VASc is a score used in atrial fibrillation to stratify clinically patients according to their risk of developing ischemic stroke or thromboembolism (Lip et al., 2010) . Caro-Codón and colleagues evaluated these scores in HCPs reporting poor predictive ability and no correlation between CHADS2, CHA2DS2-VASc and the M-CHA2DS2VASc and PE (Caro-Codón et al., 2021) . Thus, their usefulness for predicting PE in HCPs is very limited and these scores do not deserve to be further assessed in HCPs.

Among the examined scores, CHOD score seems promising in terms of predictive ability. This score was specifically developed for HCPs and is calculated on few routinely extracted elements. However only one study described its use, and no validation study was developed yet.

It is worth highlighting that included studies showed several flaws. First of all, most of these studies enrolled subject who already underwent CTPA, undermining the correct representation of the at-risk population thus introducing a potential selection bias. Moreover, most studies were retrospective, and importantly, not designed to evaluate the score predictive ability. Many of them did not report all relevant data, for evaluating the ability of scores and often lacked a multivariate analysis.

An additional limitation is represented by the study population; most of the studies included a mixed cohort, i.e. ICU and non-ICU patients, hampering the possibility to evaluate the role of score in predicting PE in critically ill patients. Furthermore, LMWH prophylaxis effect on preventing PE was not evaluated systematically. Most of the studies were issued in the early period of introduction of LMWH prophylaxis to all HCPs (Kulkarni et al., 2020) , therefore LMWH prophylaxis may not have been routinely performed in the clinical practice for HCPs. PE is substantially contributing to the severity burden of COVID-19, both on the short and the long term. As the majority of PE cases in COVID-19 do not result from DVT, new prediction rules, specifically developed and validated for estimating the risk of PE in COVID-19 are needed. Findings about CHOD score seems interesting, but future studies are needed to validate such score in clinical practice on a larger scale.

Scores should be differentiating ICU from non-ICU patients, and should consider anticoagulation prophylaxis, comorbidities exposing the HCP to an increased risk of developing PE and the time from COVID-19 diagnosis.

NP and LVR conceived and designed the study. LVR, SAM and DRD were responsible for data collection. LVR and MAC wrote the initial manuscript. NP, MAC and LVR critically revised the manuscript. All authors have read and approved the final manuscript.

Work supported by project #32, Donation Grants, INMI Spallanzani.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Marked as non-relevant/ lacking required information by reading the entire text (n = 32) -Marked as ineligible, i.e. review or case report, or reporting data for less than 10 patients. (n = 17) -Study in language other than English (n = 2) Studies included in review (n = 12)

Identification Screening Included 

Incidence of deep venous thrombosis in COVID-19 hospitalized patients during the first peak of the Italian outbreak

A risk assessment model for the identification of hospitalized medical patients at risk for venous thromboembolism: the Padua Prediction Score

Prediction of thromboembolic events and mortality by the CHADS2 and the CHA2DS2-VASc in COVID-19

Comparison of the Wells score and the revised Geneva score as a tool to predict pulmonary embolism in outpatients over age 65

Acute pulmonary embolism and COVID-19 pneumonia: a random association?

Prognostic models in acute pulmonary embolism: a systematic review and meta-analysis

Extent of pulmonary thromboembolic disease in patients with COVID-19 on CT: relationship with pulmonary parenchymal disease

Incidence, risk factors, and thrombotic load of pulmonary embolism in patients hospitalized for COVID-19 infection

Wells and Geneva Scores Are Not Reliable Predictors of Pulmonary Embolism in Critically Ill Patients: A Retrospective Study

Values of the Wells and Revised Geneva Scores Combined with D-dimer in Diagnosing Elderly Pulmonary Embolism Patients

High risk of thrombosis in patients with severe SARS-CoV-2 infection: a multicenter prospective cohort study

Infection fatality rate of COVID-19 inferred from seroprevalence data

Predicting Venous Thromboembolic Events in Patients with Coronavirus Disease

Requiring Hospitalization: an Observational Retrospective Study by the COVIDIC Initiative in a Swiss University Hospital

Wells Score to Predict Pulmonary Embolism in Patients with Coronavirus Disease 2019

The physician's estimation 'alternative diagnosis is less likely than pulmonary embolism' in the Wells rule is dependent on the presence of other required items

Confirmation of the high cumulative incidence of thrombotic complications in critically ill ICU patients with COVID-19: An updated analysis

Repurposed immunomodulatory drugs for Covid-19 in pre-ICu patients -mulTi-Arm Therapeutic study in pre-ICu patients admitted with Covid-19 -Repurposed Drugs (TACTIC-R): A structured summary of a study protocol for a randomised controlled trial

Prediction of Pulmonary Embolism in the Emergency Department: The Revised Geneva Score

Incidence and mortality of pulmonary embolism in COVID-19: a systematic review and meta-analysis

Assessment of the CHA 2 DS 2 -VASc Score in Predicting Ischemic Stroke, Thromboembolism, and Death in Patients With Heart Failure With and Without Atrial Fibrillation

Pulmonary thromboembolism in hospitalised COVID-19 patients at moderate to high risk by Wells score: a report from Lombardy, Italy

Pulmonary artery thrombi are co-located with opacifications in SARS-CoV2 induced ARDS

The PRISMA 2020 statement: An updated guideline for reporting systematic reviews

Acute pulmonary embolism in patients presenting pulmonary deterioration after hospitalization for noncritical COVID-19

New CT Index to Quantify Arterial Obstruction in Pulmonary Embolism

Pulmonary Artery Filling Defects in COVID-19 Patients Revealed Using CT Pulmonary Angiography: A Predictable Complication?

Pulmonary Embolism and Deep Vein Thrombosis in COVID-19: A Systematic Review and Meta-Analysis

Are the Wells Score and the Revised Geneva Score valuable for the diagnosis of pulmonary embolism in pregnancy?

Clinical and computed tomography characteristics of COVID-19 associated acute pulmonary embolism: A different phenotype of thrombotic disease?

Value of assessment of pretest probability of deep-vein thrombosis in clinical management

Pulmonary embolism in hospitalised patients with COVID-19

Assessing clinical probability of pulmonary embolism in the emergency ward: a simple score

Deep Vein Thrombosis in Hospitalized Patients With COVID-19 in Wuhan

Combining lung ultrasound and Wells score for diagnosing pulmonary embolism in critically ill COVID-19 patients

* Note: In CHOD score, points for heart rate are attributed when > 90bpm; in the revised Geneva score the same applies if 75-94bpm (3 points) or > 95bpm (5 points).