Leveraging the EHR4CR platform to support patient inclusion in academic studies: challenges and lessons learned.

BACKGROUND: The development of Electronic Health Records (EHRs) in hospitals offers the ability to reuse data from patient care activities for clinical research. EHR4CR is a European public-private partnership aiming to develop a computerized platform that enables the re-use of data collected from EHRs over its network. However, the reproducibility of queries may depend on attributes of the local data. Our objective was 1/ to describe the different steps that were achieved in order to use the EHR4CR platform and 2/ to identify the specific issues that could impact the final performance of the platform. METHODS: We selected three institutional studies covering various medical domains. The studies included a total of 67 inclusion and exclusion criteria and ran in two University Hospitals. We described the steps required to use the EHR4CR platform for a feasibility study. We also defined metrics to assess each of the steps (including criteria complexity, normalization quality, and data completeness of EHRs). RESULTS: We identified 114 distinct medical concepts from a total of 67 eligibility criteria Among the 114 concepts: 23 (20.2%) corresponded to non-structured data (i.e. for which transformation is needed before analysis), 92 (81%) could be mapped to terminologies used in EHR4CR, and 86 (75%) could be mapped to local terminologies. We identified 51 computable criteria following the normalization process. The normalization was considered by experts to be satisfactory or higher for 64.2% (43/67) of the computable criteria. All of the computable criteria could be expressed using the EHR4CR platform. CONCLUSIONS: We identified a set of issues that could affect the future results of the platform: (a) the normalization of free-text criteria, (b) the translation into computer-friendly criteria and (c) issues related to the execution of the query to clinical data warehouses. We developed and evaluated metrics to better describe the platforms and their result. These metrics could be used for future reports of Clinical Trial Recruitment Support Systems assessment studies, and provide experts and readers with tools to insure the quality of constructed dataset.

Association between borderline dysnatremia and mortality insight into a new data mining approach.

BACKGROUND: Even small variations of serum sodium concentration may be associated with mortality. Our objective was to confirm the impact of borderline dysnatremia for patients admitted to hospital on in-hospital mortality using real life care data from our electronic health record (EHR) and a phenome-wide association analysis (PheWAS). METHODS: Retrospective observational study based on patient data admitted to Hôpital Européen George Pompidou, between 01/01/2008 and 31/06/2014; including 45,834 patients with serum sodium determinations on admission. We analyzed the association between dysnatremia and in-hospital mortality, using a multivariate logistic regression model to adjust for classical potential confounders. We performed a PheWAS to identify new potential confounders. RESULTS: Hyponatremia and hypernatremia were recorded for 12.0% and 1.0% of hospital stays, respectively. Adjusted odds ratios (ORa) for severe, moderate and borderline hyponatremia were 3.44 (95% CI, 2.41-4.86), 2.48 (95% CI, 1.96-3.13) and 1.98 (95% CI, 1.73-2.28), respectively. ORa for severe, moderate and borderline hypernatremia were 4.07 (95% CI, 2.92-5.62), 4.42 (95% CI, 2.04-9.20) and 3.72 (95% CI, 1.53-8.45), respectively. Borderline hyponatremia (ORa = 1.57 95% CI, 1.35-1.81) and borderline hypernatremia (ORa = 3.47 95% CI, 2.43-4.90) were still associated with in-hospital mortality after adjustment for classical and new confounding factors identified through the PheWAS analysis. CONCLUSION: Borderline dysnatremia on admission are independently associated with a higher risk of in-hospital mortality. By using medical data automatically collected in EHR and a new data mining approach, we identified new potential confounding factors that were highly associated with both mortality and dysnatremia.

Efficiency and effectiveness evaluation of an automated multi-country patient count cohort system.

BACKGROUND: With the increase of clinical trial costs during the last decades, the design of feasibility studies has become an essential process to reduce avoidable and costly protocol amendments. This design includes timelines, targeted sites and budget, together with a list of eligibility criteria that potential participants need to match. The present work was designed to assess the value of obtaining potential study participant counts using an automated patient count cohort system for large multi-country and multi-site trials: the Electronic Health Records for Clinical Research (EHR4CR) system. METHODS: The evaluation focuses on the accuracy of the patient counts and the time invested to obtain these using the EHR4CR platform compared to the current questionnaire based process. This evaluation will assess the patient counts from ten clinical trials at two different sites. In order to assess the accuracy of the results, the numbers obtained following the two processes need to be compared to a baseline number, the "alloyed" gold standard, which was produced by a manual check of patient records. RESULTS: The patient counts obtained using the EHR4CR system were in three evaluated trials more accurate than the ones obtained following the current process whereas in six other trials the current process counts were more accurate. In two of the trials both of the processes had counts within the gold standard's confidence interval. In terms of efficiency the EHR4CR protocol feasibility system proved to save approximately seven calendar days in the process of obtaining patient counts compared to the current manual process. CONCLUSIONS: At the current stage, electronic health record data sources need to be enhanced with better structured data so that these can be re-used for research purposes. With this kind of data, systems such as the EHR4CR are able to provide accurate objective patient counts in a more efficient way than the current methods. Additional research using both structured and unstructured data search technology is needed to assess the value of unstructured data and to compare the amount of efforts needed for data preparation.

Increased risk of gastric cancer in relation with pernicious anaemia in patients with primary antibody deficiency: A nationwide case control study.

BACKGROUND/AIM: We aimed to assess gastrointestinal cancers risks in a large cohort of individuals with primary antibody deficiency (PAD) and their association with risk of autoimmune and inflammatory gastrointestinal diseases. METHODS: Investigating a French national database of inpatient admissions between 2010 and 2018, we identified 12,748 patients with PAD and 38,244 control non-exposed individuals. We performed multiple exposed-non-exposed studies using conditional logistic regression. RESULTS: In comparison with non-exposed patients, PAD patients had increased risk of in situ gastric carcinoma (Odds Ratio (OR) =10.5 [95 % CI 2.2; 50.5]), malignant gastric tumor (OR=3.2 [95 % CI 2.2; 4.4]) and colorectal cancer (OR=1.2 [95 % CI 1; 1.5]). PAD patients had also increased risk of pernicious anaemia (OR=8 |95 % CI 5.6; 11.5]), Crohn's disease (OR= 4.4 [95 % CI 3.5; 5.6]), ulcerative colitis (OR=2.9 [95 % CI 2.4; 3.6]) and coeliac disease (OR=13.3 [95 % CI 9.1; 19.5]). Within patients with gastric cancer, those with PAD had increased risk of pernicious anaemia (OR=8.4 [95 % CI 1.5; 215]; p = 0.01) but not of H. pylori infection. CONCLUSIONS: Risk of gastric cancer is particularly high in PAD patients and notably risk of in situ gastric carcinoma in association with pernicious anaemia. It supports indication of early endoscopic screening in these patients.

Risk-Benefit Balance Associated With Obstetric, Neonatal, and Child Outcomes After Metabolic and Bariatric Surgery.

Importance: Metabolic and bariatric surgery (MBS) is the most efficient therapeutic option for severe obesity. Most patients who undergo MBS are women of childbearing age. Data in the scientific literature are generally of a low quality due to a lack of well-controlled prospective trials regarding obstetric, neonatal, and child outcomes. Objective: To assess the risk-benefit balance associated with MBS around obstetric, neonatal, and child outcomes. Design, Setting, and Participants: The study included 53 813 women on the French nationwide database who underwent an MBS procedure and delivered a child between January 2012 and December 2018. Each women was their own control by comparing pregnancies before and after MBS. Exposures: The women included were exposed to either gastric bypass or sleeve gastrectomy. Main Outcomes and Measures: The study team first compared prematurity and birth weights in neonates born before and after maternal MBS with each other. Then they compared the frequencies of all pregnancy and child diagnoses in the first 2 years of life before and after maternal MBS with each other. Results: A total of 53 813 women (median [IQR] age at surgery, 30 [26-35] years) were included, among 3686 women who had 1 pregnancy both before and after MBS. The study team found a significant increase in the small-for-gestational-age neonate rate after MBS (+4.4%) and a significant decrease in the large-for-gestational-age neonate rate (-12.6%). The study team highlighted that compared with pre-MBS births, after MBS births had fewer occurrences of gestational hypertension (odds ratio [OR], 0.16; 95% CI, 0.10-0.23) and gestational diabetes for the mother (OR, 0.39; 95% CI, 0.34-0.45), as well as fewer birth injuries to the skeleton (OR, 0.27; 95% CI, 0.11-0.60), febrile convulsions (OR, 0.39; 95% CI, 0.21-0.67), viral intestinal infections (OR, 0.56; 95% CI, 0.43-0.71), or carbohydrate metabolism disorders in newborns (OR, 0.54; 95% CI 0.46-0.63), but an elevated respiratory failure rate (OR, 2.42; 95% CI, 1.76-3.36) associated with bronchiolitis. Conclusions and Relevance: The risk-benefit balance associated with MBS is highly favorable for pregnancies and newborns but may cause an increased risk of respiratory failure associated with bronchiolitis. Further studies are needed to better assess the middle- and long-term benefits and risks associated with MBS.

Change in Birth Rate Before and After Bariatric Surgery in France.

IMPORTANCE: Bariatric surgery worldwide has grown significantly over the past years and is performed, in a vast majority, in women of childbearing age. The impact of these procedures on birth rates remains largely unknown. OBJECTIVE: The main objective was to study the evolution of the birth rates in these women, before and after bariatric surgery and in comparison with women from the general population. The secondary objectives were to compare the birth rates before and after surgery according to the Body Mass Index (BMI) (BMI 40-50 versus >50 kg/m2). METHOD: Our analyses are based on a national medico-administrative database. All women of childbearing age and who had bariatric surgery between 2012 and 2016 were included, and we included all deliveries between 2012 and 2018 in this population. We compared the birth rates before and after bariatric surgery. National statistics (INSEE, Institut National de la Statistique et des Etudes Economiques) were used for comparison with women from the general population. FINDINGS: A total of 69,932 women were included between 2012 and 2016, with a median age at surgery of 33 years [27; 39]. Among them, 9391 (13%) had a BMI over 50 kg/m2 before surgery, 46,818 (67%) benefited from a sleeve gastrectomy (SG), and 23,376 (33%) from a gastric bypass (GBP). Birth rates tend to be smaller after surgery compared to before surgery for women under 27 years (OR: 0.92, CI95% [0.88; 0.96]), while being not different after 27 in both groups (OR: 1.00, CI95% [0.97; 1.03]). Birth rates of obese women who have benefited or will benefit from the bariatric surgery were higher than the general population, with a peak at a younger age, and then decline earlier than the general population. Trends were different according to BMI class. Birth rates were higher for women with BMI between 40 and 50 kg/m2 compared to women with BMI greater than 50 kg/m2 (OR=1.28, CI95% [1.21; 1.36]) before surgery while slightly lower after surgery (OR=0.95, CI95% [0.91; 0.99]). CONCLUSIONS AND RELEVANCE: We showed that birth rates after bariatric surgery were not different before and after bariatric surgery except for women under 27 years old who had a smaller birth rate. Interestingly, we observed an improvement of birth rates after surgery for women with a BMI of 50 kg/m2 or more.

The Adverse Drug Reactions From Patient Reports in Social Media Project: Protocol for an Evaluation Against a Gold Standard.

BACKGROUND: Social media is a potential source of information on postmarketing drug safety surveillance that still remains unexploited nowadays. Information technology solutions aiming at extracting adverse reactions (ADRs) from posts on health forums require a rigorous evaluation methodology if their results are to be used to make decisions. First, a gold standard, consisting of manual annotations of the ADR by human experts from the corpus extracted from social media, must be implemented and its quality must be assessed. Second, as for clinical research protocols, the sample size must rely on statistical arguments. Finally, the extraction methods must target the relation between the drug and the disease (which might be either treated or caused by the drug) rather than simple co-occurrences in the posts. OBJECTIVE: We propose a standardized protocol for the evaluation of a software extracting ADRs from the messages on health forums. The study is conducted as part of the Adverse Drug Reactions from Patient Reports in Social Media project. METHODS: Messages from French health forums were extracted. Entity recognition was based on Racine Pharma lexicon for drugs and Medical Dictionary for Regulatory Activities terminology for potential adverse events (AEs). Natural language processing-based techniques automated the ADR information extraction (relation between the drug and AE entities). The corpus of evaluation was a random sample of the messages containing drugs and/or AE concepts corresponding to recent pharmacovigilance alerts. A total of 2 persons experienced in medical terminology manually annotated the corpus, thus creating the gold standard, according to an annotator guideline. We will evaluate our tool against the gold standard with recall, precision, and f-measure. Interannotator agreement, reflecting gold standard quality, will be evaluated with hierarchical kappa. Granularities in the terminologies will be further explored. RESULTS: Necessary and sufficient sample size was calculated to ensure statistical confidence in the assessed results. As we expected a global recall of 0.5, we needed at least 384 identified ADR concepts to obtain a 95% CI with a total width of 0.10 around 0.5. The automated ADR information extraction in the corpus for evaluation is already finished. The 2 annotators already completed the annotation process. The analysis of the performance of the ADR information extraction module as compared with gold standard is ongoing. CONCLUSIONS: This protocol is based on the standardized statistical methods from clinical research to create the corpus, thus ensuring the necessary statistical power of the assessed results. Such evaluation methodology is required to make the ADR information extraction software useful for postmarketing drug safety surveillance. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): RR1-10.2196/11448.

Mining Adverse Drug Reactions in Social Media with Named Entity Recognition and Semantic Methods.

Suspected adverse drug reactions (ADR) reported by patients through social media can be a complementary source to current pharmacovigilance systems. However, the performance of text mining tools applied to social media text data to discover ADRs needs to be evaluated. In this paper, we introduce the approach developed to mine ADR from French social media. A protocol of evaluation is highlighted, which includes a detailed sample size determination and evaluation corpus constitution. Our text mining approach provided very encouraging preliminary results with F-measures of 0.94 and 0.81 for recognition of drugs and symptoms respectively, and with F-measure of 0.70 for ADR detection. Therefore, this approach is promising for downstream pharmacovigilance analysis.

Detection of Drug-Drug Interactions Inducing Acute Kidney Injury by Electronic Health Records Mining.

BACKGROUND AND OBJECTIVE: While risk of acute kidney injury (AKI) is a well documented adverse effect of some drugs, few studies have assessed the relationship between drug-drug interactions (DDIs) and AKI. Our objective was to develop an algorithm capable of detecting potential signals on this relationship by retrospectively mining data from electronic health records. MATERIAL AND METHODS: Data were extracted from the clinical data warehouse (CDW) of the Hôpital Européen Georges Pompidou (HEGP). AKI was defined as the first level of the RIFLE criteria, that is, an increase ≥50 % of creatinine basis. Algorithm accuracy was tested on 20 single drugs, 10 nephrotoxic and 10 non-nephrotoxic. We then tested 45 pairs of non-nephrotoxic drugs, among the most prescribed at our hospital and representing distinct pharmacological classes for DDIs. RESULTS: Sensitivity and specificity were 50 % [95 % confidence interval (CI) 23.66-76.34] and 90 % (95 % CI 59.58-98.21), respectively, for single drugs. Our algorithm confirmed a previously identified signal concerning clarithromycin and calcium-channel blockers (unadjusted odds ratio (ORu) 2.92; 95 % CI 1.11-7.69, p = 0.04). Among the 45 drug pairs investigated, we identified a signal concerning 55 patients in association with bromazepam and hydroxyzine (ORu 1.66; 95 % CI 1.23-2.23). This signal was not confirmed after a chart review. Even so, AKI and co-prescription were confirmed for 96 % (95 % CI 88-99) and 88 % (95 % CI 76-94) of these patients, respectively. CONCLUSION: Data mining techniques on CDW can foster the detection of adverse drug reactions when drugs are used alone or in combination.