De-identified Electronic Health Record Data Research Articles

Factors Associated with Completeness of Sex and Gender Fields in Electronic Health Records.

Purpose: Our purpose was to understand the completeness of sex and gender fields in electronic health record (EHR) data and patient-level factors associated with completeness of those fields. In doing so, we aimed to inform approaches to EHR sex and gender data collection. Methods: This was a retrospective observational study using 2016-2021 deidentified EHR data from a large health care system. Our sample included adults who had an encounter at any of three hospitals within the health care system or were enrolled in the health care system's Accountable Care Organization. The sex and gender fields of interest were gender identity, sex assigned at birth (SAB), and legal sex. Patient characteristics included demographics, clinical features, and health care utilization. Results: In the final study sample (N = 3,473,123), gender identity, SAB, and legal sex (required for system registration) were missing for 75.4%, 75.8%, and 0.1% of individuals, respectively. Several demographic and clinical factors were associated with having complete gender identity and SAB. Notably, the odds of having complete gender identity and SAB were greater among individuals with an activated patient portal (odds ratio [OR] = 2.68; 95% confidence interval [CI] = 2.66-2.70) and with more outpatient visits (OR = 4.34; 95% CI = 4.29-4.38 for 5+ visits); odds of completeness were lower among those with any urgent care visits (OR = 0.80; 95% CI = 0.78-0.82). Conclusions: Missingness of sex and gender data in the EHR was high and associated with a range of patient factors. Key features associated with completeness highlight multiple opportunities for intervention with a focus on patient portal use, primary care provider reporting, and urgent care settings.

Treatment patterns for patients with BRCA1/2-positive metastatic castration-resistant prostate cancer.

Patients with BRCA-positive metastatic castration-resistant prostate cancer (mCRPC) have an aggressive disease course. This study aimed to describe real-world treatment patterns among patients with BRCA-positive mCRPC. De-identified electronic health record data from the Flatiron Health-Foundation Medicine Inc. Metastatic Prostate Cancer Clinico-Genomic Database (January 01, 2011 to June 30, 2022) were used to select patients with BRCA-positive mCRPC initiating first-line (1L) therapy with an oncologist-defined advanced line of therapy (LOT) or androgen deprivation therapy (ADT) monotherapy. Treatment sequences and reasons for censoring were described in 1L, and among patients who initiated a second-line (2L) therapy. A total of 98 treated patients with BRCA-positive mCRPC were identified. The top 3 treatment regimens in 1L, overall, were ADT monotherapy (19%), enzalutamide (14%), and olaparib (13%). The main reason for censoring patients with ADT monotherapy was death (52.6%). Among 79 patients treated with an advanced LOT in 1L, 43.0% (n = 34) did not initiate a 2L therapy, of which, 29.4% died. In patients who initiated a 2L (n = 45), the most common 1L to 2L treatment sequence was olaparib to docetaxel (11.1%). The most prescribed 2L therapies were docetaxel (22.2%), olaparib (20.0%), abiraterone acetate (13.3%), and enzalutamide (11.1%). From 1L initiation, the median time-to-next-treatment was 6.2 months. Among patients with BRCA-positive mCRPC, ADT monotherapy, enzalutamide, and olaparib were most commonly used. Prognosis of BRCA-positive patients was poor, with most patients failing initial therapy resulting in a switch to a new therapy or death. These findings highlight the need for earlier and more effective treatments for patients with BRCA-positive mCRPC.

Ineffective cardiac resynchronization pacing is associated with poor outcomes in a nationwide cohort analysis

BackgroundDelivery of cardiac resynchronization therapy (CRT) requires left ventricular myocardial capture to achieve clinical benefits. ObjectiveWe sought to determine whether ineffective pacing affects survival. MethodsIneffective ventricular pacing (VP) was defined as the difference between the percentage of delivered CRT (%VP) and the percentage of EffectivCRT in CRT devices. Using the Optum de-identified electronic health record data set and Medtronic CareLink data warehouse, we identified patients implanted with applicable devices with at least 30 days of follow-up. Kaplan-Meier and Cox proportional hazards models assessed the effect of %VP and % ineffective VP on survival. ResultsAmong 7987 patients with 2.1 ± 1.0 years of follow-up, increasing ineffective VP was associated with decreasing survival: the highest observed survival was in the quartile with <0.08% ineffective VP and the lowest survival was in the quartile with >1.47% ineffective VP (85.1% vs 75.7% at 3 years; P < .001). As expected, patients with more than the median %VP of 97.7% had better survival than did patients with <97.7% VP (84.2% vs 77.8%; P < .001). However, patients who had >97.7% VP but >2% ineffective VP had similar survival to patients with <97.7% VP but ≤2% ineffective VP (81.6% vs 79.4%; P = .54). A multivariable Cox proportional hazards model demonstrated that <97.7% VP (adjusted hazard ratio 1.29; 95% confidence interval 1.14–1.46; P < .001) and >2% ineffective VP (hazard ratio 1.35; 95% confidence interval 1.18–1.54; P < .001) were both significantly associated with decreased survival. ConclusionIneffective VP is associated with decreased survival. In addition to maximizing the percentage of delivered CRT pacing, every effort should be made to minimize ineffective VP.

Characterization of hearing loss in pediatric patients with osteogenesis imperfecta

IntroductionOsteogenesis imperfecta (OI) is a common heritable disorder affecting type 1 collagen. The sequelae of OI vary, but hearing loss is a significant complication with 46–58 % of patients having some degree of hearing loss. Previous studies have suggested patients with OI may have conductive, sensorineural, or mixed hearing loss. Majority of these studies focus on the adult population. ObjectivesIdentify a relationship between OI and hearing loss in the pediatric population. MethodsThe TriNetx Analytics Network, a federated health research network that aggregates the de-identified electronic health record data of over 78 million patients across the United States, was queried for patients 18 years old or younger with a diagnosis of OI. Patients in this group with diagnosis of sensorineural, conductive, or mixed hearing loss were recorded. Patients with diagnoses of congenital cytomegalovirus, congenital inner ear malformations, and noise-induced hearing loss were excluded from analysis. ResultsOut of 3256 patients 18 years old or younger with OI, 10.07 % (95 % CI: 9.06–11.16) had a history of any form of hearing loss, 5.71 % (95 % CI: 4.94–6.57) had conductive hearing loss, 3.01 % (95 % CI: 2.45–3.66) had sensorineural hearing loss, and 1.35 % (95 % CI: 0.98–1.81) had mixed hearing loss. Relative risks for diagnosis of any type of hearing loss, conductive hearing loss, sensorineural hearing loss, and mixed hearing loss were calculated: 5.90 (95 % CI 5.32–6.53), 5.08 (95 % CI 4.42–5.84), 6.18 (95 % CI 5.09–7.51), and 13.86 (95 % CI 10.33–18.59) respectively. DiscussionThis study is the largest to date that describes a relationship between OI and conductive, sensorineural, and mixed hearing loss. Pediatric patients with OI are almost five times as likely to have any type of hearing loss. There was a significant increased risk in each subgroup, but conductive hearing loss was the most common for hearing loss in children with OI. The highest risk subtype when compared to controls was mixed hearing loss.

Hyperkalemia Recurrence Following Medical Nutrition Therapy in Patients with Stage3-4 Chronic Kidney Disease: The REVOLUTIONIZEI Real-World Study.

The REVOLUTIONIZEI study aimed to characterize the relationships between medical nutrition therapy (MNT) and hyperkalemia recurrence in patients with stage3-4 chronic kidney disease (CKD) and hyperkalemia who received MNT in real-world clinical practice. This observational cohort study used de-identified electronic health record data from patients aged ≥18years with stage3-4 CKD who received MNT between January 2019 and October 2022 and had hyperkalemia (serum potassium >5.0mmol/L) within 30days before MNT. Patients were followed for 6months or until the first censoring event (death, prescription of outpatient potassium binder, or study end). The primary outcome was the percentage of patients with ≥1hyperkalemia recurrence during follow-up. Secondary outcomes included the number of hyperkalemia recurrences per patient, time to each recurrence, and hyperkalemia-related healthcare resource utilization. Exploratory outcomes included all-cause healthcare resource utilization and mortality. The final cohort comprised 2048 patients; 1503 (73.4%) patients remained uncensored after 6months. During the 6-month follow-up period, 56.0% of patients had ≥1hyperkalemia recurrence and 37.4% had ≥1 recurrence within the first month. Patients with ≥1hyperkalemia recurrence during follow-up had a mean ± standard deviation (SD) of 2.6 ± 2.2 recurrences. The mean ± SD time to first hyperkalemia recurrence was 45 ± 46days; the time between recurrences decreased with subsequent episodes. Hyperkalemia-related hospitalizations and emergency department visits were recorded for 13.7% and 1.5% of patients, respectively. Sensitivity analyses showed that results were consistent across patient subgroups, including those with comorbid heart failure and patients receiving renin-angiotensin-aldosterone system inhibitor therapy at baseline. Most patients with stage3-4 CKD had hyperkalemia recurrence, and MNT alone was inadequate to prevent recurrence. These patients may require additional long-term treatment, such as novel potassium binders, to maintain normokalemia and prevent hyperkalemia recurrence following MNT. Infographic available for this article. INFOGRAPHIC.

Partial-linear single-index transformation models with censored data.

In studies with time-to-event outcomes, multiple, inter-correlated, and time-varying covariates are commonly observed. It is of great interest to model their joint effects by allowing a flexible functional form and to delineate their relative contributions to survival risk. A class of semiparametric transformation (ST) models offers flexible specifications of the intensity function and can be a general framework to accommodate nonlinear covariate effects. In this paper, we propose a partial-linear single-index (PLSI) transformation model that reduces the dimensionality of multiple covariates into a single index and provides interpretable estimates of the covariate effects. We develop an iterative algorithm using the regression spline technique to model the nonparametric single-index function for possibly nonlinear joint effects, followed by nonparametric maximum likelihood estimation. We also propose a nonparametric testing procedure to formally examine the linearity of covariate effects. We conduct Monte Carlo simulation studies to compare the PLSI transformation model with the standard ST model and apply it to NYU Langone Health de-identified electronic health record data on COVID-19 hospitalized patients' mortality and a Veteran's Administration lung cancer trial.

Abstract P499: Patterns of Sodium Glucose Cotransporter-2 Inhibitor Prescription by Indication in Patients With and Without Diabetes in the US Health System

Introduction: Multiple clinical practice guidelines strongly recommend sodium-glucose cotransporter-2 inhibitors (SGLT2i) for people with diabetes and atherosclerotic cardiovascular disease or chronic kidney disease as well as for people with urine albumin-to-creatinine ≥200 mg/g or heart failure (HF), regardless of diabetes status. We characterized SGLT2i prescription patterns by indication for SGLT2i therapy in patients with and without diabetes in the US in 2022. Hypothesis: The prescription of SGLT2i is low across all indications. Methods: Using deidentified electronic health record data from the Optum Labs Data Warehouse, we included adults with estimated glomerular filtration rate ≥20 ml/min/1.73 m 2 across 31 US health systems. We assessed prevalence of SGLT2i prescription by presence of diabetes and indication, and compared them with renin-angiotensin system inhibitors (RASi), which were background therapies in the landmark trials of SGLT2i. Results: We classified 760,337 patients with diabetes and 2,640,077 patients without diabetes by their indications for SGLT2i therapy (indication present, absent, or unknown due to lack of albuminuria testing). SGLT2i were prescribed in 11.2%, 11.2%, and 7.1% of patients with diabetes and indication present, absent, or unknown, respectively, and 2.5%, 0.1%, and 0.1% of patients without diabetes. In contrast, RASi were prescribed more frequently among patients with indication present (with diabetes: 31-44%; without diabetes: 16-33%). RASi were prescribed to the majority of patients prescribed SGLT2i ( Figure A ). Practice patterns varied across health systems, but &lt;30% of patients with an indication received a SGLT2i prescription in all systems. ( Figure B ). Conclusions: In this large-scale analysis of contemporary US data, prescription of SGLT2i is low, which may deprive patients of needed cardiorenal protection. Future work should investigate whether changes in costs enabled by the 2022 Inflation Reduction Act may stimulate more optimal prescription patterns.

Abstract P435: Trends and Disparities in Glycemic Control Among Youth and Adults With Type 1 Diabetes in the US Health System: 2008-2022

Introduction: Glycemic control is crucial for type 1 diabetes (T1D) management, yet contemporary data in the US are scarce. This study aims to investigate trends in glycemic control among US youth and adults with T1D from 2008 through 2022. Hypothesis: Glycemic control has improved over time. Methods: We used a well-validated algorithm to identify patients with T1D using de-identified electronic health record data from the Optum Labs Data Warehouse. We characterized trends in hemoglobin A1c (HbA1c) &lt;7% from 2008-2010 to 2020-2022, overall and by age, sex, and race/ethnicity. Results: We included 128,821 patients with T1D (17% youth under age 18 years, mean age 38 years, 48% female, 78% non-Hispanic White). Among youth, mean HbA1c decreased from 8.9% in 2008-2010 to 8.5% in 2020-2022 ( P&lt;.001 ), and the percentage with HbA1c &lt;7% increased from 8% to 16% ( P&lt;.001 ). The percentage with HbA1c &lt;7% did not differ by age or sex, and it was notably higher in Non-Hispanic White youth than racial/ethnic minority youth in 2020-2022 ( Figure A-D ). Among adults, mean HbA1c decreased from 8.2% in 2008-2010 to 8.1% in 2020-2022 ( P&lt;.001 ). Trend in percentage with HbA1c &lt;7% among adults increased significantly from 20% in 2008-2010 to 26% in 2020-2022 ( P&lt;.001 ). Adults who were older and White were more likely to have HbA1c &lt;7% compared to younger and Black or Hispanic adults in all years ( Figure E-F ). Conclusion: There have been marginal improvements in glycemic control among patients with T1D over the last 15 years. Overall, percentage with HbA1c&lt;7% remains low, particularly among youth and racial/ethnic minority patients. There is an urgent need to expand access to care and diabetes technology to enhance management in the population.

Machine Learning Methods Using Artificial Intelligence Deployed on Electronic Health Record Data for Identification and Referral of At-Risk Patients From Primary Care Physicians to Eye Care Specialists: Retrospective, Case-Controlled Study.

Identification and referral of at-risk patients from primary care practitioners (PCPs) to eye care professionals remain a challenge. Approximately 1.9 million Americans suffer from vision loss as a result of undiagnosed or untreated ophthalmic conditions. In ophthalmology, artificial intelligence (AI) is used to predict glaucoma progression, recognize diabetic retinopathy (DR), and classify ocular tumors; however, AI has not yet been used to triage primary care patients for ophthalmology referral. This study aimed to build and compare machine learning (ML) methods, applicable to electronic health records (EHRs) of PCPs, capable of triaging patients for referral to eye care specialists. Accessing the Optum deidentified EHR data set, 743,039 patients with 5 leading vision conditions (age-related macular degeneration [AMD], visually significant cataract, DR, glaucoma, or ocular surface disease [OSD]) were exact-matched on age and gender to 743,039 controls without eye conditions. Between 142 and 182 non-ophthalmic parameters per patient were input into 5 ML methods: generalized linear model, L1-regularized logistic regression, random forest, Extreme Gradient Boosting (XGBoost), and J48 decision tree. Model performance was compared for each pathology to select the most predictive algorithm. The area under the curve (AUC) was assessed for all algorithms for each outcome. XGBoost demonstrated the best performance, showing, respectively, a prediction accuracy and an AUC of 78.6% (95% CI 78.3%-78.9%) and 0.878 for visually significant cataract, 77.4% (95% CI 76.7%-78.1%) and 0.858 for exudative AMD, 79.2% (95% CI 78.8%-79.6%) and 0.879 for nonexudative AMD, 72.2% (95% CI 69.9%-74.5%) and 0.803 for OSD requiring medication, 70.8% (95% CI 70.5%-71.1%) and 0.785 for glaucoma, 85.0% (95% CI 84.2%-85.8%) and 0.924 for type 1 nonproliferative diabetic retinopathy (NPDR), 82.2% (95% CI 80.4%-84.0%) and 0.911 for type 1 proliferative diabetic retinopathy (PDR), 81.3% (95% CI 81.0%-81.6%) and 0.891 for type 2 NPDR, and 82.1% (95% CI 81.3%-82.9%) and 0.900 for type 2 PDR. The 5 ML methods deployed were able to successfully identify patients with elevated odds ratios (ORs), thus capable of patient triage, for ocular pathology ranging from 2.4 (95% CI 2.4-2.5) for glaucoma to 5.7 (95% CI 5.0-6.4) for type 1 NPDR, with an average OR of 3.9. The application of these models could enable PCPs to better identify and triage patients at risk for treatable ophthalmic pathology. Early identification of patients with unrecognized sight-threatening conditions may lead to earlier treatment and a reduced economic burden. More importantly, such triage may improve patients' lives.

Using predictive risk modelling to identify patients with hidden health needs in an Aboriginal and Torres Strait Islander health service.

In partnership with an Aboriginal and Torres Strait Islander community-controlled health service, we explored the use of a machine learning tool to identify high-needs patients for whom services are harder to reach and, hence, who do not engage with primary care. Using deidentified electronic health record data, two predictive risk models (PRMs) were developed to identify patients who were: (1) unlikely to have health checks as an indicator of not engaging with care; and (2) likely to rate their wellbeing as poor, as a measure of high needs. According to the standard metrics, the PRMs were good at predicting health checks but showed low reliability for detecting poor wellbeing. Results and feedback from clinicians were encouraging. With additional refinement, informed by clinic staff feedback, a deployable model should befeasible.

Association of Ocular Manifestations of Marfan Syndrome With Cardiovascular Complications

To evaluate associations between ocular manifestations of Marfan syndrome and cardiovascular complications. Retrospective cohort study. The TriNetX Analytics platform, a federated health research network of aggregated deidentified electronic health record data of more than 119 million patients, was used to identify patients diagnosed with Marfan syndrome. Univariate logistic regression models were used to evaluate the association of ocular manifestations of Marfan syndrome (such as retinal tears/detachment, lens dislocation, and myopia), with cardiovascular comorbidities. Additional sensitivity analyses were performed using propensity matching. Odds ratios and 95% CIs were calculated for incidence of cardiovascular comorbidities (including aortic dissection, valvular disease, and arrhythmias) following diagnosis of Marfan syndrome. A total of 19,105 patients were identified who were diagnosed with Marfan disease without ocular manifestations, and an additional 3887 Marfan patients with ocular comorbidities. Patients who were diagnosed with ocular disease included 883 with ectopic lens, 417 with retinal tear or detachment, 683 with aphakia, 534 with pseudophakia, and 2465 with myopia. Patients with any ocular manifestations of Marfan were significantly more likely to be diagnosed with all cardiovascular comorbidities modeled including aortic aneurysm and dissection (OR 2.035; P < .0001), mitral valve prolapse (OR 2.725; P < .0001), tricuspid valve disorders (OR 2.142; P < .0001), cardiac arrhythmias (OR 1.836; P < .0001), and all cardiovascular outcomes combined (OR 2.194; P < .0001). In a large and diverse cohort of patients with Marfan syndrome, ocular manifestations of the disorder appear strongly associated with cardiovascular comorbidities.

Assessing the Association Between Head and Neck Cancer and Granulomatosis with Polyangiitis.

Objective: To determine the odds of head and neck cancer (HNC) in patients with a concurrent or prior diagnosis of granulomatosis with polyangiitis (GPA). Methods and Materials: The TriNetX Analytics Network, a federated research platform that aggregates de-identified electronic health record data of over 130 million patients worldwide, was queried for patients with at least one ICD-10 encounter diagnosis of GPA. Patients within this group with an encounter diagnosis of cancer of the sinonasal, oral cavity, oropharynx, nasopharynx, and larynx concurrent or after the initial encounter diagnosis of GPA were recorded and compared to a standardized control population to determine odds ratios with a 95% confidence interval (CI). Relevant confounding variables, including human papillomavirus, Epstein Barr virus, tobacco, and alcohol exposure, were balanced between cohorts by 1:1 propensity matching. Results: Of the patients in the GPA cohort, 126 (0.48%) had an ICD-10 diagnosis of HNC. When stratifying by head and neck subsites, 20 (0.08%), 18 (0.07%), 23 (0.09%), 70 (0.27%), and 22 (0.084%) GPA patients had an ICD-10 encounter diagnosis of cancer involving the sinonasal, nasopharynx, larynx, oral cavity, and oropharynx. When comparing the experimental GPA group with the standardized control population after matching, patients in the GPA group had 1.3 times (95% CI: 1.03-1.175) greater odds of HNC when including cases diagnosed after or concurrently with the diagnosis of the vasculitis. There was no statistical difference in the odds of cancer at each anatomical subsite between the GPA and control cohort after matching. Conclusion: Our study identifies a statistically significant increase in the odds of HNC encounter diagnoses in patients with GPA.

Distinguishing cardiac catheter ablation energy modalities by applying natural language processing to electronic health records.

Aim: Catheter ablation is used to treat symptomatic atrial fibrillation (AF) and is performed using either cryoballoon (CB) or radiofrequency (RF) ablation. There is limited real world data of CB and RF in the USas healthcare codes are agnostic of energy modality. An alternative method is to analyze patients' electronic health records (EHRs) using Optum's EHR database. Objective: To determine the feasibility of using patients' EHRs with natural language processing (NLP) to distinguish CB versus RF ablation procedures. Data Source: Optum® de-identified EHRdataset, Optum® Cardiac Ablation NLP Table. Methods: This was a retrospective analysis of existing de-identified EHR data. Medical codes were used to create an ablation validation table. Frequency analysis was used to assess ablation procedures and their associated note terms. Two cohorts were created (1) index procedures, (2) multiple procedures. Possible note term combinations included (1) cryoablation (2) radiofrequency (3) ablation, or (4) both. Results: Of the 40,810 validated cardiac ablations, 3777 (9%) index ablation procedures had available and matching NLP note terms. Of these, 22% (n=844) were classified as ablation, 27% (n= 1016) as cryoablation, 49% (n=1855) as radiofrequency ablation, and 1.6% (n=62) as both. In the multiple procedures analysis, 5691 (14%) procedures had matching note terms. 24% (n=1362) were classified as ablation, 27% as cryoablation, 47% as radiofrequency ablation, and 2% as both. Conclusion: NLP has potential to evaluate the frequency of cardiac ablation by type, however, for this to be a reliable real-world data source, mandatory data entry by providers and standardized electronic health reporting must occur.

Glucagon-like peptide-1 receptor agonists and the risk of atrial fibrillation in adults with diabetes: a real-world study.

Glucagon-like peptide-1 receptor agonists (GLP-1RA) have cardiovascular benefits in type 2 diabetes, but none of the cardiovascular trials studied atrial fibrillation/atrial flutter (AF) as a primary endpoint. Data from post-marketing surveillance studies remains sparse. To examine the real-world risk of AF comparing GLP-1RA with other non-insulin glucose-lowering agents. Cohort study using de-identified electronic health record data from the Optum Labs Data Warehouse. Adult patients with diabetes who were newly prescribed add-on non-insulin glucose-lowering agents and were on metformin between 2005-2020. New users of GLP-1RA were separately compared with new users of dipeptidyl peptidase-4 inhibitors (DPP4i) and sodium-glucose cotransporter 2 inhibitors (SGLT2i), using 1:1 propensity score matching to adjust for differences in patient characteristics. The primary outcome was incident AF, defined and captured by diagnosis code for AF. Incidence rate difference (IRD) and hazard ratio (HR) were estimated in the matched cohorts. In the matched cohort of 14,566 pairs of GLP-1RA and DPP4i followed for a median of 3.8 years, GLP-1RA use was associated with a lower risk of AF (IRD, -1.0; 95% CI, -1.8 to -0.2 per 1000 person-years; HR, 0.82; 95% CI, 0.70 to 0.96). In the matched cohort of 9,424 pairs of patients on GLP-1RA and SGLT2i with a median follow-up of 2.9 years, there was no difference in the risk for AF (IRD, 0.4; 95% CI -0.7 to 1.5 per 1000 person-years; HR, 1.12; 95% CI, 0.89 to 1.42). In this real-word study, GLP-1RA was associated with a lower risk of AF compared with DPP4i, but no difference compared with SGLT2i, suggesting that cardiovascular benefits of GLP-1RA use may extend to prevention for AF in patients with diabetes. Our findings call for future randomized controlled trials to focus on the effects of GLP-1RA on AF prevention.

2838. Clinical Application and Validation of a Predictive Antimicrobial Resistance Risk Categorization Framework for Patients with Uncomplicated Urinary Tract Infection

Abstract Background Empiric antibiotic (ABX) treatment for uncomplicated urinary tract infections (uUTIs) can be ineffective due to antimicrobial resistance (AMR). Understanding the risk of AMR using data-driven approaches can inform appropriate ABX selection. We developed an AMR pathogen risk categorization framework in E. coli caused uUTI using predictive modeling and evaluated its clinical validity. Methods Eligible females with uUTI confirmed by positive E. coli urine culture treated with nitrofurantoin (NTF), trimethoprim/sulfamethoxazole (SXT), fluoroquinolones (FQs), or beta-lactams (BLs) were identified from the Optum de-identified electronic health record data set (Oct 2015–Feb 2020). We developed predictive models using machine learning to quantify AMR probability for each ABX class. A framework with 3 risk categories (low, moderate, high) was constructed using the predicted probability (PP) of non-susceptibility (NS) (Table 1). Six patient profiles from differing risk categorizations were reviewed for clinical validity by 5 clinicians (4 medical doctors, 1 pharmacist). Results Of 87,487 eligible patients, approximately half were classified as low or high risk (44.0–49.1% across ABX classes). The proportion of patients with infections due to NS organisms was 5–12-fold higher among patients classified as high or moderate vs low risk (Figure 1). After review of the patient profiles (Table 2), clinical experts confirmed the consistency of modeled risk classification for all 6 patients with their own assessment of AMR risk across all drug classes. Patient 1 was aged 20 yrs, White, West residence, no UTI or ABX history 1 year prior to her uUTI; PP of NS was low (NTF 1.5%, SXT 16.6%, FQs 4.8%, BLs 8.4%) and was classified as low risk for all ABX classes. In contrast, patient 6 was post-menopausal, Black, Midwest residence, and had UTI episodes, prior AMR, and multiple healthcare visits 1 year prior to uUTI. She had high PP of NS (NTF 10.3%, SXT 97.2%, FQs 96.4%, BLs 48.0%) and was categorized as high risk for all ABX classes. Conclusion AMR risk varied greatly between patients. Our prediction model contextualizes patients’ AMR PP to 4 commonly prescribed ABX classes in the setting of uUTIs. Clinical application of this framework could inform appropriate empiric ABX selection for patients with uUTI. Disclosures Ryan K. Shields, PharmD, MS, Allergan: Advisor/Consultant|Cidara: Advisor/Consultant|Entasis: Advisor/Consultant|GSK: Advisor/Consultant|Melinta: Advisor/Consultant|Melinta: Grant/Research Support|Menarini: Advisor/Consultant|Merck: Advisor/Consultant|Merck: Grant/Research Support|Pfizer: Advisor/Consultant|Roche: Grant/Research Support|Shionogi: Advisor/Consultant|Shionogi: Grant/Research Support|Utility: Advisor/Consultant|Venatorx: Advisor/Consultant|Venatorx: Grant/Research Support Wendy Y. Cheng, MPH, PhD, ORCID: 0000-0002-8281-2496, Analysis Group, Inc.: Wendy Y. Cheng is an employee of Analysis Group, Inc., a consulting company that received funding from GSK to conduct this study Kalé Kponee-Shovein, ScD, Analysis Group, Inc: Employee of Analysis Group, Inc., which received funding from GSK to conduct the study Fernando Kuwer, MSc, Analysis Group, Inc.: Employee of Analysis Group, Inc., which received funding from GSK to conduct the study Chi Gao, ScD, Analysis Group, Inc.: Employee of Analysis Group, Inc., which received funding from GSK to conduct the study Ashish V. Joshi, PhD, GSK: Employee|GSK: Stocks/Bonds Fanny S. Mitrani-Gold, MPH, GSK: Employee|GSK: Stocks/Bonds Patrick Schwab, PhD, GSK: Employment|GSK: Stocks/Bonds Diogo Ferrinho, PharmD, GSK: Employee|GSK: Stocks/Bonds Malena Mahendran, MS, Analysis Group, Inc.: Malena Mahendran is an employee of Analysis Group, Inc., a consulting company that received funding from GSK to conduct this study Daniel Indacochea, PhD, Analysis Group, Inc.: Employee of Analysis Group, Inc., which received funding from GSK to conduct the study Lisa Pinheiro, MFin, Analysis Group, Inc.: Employee of Analysis Group, Inc., which received funding from GSK to conduct the study Jimmy Royer, PhD, Analysis Group, Inc.: Employee of Analysis Group, Inc., which received funding from GSK to conduct the study Madison T. Preib, MPH, GSK: Employee|GSK: Stocks/Bonds Jennifer Han, MD, GSK: Employment|GSK: Stocks/Bonds Richard Colgan, MD, GSK: Advisor/Consultant

Post-treatment level of LDL cholesterol and all-cause mortality in patients with atherosclerotic cardiovascular disease: evidence from real-world setting.

This study aimed to evaluate the safety of the currently recommended target of LDL cholesterol (LDL-C) control on mortality in patients with atherosclerotic cardiovascular disease (ASCVD). Using deidentified electronic health record data, we conducted a multicentre retrospective cohort study involving individuals with documented ASCVD who had received statin treatment for at least 3 months across China. The primary outcomes assessed encompassed all-cause mortality, CV mortality, and non-CV mortality. Relationships between post-treatment LDL-C concentrations and outcomes were evaluated using restricted cubic spline curves based on Cox proportional hazards regression analyses. Additionally, competitive risk models were employed to explore associations between LDL-C levels and cause-specific mortality. Among 33 968 participants, we identified nearly linear associations of post-treatment LDL-C level with all-cause mortality and CV mortality during a median follow-up of 47 months. Notably, patients who achieved the recommended target of LDL-C (<1.4 mmol/L) were at significantly lower risks of all-cause mortality [hazard ratio (HR), 0.77; 95% confidence interval (CI), 0.69-0.86] and CV mortality (subdistribution HR, 0.68; 95% CI, 0.58-0.79), compared with those with LDL-C ≥ 3.4 mmol/L. This survival benefit was consistent in patients with different intensities of LDL-C reduction and other subgroup analyses. And no correlation was found between post-treatment LDL-C concentration and non-CV mortality. Our findings supported the safety of currently recommended target of LDL-C control and the 'lower is better' principle in patients with ASCVD.

Associations of comorbid substance use disorders with clinical outcomes in schizophrenia using electronic health record data.

Schizophrenia and comorbid substance use disorders (SUDs) are associated with poor treatment outcomes but differences between the associations of different SUDs with clinical outcomes are poorly characterized. This study examines the associations of comorbid SUDs with clinical outcomes in schizophrenia using a largescale electronic health record (EHR) database. Real-world data (RWD) analysis using the NeuroBlu database; de-identified EHR data were analysed. Multivariable logistic regression, Poisson and CoxPH models were used to compare the associations of specific comorbid SUDs with outcome variables. Comorbid SUD was significantly different on all outcome measures compared to no SUD (U=1.44e7-1.81e7, all ps<.001), except number of unique antipsychotics (U=1.61e7, p=.43). Cannabis (OR=1.58, p<.001) and polysubstance (OR=1.22, p=.007) use disorders were associated with greater CGI-S. Cannabis (IRR=1.13, p=.003) and polysubstance (IRR=1.08, p=.003) use disorders were associated with greater number of unique antipsychotics prescribed, while cocaine (HR=1.87, p<.001), stimulants (HR=1.64, p=.024), and polysubstance (HR=1.46, p<.001) use disorders were associated with a shorter time to antipsychotic discontinuation. Conversely, alcohol use (IRR=0.83, p<.001), cocaine use (IRR=0.61, p<.001), opioid use (IRR=0.61, p<.001), stimulant use (IRR=0.57, p<.001) and polysubstance use (IRR=0.87, p<.001) disorders were associated fewer inpatient days. Comorbid SUDs were generally associated with greater CGI-S and poorer clinical outcomes in patients with schizophrenia. Treatment strategies should target not only schizophrenia symptoms but also comorbid SUD to improve management of both conditions.

Increased risk of osteomyelitis and septic arthritis in osteogenesis imperfecta.

We investigated the prevalence of osteomyelitis and septic arthritis in individuals with osteogenesis imperfecta (OI) as compared to the general population. We performed a retrospective cohort study utilizing the TriNetX Analytics platform, a federated, aggregated electronic health record (EHR) research network containing national, deidentified EHR data. We queried patients with OI, based on encounter diagnoses. Patients in this group with any occurrence of osteomyelitis or septic arthritis were recorded. A control cohort was established to compare the prevalence in patients without OI. Of 8444 individuals with OI, 433 (5.13%) had encounter diagnoses for osteomyelitis and 61 (0.72%) had encounter diagnoses for septic arthritis. In comparison, of 79 176 436 patients without OI, 352 009 (0.44%) had encounter diagnoses for osteomyelitis, while 106 647 (0.13%) had encounter diagnoses for septic arthritis. The relative risk for osteomyelitis in OI patients was 11.53 (95% CI: 10.52-12.64), while the relative risk for septic arthritis was 5.36 (95% CI: 4.18-6.89). The relative risk for osteomyelitis in pediatric OI patients was 30.55 (95% CI: 24.35-38.28). To our knowledge, this is the first study investigating musculoskeletal infections in patients with OI, as well as the first to report the overall prevalence in the general population. Clinicians may benefit from a high index of suspicion for musculoskeletal infections in OI patients with corresponding symptoms. Further study is warranted to investigate if modifications to conventional diagnostic pathways and criteria are valuable in this population. Level of evidence: Retrospective Cohort Study - Level II.

Pre-appendectomy hyponatremia is associated with increased rates of complicated appendicitis.

BackgroundAcute appendicitis is one of the most common surgical emergencies worldwide. Preoperative assessment of the risk of complicated appendicitis may aid in treatment planning. We sought to investigate the association between pre-appendectomy hyponatremia and diagnosis of complicated appendicitis. MethodsThe TriNetX platform, a federated health research network that aggregates de-identified electronic health record data of over 90 million patients across the United States, was queried for patients who underwent appendectomy starting January 2019 and who had at least one sodium value from the preoperative period. The study population was stratified into three age groups: pediatric (age < 18), adult (age 18–64), and older adult (age ≥ 65). These groups were subdivided into patients with preoperative hyponatremia (<135 mmol/L) and normonatremia (135–145 mmol/L). ResultsAmong the 61,245 patients who met inclusion criteria, 17,546 were included for analysis following propensity score matching. The odds of complicated appendicitis were highest in pediatric patients (age < 18) with pre-appendectomy hyponatremia (odds ratio [OR] = 2.91, 95 % CI [2.53, 3.35]). Patients age 18–64 and aged ≥ 65 with preoperative hyponatremia also demonstrated increased odds of a complicated appendicitis diagnosis, but to a lesser extent (OR = 2.11, 95 % CI [1.92, 2.32] (OR = 1.49, 95 % CI [1.25, 1.77], respectively). ConclusionsIn a large analysis of matched patients with acute appendicitis, we found an association between immediate preoperative hyponatremia and complicated appendicitis. Future studies are indicated to further evaluate the role of hyponatremia as a potential diagnostic marker for complicated appendicitis in all age groups. Key messageThis study suggests a role of hyponatremia as one of multiple variables to incorporate into future clinical decision tools for complicated acute appendicitis.

Risk of New Retinal Vascular Occlusion After mRNA COVID-19 Vaccination Within Aggregated Electronic Health Record Data

New-onset retinal vascular occlusion (RVO) occurring acutely after messenger RNA (mRNA) COVID-19 vaccination has been described in recent literature. Because RVO can cause vision loss or blindness, an epidemiologic investigation evaluating this potential association is of great importance to public health. To investigate how often patients are diagnosed with new RVO acutely after the mRNA COVID-19 vaccine compared with influenza and tetanus, diphtheria, pertussis (Tdap) vaccines. A retrospective population-based cohort design using the TriNetX Analytics platform, a federated, aggregated electronic health record (EHR) research network containing the deidentified EHR data of more than 103 million patients, was used to examine aggregate EHR data. Data were collected and analyzed on October 20, 2022. Data on patients within the TriNetX Analytics platform were searched for the presence of vaccination Common Procedural Technology codes, and instances of newly diagnosed RVO within 21 days of vaccination were recorded and reported. Propensity score matching based on demographic characteristics (age, sex, race and ethnicity) and comorbidities (diabetes, hypertension, and hyperlipidemia) was performed between vaccination groups for evaluation of relative risks (RRs). The appearance of a new-encounter diagnosis of RVO within 21 days of the mRNA COVID-19 vaccination was the primary outcome. Historical comparison cohorts of patients receiving influenza and Tdap vaccinations allowed for evaluation of the RRs for RVO. Of 3 108 829 patients (mean [SD] age at vaccination, 50.7 [20.4] years; 56.4% women) who received the mRNA COVID-19 vaccine, 104 (0.003%; 95% CI, 0.003%-0.004%) patients had a new diagnosis of RVO within 21 days of vaccination. After propensity score matching, the RR for new RVO diagnosis after the first dose of COVID-19 vaccination was not significantly different from that after influenza (RR, 0.74; 95% CI, 0.54-1.01) or Tdap (RR, 0.78; 95% CI, 0.44-1.38) vaccinations, but was greater when compared with the second dose of the COVID-19 vaccination (RR, 2.25; 95% CI, 1.33-3.81). The findings of this study suggest that RVO diagnosed acutely after mRNA COVID-19 vaccination occurs extremely rarely at rates similar to those of 2 different historically used vaccinations, the influenza and Tdap vaccines. No evidence suggesting an association between the mRNA COVID-19 vaccination and newly diagnosed RVO was found.