In this French nationwide study, the total number of patients hospitalized with PE in 2020 increased by about 16% compared with 2019, and mortality in hospitalized PE patients increased to 10.3% (+ 1.2%). Thus, overall hospitalization with PE increased in France during the COVID-19 pandemic, contrary to Danish reports  but similar to the UK . While the Danish study showed a significant decline in hospitalizations for PE during the first wave and lockdown, the magnitude of the first wave was lower in Denmark and the lockdown was shorter than in France. In the U.K on the contrary, there was an estimated adjusted relative risk of PE of 1.5 during the 5 first months of the pandemic compared to the “pre-COVID-19 period” from Feb 1, 2018, to the outbreak of the pandemic. It is worth noting that these two studies only investigated the 1st wave of the epidemic, whereas our results include the first 2 waves of the epidemic in France in 2020.
The increases in PE hospitalizations in France were mostly linked to the two COVID-19 waves, which were also associated with PE hospitalizations in COVID-19 patients. However, it is important to note that our analyses of PE in non-COVID-19 patients show that there was a 2.7% increase in 2020 compared with the previous three years. The frequency of patients hospitalized with PE needing ICU care also increased compared to previous years, but for COVID-19 patients only. Our findings relative to COVID-19 hospitalizations are consistent with official French government figures available from May 2020 onwards. There is a slight discrepancy in the peak of COVID-19, probably due to time elapsed between the first symptoms and severe stage of the disease requiring hospitalization, as described in Huang et al. (Additional file 1: Figure S1) .
Among hospitalized COVID-19 patients, the observed overall frequency of PE was 3.7% (2.8% in non-ICU and 8.8% in ICU). Our data suggest that 25% of COVID-19-associated PE cases were diagnosed secondarily during the hospitalization, which is much higher than the 15 to 18% observed in the years before the pandemic. It should also be underlined that COVID-19 diagnosis has the highest odds ratio among the risk factors analyzed in hospitalized PE in 2020.
The threefold higher frequency of PE in the ICU reflects that PE is a sign of severe COVID-19 and that ICU hospitalization itself increases the risk of PE. The proportion of CTPA leading to PE diagnosis in COVID-19 patients do not suggest an excessive rate of screening in ICU patients (data not shown). However, we observed lower frequencies than in several recent studies [12, 17], (2.6% to 8.9% in non-ICU and 13.7% to nearly 33% in ICU) [2, 4]. This might reflect an over-representation of the ICU population in these studies. Higher frequencies in small cohorts have already been discussed in a meta-analysis by Gallastegui et al. , who described frequencies similar to ours in cohorts with more than 400 patients. Our results in COVID-19-associated PE in ICU and non-ICU patients are in line with a recent meta-analysis . Finally, contrary to the higher frequencies reported by Jevnikar et al. , our data did not result from systematic CTPA on admission and might better reflect clinically significant PE leading to CTPA indication, and linked to clinical symptoms during the whole hospital stay.
The high frequency of PE in COVID-19 patients is due mainly to COVID-19 itself and the associated inflammatory burden. However, there are some significant differences in the PE risk factors prevalent in COVID-19 patients with PE. For instance there was a higher frequency of obesity, which is is a factor related to COVID-19 disease severity, leading both to a higher risk of PE (more sedentary lifestyle) and to more likely transfer to ICU. The frequency of cancer and recent surgery were lower in COVID-19 PE than in non-COVID-19 PE patients. These lower frequencies might be indirectly linked to more preventive behaviors in patients with cancer or recent surgery, resulting in a fewer cases of COVID-19 and, ultimately, less COVID-19-associated PE. The odds ratio associated with the usual PE risk factors derived from logistic regression of hospitalized PE were very similar to 2019 (Additional file 1: Table S1).
The very modest decrease in PE frequency in COVID-19 patients between the two waves of the epidemic may appear surprising given the enhanced preventive anticoagulation policy that was associated with a reduction in inflammatory biomarkers . One hypothesis could be that greater attention was paid to the risk of PE, which might have resulted in a more inclusive indication for CTPA, as already reported . An overdiagnosis of PE when there were perfusion defects unrelated to PE in dual energy CT angiography is also a possible explanation . The potential overdiagnosis bias should be considered alongside a potential underdiagnosis bias, which may have occurred both in non-COVID-19 patients and in less severe non-ICU COVID-19 patients. Indeed those groups might receive less medical consideration for events that are not hemodynamically significant. This seems particularly plausible in the context of the COVID-19 epidemic and the subsequent need for CTPA in severe COVID-19 patients.
Furthermore, our database shows that mortality was significantly greater for COVID-19 patients with PE versus COVID-19 patients without PE. This result contrasts with a previous study  which found no difference. This might be due to the greater power of our national data or to national differences in patient management.
As for PE in hospitalized non-COVID-19 patients, there was a surprising notable increase in patients compared to 2019. Our nationwide data illustrate a phenomenon that has been described on a local scale in Japan . In contrast, studies based on the same French nationwide database did not observe this phenomenon for other vascular diseases (such as strokes  or myocardial infarction ) during the first wave of the epidemic. In fact, they found a significant and sustained decrease in hospital admissions in 2020, driven by a sharp drop in admission during the first lockdown and a lack of compensation in the post-lockdown period. The overall increase in non-COVID-19 patients hospitalized with PE observed here is not due to changes in the usual PE risk factors. Indeed, in 2020, the frequencies of obesity, cancer and recent surgery in these patients do not differ significantly from the corresponding prevalence from 2017 to 2019. However, this phenomenon might reflect a decrease in physical activity [26, 27] associated with pandemic restrictions and lockdown, and subsequent increases in the risk of thromboembolic diseases. Other potential risk factors include an increase in smoking  or poor adherence to usual treatment in non-COVID-19 patients receiving less medical attention. On the other hand, an increase in the use of CTPA due to the suspicion of COVID-19, even in non-COVID-19 patients, may have identified PE cases which would have been overlooked otherwise.
The lockdown had only a slight influence on the monthly distribution of non-COVID-19 hospitalized PE when compared with COVID-19 patients, and the effect appears to be smaller than the reported decreases in Denmark  and Austria . There was, however, an early downturn followed by a rapid increase in hospitalized non-COVID-19 patients with PE which deserves attention. The population may have been reluctant to seek medical care during the early phase of the pandemic and lockdown (in March) due to public health policy and fear of contamination . It could also be a delayed effect of the restrictions and lockdown. Even if the lockdowns had clear beginning and end dates, people were certainly doing less physical activity, leading to a more sedentary lifestyle. In addition, the less marked slowdown and catch-up during the second wave could indicate more effective hospital reorganization and the habituation of patients to the pandemic and lockdown, meaning that they were more inclined to seek medical care and maintain a certain level physical activity. Differences with Denmark and Austria may be explained by (1) epidemiological differences including shorter lockdowns and less dramatic COVID-19 peaks, and (2) hypothetical differences in care-seeking behaviors (more likely to consult for potential cardiac symptoms than in France).
The overall increase in PE will possibly lead to increased chronic complications of PE, such as new cases of chronic thromboembolic pulmonary hypertension (CTEPH). CTEPH is both an interesting retrospective marker of COVID-19-associated PE and a conditions that will require particular consideration when diagnosing chronic dyspnea in the near future. Unfortunately, our study does not allow us to assess or predict this risk.
Further investigations will be needed to clarify the relationship between deep vein thrombosis (DVT) and PE in COVID-19 and in comparison with other ARDS. As suggested in metanalyses on DVT and PE [24, 31], there is a major to identify the proportion of PE resulting from DVT embolism and the proportion resulting from COVID-19 pulmonary vasculopathy. The endothelial effect of ARDS, which is well-known, is more correlated with duration of disease than with a specific cause [32, 33]). In COVID-19, the endothelial effect is also attributed to non-specific mechanisms such as severe hypoxemia, which in turn predisposes to thrombosis by increasing blood viscosity, complement activation, and cytokine storms, resulting in microthrombosis . However, it seems that antiphospholipid antibodies may also play a role in COVID-19 coagulopathy  as well as fibrinolysis resistance and platelet aggregation .
This article provides a nationwide report covering all of 2020, including the first two COVID-19 waves and associated lockdowns. It distinguishes COVID-19 vs non-COVID-19 patients on a monthly basis and provides a comparison with the three previous years. This work is unique for the large number of patients included and the comprehensiveness of the data. In France, the national hospital database includes information from all private and public French hospitals; it is used for the allocation of hospital budgets and encourages high levels of data coherence, accuracy, and exhaustiveness. The accuracy of PMSI for these diagnoses has been previously tested in previous literature, for instance for obesity  or for EP, with a positive predictive value of 99% for PE diagnosis, indicating that these codes can be used in comparative pharmacoepidemiological studies . Our study also adds valuable data to the pending debate over the best anticoagulation policy in COVID-19 hospitalized patients, and particularly the balance between safety and efficacy .”
Although this study includes data collected on a national level with almost 80 000 adult patients of all ages hospitalized with PE and more than 250 000 hospitalized with COVID-19, we recognize that it has several limitations. One limitation is the potential for biases related to misclassification or under-detection. However, the reliability of the detection of PE and COVID-19 in our medico-administrative database is likely very high considering its severity and the impact on patient management. Moreover, the quality and completeness of the coding are evaluated locally by the medical information departments of the individual institutions and nationally by the national health insurance system. Since the PMSI only includes data collected from hospitalized patients, we have no data from outpatients and or from patients deceased at home. For instance, our data does not allow us to provide more details about the anticoagulant treatment (prophylactic or curative) received by the patients, and further investigation by using another database which would include drugs prescribed in hospital or in an ambulatory setting would be very interesting. There are also some limits due to the study design, for instance we cannot guarantee standardized CTPA screening policy and the PMSI does not provide data to study the method use to diagnose PE. For some variables, information may not have been completed in the discharge abstract when there was no direct impact on patient care during hospitalization (e.g. tobacco use). Nevertheless, this is most often the case for day hospitalizations, and is less likely to occur in the data for inpatient hospitalizations considered in this study.
Finally, PE during pregnancy could not be taken into account seeing as data for 2021 were not yet available to identify deliveries. However, we were able to identify postpartum stays for women who gave birth in 2019 and 2020. It was not possible to take into account the effect of surgery in the logistic regression either since surgical activity was heavily affected in this very particular context, in particular during the lockdown periods.
In conclusion, the overall number of patients hospitalized with PE in France increased in 2020 compared with the previous three years, and not only because of COVID-19. The increase was visible throughout the studied months of the COVID-19 pandemic, despite a sharp decrease in planned surgeries during the lockdowns. This unique nationwide database shows a high frequency of PE among COVID-19 hospitalized patients, thus confirming the results obtained in some meta-analyses with smaller cohorts. In addition, it shows a higher mortality rate among COVID-19 patients with PE compared to those without PE, which adds information to previous reports. This study also indicates that there was an increase in PE in non-COVID-19 patients, particularly during the waves of COVID-19 and associated lockdowns.
Our results suggest that there is a need for proactive public policy concerning PE prevention in both COVID-19 and non-COVID-19 patients. This could include the promotion of physical activity as well as medical awareness, which would imply drawing even more attention to the burden PE. The apparent modest efficacy of treatments and anticoagulation approaches on PE in COVID-19 is also another strong rational for the importance of extensive SARS-COV2 vaccination.