Ventilator-induced diaphragm dysfunction (VIDD) is a frequent and under-recognized consequence of prolonged mechanical ventilation in intensive-care unit (ICU) patients. Identifying eligible candidates for clinical trials targeting VIDD remains a major operational challenge. This study evaluates the use of large language models (LLMs) to automate patient prescreening from ICU discharge summaries and estimate recruitment capacity for a future phase 2 trial. We developed an LLM-based prescreening pipeline to assess trial eligibility criteria from ICU discharge summaries, which was deployed to screen all 2024 ICU stays. Stays that were flagged as potentially eligible underwent expert adjudication. An enriched set of 50 ICU stays was independently annotated by six clinicians to define a reference standard, which was used to evaluate criterion-level model performances using F1-scores. The best-performing model was GPT-OSS:120B with a criterion-level F1-score of 0.82. When applied to consecutive 1,342 ICU stays from Montpellier University Hospital in 2024, the selected model identified 532 patients with ≥ 3 days of mechanical ventilation. After applying exclusion criteria, 185 patients remained potentially eligible. Expert review confirmed 133 patients as eligible, resulting in a positive predictive value of 72

Patient satisfaction feedback is crucial for hospital service quality, but manual reviews are time-consuming, and traditional natural language processing methods remain inadequate. Large Language Models (LLMs) show promise but are prone to extrinsic faithfulness hallucinations—fabricated or illogical outputs that limit their reliability in healthcare. This study aimed to evaluate the Self-Logical Consistency Assessment (SLCA), an original method designed to enhance LLM feedback classification reliability by enforcing a logically-structured chain of thought. SLCA uses two validation steps: self-consistency (identifying the most coherent response) and logical consistency (ensuring alignment with the original statement and expert classifications). We evaluated SLCA using GPT-4 and Llama-3.1 405B on 12,600 classifications from 100 patient feedback samples to assess hallucinations, and tested its performance on a 49,140-classification benchmark derived from 1,170 feedbacks. SLCA reduced hallucinations among detected categories from 15.80% (168/1063) to 0.51% (4/786) with GPT-4 and from 7.17% (51/711) to 1.67% (10/599) with Llama-3.1, with residual errors confined to the emergency feedback category. On the benchmark, SLCA achieved precision-recall scores of 0.86-0.78 for GPT-4 and 0.84-0.58 for Llama-3.1. These results demonstrate SLCA’s ability to achieve human-level performance across LLMs. SLCA offers a scalable, explainable solution for improving LLM classification reliability in healthcare. Its capacity to enhance performance without fine-tuning or additional training data positions it as a valuable tool for analyzing patient feedback and supporting hospital service quality improvement.

Introduced in the early 2010s, Electronic Health Records (EHRs) have become ubiquitous in hospitals. Despite clear benefits, they remain unpopular among healthcare professionals and present significant challenges. Positioned at the intersection of Health Information Systems studies, Computer Supported Collaborative Work (CSCW), Service Design, and Participatory Design (PD), our research investigates how involving users in the co-design of new EHR components within a dedicated hospital space can transform healthcare practices. Through participatory co-design methodologies, including ethnographic observation, collaborative workshops, and realistic simulations, we identify the material and interactional elements essential for rebalancing power dynamics between users and designers. This project contributes to rethinking traditional EHR design approaches, embedding design practice into systemic transformation to genuinely meet healthcare professionals' needs.

ERIOS, is a collaborative project between Dedalus, a health software company, Montpellier University Hospital Center (CHU), and the University of Montpellier. This initiative aims to incorporate research and development (R&D) directly within the hospital, focusing on co-creating components of the Electronic Health Record (EHR) alongside end-users. The project was initiated with two initial use cases, which led to the development of components for dynamic temporal visualization, now integrated into specific dashboards. The application of academic recommendations regarding user engagement methodology and human-computer interactions significantly enhanced our ability to meet user needs.

IntroductionLa santé numérique est en plein essor et appuie la transition vers une médecine personnalisée. L'implémentation d'outils numériques en pratique clinique nécessite un travail multidisciplinaire. Dans ce contexte, la Société de Pathologie Infectieuse de Langue Française a créé en 2022 le groupe d'infectiologie digitale (GID), dont l'objectif était de présenter les spécificités de l'infectiologie numérique et d'identifier des opportunités stratégiques.MéthodesLe GID s'est réuni de manière trimestrielle pendant un an et demi, avec des réunions rapprochées par sous-groupe. Un livre blanc a été rédigé.RésultatsL'infectiologie est une discipline transversale intégrant des données multimodales et multi-échelles, qui bénéficie du développement d'outils numériques pour le soin, la recherche, et l’éducation. Après quelques définitions, des spécificités de l'infectiologie sont soulignées, comme les épidémies et l’émergence d'agents pathogènes, l'importance du lien ville-hôpital, l'illectronisme chez une partie des patients, ou encore les enjeux de confidentialité dans le contexte des maladies transmissibles. Des cas d'usages stratégiques ainsi que des points d'attention sont présentés, notamment pour accompagner le développement d'outils numériques en infectiologie. Les enjeux et les limites liés à la réutilisation secondaire des données et au développement de l'intelligence artificielle sont soulignés, et des pistes d'utilisation du numérique pour l’éducation sont évoquées.ConclusionIl est essentiel de combler le fossé entre la pratique médicale, les patients, les experts en numérique et les institutions dans ce domaine en pleine expansion. Le développement d'initiatives similaires aux niveaux national et européen permettrait de combiner les efforts pour développer des outils validés et utiles pour les patients.

Patient satisfaction feedback is crucial for hospital service quality, but human-led reviews are time-consuming and traditional natural language processing remains ineffective. Large Language Models (LLM) offer potential, but their tendency to generate illogical thoughts limits their use in healthcare. Here we describe Self-Logical Consistency Assessment (SLCA), a method ensuring a reproducible LLM classification explained by a logically-structured chain of thought. In an analysis targeting extrinsic faithfulness hallucinations, SLCA mitigated the 16% GPT-4 hallucination rate, leaving only three residual cases across 12,600 classifications from 100 diverse patient feedbacks. In a benchmark designed to evaluate classification accuracy, SLCA applied to GPT-4 outperformed best algorithms, with a 88% precision rate and a 71% recall rate across 49,140 classifications from 1,170 sampled patient feedbacks. This method provides a reliable, scalable solution for improving hospital services and shows potential for accurate, explainable text classifications without fine-tuning. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement The study was conceived, funded, and executed entirely by Montpellier Hospital University Centre. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: This study complies with French regulations relating to data protection laws. Patients were informed about the usage of their data and had the option to withdraw access at any time. The ethical approval of this work was been given by the Ethical and Scientific Committee of the Montpellier University Hospital Centre (registration number : A015/2024-05-050/001) I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes The datasets used during the current study are available from the Corresponding Authors on reasonable request. Pseudonymised patient feedback may contain personal health information and their access can only be granted with traceability according to the French National Commission on Informatics and Liberty and Montpellier University Hospital Centre policies.

The early administration of appropriate antibiotic therapy is crucial for the survival of patients with bacteremia. Current research focuses on improving analytical times through technology, whereas there have been very few efforts to improve postanalytical times even though they represent 40% of the time between blood taking and appropriate treatment administration. One of the clues is the efficiency and appropriateness of the result communication system. Here, we review all delays in the entire process with the aim of improving time to appropriate treatment administration. We discuss causes for long times to adjust treatment once microbiological results are released. We argue that the pervasive health information system in this organization serves as both a bottleneck and a rigid framework. Finally, we explore how next-generation hospital information systems should be designed to effectively assist the doctors in treating patients with bacteremia.

Although off‐the‐shelf enterprise systems (ES) have been widely adopted in organisations, the extant literature repeatedly documents ES failures caused by misfits between organisational processes and the ES. Although some misfits can be identified early in the ES lifecycle, others emerge in the onward and upward phase (i.e., after the implementation) and, hence, must be resolved reactively. Prior research on misfits and resolution strategies has primarily focused on the implementation phase, often assuming that close‐to‐perfect information on the misfit's nature and characteristics is available. However, no study has examined how to effectively complete a shared diagnosis and resolution of misfits when diverging individual user perceptions are taken as the starting point. Such situations may be particularly pronounced in pluralistic organisations, where a variety of interdependent processes and potentially competing perceptions of processes are prevalent. The main objective of this study is to address this gap. To this end, we propose a pragmatic method for the diagnosis and resolution of misfits between organisational processes and enterprise systems, which builds on an actionable conceptualization of misfits. This method builds on theoretical concepts of affordances, affordance actualization, user participation, and change agentry. To demonstrate the feasibility and effectiveness of the proposed method, we conducted an action research study in a university hospital. Our analysis focused on a specific misfit involving the hospital's ES‐supported clinical processes. The findings suggest that the method effectively diagnoses and resolves misfits and optimises the resources required for their resolution through efficient management of user participation. We conclude with a discussion of the theoretical and practical contributions of our work.

OBJECTIVES:The objective of this prospective, single-center study was to explore the mid-term outcomes 6 to 9 months after hospitalization in an Intensive Care Unit (ICU) for severe COVID-19 infection. METHODS:Patients systematically underwent biological tests, pulmonary function tests, chest computed tomography (CT) scan, and psychological tests. RESULTS:Among 86 patients, including 71 (82.6%) men, median age of 65.8 years (56.7; 72.4), 57 (71.3%) patients presented post-COVID-19 asthenia, 39 (48.1%) muscle weakness, and 30 (36.6%) arthralgia. Fifty-two (64.2%) patients had a decreased diffusion capacity for carbon monoxide (DLCO) <80% and 16 (19.8%) had DLCO <60%. Chest CT-scans showed ground glass opacities in 35 (40.7%) patients, and reticular changes in 28 patients (33.7%), including fibrosis-like changes in 18 (21.7%) patients. Reticular changes and DLCO <60% were associated with length of stay in ICU, and reticular changes with higher maximal CRP level. The psychological questionnaires found 37.7% suffered from depression, 23.5% from anxiety, 42.4% from insomnia, and 9.4% from post-traumatic stress. Being female was associated with a higher frequency of depression and anxiety, with depression scores being associated with obesity. CONCLUSIONS:Many patients hospitalized in ICU for severe COVID-19 infection have mid-term sequelae. Additional studies on the prognostic factors seem necessary.

OBJECTIVE:To compare the estimated glomerular filtration rate (eGFR) using the creatinine equation (eGFRcreat) or the cystatin C equation (eGFRcys) in people with HIV (PWH) under antiretroviral drugs. We specifically included patients with an eGFRcreat around 60 ml/min per 1.73 m2 to evaluate agreement on stage 2 and 3 chronic kidney disease (CKD) classification.DESIGN:eGFRcreat, eGFRcys and resulting CKD staging were determined in 262 consecutive patients with HIV-1 (PWH) with a suppressed viral load (<200 copies/ml) under antiretroviral drugs and having impaired renal function (eGFRcreat between 45 and 80 ml/min per 1.73 m2). Antiretroviral drugs regimens were classified into eight groups: cobicistat (COBI)+elvitegravir (EVG), ritonavir (RTV)+protease inhibitor, dolutegravir (DTG), DTG+rilpivirine (RPV), RPV, raltegravir (RAL), bictegravir (BIC), and other antiretroviral drugs.RESULTS:Mean eGFRcys was higher than mean eGFRcreat (77.7 ± 0.5 vs. 67.9  ± 7.9 ml/min per 1.73 m2, P < 0.0001). The differences were significant in five treatment groups with COBI/EVG; DTG; DTG+RPV; RPV; RAL. CKD classification was modified for 51% of patients when using eGFRcys instead of eGFRcreat, with reclassification to less severe stages in 37% and worse stages in 14%.CONCLUSION:This study highlighted significant differences in eGFR depending on the renal marker used in PWH, having a significant impact on CKD classification. eGFRcys should be an additive tool for patients having eGFRcreat around 60 ml/min per 1.73 m2 for better identification of renal impairment.

•La façon dont les informations médicales du dossier patient informatisé (DPI) sont perceptibles facilite la compréhension d'une situation et permet d'anticiper l’évolution du patient, jouant un rôle majeur pour sécuriser le raisonnement médical ;•Des besoins spécifiques existent pour le raisonnement en infectiologie : dans l'analyse des situations cliniques, le bon usage individuel et collectif des antibiotiques, la gestion des maladies contagieuses et des épidémies ;•Les cliniciens sont dépendants en tant qu'utilisateurs finaux de la façon dont les développeurs ont compris leurs besoins et de la stratégie produit de l’éditeur ;•Le transfert des connaissances utiles de la pratique professionnelle des infectiologues vers les éditeurs n'est pas organisé et ne peut être l'objet de rencontres sporadiques et individuelles ;•Des méthodes de conception centrées sur l'utilisateur et l'organisation du travail, impliquant des acteurs spécifiques peuvent améliorer significativement le DPI.

We analyzed 324,734 SARS-CoV-2 variant screening tests from France enriched with 16,973 whole-genome sequences sampled during September 1, 2021-February 28, 2022. Results showed the estimated growth advantage of the Omicron variant over the Delta variant to be 105% (95% CI 96%-114%) and that of the BA.2 lineage over the BA.1 lineage to be 49% (95% CI 44%-52%). Quantitative PCR cycle threshold values were consistent with an increased ability of Omicron to generate breakthrough infections. Epidemiologic modeling shows that, in spite of its decreased virulence, the Omicron variant can generate important critical COVID-19 activity in hospitals in France. The magnitude of the BA.2 wave in hospitals depends on the level of relaxing of control measures but remains lower than that of BA.1 in median scenarios.

Background Primary Sjögren’s syndrome (pSS) is an autoimmune disease with increased risk of infections. Here, we assessed whether pSS patients were at higher risk of hospitalization for community and opportunistic infections.Methods We selected newly hospitalized pSS patients between 2011 and 2018, through a nationwide population-based retrospective study using the French Health insurance database. We compared the incidence of hospitalization for several types of infections (according to International Classification for Disease codes, ICD-10) between pSS patients and an age- and sex-matched (1:10) hospitalized control group. We calculated adjusted Hazard Ratios (aHR, 95% CI) adjusted on socio-economic status, past cardiovascular or lung diseases and blood malignancies factors.Results We compared 25 661 pSS patients with 252 543 matched patients. The incidence of hospitalizations for a first community infection was increased in pSS patients [aHR of 1.29 (1.22–1.31), p < .001]. The incidence of hospitalization for bronchopulmonary infections was increased in pSS patients [aHR of 1.50 (1.34–1.69), p < .001, for pneumonia]. Hospitalizations for pyelonephritis and intestinal infections were increased [aHR of 1.55 (1.29–1.87), p < .001 and 1.18 (1.08–1.29), p < .001, respectively]. Among opportunistic infections, only zoster, and mycobacteria infections (tuberculosis and non-tuberculous) were at increased risk of hospitalization [aHR of 3.32 (1.78–6.18), p < .001; 4.35 (1.41–13.5), p = .011 and 2.54 (1.27–5.06), p = .008, respectively].Conclusions pSS patients are at higher risk of hospitalization for infections. The increased risk of hospitalization for mycobacterial infections illustrates the potential bilateral relationship between the two conditions. Vaccination against respiratory pathogens and herpes zoster virus may help prevent some hospitalizations in pSS patients.KEY MESSAGESPrimary Sjögren’s syndrome (pSS) increases hospitalization risk for community infections: bronchopulmonary, skin, dental, ear–nose–throat, intestinal infections and pyelonephritis.Hospitalizations for zoster and mycobacterial infections are also increased in this population.Dedicated preventive measures and vaccination campaigns could decrease the burden of infections in pSS patients.

Identifié en Chine en décembre 2019, le Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-COV2) s'est rapidement propagé au niveau mondial. Si les études se sont initialement concentrées sur la prise en charge de la phase aiguë de la COrona VIrus Disease (COVID), l'objectif de ce travail est de s'intéresser aux conséquences à distance d'une hospitalisation pour COVID sévère. Nous avons réalisé une étude prospective, monocentrique, incluant des patients 6 à 9 mois après leur hospitalisation en soins critiques (soins intensifs ou réanimation) pour une infection confirmée à SARS-COV2. Les patients étaient évalués au cours d'une hospitalisation de jour en médecine interne. L'entretien débutait par un recueil des antécédents du patient, des évènements et des symptômes post-COVID. L'examinateur procédait ensuite à un examen clinique détaillé et un test de marche des 6 minutes (TM6). Les patients réalisaient de manière systématique des explorations fonctionnelles respiratoires (EFR), une tomodensitométrie (TDM) thoracique non injectée, une échographie cardiaque trans-thoracique (ETT), et un bilan biologique complet. Une batterie de tests était réalisée, explorant la qualité de vie et les séquelles psychologiques. Quatre-vingt-six patients, dont 71 (82,6 %) hommes, d'âge médian 65,8 ans (56,7;72,4), ont été évalués dans un délai moyen de 7 mois (min 3,4; max 14,9). L'hypertension artérielle (46,5 %), le diabète (34,9 %) et la dyslipidémie (39,5 %) étaient les antécédents médicaux les plus représentés. Douze (14,0 %) patients avaient une pathologie pulmonaire sous-jacente, principalement une broncho-pneumopathie chronique obstructive (BPCO). L'indice de comorbidités de Charlson médian était de 1 (0,0; 2,0). La durée médiane du séjour en soins critiques était de 10,0 (6,0;17,0) jours. Cinquante-quatre (62,8 %) patients ont eu une ventilation invasive et 67 (77,9 %) patients ont reçu un traitement par corticothérapie. Cinquante-sept (71,3 %) patients ont présenté une asthénie post-COVID, 39 (48,1 %) une faiblesse musculaire, 30 (36,6 %) des arthralgies. Quinze patients ont développé un déséquilibre du diabète et 9 un déséquilibre de la pression artérielle. Dix-sept (21,3 %) patients avaient une distance au TM6 < 80 % de la théorique. Quarante-six patients53,5 %) avaient un score de dyspnée Medical Reaserch Council (MRC) de 0 et 24 (27,9 %) avaient un score MRC de 1. Treize (15,5 %) patients avaient une auscultation anormale à type de crépitants secs des bases. Cinquante-deux (64,2 %) patients ont présenté un déficit de la diffusion du CO défini par une DLCO < 80 % et 16 (19,8 %) avaient une DLCO < 60 %. Le scanner thoracique montrait chez 35 (40,7 %) patients des lésions de verre dépoli et 18 (21,7 %) de la fibrose. Trente-deux patients (37,6 %) avaient un score PHQ-9 significatif pour une dépression, 20 (23,5 %) un score GAD-7 significatif pour un trouble anxieux et 8 (9,4 %) patients avaient un score PCL5 supérieur au seuil évoquant un état de stress post-traumatique. Le niveau de qualité de vie à distance de l'épisode aigu était satisfaisant avec une médiane de l'EQ-5D-3L à 0,89 (0,64;1,00). En analyse univariée, la fibrose sur le scanner de réévaluation et une DLCO < 60 % étaient associées à la durée du séjour en soins critiques et au niveau maximal de CRP pendant l'épisode aigu. En analyse multivariée, la fibrose et la DLCO étaient associées à la durée de séjour. En analyse uni et multivariée, le PHQ-9 et le GAD-7 étaient associés au sexe féminin. À ce jour, rares sont les études qui ont étudié de manière systématisée les séquelles somatiques et psychologiques à long terme de la COVID chez les patients qui ont été pris en charge en soins critiques. Conformément à la littérature, la diminution de la DLCO est l'anomalie la plus fréquemment retrouvée aux EFR, les plages de verre dépoli persistent sur le scanner et des lésions fibrotiques sont retrouvées chez plus d'un patient sur 5 [1], [2]. Ces anomalies contrastent avec une plainte respiratoire modérée et un examen clinique rassurant. Les symptômes généraux (asthénie, faiblesse musculaire et arthralgies) sont au cœur de la plainte fonctionnelle. Les déséquilibres de la tension artérielle et du diabète observés justifient un suivi accru des pathologies chroniques au décours de l'épisode aigu. Enfin, nous avons noté une prévalence plus importante des troubles psychologiques que celles relevées rétrospectivement (dossier informatisé) dans une autre étude [3], renforçant, selon nous, la nécessité d'un dépistage systématique des troubles psychiatriques. De nombreux patients présentent des séquelles à distance d'une COVID sévère nous incitant à un dépistage systématique afin de proposer une prise en charge adaptée.

We assessed the expression of CD169, a type I interferon-inducible receptor, on monocytes (monocyte CD169 [mCD169]) in 53 adult patients admitted to the hospital during the coronavirus disease 2019 (COVID-19) outbreak for a suspicion of severe acute respiratory syndrome coronavirus 2 infection. Monocyte CD169 was strongly overexpressed in 30 of 32 (93.7%) confirmed COVID-19 cases, compared with 3 of 21 (14.3%) patients in whom the diagnosis of COVID-19 was finally ruled out. Monocyte CD169 was associated with the plasma interferon-alpha level and thrombocytopenia. Monocyte CD169 testing may be helpful for the rapid triage of suspected COVID-19 patients during an outbreak.

In the context of social events reopening and economic relaunch, sanitary surveillance of SARS-CoV-2 infection is still required. Here, we evaluated the diagnostic performances of a rapid, extraction-free and connected reverse-transcription loop-mediated isothermal amplification (RT-LAMP) assay on saliva. Nasopharyngeal (NP) swabs and saliva from 443 outpatients were collected simultaneously and tested by reverse-transcription quantitative PCR (RT-qPCR) as reference standard test. Seventy-one individuals (16.0%) were positive by NP and/or salivary RT-qPCR. Sensitivity and specificity of salivary RT-LAMP were 85.9% (95%CI 77.8-94.0%) and 99.5% (98.7-100%), respectively. Performances were similar for symptomatic and asymptomatic participants. Moreover, SARS-CoV-2 genetic variants were analyzed and no dominant mutation in RT-LAMP primer region was observed during the period of the study. We demonstrated that this RT-LAMP test on self-collected saliva is reliable for SARS-CoV-2 detection. This simple connected test with optional automatic results transfer to health authorities is unique and opens the way to secure professional and social events in actual context of economics restart.

The implementation of rapid diagnostic tests (RDTs) may enhance the efficiency of severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) testing, as RDTs are widely accessible and easy to use. The aim of this study was to evaluate the performance of a diagnosis strategy based on a combination of antigen and immunoglobulin M (IgM) or immunoglobulin G (IgG) serological RDTs. Plasma and nasopharyngeal samples were collected between 14 March and 11 April 2020 at hospital admission from 45 patients with reverse transcription polymerase chain reaction (RT‐PCR) confirmed COVID‐19 and 20 negative controls. SARS‐CoV‐2 antigen (Ag) was assessed in nasopharyngeal swabs using the Coris Respi‐Strip. For IgM/IgG detection, SureScreen Diagnostics and Szybio Biotech RDTs were used in addition to laboratory assays (Abbott Alinity i SARS‐CoV‐2 IgG and Theradiag COVID‐19 IgM enzyme‐linked immunosorbent assay). Using the Ag RDT, 13 out of 45 (29.0%) specimens tested positive, the sensitivity was 87.0% for cycle threshold (Ct) values ≤25% and 0% for Ct values greater than 25. IgG detection was associated with high Ct values and the amount of time after the onset of symptoms. The profile of isolated IgM on RDTs was more frequently observed during the first and second week after the onset of symptoms. The combination of Ag and IgM/IgG RDTs enabled the detection of up to 84.0% of COVID‐19 confirmed cases at hospital admission. Antigen and antibody‐based RDTs showed suboptimal performances when used alone. However when used in combination, they are able to identify most COVID‐19 patients admitted in an emergency department.

We read with great interest the studies by Martin and colleagues on the dynamics of SARS-CoV-2 reverse transcriptase polymerase chain reaction (RT-PCR) positivity and seroprevalence among high-risk healthcare workers and hospital staff [[1]Martin C. Montesinos I. Dauby N. Gilles C. Dahma H. Van Den Wijngaert S. et al.Dynamics of SARS-CoV-2 RT-PCR positivity and seroprevalence among high-risk healthcare workers and hospital staff.J Hosp Infect. 2020; 106: 102-106Abstract Full Text Full Text PDF PubMed Scopus (77) Google Scholar], and Zheng and colleagues on characteristics and transmission dynamics of COVID-19 in healthcare workers at a London teaching hospital [[2]Zheng C. Hafezi-Bakhtiari N. Cooper V. Davidson H. Habibi M. Riley P. et al.Characteristics and transmission dynamics of COVID-19 in healthcare workers at a London teaching hospital.J Hosp Infect. 2020; 106: 325-329Abstract Full Text Full Text PDF PubMed Scopus (54) Google Scholar]. Taken together, these two studies underscored the efficacy of personal protective equipment (PPE), the acquisition of infections predominantly around lockdown time, and possible extraprofessional exposures as the source of infections. We report here a seroprevalence study of differentially exposed healthcare workers and hospital personnel to COVID-19 patients, which showed similar results, but, in contrast, a significant increased risk of COVID-19 in staff working in high-risk COVID-19 units. With the expansion of infections in France, authorities implemented a national lockdown on 17 March 2020 which lasted until 11 May 2020. Measures implemented at our hospital and their timelines are detailed in Figure 1. Between 21 April and 3 June 2020, we included 647 healthcare and hospital personnel volunteers from highly, mildly and unexposed COVID-19 units who had physically been present during the lockdown. Highly exposed volunteers had worked in the medical, intensive care and screening COVID-19 units, mildly exposed in the non-COVID-19 medical units, and unexposed personnel from the administration or laboratories. For highly and mildly exposed healthcare workers, eligibility implied being in contact with patients or their immediate environment (i.e. cleaning agents). After completing an investigator-led questionnaire, a blood sample for serological determination was collected, using the anti-SARS-CoV-2 IgG antibobies with the ID Screen® SARS-CoV-2-N IgG Indirect assay (ID.Vet®). Highly exposed participants had a systematic concomitant nasopharyngeal swab for SARS-CoV-2 RT-PCR test; participants in other groups were swabbed only if seropositive. Seropositive participants were investigated by two specialists to determine whether SARS-CoV-2 acquisition was most likely professional or extraprofessional. We aimed to include at least 156 subjects per exposure group, and to compare rates in each group using an exact logistic regression adjusted on age, gender and profession (doctor/resident, paramedical or other). We included 261 highly and 227 mildly exposed (representing 70% and 93% of the corresponding eligible workforce, respectively), and 159 unexposed volunteers. Mean age was 38.3 (standard deviation (SD) ± 11.0), 496 (77 %) were women. Mean days of work in the highly exposed was 21.5 days (± 11.0). Thirteen personnel tested positive for SARS-CoV-2 NP IgG (prevalence rate of 2.01 % (95% confidence Interval: 0.93–3.09)); 10 people (3.91 % (1.53–6.28)) in the highly exposed group, two (0.8 % (0.0–2.1)) in the mildly exposed group and one (0.63 % (0.00–1.86])) in the unexposed group (P=0.022). After adjustment, the odds ratio (OR) of being positive for SARS-CoV-2 in the highly exposed group was 4.43 (95% CI 1.15–17.06) vs mildly and unexposed groups (P=0.031). One highly exposed healthcare worker had a positive SARS-CoV-RT-PCR at study entry, with a positive COVID-19 serology. The study protocol was reviewed and approved by the University Hospital of Montpellier Institutional Review board (RB ID: 202000465), and registered on clinicalTrials.gov under the ID: NCT04376944. All participants consented to the study procedures and objectives. The participants were not involved in the design, or conduct, or reporting, or dissemination plans of our research. Seven seropositive cases had exposures prior or within 14 days of lockdown (Figure 1). Extra professional exposure was deemed most probable for six cases. Four of the seven most probable professional acquisition occurred in workers who recalled unprotected contact with a COVID-19 case prior to full implementation of PPE in a non COVID-19 department, and in one worker from the medical COVID-19 department intermittently using a surgical mask in presence of colleagues. Two workers did not report any known specific exposure. Despite an increased risk of acquiring COVID-19 in highly exposed personnel, seroprevalence was low, reflecting the efficacy of PPE and barrier procedures, in line with two non-comparative studies carried out in highly exposed healthcare workers [[3]Fusco F.M. Pisaturo M. Iodice V. Bellopede R. Tambaro O. Parrella G. et al.COVID-19 among healthcare workers in a specialist infectious diseases setting in Naples, Southern Italy: results of a cross-sectional surveillance study.J Hosp Infect. 2020; 105: 596-600Abstract Full Text Full Text PDF PubMed Scopus (66) Google Scholar,[4]Lahner E. Dilaghi E. Prestigiacomo C. Alessio G. Marcellini L. Simmaco M. et al.Prevalence of Sars-Cov-2 Infection in Health Workers (HWs) and Diagnostic Test Performance: The Experience of a Teaching Hospital in Central Italy.Int J Environ Res Public Health. 2020; 17Crossref PubMed Scopus (83) Google Scholar]. Most nosocomial COVID-19 infections occurred at the beginning of the lockdown, a period in which recommendations were being fully upgraded and implemented, and extraprofessional acquisition more probable due to the high community viral circulation. It was also a time during which medical and paramedical teams were still inexperienced and stressed, which could enhance mistakes when using new protective gear [[5]Tomas M.E. Kundrapu S. Thota P. Sunkesula V.C. Cadnum J.L. Mana T.S. et al.Contamination of Health Care Personnel During Removal of Personal Protective Equipment.JAMA Intern Med. 2015; 175: 1904-1910Crossref PubMed Scopus (206) Google Scholar]. Also, some data suggest possible airborne transmission of SARS-CoV-2 in enclosed environments, against which surgical masks may lack efficacy [[6]Cai J. Sun W. Huang J. Gamber M. Wu J. He G. Indirect Virus Transmission in Cluster of COVID-19 Cases, Wenzhou, China, 2020.Emerg Infect Dis. 2020; 26: 1343-1345Crossref PubMed Google Scholar]. Our findings contrast with a Chinese study of 420 healthcare professionals deployed to Wuhan for direct care of COVID-19 patients, for whom serology and SARS-CoV-2 RT-PCR of nasopharyngal swabs on return revealed no infection [[7]Liu M. Cheng S.Z. Xu K.W. Yang Y. Zhu Q.T. Zhang H. et al.Use of personal protective equipment against coronavirus disease 2019 by healthcare professionals in Wuhan, China: cross sectional study.BMJ. 2020; 369: m2195Crossref PubMed Scopus (163) Google Scholar]. However, in this study, Liu and colleagues described full measures implemented prior to the personnel's arrival, but also very strict extraprofessional rules which may be difficult to implement in other parts of the world and over an extended period. In conclusion, SARS-CoV-2 nosocomial transmissions to healthcare workers occur in high-risk settings, but PPE procedures are effective in reducing acquisition. Many cases were likely due to extraprofessional exposures and incomplete compliance with procedures. Following strict PPE procedures at work and outside are essential to reduce nosocomial acquisition of SARS-CoV-2. The authors would like to thank all participants in the study, as well as the investigators and staff who helped to organize this study: Celine Fernandez, Christine Tramoni (Infectious Diseases Departement); Géraud Philippe, Rouges Cécile, Crantelle Laura, Gabillaud Isabelle, Degline Stéphanie, Combelasse Sandrine, Mouric Hélène, Nyiramigisha Espérance, Stéphanie Marchand, Duchamp Elisabeth, Cerret Florent, Boukouche Malika, Bouta Najima (Clinical Investigation Centre); Dr Fouad Belafia, Professor Samir Jaber (Intensive Care Unit), Dereure Maëlle (Medical Informatics Department). All authors contributed to the investigation, data analysis and interpretation, and final approval of article. M.B. and O.V. carried out most of the investigation and the first draft of the article, F.G., M.C.P., G.M. and A.M. conceived of and planned the study. The authors report no financial relationships with any organizations that might have an interest in the submitted work in the previous three years, and no other relationships or activities that could appear to have influenced the submitted work. The authors report no support from any organization for the submitted work.