Coordination and Referral Platforms for Children Aged 7 and Under with Suspected Neurodevelopmental Outcome: What Data for Research on Care and Developmental Trajectories?

At the core of the child's medical, social, and educational pathway, coordination and referral platforms (CRPs) for neurodevelopmental disorders (NDDs) have been gradually deployed in France since 2018 and support the early detection of NDDs in children. The 112 nationwide CRPs do not benefit from a common electronic health record system. Our aim was to propose an HER model for CRP to enable real-life data reuse, optimize care pathway management and conduct pre-screening for research. CRP data were collected (n=34) into an application enriched by a NLP tool extracting standardized scales for NDDs assessments from medical and paramedical professionals. NLP tool evaluation presented a precision of 86.4% and recall of 90.5%. CRP support was provided to 195 children included between 1 September 2022 and 31 August 2023, aged 4 years, with a sex ratio of 2.8, with delays reported in language (75%) and concerned by global developmental delays (16%). Children's ND phenotype and care pathway description could be automated by a harmonized and structured EHR. While many clinical situations are at an impasse, real-life data-driven evidence is particularly relevant in the context of NDDs, where early intervention plays such a key role in children's development and prognosis. A harmonized and enriched CRP database could benefit both individual and public health levels with pathway monitoring, intervention proposals and research pre-screenings.

Performance of a NLP Tool for Text Classification from Orthopaedic Operative Reports, Using Data from the Large Network of Clinical Data Warehouses of the West of France: The HACRO-HUGORTHO Project.

Electronic health data concerning implantable medical devices (IMD) opens opportunities for dynamic real-world monitoring to assess associated risks related to implanted materials. Due to population ageing and expanding demands, total hip, knee, and shoulder arthroplasties are increasing. Automating the collection and analysis of orthopedic device features could benefit physicians and public health policies enabling early issue detection, IMD monitoring and patient safety assessment. A machine learning tool using natural language processing (NLP) was developed for the automated extraction of operation information from medical reports in orthopedics. A corpus of 959 orthopaedic operative reports from 5 centres was manually annotated using the Prodigy software® with a strong inter-annotator agreement of 0.80. Data to extract concerned key clinical and procedure information (n= 9) selected by a multidisciplinary group based on the French health authority checklist. Performances parameters of the NLP model estimated an overall strong precision and recall of respectively 97.0 and 96.0 with a F1-score 96.3. Systematic monitoring of orthopedic devices could be ensured by an automated tool, leveraging clinical data warehouses. Traceability of medical devices with implantation modalities will allow detection of implant factors leading to complications. The evidence from real-world data could provide concrete and dynamic insights to surgeons and infectious disease specialists concerning implant follow-up, guiding therapeutic decision-making, and informing public health policymakers. The tool will be applied on clinical data warehouses to automate information extraction and presentation, providing feedback on mandatory information completion and contents of operative reports to support improvements, and thereafter implant research projects.

Prescreening in oncology trials using medical records. Natural language processing applied on lung cancer multidisciplinary team meeting reports.

Defining profiles of patients that could benefit from relevant anti-cancer treatments is essential. An increasing number of specific criteria are necessary to be eligible to specific anti-cancer therapies. This study aimed to develop an automated algorithm able to detect patient and tumor characteristics to reduce the time-consuming prescreening for trial inclusions without delay. Hence, 640 anonymized multidisciplinary team meetings (MTM) reports concerning lung cancers from one French teaching hospital data warehouse between 2018 and 2020 were annotated. To automate the extraction of eight major eligibility criteria, corresponding to 52 classes, regular expressions were implemented. The RegEx's evaluation gave a F1-score of 93% in average, a positive predictive value (precision) of 98% and sensitivity (recall) of 92%. However, in MTM, fill rates variabilities among patient and tumor information remained important (from 31% to 100%). Genetic mutations and rearrangement test results were the least reported characteristics and also the hardest to automatically extract. To ease prescreening in clinical trials, the PreScIOUs study demonstrated the additional value of rule based and machine learning based methods applied on lung cancer MTM reports.

Management of death-related noncompleters in cluster randomized trials carried out in nursing homes: a methodological review.

BACKGROUND: A noncompleter is defined as a participant who leaves a trial before the end of the planned follow-up. Research in nursing homes is highly exposed to this problem because of high death rates. OBJECTIVES: The aim of this trial is to assess the statistical management of noncompleters in cluster randomized trials carried out in nursing homes. STUDY DESIGN AND SETTING: A methodological review of published cluster randomized trials. RESULTS: We selected 37 articles. For 22 (59%) trials, the design was closed-cohort (i.e., participants included all at the same time when randomizing clusters). In those 22 closed-cohort trials, the median follow-up was 6.5 months (interquartile range 4-12). The median noncompleter rate was 19.5% and the median noncompletions due to death was 73.2%. In only one trial were the baseline characteristics of completers and noncompleters compared. Strategies to deal with noncompleters were an inflation of the planned sample size (11 trials), the use of repeated measurements of the outcome (12 trials), and the use of imputation methods when analyzing data (7 trials). CONCLUSION: In cluster randomized trials of nursing homes, noncompleters are managed as for any missing data, but they are essentially due to death. Methodological and statistical developments and guidance are needed.

Pilot Study of an e-Cohort to Monitor Adverse Event for Patient with Hip Prostheses from Clinical Data Warehouse.

Hip arthroplasty represents a large proportion of orthopaedic activity, constantly increasing. Automating monitoring from clinical data warehouses is an opportunity to dynamically monitor devices and patient outcomes allowing improve clinical practices. Our objective was to assess quantitative and qualitative concordance between claim data and device supply data in order to create an e-cohort of patients undergoing a hip replacement. We performed a single-centre cohort pilot study, from one clinical data warehouse of a French University Hospital, from January 1, 2010 to December 31, 2019. We included all adult patients undergoing a hip arthroplasty, and with at least one hip medical device provided. Patients younger than 18 years or opposed to the reuse of their data were excluded from the analysis. Our primary outcome was the percentage of hospital stays with both hip arthroplasty and hip device provided. The patient and stay characteristics assessed in this study were: age, sex, length of stay, surgery procedure (replacement, repositioning, change, or reconstruction), medical motif for surgery (osteoarthritis, fracture, cancer, infection, or other) and device provided (head, stem, shell, or other). We found 3,380 stays and 2,934 patients, 96.4% of them had both a hip surgery procedure and a hip device provided. These data from different sources are close enough to be integrated in a common clinical data warehouse.

Prescreening in Oncology Using Data Sciences: The PreScIOUS Study.

The development of precision medicine in oncology to define profiles of patients who could benefit from specific and relevant anti-cancer therapies is essential. An increasing number of specific eligibility criteria are necessary to be eligible to targeted therapies. This study aimed to develop an automated algorithm based on natural language processing to detect patients and tumor characteristics to reduce the time-consuming prescreening for trial inclusions. Hence, 640 anonymized multidisciplinary team meeting (MTM) reports concerning lung cancer were extracted from one teaching hospital data warehouse in France and annotated. To automate the extraction of 52 bioclinical information corresponding to 8 major eligibility criteria, regular expressions were implemented and evaluated. The performance parameters were satisfying: macroaverage F1-score 93%; rates reached 98% for precision and 92% for recall. In MTM, fill rates variabilities among patients and tumors information remained important (from 31.4% to 100%). The least reported characteristics and the most difficult to automatically collect were genetic mutations and rearrangement test results.