Search

Keeping track of data semantics and data changes in the databases is essential to support retrospective studies and the reproducibility of longitudinal clinical analysis by preventing false conclusions from being drawn from outdated data. A knowledge model combined with a temporal model plays an essential role in organizing the data and improving query expressiveness across time and multiple institutions. This paper presents a modelling framework for temporal relational databases using an ontology to derive a shareable and interoperable data model. The framework is based on: OntoRela an ontology-driven database modelling approach and Unified Historicization Framework a temporal database modelling approach. The method was applied to hospital organizational structures to show the impact of tracking organizational changes on data quality assessment, healthcare activities and data access rights. The paper demonstrated the usefulness of an ontology to provide a formal, interoperable, and reusable definition of entities and their relationships, as well as the adequacy of the temporal database to store, trace, and query data over time., Competing Interests: Declaration of conflicting interestsThe author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Background: Health care professionals are caught between the wish of patients to speed up health-related communication via emails and the need for protecting health information., Objective: We aimed to analyze the demographic characteristics of patients providing an email, and study the distribution of emails' domain names., Methods: We used the information system of the European Hospital Georges Pompidou (HEGP) to identify patients who provided an email address. We used a 1:1 matching strategy to study the demographic characteristics of the patients associated with the presence of an email, and described the characteristics of the emails used (in terms of types of emails-free, business, and personal)., Results: Overall, 4.22% (41,004/971,822) of the total population of patients provided an email address. The year of last contact with the patient is the strongest driver of the presence of an email address (odds ratio [OR] 20.8, 95% CI 18.9-22.9). Patients more likely to provide an email address were treated for chronic conditions and were more likely born between 1950 and 1969 (taking patients born before 1950 as reference [OR 1.60, 95% CI 1.54-1.67], and compared to those born after 1990 [OR 0.56, 95% CI 0.53-0.59]). Of the 41,004 email addresses collected, 37,779 were associated with known email providers, 31,005 email addresses were associated with Google, Microsoft, Orange, and Yahoo!, 2878 with business emails addresses, and 347 email addresses with personalized domain names., Conclusions: Emails have been collected only recently in our institution. The importance of the year of last contact probably reflects this recent change in contact information collection policy. The demographic characteristics and especially the age distribution are likely the result of a population bias in the hospital: patients providing email are more likely to be treated for chronic diseases. A risk analysis of the use of email revealed several situations that could constitute a breach of privacy that is both likely and with major consequences. Patients treated for chronic diseases are more likely to provide an email address, and are also more at risk in case of privacy breach. Several common situations could expose their private information. We recommend a very restrictive use of the emails for health communication., (©Vincent Looten, Antoine Neuraz, Nicolas Garcelon, Anita Burgun, Gilles Chatellier, Bastien Rance. Originally published in the Journal of Medical Internet Research (http://www.jmir.org), 24.02.2021.)

Although the consequences of the General Data Protection Regulation (GDPR) have been widely discussed, the violations have not been described in medical literature. In this study, we focus our analyses on the data breach notifications, in France, defined in the article 4 of GDPR as "a breach of security resulting, accidentally or unlawfully, in the destruction, loss, alteration, unauthorized disclosure of personal data transmitted, stored or otherwise processed, or unauthorized access to such data." Among 3,824 data breach notifications reported between May 2018 and February 2020, 244 (6.4%) is related to the health sector. Loss of confidentiality is the most important breach (80.7%) in this sector, followed by the loss of availability (27.5%). Malicious cause occurred in 58.2% of them. We hypothesized a phenomenon of underreported data breach incidents in health due to a mismatch between cybersecurity and data privacy issues.

In 2017, French institutions reformed their data access policy regarding the national insurance and administrative databases systems (French abbreviation: SNDS), including claims data from hospitalization discharge summaries for the entire population follow-up encompassing over 10 years. Our study aimed to quantitatively and qualitatively describe such authorization before and after the reform. We extracted data access demands for French National Health Data Institute (INDS) data before and after the reforms. We included only studies that needed data extracted from the SNDS database and authorization of the regulator. We inferred the number of projects accepted pre- and post-reform, and we describe the types of studies, their topics, and the types of data used. We included 802 data access demands between January 1st 2008 and September 21st 2019. The median of data access demands by year increased from 21.5 to 203. This increase was lower in the studies included insurance data (21.5 to 70). The evolution is driven by the activity of Private companies and contract-research organization. The number of studies on Hematology and oncology and internal medicine increased respectively by 1.7 and 1.4 factors. Data access of claims data refers to the "accessible" dimension of the FAIR guiding principles. However, extrinsic factors influence the accessibility of claims data such as human factors (e.g. data scientist with experience in claims data) and economic factors (e.g. data infrastructure HIPAA and GDPR compliant).

Objective: To identify common temporal evolution profiles in biological data and propose a semi-automated method to these patterns in a clinical data warehouse (CDW)., Materials and Methods: We leveraged the CDW of the European Hospital Georges Pompidou and tracked the evolution of 192 biological parameters over a period of 17 years (for 445,000 + patients, and 131 million laboratory test results)., Results: We identified three common profiles of evolution: discretization, breakpoints, and trends. We developed computational and statistical methods to identify these profiles in the CDW. Overall, of the 192 observed biological parameters (87,814,136 values), 135 presented at least one evolution. We identified breakpoints in 30 distinct parameters, discretizations in 32, and trends in 79., Discussion and Conclusion: our method allowed the identification of several temporal events in the data. Considering the distribution over time of these events, we identified probable causes for the observed profiles: instruments or software upgrades and changes in computation formulas. We evaluated the potential impact for data reuse. Finally, we formulated recommendations to enable safe use and sharing of biological data collection to limit the impact of data evolution in retrospective and federated studies (e.g. the annotation of laboratory parameters presenting breakpoints or trends)., (Copyright © 2018. Published by Elsevier B.V.)

Hospital admission of patients with sickle-cell disease (SCD) presenting with a vaso-occlusive crisis (VOC) can be justified by pain refractory to usual outpatient care and/or the occurrence of a complication. Yet, the trajectories of vital parameters and standard biomarkers throughout a non-complicated VOC has not been established. In this observational cohort study, we describe the course of routine parameters throughout 329 hospital stays for non-complicated VOC. We used a new spline-based approach to study and visualize non-specific time-dependent variables extracted from the hospital clinical data warehouse. We identified distinct trends during the VOC for hemoglobin level, leukocytes count, C-Reactive Protein (CRP) level and temperature. Hemoglobin decreased after admission and rarely returned to steady state levels before discharge. White blood cell counts were elevated at admission before immediately decreasing, whereas eosinophils increased slowly throughout the first five days of the stay. In over 95% of non-complicated VOC-related stays, the CRP value was below 100 mg/L within the first day following admission and above normal after 48 hours, and the temperature was below 38 °C throughout the entire stay. Knowing the typical trajectories of these routine parameters during non-complicated VOC may urge the clinicians to be more vigilant in case of deviation from these patterns., Competing Interests: The authors declare no conflict of interest.

A significant part of medical knowledge is stored as unstructured free text. However, clinical narratives are known to contain duplicated sections due to clinicians' copy/paste parts of a former report into a new one. In this study, we aim at evaluating the duplications found within patient records in more than 650,000 French clinical narratives. We adapted a method to identify efficiently duplicated zones in a reasonable time. We evaluated the potential impact of duplications in two use cases: the presence of (i) treatments and/or (ii) relative dates. We identified an average rate of duplication of 33%. We found that 20% of the document contained drugs mentioned only in duplicated zones and that 1.45% of the document contained mentions of relative dates in duplicated zone, that could potentially lead to erroneous interpretation. We suggest the systematic identification and annotation of duplicated zones in clinical narratives for information extraction and temporal-oriented tasks.

We explore the impact of data source on word representations for different NLP tasks in the clinical domain in French (natural language understanding and text classification). We compared word embeddings (Fasttext) and language models (ELMo), learned either on the general domain (Wikipedia) or on specialized data (electronic health records, EHR). The best results were obtained with ELMo representations learned on EHR data for one of the two tasks(+7% and +8% of gain in F1-score).

Methodological reproducibility refers to the ability to reproduce exactly the same results by reproducing the same study protocol on the same data. The aim of this study was to assess the methodological reproducibility of studies published in the Revue d'Epidémiologie et de Santé Publique between 2008 and 2017 and using data from the national health data system. Our results suggest that only 49% of the studies could be replicated without the help of the authors. The results may reveal a lack of concern about the purpose of public health research. It is difficult to attribute responsibility for this default of reproducibility solely to researchers, so we hypothesize an instituted ethical misconduct., (© 2019 médecine/sciences – Inserm.)

Background: Choosing the most performing method in terms of outcome prediction or variables selection is a recurring problem in prognosis studies, leading to many publications on methods comparison. But some aspects have received little attention. First, most comparison studies treat prediction performance and variable selection aspects separately. Second, methods are either compared within a binary outcome setting (where we want to predict whether the readmission will occur within an arbitrarily chosen delay or not) or within a survival analysis setting (where the outcomes are directly the censored times), but not both. In this paper, we propose a comparison methodology to weight up those different settings both in terms of prediction and variables selection, while incorporating advanced machine learning strategies., Methods: Using a high-dimensional case study on a sickle-cell disease (SCD) cohort, we compare 8 statistical methods. In the binary outcome setting, we consider logistic regression (LR), support vector machine (SVM), random forest (RF), gradient boosting (GB) and neural network (NN); while on the survival analysis setting, we consider the Cox Proportional Hazards (PH), the CURE and the C-mix models. We also propose a method using Gaussian Processes to extract meaningfull structured covariates from longitudinal data., Results: Among all assessed statistical methods, the survival analysis ones obtain the best results. In particular the C-mix model yields the better performances in both the two considered settings (AUC =0.94 in the binary outcome setting), as well as interesting interpretation aspects. There is some consistency in selected covariates across methods within a setting, but not much across the two settings., Conclusions: It appears that learning withing the survival analysis setting first (so using all the temporal information), and then going back to a binary prediction using the survival estimates gives significantly better prediction performances than the ones obtained by models trained "directly" within the binary outcome setting.

Purpose: Heart failure (HF) is a common, serious, and still poorly known illness, which might benefit from studies in claims databases. However, to provide reliable estimates, HF patients must be adequately identified. This validation study aimed to estimate the diagnostic accuracy of the International Classification of Diseases, Tenth Revision (ICD-10) codes I50.x, heart failure, in the French hospital discharge diagnoses database., Methods: This study was performed in two university hospitals, comparing recorded discharge diagnoses and electronic health records (EHRs). Patients with discharge ICD-10 codes 150.x were randomly selected. Their EHRs were reviewed to classify HF diagnosis as definite, potential, or miscoded based on the European Society of Cardiology diagnostic criteria, from which the codes' positive predictive value (PPV) was computed. To estimate sensitivity, patients with an EHR HF diagnosis were identified, and the presence of the I50.x codes was sought for in the hospital discharge database., Results: Two hundred possible cases of HF were selected from the hospital discharge database, and 229 patients with an HF diagnosis were identified from the EHR. The PPV of I50.x codes was 60.5% (95% CI, 53.7%-67.3%) for definite HF and 88.0% (95% CI, 83.5%-92.5%) for definite/potential HF. The sensitivity of I50.x codes was 64.2% (95% CI, 58.0%-70.4%). PPV results were similar in both hospitals; sensitivity depended on the source of EHR: Departments of cardiology had a higher sensitivity than had nonspecialized wards., Conclusions: Diagnosis codes I50.x in discharge summary databases accurately identify patients with HF but fail to capture some of them., (© 2018 John Wiley & Sons, Ltd.)

Signs and symptoms of pulmonary embolism (PE) are not specific and this can lead to a diagnostic delay. Little is known about the determinants of this delay and its prognostic implication. We conducted a retrospective analysis of a prospective cohort involving 514 patients with a first episode of PE. The diagnostic delay was defined as a time from first symptom onset to diagnosis of >3 days, corresponding of the median time in the population. Multivariable logistic regression analysis was performed to identify determinants of diagnostic delay. Prognostic implication was measured as the occurrence of 30-day all-cause mortality, haemodynamic collapse or recurrent PE. A total of 240 (47%) among 514 patients had a time from first symptom to diagnosis > 3 days. Previous deep vein thrombosis (OR 0.55, 95% Confidence Interval (CI), 0.32-0.93), immobilization (OR 0.52, 95% CI, 0.28-0.96), surgery (OR 0.31, 95% CI, 0.16-0.62), chest pain (OR 0.58, 95% CI, 0.39-0.86), syncope (OR 0.48, 95% CI, 0.23-1.01), dyspnea (OR 2.48, 95% CI, 1.57-3.91) and hemoptysis (OR 3.57, 95% CI, 1.40-9.07) were associated with diagnostic delay. Twenty-two patients (4.3%, 95%CI, 2.8-6.5) experienced an outcome event within 30 days. Among them, 15 patients (6.2% 95%CI, 3.7-10.3) had a diagnostic delay and 7 (2.6%, 95% CI 1.1-5.4) did not (p = 0.039). In this cohort, diagnostic delay is associated with the absence of major risk factors for PE or clinical features such as chest pain or syncope and the presence of dyspnea or hemoptysis. Diagnostic delay is associated with a worse 30-day prognosis., (Copyright © 2018 Elsevier Ltd. All rights reserved.)

Objective: This study was designed to further our knowledge of the management of pregnant women based on the national health data system (SNDS)., Material and Methods: Women covered by the national health insurance general scheme or a local mutualist section, who delivered in 2015., Results: Among the 672,182 women included (mean age: 31 years, SD 5.3), 0.3% were under the age of 18 years, 4% lived in a French overseas department (<18 years: 21%), 17% had complementary universal health insurance coverage (<18 years: 75%), 1.2% presented a mental illness, 0.6% had a cancer, and 0.4% had cardiovascular disease. At least one outpatient visit with a gynaecologist or midwife was detected for 93% of women (first trimester (T1): 75%), specific or nonspecific pelvic ultrasound was performed in 98% (T1: 92%), blood glucose assay was performed in 78% (T1: 61%), and an oral glucose tolerance test was performed in 58%. Before delivery, 0.2% of women had at least one admission to the intensive care unit and 22% had at least one hospital stay (<18 years: 38%), for which the principal diagnoses were: false labour (4.5%), threatened preterm labour (2.5%), surveillance of high-risk pregnancy (2.6%), diabetes (2.6%), and hypertension (0.7%). The preterm delivery rate was 6.7% (<18 years: 14%, ≥40 years: 9%). Although 20% of deliveries were performed by caesarean section, 16% of vaginal deliveries required instrumental extraction., Discussion: SNDS data enrich the data derived from periodic national perinatal surveys, such as the poor follow-up of adolescent girls. These data can promote the elaboration and monitoring of annual indicators., (Copyright © 2018 Elsevier Masson SAS. All rights reserved.)

Open data available largely and with minimal constraints to the general public and journalists are needed to help rebuild trust between citizens and the health system. By opening data, we can expect to increase the democratic accountability, the self-empowerment of citizens. This article aims at assessing the quality and reusability of the Transparency - Health database with regards to the FAIR principles. More specifically, we observe the quality of the identity of the French medical doctors in the Transp-db. This study shows that the quality of the data in the Transp-db does not allow to identity with certainty those who benefit from an advantage or remuneration to be confirmed, reducing noticeably the impact of the open data effort.

Purpose: Macular edema is the leading cause of vision loss in bilateral chronic noninfectious posterior uveitis, and is currently being treated using corticosteroids, immunosuppressive agents, and biotherapies. The aim of this trial was to assess and compare the efficacy and safety of corticosteroids and interferon-α (IFN-α) in adults with such conditions., Design: Randomized controlled trial., Methods: Subjects: Adult patients with bilateral posterior autoimmune noninfectious and nontumoral uveitis complicated by macular edema in at least 1 eye., Intervention: Patients received either subcutaneous IFN-α2a, systemic corticosteroids, or no treatment for 4 months. The efficacy and safety were assessed for up to 4 months., Main Outcome Measures: The main endpoint was the change of the central foveal thickness (CFT) obtained by optical coherence tomography., Results: Forty-eight patients were included. In intention-to-treat analysis, the median CFT change showed no significant difference. However, the per-protocol analysis showed a significant difference between groups for both eyes (OD and OS), and for the worse and better eyes. Statistically significant difference was found between the control and corticosteroid groups for the OD (P = .0285), and between the control and IFN-α groups for the OD (P = .0424) and worse eye (P = .0354). Serious adverse events occurred in 2 patients in the IFN group, in 1 patient in the corticosteroid group, and in 2 patients in the control group and were completely resolved after switch., Conclusions: IFN-α and systemic corticosteroids, compared with no treatment, were associated with significant anatomic and visual improvement shown in the per-protocol study., (Copyright © 2017 Elsevier Inc. All rights reserved.)

Introduction: The prevalence rate of congestive heart failure is approximately 2% in high-income countries. The aim of this study was to assess the overall benefit of ultrafiltration therapy in patients with acute or persistent congestive heart failure., Methods: We conducted a health technology assessment following the EUnetHTA guidelines, with systematic literature review from bibliographic medical databases, independent experts and manufacturer interviews., Results: Thirteen clinical trials and five meta-analyses were examined. In the most recent one, 608 patients were included, of which 304 received ultrafiltration therapy and 304 received intravenous loop diuretics. Ultrafiltration therapy seems to be more beneficial regarding the fluid removal and the body weight reduction, (mean difference respectively 1.44kg, IC95% [0.29; 2.59], P-value=0.01 and 1.28L [0.43; 2.12], P-value=0.003). No difference has been showed in overall mortality, renal function, hospital readmission or safety. Medico-economic studies are incomplete and contradictory., Conclusion: Ultrafiltration therapy seems to be effective, most likely for patients ineligible or resistant to intravenous diuretics. But most topics remain uncertain, mainly impact on overall mortality, safety and cost-effectiveness. Given these knowledge-gaps, the generalization of ultrafiltration therapy should be examined cautiously, and conditional upon a large-scale systematic evaluation., (Copyright © 2016. Published by Elsevier SAS.)

The treatment of acute hematogenous osteomyelitis has evolved in recent years to a shorter parenteral treatment with an early switch to the oral route. Current publications recommend a 2- to 4-day parenteral treatment before the oral switch. We retrospectively analyzed a series of 45 children aged 1 to 11 years and treated in our department for acute osteomyelitis without severity criterion. Nineteen of 45 patients were treated by an exclusive ambulatory oral treatment by amoxicillin and clavulanic acid. Twenty six of 45 patients had a 2- to 4-day parenteral treatment before the oral switch. The minimum follow-up was 6 months. The primary endpoint was a clinical, radiographic, and biologic healing, 6 months after the beginning of the treatment. The secondary endpoints evaluated were the length of hospitalization, the total duration of treatment, and the type of antibiotic used. On the primary endpoint, we did not find any significant difference between the 2 treatments (P = 0.38). On the duration of treatment, we found a significant difference (P = 0.049) in favor of oral treatment. The ambulatory oral treatment by amoxicillin and clavulanic acid seems to be a valid alternative to the classical parenteral then oral sequence in the treatment of acute hematogenous osteomyelitis in children without severity criterion.

Background: As a part of a larger Health Technology Assessment (HTA), the measurement error of a device used to monitor the hemoglobin concentration of a patient undergoing surgery, as well as its decision consequences, were to be estimated from published data., Methods: A Bayesian hierarchical model of measurement error, allowing the meta-analytic estimation of both central and dispersion parameters (under the assumption of normality of measurement errors) is proposed and applied to published data; the resulting potential decision errors are deduced from this estimation. The same method is used to assess the impact of an initial calibration., Results: The posterior distributions are summarized as mean ± sd (credible interval). The fitted model exhibits a modest mean expected error (0.24 ± 0.73 (-1.23 1.59) g/dL) and a large variability (mean absolute expected error 1.18 ± 0.92 (0.05 3.36) g/dL). The initial calibration modifies the bias (-0.20 ± 0.87 (-1.99 1.49) g/dL), but the variability remains almost as large (mean absolute expected error 1.05 ± 0.87 (0.04 3.21) g/dL). This entails a potential decision error ("false positive" or "false negative") for about one patient out of seven., Conclusions: The proposed hierarchical model allows the estimation of the variability from published aggregates, and allows the modeling of the consequences of this variability in terms of decision errors. For the device under assessment, these potential decision errors are clinically problematic.

Objective: To assess work productivity (WP) loss during the first 3 years of disease in a cohort of patients with early arthritis (EA) diagnosed between 2002 and 2005., Methods: The ESPOIR (Etude et Suivi des Polyarthrites Indifférenciées Récentes) cohort included 813 EA patients; we included those of working age at baseline in the present study. WP loss was assessed by 3 components: sick leave, permanent disability, and early retirement. The proportion of affected patients and the mean number of days off work were assessed for each component. WP costs were estimated and determinants of positive and extreme costs were assessed by logistic regression models., Results: Among the 664 patients included, 81.6% were in the workforce at baseline. During the first 3 years of disease, 45% reported at least 1 sick leave day and 11% reported permanent disability. Only a few patients (1%) reported early retirement. The mean number of days on sick leave due to EA decreased regularly from 44 to 13, whereas the mean number of days on permanent disability tripled from 10 to 33. The mean annual cost was 1,333 (95% confidence interval 1,075-1,620). Sick leave longer than 30 days due to EA before inclusion and a decrease in mental and physical scores of the Medical Outcomes Study Short Form 36 at inclusion were independent determinants of positive and extreme costs in multivariate models., Conclusion: WP loss is substantial in EA patients and is due to permanent disability before the third year of disease. Work absence and poor mental and physical health status at baseline are major determinants of WP costs., (Copyright © 2014 by the American College of Rheumatology.)