Electronic Medical Records Research Articles

Age-stratified association between preconception body mass index and risk of macrosomia at delivery

Objective: To investigate the impact of preconception body mass index (BMI) on neonatal birth weight and the risk of macrosomia in pregnant women across various age groups. Methods: A cohort study was conducted, selecting pregnant women who underwent their initial prenatal assessment at Beijing Obstetrics and Gynecology Hospital from September 1st, 2018 to March 31st, 2020. Relevant data were collected from the hospital's electronic medical record system. Logistic regression nested cubic spline was used to analyze the nonlinear association between preconception BMI and neonatal birth weight. Binary logistic regression was also employed to assess the association between preconception BMI and macrosomia risk. Results: (1) A total of 13 015 pregnant women were examined, revealing an incidence of macrosomia of 6.33% (824/13 015). The preconception BMI of pregnant women in the macrosomia group was significantly higher than that in the non-macrosomia group [(23.1±3.4) vs (21.6±3.1) kg/m2], and the age was significantly higher than that in the non-macrosomia group [(32.1±3.6) vs (31.7±3.7) years], the differences were statistically significant (all P<0.05). (2) Preconception BMI was positively correlated with neonatal birth weight. Pregnant women with preconception BMI of 15.0 kg/m2, 20.0 kg/m2, and 25.0 kg/m2 had decreased birth weight of 121 g (95%CI: 35-183 g) and increased birth weights of 78 g (95%CI: 54-102 g) and 182 g (95%CI: 151-213 g), respectively, compared to those with a preconception BMI of 18.0 kg/m2. (3) For each 1.0 kg/m2 increase in preconception BMI, the risk of macrosomia increased by 14% (OR=1.14, 95%CI: 1.11-1.16; P<0.001). When stratified by age, it was observed that elevated preconception BMI significantly increased the incidence of macrosomia in women aged 27-38 years. Among them, the risk of delivering macrosomia among 37 years old pregnant women was most affected by preconception BMI (OR=1.33, 95%CI: 1.17-1.51; P<0.001). (4) The stability and sensitivity analysis results showed that the preconception BMI of pregnant women with a preconception BMI of 18.0-<25.0 kg/m2 had a significant impact on the risk of macrosomia (OR=1.23, 95%CI: 1.17-1.29; P<0.001), while the preconception BMI of other preconception BMI stratification pregnant women had no significant impact on the risk of macrosomia (all P>0.05). Hypertension disorders in pregnancy, gestational diabetes mellitus and abnormal blood lipid during pregnancy were not the mediators associated with preconception BMI and macrosomia. After excluding three factors respectively, the impact of preconception BMI on the risk of macrosomia was the same as before (OR=1.14, 95%CI: 1.11-1.16; P<0.001). Conclusions: Preconception BMI is linked to neonatal birth weight and the risk of macrosomia, which is influenced by the pregnant woman's age. Both factors should be considered when evaluating the risk of macrosomia in clinical practice.

Application of an Externally Developed Algorithm to Identify Research Cases and Controls from Electronic Health Record Data: Failures and Successes.

The use of Electronic Health Records (EHRs) in research demands robust, interoperable systems. By linking biorepositories to EHR algorithms, researchers can efficiently identify cases and controls for large observational studies (e.g., Genome-Wide Association Studies (GWAS)). This is critical for ensuring efficient and cost-effective research. However, the lack of standardized metadata and algorithms across different EHRs complicates their sharing and application. Our study presents an example of a successful implementation and validation process. To implement and validate a rule-based algorithm from a tertiary medical center in Tennessee to classify cases and controls from a research study on rotator cuff tear nested within a tertiary medical center in North Texas and to assess the algorithm's performance. We applied a phenotypic algorithm (designed and validated in a tertiary medical center in Tennessee) using EHR data from 492 patients enrolled in case-control study recruited from a tertiary medical center in North Texas. The algorithm leveraged ICD (International Classification of Diseases) and CPT (Current Procedural Terminology) codes to identify case and control status for degenerative rotator cuff tears. A manual review was conducted to compare the algorithm's classification with a previously recorded gold standard documented by clinical researchers. Initially the algorithm identified 398 (80.9%) patients correctly as cases or controls. After fine-tunning and corrections of errors in our gold standard dataset, we calculated a sensitivity of 0.94 and specificity of 0.76. The implementation of the algorithm presented challenges due to the variability in coding practices between medical centers. To enhance performance, we refined the algorithm's data dictionary by incorporating additional codes. The process highlighted the need for meticulous code verification and standardization in multi-center studies. Sharing case-control algorithms boosts EHR research. Our rule-based algorithm improved multi-site patient identification and revealed 12 data entry errors, helping validate our results.

Lipoprotein(a) Atherosclerotic Cardiovascular Disease Risk Score Development and Prediction in Primary Prevention From Real-World Data.

Lp(a; Lipoprotein[a]) is a predictor of atherosclerotic cardiovascular disease (ASCVD); however, there are few algorithms incorporating Lp(a), especially from real-world settings. We developed an electronic health record (EHR)-based risk prediction algorithm including Lp(a). Utilizing a large EHR database, we categorized Lp(a) cut points at 25, 50, and 75 mg/dL and constructed 10-year ASCVD risk prediction models incorporating Lp(a), with external validation in a pooled cohort of 4 US prospective studies. Net reclassification improvement was determined among borderline-intermediate risk patients. We included 5902 patients aged ≥18 years (mean age 48.7±16.7 years, 51.2% women, and 7.7% Black). Our EHR model included Lp(a), age, sex, Black race/ethnicity, systolic blood pressure, total and high-density lipoprotein cholesterol, diabetes, smoking, and hypertension medication. Over a mean follow-up of 6.8 years, ASCVD event rates (per 1000 people-years) ranged from 8.7 to 16.7 across Lp(a) groups. A 25 mg/dL increment in Lp(a) was associated with an adjusted hazard ratio of 1.23 (95% CI, 1.10-1.37) for composite ASCVD. Those with Lp(a) ≥75 mg/dL had an 88% higher risk of ASCVD (hazard ratio, 1.88 [95% CI, 1.30-2.70]) and more than double the risk of incident stroke (hazard ratio, 2.55 [95% CI, 1.54-4.23]). C-statistics for our EHR and EHR+Lp(a) models in our EHR training data set were 0.7475 and 0.7556, respectively, with external validation in our pooled cohort (n=21 864) of 0.7350 and 0.7368, respectively. Among those at borderline/intermediate risk, the net reclassification improvement was 21.3%. We show the feasibility of developing an improved ASCVD risk prediction model incorporating Lp(a) based on a real-world adult clinic population. The inclusion of Lp(a) in ASCVD prediction models can reclassify risk in patients who may benefit from more intensified ASCVD prevention efforts.

Efficacy of a clinical decision support tool to promote guideline concordant evaluations in patients with high-risk microscopic hematuria: a cluster randomized quality improvement project.

We aimed to determine whether implementation of clinical decision support (CDS) tool integrated into the electronic health record (EHR) of a multi-site academic medical center increased the proportion of patients with American Urological Association (AUA) "high risk" microscopic hematuria (MH) who receive guideline concordant evaluations. We conducted a two-arm cluster randomized quality improvement project in which 202 ambulatory sites from a large health system were randomized to either have their physicians receive at time of test results an automated CDS alert for patients with 'high-risk' MH with associated recommendations for imaging and cystoscopy (intervention) or usual care (control). Primary outcome was met if a patient underwent both imaging and cystoscopy within 180 days from MH result. Secondary outcomes assessed individual completion of imaging, cystoscopy or placement of imaging orders. There were 917 patients randomized to intervention (n=476) or control (n=441) arms between October-December 2021. The percentage of eligible patients for whom the alert correctly triggered in the intervention arm was 83%. Primary outcome was achieved in 0.6% vs. 0.9% (RR: 0.69; 95% CI 0.15, 3.10) of patients in the intervention and control arms, respectively. Patients in the intervention and control groups had similar rates of completed imaging (17.7% vs. 14.7%) and cystoscopy (1.5% vs. 0.9%). Those in the intervention arm had a higher likelihood of CT urogram order (5.5% vs. 1.1%, p=0.003) and a non-significant increase in urology evaluation (11.1% vs. 7.5%, p=0.09). Implementing an EHR-integrated CDS tool to promote evaluation of patients with high-risk MH did not lead to improvements in patient completion of a full guideline-concordant evaluation. The development of an algorithm to trigger a CDS alert was demonstrated to be feasible and effective. Further multi-level assessment of barriers to evaluation are necessary to continue to improve the approach to evaluating high risk patients with MH.

Association of an HIV-Prediction Model with Uptake of Pre-Exposure Prophylaxis (PrEP).

Global efforts aimed at ending human immunodeficiency virus (HIV) incidence have adapted and evolved since the turn of the century. The utilization of machine learning incorporated into an electronic health record (EHR) can be refined into prediction models that identify when an individual is at greater HIV infection risk. This can create a novel and innovative approach to identifying patients eligible for preventative therapy. This study's aim was to evaluate the effectiveness of an HIV prediction model in clinical workflows. Outcomes included pre-exposure prophylaxis (PrEP) prescriptions generated and the model's ability to identify eligible patients. A prediction model was developed and implemented at the safety-net hospital in Dallas County. Patients seen in primary care clinics were evaluated between July 2020 to June 2022. The prediction model was incorporated into an existing best practice advisory (BPAs) used to identify potentially eligible PrEP patients. The prior, basic BPA (bBPA) displayed if a prior sexually transmitted infection was documented and the enhanced BPA (eBPA) incorporated the HIV prediction model. A total of 3,218 unique patients received the BPA during the study time period, with 2,346 ultimately included for evaluation. There were 678 patients in the bBPA group and 1,666 in the eBPA group. PrEP prescriptions generated increased in the post-implementation group within the 90-day follow-up period (bBPA:1.48 v. eBPA:3.67 prescriptions per month, p<0.001). Patient demographics also differed between groups, resulting in a higher median age (bBPA:36[IQR 24] v. eBPA:52[QR 19] years, p<0.001) and an even distribution between birth sex in the post-implementation group (female sex at birth bBPA:62.2% v. eBPA:50.2%, p=<0.001). The implementation of a HIV prediction model yielded a higher number of PrEP prescriptions generated and was associated with the identification of twice the number of potentially eligible patients.

Optimizing laboratory medication safety monitoring in patients with juvenile idiopathic arthritis to advance value-based care.

Most children with juvenile idiopathic arthritis (JIA) are treated with medications that require safety monitoring labs. Recommended testing includes a creatinine level. However, 87.5% of our visits had a basic metabolic panel (BMP) done. The difference in charges for ordering a serum creatinine versus a BMP is $31, and the difference in matched net revenue per unit is $12.09. Our aim was to increase the ordering of a creatinine rather than a BMP in patients with non-systemic JIA needing methotrexate, leflunomide, and/or biologic medication safety lab monitoring in the Rheumatology Clinic from 12.5 to 80% in 8months. We utilized quality improvement tools, incorporated this project in our pre-visit planning process, provided physician education, and made updates to our electronic medical record (EMR) order sets. We tracked the percent of patients with JIA taking the pertinent medications who had a creatinine checked rather than a BMP on a run chart to review performance over time. Our baseline median was 12.5% of visits having a creatinine alone checked instead of a BMP (Fig.1). We had two shifts in our data with a final center line of 99%. There were no unintended consequences noted during our project. Our project optimized medication safety monitoring in patients with JIA by eliminating unnecessary tests to save costs and advance value-based care. Our interventions of education and EMR modifications allowed for standardization of laboratory test ordering. We plan to expand this project to other medications and other diseases to further decrease costs without sacrificing patient safety. Key Points • Most of our patients with JIA that require medication safety monitoring had a basic metabolic panel obtained rather than the recommended serum creatinine. • We significantly increased ordering of a clinically recommended and more cost-effective serum creatinine alone for screening of renal function through quality improvement methodology, physician education, pre-visit planning, and electronic medical record modifications.

Consumer Access and Utilization of Patient Portals for Electronic Health Records: A Cross-Sectional Study in Saudi Arabia.

Consumer Access and Utilization of Patient Portals for Electronic Health Records: A Cross-Sectional Study in Saudi Arabia.

The impact of imprecise case definitions in electronic health record research: a melanoma case-study from the Million Veteran Program.

Cases for a disease can be defined broadly using diagnostic codes, or narrowly using gold-standard confirmation that often is not available in large administrative datasets. These different definitions can have significant impacts on the results and conclusions of studies. We conducted this study to assess how using melanoma phecodes versus histologic confirmation for invasive or in situ melanoma impacts the results of a genome-wide association study (GWAS) using the Million Veteran Program. Melanoma status was determined three ways: (1) by the presence of two or more phecodes, (2) histologically-confirmed invasive melanoma, and (3) histologically-confirmed melanoma in situ. We conducted a GWAS for variants with minor allele frequencies of 1% or greater. There were 45,665 cases in the phecode cohort, 5364 cases in the confirmed invasive melanoma cohort, and 4792 cases in the confirmed melanoma in situ cohort. There were 20,457 variants significant at the genome-wide level in the phecode cohort, 2582 in the invasive melanoma cohort, and 1989 in the melanoma in situ cohort. Most of the variants identified in the phecode cohort did not replicate in the histologically-confirmed cohorts. The different case definitions led to large differences in sample size and variants associated at the genome-wide level. Unvalidated and imprecise case definitions can lead to less accurate results. Investigators should use validated phenotypes when gold-standard definitions are not available.

ARCH: Large-scale knowledge graph via aggregated narrative codified health records analysis.

Electronic health record (EHR) systems contain a wealth of clinical data stored as both codified data and free-text narrative notes (NLP). The complexity of EHR presents challenges in feature representation, information extraction, and uncertainty quantification. To address these challenges, we proposed an efficient Aggregated naRrative Codified Health (ARCH) records analysis to generate a large-scale knowledge graph (KG) for a comprehensive set of EHR codified and narrative features. Using data from 12.5 million Veterans Affairs patients, ARCH first derives embedding vectors and generates similarities along with associated p-values to measure the strength of relatedness between clinical features with statistical certainty quantification. Next, ARCH performs a sparse embedding regression to remove indirect linkage between features to build a sparse KG. Finally, ARCH was validated on various clinical tasks, including detecting known relationships between entity pairs, predicting drug side effects, disease phenotyping, as well as sub-typing Alzheimer's disease patients. ARCH produces high-quality clinical embeddings and KG for over 60,000 codified and narrative EHR concepts. The KG and embeddings are visualized in the R-shiny powered web-API3. ARCH achieved high accuracy in detecting EHR concept relationships, with AUCs of 0.926 (codified) and 0.861 (NLP) for similar EHR concepts, and 0.810 (codified) and 0.843 (NLP) for related pairs. It detected drug side effects with a 0.723 AUC, which improved to 0.826 after fine-tuning. Using both codified and NLP features, the detection power increased significantly. Compared to other methods, ARCH has superior accuracy and enhances weakly supervised phenotyping algorithms' performance. Notably, it successfully categorized Alzheimer's patients into two subgroups with varying mortality rates. The proposed ARCH algorithm generates large-scale high-quality semantic representations and knowledge graph for both codified and NLP EHR features, useful for a wide range of predictive modeling tasks.

How Often Is Rifampin Therapy Initiated and Completed in Patients With Periprosthetic Joint Infections?

Rifampin therapy is indicated for the treatment of staphylococcal periprosthetic joint infection (PJI) in patients who have undergone debridement, antibiotics, and implant retention (DAIR) or one-stage revision as per the Infectious Diseases Society of America (IDSA) guideline. Given the well-established effectiveness of rifampin as adjunctive therapy in staphylococcal PJI, it is crucial to evaluate its utilization in practice and identify factors that contribute to its underuse or incomplete administration, as these deviations may undermine treatment efficacy and patient outcomes. Among patients who met clear indications for rifampin use having undergone DAIR or one-stage revision for staphylococcal PJI, (1) what proportion of patients did not receive it? (2) What proportion of patients started it but did not complete the planned course? (3) Where documented in the medical record, what were the common reasons for not using it or prematurely discontinuing it, and in what percentage of the patients' charts was no reason given? (4) What proportion of patients were taking a medication that put them at risk for a drug-drug interaction (DDI)? Using an institutional database, patients who underwent DAIR or revision arthroplasty for PJI from January 2013 to April 2023 were identified (n = 935). We defined clear indications for rifampin use as a patient diagnosed with staphylococcal PJI treated via DAIR or one-stage revision, as set forth by the clinical practice guidelines of the IDSA. Based on those indications, we surveyed patients' electronic medical records (EMRs) and calculated the proportion of patients who did not receive the drug in situations where it would have seemed appropriate that they receive it. We then determined the planned length of antibiotic management and calculated the proportion of patients whose planned courses of rifampin were not completed based on what was documented in each patient's chart. Next, we performed a review of the EMR and recorded qualitatively the reasons why patients either did not receive rifampin or why the planned course was not completed, as well as the percentage of patients whose medical records did not indicate why the drug was not used or was discontinued before the planned course had been administered. Last, we retrospectively reviewed the medications that the patients were taking at the time of indicated adjunctive rifampin initiation and evaluated whether there were DDIs and the potential severity of those interactions. A total of 9% (87 of 935) of patients who underwent DAIR or revision arthroplasty for PJI met IDSA criteria, suggesting that rifampin use would have been appropriate. Of those meeting IDSA criteria, 49% (43 of 87) started rifampin. Of those who started on rifampin for staphylococcal PJI, 19% (8 of 43) discontinued it before the planned 6-week course was complete as documented in the EMR. Among the patients who did not receive rifampin when it appeared that they ought to have based on IDSA criteria, no reason for this decision was documented in 70% (31 of 44); among those for whom a reason was documented, it was DDI in 8 of 13, concern for hepatotoxicity in 1 of 13, history of a side effect in 1 of 13, and other reasons in 3 of 13. Among the patients whose courses were terminated before the planned course was completed, it was a gastrointestinal issue in 5 of 8, acute kidney injury in 1 of 8, or other reasons in 2 of 8. Only about one-half of patients who met IDSA criteria for adjunctive rifampin therapy for staphylococcal PJI after DAIR or one-stage revision at a large tertiary referral academic center were started on treatment. For most patients who did not receive rifampin, there is no documentation for why not. Given the evidence of superior outcomes for patients treated with adjunctive rifampin, it is imperative that all patients meeting criteria be started on the medication unless there is a clear contraindication, which should be documented. Future studies should further evaluate the clinical benefit of rifampin when treating staphylococcal PJI and which DDIs would be considered true contraindications of treatment. This could be achieved with a prospective study that evaluates successful long-term infection-free survival after treatment with adjunctive rifampin while accounting for and stratifying the potential clinical risk of these DDIs. Level IV, therapeutic study.

The Influence of Structural Factors and Telemedicine on Missed Appointments Among Medicaid-Insured Patients.

Missed clinic appointments disproportionately affect Medicaid-insured patients and residents of socioeconomically deprived neighborhoods. The role of the recent telemedicine expansion in reducing these disparities is unclear. We analyzed the relationship between census tract (CT) poverty level, residential segregation, missed appointments, and the role of telemedicine. This retrospective cohort study used electronic health records (EHR) data merged with population datasets and restricted to adult patients (≥ 18years) with completed or missed internal or family medicine outpatient clinic visits (03/2020-12/2022). Using generalized estimating equations, we analyzed the association between missed appointments, CT poverty level, CT residential segregation (operationalized using the isolation index), and appointment modality. Sample size was 125,229 appointments for 68,471 unique patients (Hispanic [46.6%], White [9.4%], Black [18.7%], Asian [2.6%], Native American [1.9%], and "other race/ethnicity" [0.8%]; 18-39years [38.2%], 40-64years [56.4%], and > = 65years [5.36]). There was an increased likelihood of missed appointments with increasing Hispanic isolation index (OR 1.12; 95% CI 1.03, 1.23) and decreased likelihood with increasing Black isolation index (OR 0.71; 95% CI 0.61, 0.83). The protective effect of telemedicine on missed appointments decreased with increasing Black and Hispanic isolation index. Our results suggest that further studies are needed to understand the potential impact of telemedicine on healthcare use inequities among residents of residentially segregated areas. Targeted interventions that aim to identify and address structural factors that could limit the benefits of telemedicine use are also needed.

Surgical Scheduling Errors During Manual Data Transfer.

Retrospective studies examining errors within a surgical scheduling setting do not fully represent the effects of human error involved in transcribing critical patient health information (PHI). These errors can negatively impact patient care and reduce workplace efficiency due to insurance claim denials and potential sentinel events. Previous reports underscore the burden physicians face with prior authorizations which may lead to serious adverse events or the abandonment of treatment due to these delays. This study simulates the process of PHI transfer during surgical scheduling to examine the error rate of experienced schedulers when manually transferring PHI from surgical forms into electronic health records (EHR). Participants (n=50) manually input PHI from four surgical scheduling forms into a simulated EHR form. Eight critical data points were identified and defined as data that delay claim approvals and payments. Subjects were randomly assigned to either a control (18minutes) or experimental (10minutes) group. Transcription errors were flagged to measure the percentage of incorrectly inputted data fields. Two-tailed t-tests were used to determine statistical significance (P <.05). 100% of subjects in both cohorts had at least one or more errors in every form. The 10-minute cohort had a higher average "critical errors" rate than the 18-minute cohort (P =.03). Of the 200 forms completed, 171 forms contained 1 or more "critical errors," resulting in a potential 85.5% delay or denial in authorization or payments. The highest incidence of critical errors across all fields occurred with ICD-10 codes, CPT codes, authorization number, procedure, and insurance ID number. As critical errors fields of authorization number and insurance ID often lead to automatic denials, not only are they more susceptible to transcription error due to alphanumeric values but more indicative of delays in treatment. These findings reveal a clear "pain point" in the routine scheduling process that leads to authorization and payment denials. With various touch points of manual data transfer in surgical scheduling, data degradation due to human error may compound at each step. Health care institutions should consider adopting digital solutions and investing in training programs to optimize clinical practice efficiency and reduce the possibility of inaccurate manual PHI transfer. Future case studies on denied payments will help further elucidate the economic impact on practices, as well as inform strategic decisions by those who directly handle health care management.

Random Survival Forest Machine Learning for the Prediction of Cardiovascular Events Among Patients With a Measured Lipoprotein(a) Level: A Model Development Study.

Established risk models may not be applicable to patients at higher cardiovascular risk with a measured Lp(a) (lipoprotein[a]) level, a causal risk factor for atherosclerotic cardiovascular disease. This was a model development study. The data source was the Nashville Biosciences Lp(a) data set, which includes clinical data from the Vanderbilt University Health System. We included patients with an Lp(a) measured between 1989 and 2022 and who had at least 1 year of electronic health record data before measurement of an Lp(a) level. The end point of interest was time to first myocardial infarction, stroke/TIA, or coronary revascularization. A random survival forest model was derived and compared with a Cox proportional hazards model derived from traditional cardiovascular risk factors (ie, the variables used to estimate the Pooled Cohort Equations for the primary prevention population and the variables used to estimate the Second Manifestations of Arterial Disease and Thrombolysis in Myocardial Infarction Risk Score for Secondary Prevention scores for the secondary prevention population). Model discrimination was evaluated using Harrell C-index. A total of 4369 patients were included in the study (49.5% were female, mean age was 51 [SD, 18] years, and mean Lp(a) level was 33.6 [38.6] mg/dL, of whom 23.7% had a prior cardiovascular event). The random survival forest model outperformed the traditional risk factor models in the test set (c-index, 0.82 [random forest] versus 0.69 [primary prevention] versus 0.80 [secondary prevention]). These results were similar when restricted to a primary prevention population and under various strategies to handle competing risk. A Cox proportional hazard model based on the top 25 variables from the random forest model had a c-index of 0.80. A random survival forest model outperformed a model using traditional risk factors for predicting cardiovascular events in patients with a measured Lp(a) level.

Informing ICU Digital Twins: Dynamic Assessment of Cardiorespiratory Failure Trajectories in Patients with Sepsis.

Understanding clinical trajectories of sepsis patients is crucial for prognostication, resource planning, and to inform digital twin models of critical illness. This study aims to identify common clinical trajectories based on dynamic assessment of cardiorespiratory support using a validated electronic health record data that covers retrospective cohort of 19,177 patients with sepsis admitted to ICUs of Mayo Clinic Hospitals over eight-year period. Patient trajectories were modeled from ICU admission up to 14 days using an unsupervised machine learning two-stage clustering method based on cardiorespiratory support in ICU and hospital discharge status. Of 19,177 patients, 42% were female with a median age of 65 (IQR, 55-76) years, APACHE III score of 70 (IQR, 56-87), hospital length of stay (LOS) of 7 (IQR, 4-12) days, and ICU LOS of 2 (IQR, 1-4) days. Four distinct trajectories were identified: fast recovery (27% with a mortality rate of 3.5% and median hospital LOS of 3 (IQR, 2-15) days), slow recovery (62% with a mortality rate of 3.6% and hospital LOS of 8 (IQR, 6-13) days), fast decline (4% with a mortality rate of 99.7% and hospital LOS of 1 (IQR, 0-1) day), and delayed decline (7% with a mortality rate of 97.9% and hospital LOS of 5 (IQR, 3-8) days). Distinct trajectories remained robust and were distinguished by Charlston comorbidity index, Apache III scores, day 1 and day 3 SOFA (p < 0.001 ANOVA). These findings provide a foundation for developing prediction models and digital twin decision support tools, improving both shared decision-making and resource planning.

Longitudinal evolution of the transdiagnostic prodrome to severe mental disorders: a dynamic temporal network analysis informed by natural language processing and electronic health records.

Modelling the prodrome to severe mental disorders (SMD), including unipolar mood disorders (UMD), bipolar mood disorders (BMD) and psychotic disorders (PSY), should consider both the evolution and interactions of symptoms and substance use (prodromal features) over time. Temporal network analysis can detect causal dependence between and within prodromal features by representing prodromal features as nodes, with their connections (edges) indicating the likelihood of one feature preceding the other. In SMD, node centrality could reveal insights into important prodromal features and potential intervention targets. Community analysis can identify commonly occurring feature groups to define SMD at-risk states. This retrospective (2-year) cohort study aimed to develop a global transdiagnostic SMD network of the temporal relationships between prodromal features and to examine within-group differences with sub-networks specific to UMD, BMD and PSY. Electronic health records (EHRs) from South London and Maudsley(SLaM) NHS Foundation Trust were included from 6462 individuals with SMD diagnoses (UMD:2066; BMD:740; PSY:3656). Validated natural language processing algorithms extracted the occurrence of 61 prodromal features every three months from two years to six months before SMD onset. Temporal networks of prodromal features were constructed using generalised vector autoregression panel analysis, adjusting for covariates. Edge weights (partial directed correlation coefficients, z) were reported in autocorrelative, unidirectional and bidirectional relationships. Centrality was calculated as the sum of (non-autoregressive) connections leaving (out-centrality, cout) or entering (in-centrality, cin) a node. The three sub-networks (UMD, BMD, PSY) were compared using permutation analysis, and community analysis was performed using Spinglass. The SMD network revealed strong autocorrelations (0.04 ≤ z ≤ 0.10), predominantly positive connections, and identified aggression (cout = 0.103) and tearfulness (cin = 0.134) as the most central features. Sub-networks for UMD, BMD, and PSY showed minimal differences, with 3.5% of edges differing between UMD and PSY, 0.8% between UMD and BMD, and 0.4% between BMD and PSY. Community analysis identified one positive psychotic community (delusional thinking-hallucinations-paranoia) and two behavioural communities (aggression-cannabis use-cocaine use-hostility, aggression-agitation-hostility) as the most common. This study represents the most extensive temporal network analysis conducted on the longitudinal interplay of SMD prodromal features. The findings provide further evidence to support transdiagnostic early detection services across SMD, refine assessments to detect individuals at risk and identify central features as potential intervention targets.

Association of COVID-19 Vaccinations with Osteoarthritis Flares: A Case-Crossover Study.

People with osteoarthritis (OA) commonly experience flares. Whether COVID-19 vaccination triggers OA flares is unknown. Adults with OA enrolled in a COVID-19 Rheumatology Registry were invited to participate in a case-crossover study. Vaccine data were ascertained from self-report and electronic health records (EHR). OA was identified using coding algorithms and validated via EHR. Participants reported flare and non-flare periods. Vaccine exposures in the 2-, 7-, and 14-day "lookback windows" prior to OA flares were compared to vaccine exposures during similar non-flare ("control") periods. 279 participants had validated OA, and 136 (49%) contributed at least one flare and one control period. Mean age was 68 years [SD ±8], 82% female, 87% White, 62% knee OA, 56% hip OA, 37% hand/wrist OA and 60% had >= one anatomic location of OA. 525 COVID-19 vaccine doses were recorded, and participants reported 374 OA flares: 30% were mild, 55% were moderate, and 14% were severe. OA flares were not associated with COVID-19 vaccination 2 or 7 days prior (OR 0.69 [95% CI: 0.28, 1.66], OR 0.54 [95% CI: 0.27, 1.07], respectively). In the 14-day lookback window, fewer flares occurred after vaccination (OR 0.57 [95% CI: 0.34, 0.97], p=0.039). Analyses stratified on sex, age, knee or hand OA, vaccine brand, and dose showed no increased association between COVID-19 vaccination and OA flares. The lack of positive association between COVID-19 vaccination and OA flare in any primary or secondary analysis provides reassurance regarding the use of COVID-19 vaccines in people with OA.

Combining a Risk Factor Score Designed From Electronic Health Records With a Digital Cytology Image Scoring System to Improve Bladder Cancer Detection: Proof-of-Concept Study.

To reduce the mortality related to bladder cancer, efforts need to be concentrated on early detection of the disease for more effective therapeutic intervention. Strong risk factors (eg, smoking status, age, professional exposure) have been identified, and some diagnostic tools (eg, by way of cystoscopy) have been proposed. However, to date, no fully satisfactory (noninvasive, inexpensive, high-performance) solution for widespread deployment has been proposed. Some new models based on cytology image classification were recently developed and bring good perspectives, but there are still avenues to explore to improve their performance. Our team aimed to evaluate the benefit of combining the reuse of massive clinical data to build a risk factor model and a digital cytology image-based model (VisioCyt) for bladder cancer detection. The first step relied on designing a predictive model based on clinical data (ie, risk factors identified in the literature) extracted from the clinical data warehouse of the Rennes Hospital and machine learning algorithms (logistic regression, random forest, and support vector machine). It provides a score corresponding to the risk of developing bladder cancer based on the patient's clinical profile. Second, we investigated 3 strategies (logistic regression, decision tree, and a custom strategy based on score interpretation) to combine the model's score with the score from an image-based model to produce a robust bladder cancer scoring system. We collected 2 data sets. The first set, including clinical data for 5422 patients extracted from the clinical data warehouse, was used to design the risk factor-based model. The second set was used to measure the models' performances and was composed of data for 620 patients from a clinical trial for which cytology images and clinicobiological features were collected. With this second data set, the combination of both models obtained areas under the curve of 0.82 on the training set and 0.83 on the test set, demonstrating the value of combining risk factor-based and image-based models. This combination offers a higher associated risk of cancer than VisioCyt alone for all classes, especially for low-grade bladder cancer. These results demonstrate the value of combining clinical and biological information, especially to improve detection of low-grade bladder cancer. Some improvements will need to be made to the automatic extraction of clinical features to make the risk factor-based model more robust. However, as of now, the results support the assumption that this type of approach will be of benefit to patients.

Automated extraction of post-stroke functional outcomes from unstructured electronic health records.

Population level tracking of post-stroke functional outcomes is critical to guide interventions that reduce the burden of stroke-related disability. However, functional outcomes are often missing or documented in unstructured notes. We developed a natural language processing (NLP) model that reads electronic health records (EHR) notes to automatically determine the modified Rankin Scale (mRS). We included consecutive patients (⩾18 years) with acute stroke admitted to our center (2015-2024). mRS scores were obtained from the Get With the Guidelines registry and clinical notes (if documented), and used as the gold standard to compare against NLP-generated scores. We used text-based features from notes, along with age, sex, discharge status, and outpatient follow-up to train a logistic regression for prediction of good (0-2) versus poor (3-6) mRS, and a linear regression for the full range of mRS scores. The models were trained for prediction of mRS at hospital discharge and post-discharge. The models were externally validated in a dataset of patients with brain injuries from a different healthcare center. We included 5307 patients, 5006 in train and test and 301 in validation; average age was 69 (SD 15) and 65 (SD 17) years, respectively; 47% female. The logistic regression achieved an area under the receiver operating curve (AUROC) of 0.94 [CI 0.93-0.95] (test) and 0.94 [0.91-0.96] (validation), and the linear model a root mean squared error (RMSE) of 0.91 [0.87-0.94] (test) and 1.17 [1.06-1.28] (validation). The NLP-based model is suitable for use in large-scale phenotyping of stroke functional outcomes and population health research.

Review of practices concerning the delivery of an international implant card following the implantation of a medical device.

The European regulation 2017/745 requires the delivery of an international implant card to the patient following the implantation of implantable medical devices (IMD). Currently, there are no procedure to describe the process for providing traceability documents to patients in our institution. The objective of this work was to evaluate the practices of the various departments concerned in our institution and to think about actions to meet the current regulation. The evaluation of practices concerning the delivery of traceability documents to patients was evaluated through a questionnaire with 13 departments. Then, for each specialty, three Electronic Patient Record (EPR) were audited. Finally, IMD's suppliers were asked about their ability to provide an international implant card. 85% (11/13) of the departments sent an IMD traceability document to patients, and in 91% (10/11) of cases, this transmission was done with the international implant card or the operative report. However, 63% of the operative report analyzed did not contain the traceability data of the IMD. Finally, only 60% of the suppliers contactes accompanied their IMD with an international implant card. While waiting for the suppliers to send an international implant card, it was proposed to the departments to give patients the IMD traceability sheet from the IMD traceability software. However, this document does not allow us to be in full compliance with the European regulation to the absence of regulatory information (UDI), address and manufacturer's website).

A network-based systems genetics framework identifies pathobiology and drug repurposing in Parkinson’s disease

Parkinson’s disease (PD) is the second most prevalent neurodegenerative disorder. However, current treatments only manage symptoms and lack the ability to slow or prevent disease progression. We utilized a systems genetics approach to identify potential risk genes and repurposable drugs for PD. First, we leveraged non-coding genome-wide association studies (GWAS) loci effects on five types of brain-specific quantitative trait loci (xQTLs, including expression, protein, splicing, methylation and histone acetylation) under the protein–protein interactome (PPI) network. We then prioritized 175 PD likely risk genes (pdRGs), such as SNCA, CTSB, LRRK2, DGKQ, and CD44, which are enriched in druggable targets and differentially expressed genes across multiple human brain-specific cell types. Integrating network proximity-based drug repurposing and patient electronic health record (EHR) data observations, we identified Simvastatin as being significantly associated with reduced incidence of PD (hazard ratio (HR) = 0.91 for fall outcome, 95% confidence interval (CI): 0.87–0.94; HR = 0.88 for dementia outcome, 95% CI: 0.86–0.89) after adjusting for 267 covariates. In summary, our network-based systems genetics framework identifies potential risk genes and repurposable drugs for PD and other neurodegenerative diseases if broadly applied.