Longitudinal Electronic Health Record Data Research Articles

Causal Effects of Stochastic PrEP Interventions on HIV Incidence Among Men Who Have Sex With Men.

Antiretroviral preexposure prophylaxis (PrEP) is highly effective in preventing human immunodeficiency virus (HIV) infection, but uptake has been limited and inequitable. Although interventions to increase PrEP uptake are being evaluated in clinical trials among men who have sex with men (MSM), those trials cannot evaluate effects on HIV incidence. Estimates from observational studies of the causal effects of PrEP-uptake interventions on HIV incidence can inform decisions about intervention scale-up. We used longitudinal electronic health record data from HIV-negative MSM accessing care at Fenway Health, a community health center in Boston, Massachusetts, from January 2012 through February 2018, with 2 years of follow-up. We considered stochastic interventions that increased the chance of initiating PrEP in several high-priority subgroups. We estimated the effects of these interventions on population-level HIV incidence using a novel inverse-probability weighted estimator of the generalized g-formula, adjusting for baseline and time-varying confounders. Our results suggest that even modestincreases in PrEP initiation in high-priority subgroups of MSM could meaningfully reduce HIV incidence in the overall population of MSM. Interventions tailored to Black and Latino MSM should be prioritized to maximize equity and impact.

Digital Health Transformers and Opportunities for Artificial Intelligence-Enabled Nephrology.

Digital Health Transformers and Opportunities for Artificial Intelligence-Enabled Nephrology.

Air pollution and dementia using longitudinal electronic health record data from the Indian Health Service

AbstractBackgroundAir pollution has been identified as an important risk factor for dementia in previous studies. As a unique population with a history of colonization, American Indians have long been disproportionately affected by air pollution. However, few studies have investigated the effects of air pollution on the risk of developing dementia in American Indians. This study aims to address this important knowledge gap.MethodCounty‐level average air pollution data in 2003‐2007 were downloaded from the land‐use regression models developed by the Center for Air, Climate, & Energy Solutions (CACES) and merged with data from the Indian Health Service (IHS) National Data Warehouse and related electronic health record databases between fiscal year (FY) 2007‐2013. We assigned air pollution exposure to IHS users based on their county of residence and calculated 5‐year averages of PM2.5, O3, CO, and NO2 in each county. A total of 14,910 American Indians who were 65+ years old in FY2007, used IHS services at least once each year, and were dementia free between FY2007‐2009 were included in the main analysis. Cox proportional hazard models with age as the time scale were employed to examine the association of air pollution with dementia incidence, adjusting for potential confounders.ResultThe PM2.5 levels in the IHS counties were lower than those in all US counties during 2003‐2013, while the levels of other air pollutants were similar to the US (Figure 1). More dementia incident cases were found in IHS counties with higher O3 levels (Figure 2). In multivariable Cox regression models (Table1), we found a significant positive association between dementia and O3 with a hazard ratio (HR) of 1.26 (95% CI: 1.02‐1.55) per 1 ppb standardized O3, while the associations between dementia and PM2.5, and CO were not statistically significant. We found similar results in sensitivity analyses using longer lags between air pollution and dementia or replacing CO by NO2 in the regression model.ConclusionOur study suggests that exposure to O3 is associated with the higher risk of dementia in American Indians. Future efforts to reduce exposure to air pollution might help lower dementia risk in American Indian communities.

Effects of Androgen Deprivation Therapy on Alzheimer's disease and Related Dementia Risk in the Million Veteran Program

AbstractBackgroundAndrogen deprivation therapy (ADT) for prostate cancer depletes circulating testosterone levels and has been shown to increase risk for Alzheimer’s disease (AD) and related dementias (ADRD). It is unclear whether ADT acts independently to increase ADRD risk, or whether effects of this treatment are exacerbated by the genetic risk for AD. We hypothesized that the effect of ADT on ADRD risk will be moderated by AD polygenic risk.MethodClinical and genetic data were analyzed from 87,060 United States Veterans, who had a diagnosis of prostate cancer and were participants in the Million Veterans Program (MVP) biorepository study. Clinical conditions, including diagnosis codes for ADRD, cancer, Charlson Comorbidity Index conditions, laboratory values (Gleason score, Prostate Specific Antigen levels), cancer treatments, and vital status were obtained from the Veterans Affairs (VA) longitudinal electronic health record data. A polygenic risk score (PRS) for AD was derived based on summary statistics from the Kunkle et al (2019) GWAS of Alzheimer’s disease. The impact of ADT on ADRD was assessed using a time‐varying Fine‐Gray competing risks regression. Death from any cause was considered as a competing event to the development of ADRD.ResultThere were 17,092 (19.6%) patients who received ADT. 72% of the cohort was of white/European ancestry. The average age at prostate cancer diagnosis was 66 years. The median duration of ADT was 360 days. ADT users were more likely to be African American, and had higher PSA levels and Gleason Scores. ADT exposure and the AD PRS each showed significant main effects on ADRD risk, HRs = 1.16 (95% CI: 1.04 ; 1.29) and 1.10 (95% CI: 1.06 ; 1.15) respectively; however, we did not observe a significant interaction effect. Sensitivity analyses indicate that the association was not confounded by including Gleason score, PSA level, income, or education, and results were not altered significantly by excluding or focusing on the effect of the APOE:E4 locus.ConclusionADT significantly increases risk of ADRD, independent of underlying AD genetic risk factors. Further research is necessary in order to determine the mechanism by which androgen depletion contributes to the pathophysiology of ADRD.

Abstract 13653: Leveraging Natural Language Processing and Machine Learning to Predict Worsening Heart Failure Events

Background: Prior risk models in patients with heart failure (HF) have focused on hospitalizations for worsening HF (WHF) and have not evaluated for differences in predictors by left ventricular ejection fraction (LVEF). We used natural language processing (NLP) and machine learning methods with access to longitudinal electronic health record (EHR) data to develop risk prediction models for WHF events across practice settings and by LVEF category. Methods: We identified all adults with HF and known LVEF on January 1 st of each year from 2011-2019 in an integrated health care system. WHF events within 1 year were defined as any hospitalization, emergency department, or outpatient encounter with ≥1 symptom, ≥2 objective findings including ≥1 sign, and ≥1 change in HF-related therapy. Signs and symptoms were ascertained using rule-based NLP. We conducted boosted decision tree-based ensemble models for any WHF event within each LVEF category: HF with reduced EF (HFrEF; LVEF ≤40%), HF with mildly reduced EF (HFmrEF; LVEF 41-49%), and HF with preserved EF (HFpEF; LVEF ≥50%). We evaluated model discrimination using area under the curve (AUC) and model calibration using Brier scores. Results: Among 359,298 patients from 2011-2019, 65,838 (18%) had HFrEF, 52,491 (15%) had HFmrEF, and 240,969 (67%) had HFpEF. Mean age was 75±12, 47% were women, and 37% were minorities including 10% Black, 11% Asian/Pacific Islander, and 12% of Hispanic ethnicity. WHF events occurred in 22% of patients with HFrEF, 17% with HFmrEF, and 16% with HFpEF. The models displayed an AUC of 0.75 and Brier score of 0.15 for HFrEF and an AUC of 0.77 and Brier scores of 0.12 for both HFmrEF and HFpEF. Clinical predictors were similar across LVEF categories ( Table ). Conclusions: Longitudinal EHR data can be leveraged using NLP and machine learning for accurate risk estimation that reliably identifies clinical predictors across a range of LVEF. These findings may provide novel insight into the natural history of HF.

An integrated LSTM-HeteroRGNN model for interpretable opioid overdose risk prediction

Opioid overdose (OD) has become a leading cause of accidental death in the United States, and overdose deaths reached a record high during the COVID-19 pandemic. Combating the opioid crisis requires targeting high-need populations by identifying individuals at risk of OD. While deep learning emerges as a powerful method for building predictive models using large scale electronic health records (EHR), it is challenged by the complex intrinsic relationships among EHR data. Further, its utility is limited by the lack of clinically meaningful explainability, which is necessary for making informed clinical or policy decisions using such models. In this paper, we present LIGHTED, an integrated deep learning model combining long short term memory (LSTM) and graph neural networks (GNN) to predict patients' OD risk. The LIGHTED model can incorporate the temporal effects of disease progression and the knowledge learned from interactions among clinical features. We evaluated the model using Cerner's Health Facts database with over 5 million patients. Our experiments demonstrated that the model outperforms traditional machine learning methods and other deep learning models. We also proposed a novel interpretability method by exploiting embeddings provided by GNNs to cluster patients and EHR features respectively, and conducted qualitative feature cluster analysis for clinical interpretations. Our study shows that LIGHTED can take advantage of longitudinal EHR data and the intrinsic graph structure of EHRs among patients to provide effective and interpretable OD risk predictions that may potentially improve clinical decision support.

Prevalence of High Weight Status in Children Under 2 Years in NHANES and Statewide Electronic Health Records Data in North Carolina and South Carolina.

We evaluated the prevalence of high weight status in children ages 0 to 24 months (m) using data from electronic health records (EHR) and NHANES. We also examined relationships between weight status during infancy and obesity at 24 months of age. EHR data from 4 institutions in North and South Carolina included patients born January 1, 2013-October 10, 2017 (N=147,290). NHANES data included study waves from 1999 to 2018 (unweighted N=5121). We calculated weight-for-length (WFL), weight-for-age (WFA), and body mass index (BMI), excluding implausible values, and categorized weight status (<85th, 85th to <95th, or ≥95th percentile), assessing prevalence at birth, 6, 12, 18, and 24 months. Utilizing individual, longitudinal EHR data, we used separate regression models to assess obesity risk at 24 months based on anthropometrics at birth, 6, 12, and 18 months, adjusting for sex, race/ethnicity, insurance, and health system. Prevalence of BMI ≥95th percentile in EHR data at 6, 12, 18, and 24 months were 9.7%, 15.7%, 19.6%, and 20.5%, respectively. With NHANES the prevalence was 11.6%, 15.0%, 16.0%, and 8.4%. For both, the prevalence of high weight status was higher in Hispanic children. In EHR data, high weight status at 6, 12, and 18 months was associated with obesity at 24 months, with stronger associations as BMI category increased and as age increased. High weight status is common in infants and young children, although lower at 24 months in NHANES than EHR data. In EHR data, high BMI at 6, 12, and 18 months was associated with increased risk of obesity at 24 months.

Sociodemographic variables associated with risk for diabetic retinopathy

Background:Several systemic and sociodemographic factors have been associated with the development and progression of diabetic retinopathy (DR). However, there is limited investigation of the potential role sociodemographic factors may play in augmenting systemic risk factors of DR. We hypothesize that age, sex, race, ethnicity, income, and insurance payor have an impact on hemoglobin A1c (HbA1c), body mass index, and systolic blood pressure, and therefore an upstream effect on the development of DR and vision-threatening forms of DR (VTDR).Methods:Multivariable analysis of longitudinal electronic health record data at a large academic retina clinic was performed. Sociodemographic factors included race, ethnicity, income, and insurance payor. Systemic risk factors for DR included hemoglobin A1c (HbA1c), systolic blood pressure (SBP), and body mass index (BMI). VTDR was identified from encounter diagnostic codes indicating proliferative retinopathy or diabetic macular edema. Patient-reported primary address zip codes were used to approximate income level, stratified into quartiles.Results:From 2016 to 2018, 3,470 patients with diabetes totaled 11,437 visits were identified. Black patients had higher HbA1c and SBP compared to White patients. White patients had higher BMI and SBP compared to patients of unknown/other race and greater odds of VTDR than the latter. Patients of Hispanic ethnicity had significantly higher SBP than non-Hispanic patients. Low-income patients had higher BMI and SBP than high-income patients and greater odds of VTDR than the latter. Medicaid recipients had greater odds of VTDR than those with Blue Care Network (BCN) and Blue Cross Blue Shield (BCBS) insurance. Medicaid and Medicare recipients had higher SBP compared to BCBS recipients. Finally, both higher HbA1c and SBP had greater odds of VTDR. There were no differences in odds of VTDR between White and Black patients or between Hispanic and non-Hispanic patients.Conclusion:Significant associations exist between certain sociodemographic factors and well-known risk factors for DR. Income and payor were associated with increased severity of systemic risk factors and presence of VTDR. These results warrant further investigation of how risk factor optimization and disease prevention may be further improved by targeted intervention of these modifiable sociodemographic factors.

Validation of risk prediction models applied to longitudinal electronic health record data for the prediction of major cardiovascular events in the presence of data shifts.

Deep learning has dominated predictive modelling across different fields, but in medicine it has been met with mixed reception. In clinical practice, simple, statistical models and risk scores continue to inform cardiovascular disease risk predictions. This is due in part to the knowledge gap about how deep learning models perform in practice when they are subject to dynamic data shifts; a key criterion that common internal validation procedures do not address. We evaluated the performance of a novel deep learning model, BEHRT, under data shifts and compared it with several ML-based and established risk models. Using linked electronic health records of 1.1 million patients across England aged at least 35 years between 1985 and 2015, we replicated three established statistical models for predicting 5-year risk of incident heart failure, stroke, and coronary heart disease. The results were compared with a widely accepted machine learning model (random forests), and a novel deep learning model (BEHRT). In addition to internal validation, we investigated how data shifts affect model discrimination and calibration. To this end, we tested the models on cohorts from (i) distinct geographical regions; (ii) different periods. Using internal validation, the deep learning models substantially outperformed the best statistical models by 6%, 8%, and 11% in heart failure, stroke, and coronary heart disease, respectively, in terms of the area under the receiver operating characteristic curve. The performance of all models declined as a result of data shifts; despite this, the deep learning models maintained the best performance in all risk prediction tasks. Updating the model with the latest information can improve discrimination but if the prior distribution changes, the model may remain miscalibrated.

Unmet need in major depressive disorder and acute suicidal ideation or behavior: findings from a longitudinal electronic health record data analysis

AimsUsing a national electronic health records (EHR) database, the current study describes treatments, depression severity, and health care resource utilization (HRU) among patients with major depressive disorder (MDD) and acute suicidal ideation or behavior (MDSI) prior to, during, and following a suicide-related event in the United States.Materials and methodsThis retrospective matched cohort study used data collected from the Optum EHR de-identified database for patients with diagnosis codes for MDD and acute suicidal ideation or behavior and a propensity score–matched cohort of patients without MDD or a suicide-related event. The study period was 31 October 2015–30 September 2019. MDD-related treatments and 9-item Patient Health Questionnaire (PHQ-9) scores, when available, were assessed at the first health care encounter for a suicide-related event (index period), 12 months before (pre-period), and 6 months after (post-period). All-cause and MDD-related HRU were assessed during the post-period.ResultsThe mean (standard deviation) age of patients with MDSI was 39 (16) years; 55.0% were female. Index events occurred as follows: inpatient stay, 38.9%; observation unit stay, 4.6%; emergency department (ED) visit, 46.5%; and outpatient visit, 10.1%. Antidepressants and psychotherapy were the most common pharmacologic and nonpharmacologic treatments, respectively, prescribed during the pre- (31.3%, 9.5%, respectively) and index (41.2%, 18.7%, respectively) periods. Post-period data (n = 40,261) revealed only 43.4% received an antidepressant and 20.5% had psychotherapy after the suicide-related event. Few patients had PHQ-9 scores recorded during the pre- (4.4%), index (1.3%), and post- (7.6%) periods. During the post-period, 11.8%, 5.0%, and 33.1% of patients had ≥1 all-cause inpatient stay, observation unit stay, and ED visit, respectively; 61.0% had ≥1 all-cause and 33.4% ≥1 MDD-related outpatient visit. Most patients with MDSI and an inpatient encounter or ED visit were discharged to home or self-care (65.4%). Odds of an all-cause hospital encounter during the post-period were higher for patients with versus without MDSI (by 30.1, 33.5, and 33.9 times for inpatient stay, ED visit, and observation unit stay, respectively).ConclusionThis analysis highlights an opportunity to improve outcomes for this vulnerable population. More complete data on patient outcomes is needed to inform strategies designed to optimize screening and treatment.

Comparison of a Target Trial Emulation Framework vs Cox Regression to Estimate the Association of Corticosteroids With COVID-19 Mortality

Communication and adoption of modern study design and analytical techniques is of high importance for the improvement of clinical research from observational data. To compare a modern method for statistical inference, including a target trial emulation framework and doubly robust estimation, with approaches common in the clinical literature, such as Cox proportional hazards models. This retrospective cohort study used longitudinal electronic health record data for outcomes at 28-days from time of hospitalization within a multicenter New York, New York, hospital system. Participants included adult patients hospitalized between March 1 and May 15, 2020, with COVID-19 and not receiving corticosteroids for chronic use. Data were analyzed from October 2021 to March 2022. Corticosteroid exposure was defined as more than 0.5 mg/kg methylprednisolone equivalent in a 24-hour period. For target trial emulation, exposures were corticosteroids for 6 days if and when a patient met criteria for severe hypoxia vs no corticosteroids. For approaches common in clinical literature, treatment definitions used for variables in Cox regression models varied by study design (no time frame, 1 day, and 5 days from time of severe hypoxia). The main outcome was 28-day mortality from time of hospitalization. The association of corticosteroids with mortality for patients with moderate to severe COVID-19 was assessed using the World Health Organization (WHO) meta-analysis of corticosteroid randomized clinical trials as a benchmark. A total of 3298 patients (median [IQR] age, 65 [53-77] years; 1970 [60%] men) were assessed, including 423 patients who received corticosteroids at any point during hospitalization and 699 patients who died within 28 days of hospitalization. Target trial emulation analysis found corticosteroids were associated with a reduced 28-day mortality rate, from 32.2%; (95% CI, 30.9%-33.5%) to 25.7% (95% CI, 24.5%-26.9%). This estimate is qualitatively identical to the WHO meta-analysis odds ratio of 0.66 (95% CI, 0.53-0.82). Hazard ratios using methods comparable with current corticosteroid research range in size and direction, from 0.50 (95% CI, 0.41-0.62) to 1.08 (95% CI, 0.80-1.47). These findings suggest that clinical research based on observational data can be used to estimate findings similar to those from randomized clinical trials; however, the correctness of these estimates requires designing the study and analyzing the data based on principles that are different from the current standard in clinical research.

Oncology Drug Effectiveness from Electronic Health Record Data Calibrated Against RCT Evidence: The PARSIFAL Trial Emulation.

BackgroundThe use of electronic health records (EHR) data to assess drug effectiveness in clinical oncology practice is of great interest to regulators, clinicians, and payers. However, the utility of EHR data in clinical effectiveness studies may be limited by missing data, unmeasured confounding, and imperfect outcome surveillance. This study sought to emulate and compare the results of a randomized controlled trial investigating the efficacy of palbociclib with fulvestrant vs letrozole in advanced breast cancer.MethodsThis was a cohort study using longitudinal EHR data derived from outpatient oncology practices in the United States. Eligibility criteria from the PARSIFAL trial were emulated as closely as possible. Patients were included if they had hormone-positive, human epidermal growth factor receptor – 2 (HER-2) negative metastatic breast cancer and had no record of prior treatment for metastatic disease. Patients initiating first-line treatment with palbociclib and fulvestrant following their first record of metastasis were compared to those initiating palbociclib and letrozole on the same day. Treatments were ascertained by oncology medication ordering records in the data source. The primary outcome was death as recorded in the oncologists’ EHR systems.ResultsThere were 1886 eligible women in the study cohort. Although the 3-year survival was meaningfully lower in clinical practice (59%) compared to the randomized trial (78%), the relative effect size was a hazard ratio (HR) of 1.07 (95% CI: 0.86–1.35), similar to the randomized trial (HR = 1.00; 95% CI: 0.68–1.48).ConclusionDespite common challenges encountered in EHR-based studies, it is possible to achieve similar conclusions to emulated randomized trials with the application of analytic approaches that address missing data, confounding, and selection bias. This is a promising finding in light of other emulations and ongoing efforts to improve data from clinical practice and causal analytics.

Glycemic Control Trajectories and Risk of Perinatal Complications Among Individuals With Gestational Diabetes

Glycemic control is the cornerstone of gestational diabetes management. Glycemic control trajectories account for differences in longitudinal patterns throughout pregnancy; however, studies on glycemic control trajectories are scarce. To examine whether glycemic control trajectories from gestational diabetes diagnosis to delivery were associated with differential risk of perinatal complications. This population-based cohort study included individuals with gestational diabetes with longitudinal electronic health record data from preconception to delivery who received prenatal care at Kaiser Permanente Northern California (KPNC) and were enrolled in KPNC's telemedicine-based gestational diabetes care program between January 2007 and December 2017. Data analysis was conducted from September 2021 to January 2022. Glycemic control trajectories were derived using latent class modeling based on the American Diabetes Association's recommended self-monitoring of blood glucose measurements. Optimal glycemic control was defined as at least 80% of all measurements meeting the targets at KPNC clinical settings. Multivariable Poisson regression models were used to estimate the associations of glycemic control trajectories with cesarean delivery, preterm birth, shoulder dystocia, large- and small-for-gestational-age, and neonatal intensive care unit admission and stay of 7 days or longer. Among a total of 26 774 individuals (mean [SD] age, 32.9 [5.0] years; 11 196 Asian or Pacific Islander individuals [41.8%], 1083 Black individuals [4.0%], 7500 Hispanic individuals [28.0%], and 6049 White individuals [22.6%]), 4 glycemic control trajectories were identified: stably optimal (10 528 individuals [39.3%]), rapidly improving to optimal (9151 individuals [34.2%]), slowly improving to near-optimal (4161 individuals [15.5%]), and slowly improving to suboptimal (2934 individuals [11.0%]). In multivariable models with the rapidly improving to optimal trajectory group as the reference group, glycemic control trajectories were associated with perinatal complications with a gradient across stably optimal to slowly improving to suboptimal. For individuals in the stably optimal trajectory group, there were lower risks of cesarean delivery (adjusted relative risk [aRR], 0.93 [95% CI, 0.89-0.96]), shoulder dystocia (aRR, 0.75 [95% CI, 0.61-0.92]), large-for-gestational age (aRR, 0.74 [95% CI, 0.69-0.80]), and neonatal intensive care unit admission (aRR, 0.90 [95% CI, 0.83-0.97]), while for patients in the slowly improving to suboptimal glycemic control trajectory group, risks were higher for cesarean delivery (aRR, 1.18 [95% CI, 1.12-1.24]; (P for trend < .001), shoulder dystocia (aRR, 1.41 [95% CI, 1.12-1.78]; P for trend < .001), large-for-gestational-age (aRR, 1.42 [95% CI, 1.31-1.53]; P for trend < .001), and neonatal intensive care unit admission (aRR, 1.33 [95% CI, 1.20-1.47]; P for trend < .001). The risk of small-for-gestational-age was higher in patients in the stably optimal group (aRR, 1.10 [95% CI, 1.02-1.20]) and lower in the slowly improving to suboptimal group (aRR, 0.63 [95% CI, 0.53-0.75]). These findings suggest that slowly improving to near-optimal and slowly improving to suboptimal glycemic control trajectories were associated with increased risk of perinatal complications. Future interventions should help individuals achieve glycemic control early after gestational diabetes diagnosis and throughout pregnancy to decrease the risk of perinatal complications.

Keeping synthetic patients on track: feedback mechanisms to mitigate performance drift in longitudinal health data simulation.

Synthetic data are increasingly relied upon to share electronic health record (EHR) data while maintaining patient privacy. Current simulation methods can generate longitudinal data, but the results are unreliable for several reasons. First, the synthetic data drifts from the real data distribution over time. Second, the typical approach to quality assessment, which is based on the extent to which real records can be distinguished from synthetic records using a critic model, often fails to recognize poor simulation results. In this article, we introduce a longitudinal simulation framework, called LS-EHR, which addresses these issues. LS-EHR enhances simulation through conditional fuzzing and regularization, rejection sampling, and prior knowledge embedding. We compare LS-EHR to the state-of-the-art using data from 60 000 EHRs from Vanderbilt University Medical Center (VUMC) and the All of Us Research Program. We assess discrimination between real and synthetic data over time. We evaluate the generation process and critic model using the area under the receiver operating characteristic curve (AUROC). For the critic, a higher value indicates a more robust model for quality assessment. For the generation process, a lower value indicates better synthetic data quality. The LS-EHR critic improves discrimination AUROC from 0.655 to 0.909 and 0.692 to 0.918 for VUMC and All of Us data, respectively. By using the new critic, the LS-EHR generation model reduces the AUROC from 0.909 to 0.758 and 0.918 to 0.806. LS-EHR can substantially improve the usability of simulated longitudinal EHR data.

Impact of malnutrition on clinical outcomes in patients diagnosed with COVID-19.

BackgroundCoronavirus disease 2019 (COVID‐19) is now the third leading cause of death in the United States. Malnutrition in hospitalized patients increases risk of complications. However, the effect of malnutrition on outcomes in patients infected is unclear. This study aims to identify the impact of malnutrition on mortality and adverse hospital events in patients hospitalized with COVID‐19.MethodsThis study used data from the National COVID Cohort Collaborative (N3C), a COVID‐19 repository containing harmonized, longitudinal electronic health record data from US health systems. Malnutrition was categorized into three groups based on condition diagnosis: (1) none documented, (2) history of malnutrition, and (3) hospital‐acquired malnutrition. Multivariable logistic regression was performed to determine whether malnutrition was associated with mortality and adverse events, including mechanical ventilation, acute respiratory distress syndrome, extracorporeal membrane oxygenation, and hospital‐acquired pressure injury, in hospitalized patients with COVID‐19.ResultsOf 343,188 patients hospitalized with COVID‐19, 11,206 had a history of malnutrition and 15,711 had hospital‐acquired malnutrition. After adjustment for potential confounders, odds of mortality were significantly higher in patients with a history of malnutrition (odds ratio [OR], 1.71; 95% confidence interval [CI], 1.63–1.79; P < 0.001) and hospital‐acquired malnutrition (OR, 2.5; 95% CI, 2.4–2.6; P < 0.001). Adjusted odds of adverse hospital events were also significantly elevated in both malnutrition groups.ConclusionsResults indicate the risk of mortality and adverse inpatient events in adults with COVID‐19 is significantly higher in patients with malnutrition. Prevention, diagnosis, and treatment of malnutrition could be a key component in improving outcomes in these patients.

COVID-19 Surveillance in the Biobank at the Colorado Center for Personalized Medicine: Observational Study.

BackgroundCharacterizing the experience and impact of the COVID-19 pandemic among various populations remains challenging due to the limitations inherent in common data sources, such as electronic health records (EHRs) or cross-sectional surveys.ObjectiveThis study aims to describe testing behaviors, symptoms, impact, vaccination status, and case ascertainment during the COVID-19 pandemic using integrated data sources.MethodsIn summer 2020 and 2021, we surveyed participants enrolled in the Biobank at the Colorado Center for Personalized Medicine (CCPM; N=180,599) about their experience with COVID-19. The prevalence of testing, symptoms, and impacts of COVID-19 on employment, family life, and physical and mental health were calculated overall and by demographic categories. Survey respondents who reported receiving a positive COVID-19 test result were considered a “confirmed case” of COVID-19. Using EHRs, we compared COVID-19 case ascertainment and characteristics in EHRs versus the survey. Positive cases were identified in EHRs using the International Statistical Classification of Diseases, 10th revision (ICD-10) diagnosis codes, health care encounter types, and encounter primary diagnoses.ResultsOf the 25,063 (13.9%) survey respondents, 10,661 (42.5%) had been tested for COVID-19, and of those, 1366 (12.8%) tested positive. Nearly half of those tested had symptoms or had been exposed to someone who was infected. Young adults (18-29 years) and Hispanics were more likely to have positive tests compared to older adults and persons of other racial/ethnic groups. Mental health (n=13,688, 54.6%) and family life (n=12,233, 48.8%) were most negatively affected by the pandemic and more so among younger groups and women; negative impacts on employment were more commonly reported among Black respondents. Of the 10,249 individuals who responded to vaccination questions from version 2 of the survey (summer 2021), 9770 (95.3%) had received the vaccine. After integration with EHR data up to the time of the survey completion, 1006 (4%) of the survey respondents had a discordant COVID-19 case status between EHRs and the survey. Using all longitudinal EHR and survey data, we identified 11,472 (6.4%) COVID-19-positive cases among Biobank participants. In comparison to COVID-19 cases identified through the survey, EHR-identified cases were younger and more likely to be Hispanic.ConclusionsWe found that the COVID-19 pandemic has had far-reaching and varying effects among our Biobank participants. Integrated data assets, such as the Biobank at the CCPM, are key resources for population health monitoring in response to public health emergencies, such as the COVID-19 pandemic.

Predicting emergency department utilization among children with asthma using deep learning models

Pediatric asthma is a leading cause of emergency department (ED) utilization, which is expensive and often preventable. Therefore, development of ED utilization predictive models that can accurately predict patients at high-risk of frequent ED use and subsequently steering their treatment pathway towards more personalized interventions, has high clinical utility. In this paper, we investigate the extent to which deep learning models, specifically recurrent neural networks (RNNs), coupled with routinely collected electronic health record (EHR) clinical data can predict the frequency of emergency department utilization among children with asthma.We use retrospective longitudinal EHR data of 87,413 children with asthma aged 0–18 years, who were attributed to one or more healthcare facility for at least 2 consecutive years between 2000–2013. The models were trained for the task of predicting the frequency of emergency department visits in the next 12 months. We compared prediction results of three recurrent neural network (RNN) models: bidirectional long short-term memory (BiLSTM), bidirectional gated recurrent unit (BiGRU), and reverse time attention model (RETAIN), to a baseline multinomial logistic regression model. We assessed the predictive accuracy of the models using receiver operating characteristic curve (AUC–ROC), precision–recall curve (AUC-PR), and F1-score.The results indicated that all RNN models have similar performances reaching AUC–ROC: 0.85, AUC-PR: 0.74, and F1-score: 0.61, compared to AUC–ROC: 0.81, AUC-PR: 0.69, and F1-score: 0.56 for a baseline multinomial logistic regression.Predictive models created from large routinely available EHR data using RNN models can accurately identify children with asthma at high-risk of repeated ED visits, without interacting with the patient or collecting information beyond the patient’s EHR.

Perinatal Complications in Individuals in California With or Without SARS-CoV-2 Infection During Pregnancy

Additional research from population-based studies is needed to inform the treatment of SARS-CoV-2 infection during pregnancy and to provide health risk information to pregnant individuals. To assess the risk of perinatal complications associated with SARS-CoV-2 infection and to describe factors associated with hospitalizations. This population-based cohort study included 43 886 pregnant individuals with longitudinal electronic health record data from preconception to delivery who delivered at Kaiser Permanente Northern California between March 1, 2020, and March 16, 2021. Individuals with diagnostic codes for COVID-19 that did not have a confirmatory polymerase chain reaction test for SARS-CoV-2 were excluded. SARS-CoV-2 infection detected by polymerase chain reaction test (from 30 days before conception to 7 days after delivery) as a time varying exposure. Severe maternal morbidity including 21 conditions (eg, acute myocardial infarction, acute renal failure, acute respiratory distress syndrome, and sepsis) that occurred at any time during pregnancy or delivery; preterm birth; pregnancy hypertensive disorders; gestational diabetes; venous thromboembolism (VTE); stillbirth; cesarean delivery; and newborn birth weight and respiratory conditions. Standardized mean differences between individuals with and without SARS-CoV-2 were calculated. Cox proportional hazards regression was used to estimate the hazard ratios (HRs) and 95% CIs for the association between SARS-CoV-2 infection and perinatal complications and hospitalization and to consider the timing of SARS-CoV-2 infection relative to outcomes. In this study of 43 886 pregnant individuals (mean [SD] age, 30.7 [5.2] years), individuals with a SARS-CoV-2 infection (1332 [3.0%]) were more likely to be younger, Hispanic, multiparous individuals with a higher neighborhood deprivation index and obesity or chronic hypertension. After adjusting for demographic characteristics, comorbidities, and smoking status, individuals with SARS-CoV-2 infection had higher risk for severe maternal morbidity (HR, 2.45; 95% CI, 1.91-3.13), preterm birth (<37 weeks; HR, 2.08; 95% CI, 1.75-2.47), and VTE (HR, 3.08; 95% CI, 1.09-8.74) than individuals without SARS-CoV-2. SARS-CoV-2 infection was also associated with increased risk of medically indicated preterm birth (HR, 2.56; 95% CI, 2.06-3.19); spontaneous preterm birth (HR, 1.61; 95% CI, 1.22-2.13); and early (HR, 2.52; 95% CI, 1.49-4.24), moderate (HR, 2.18; 95% CI, 1.25-3.80), and late (HR, 1.95; 95% CI, 1.61-2.37) preterm birth. Among individuals with SARS-CoV-2 infection, 76 (5.7%) had a hospitalization; pregestational diabetes (HR, 7.03; 95% CI, 2.22-22.2) and Asian or Pacific Islander (HR, 2.33; 95% CI, 1.06-5.11) and Black (HR, 3.14; 95% CI, 1.24-7.93) race and ethnicity were associated with an increased risk of hospitalization. In this cohort study, SARS-CoV-2 infection was associated with increased risk of severe maternal morbidity, preterm birth, and VTE. The study findings inform clinicians and patients about the risk of perinatal complications associated with SARS-CoV-2 infection in pregnancy and support vaccination of pregnant individuals and those planning conception.

Characterization of Symptoms and Symptom Clusters for Type 2 Diabetes Using a Large Nationwide Electronic Health Record Database.

A variety of symptoms may be associated with type 2 diabetes and its complications. Symptoms in chronic diseases may be described in terms of prevalence, severity, and trajectory and often co-occur in groups, known as symptom clusters, which may be representative of a common etiology. The purpose of this study was to characterize type 2 diabetes-related symptoms using a large nationwide electronic health record (EHR) database. We acquired the Cerner Health Facts, a nationwide EHR database. The type 2 diabetes cohort (n = 1,136,301 patients) was identified using a rule-based phenotype method. A multistep procedure was then used to identify type 2 diabetes-related symptoms based on International Classification of Diseases, 9th and 10th revisions, diagnosis codes. Type 2 diabetes-related symptoms and co-occurring symptom clusters, including their temporal patterns, were characterized based the longitudinal EHR data. Patients had a mean age of 61.4 years, 51.2% were female, and 70.0% were White. Among 1,136,301 patients, there were 8,008,276 occurrences of 59 symptoms. The most frequently reported symptoms included pain, heartburn, shortness of breath, fatigue, and swelling, which occurred in 21-60% of the patients. We also observed over-represented type 2 diabetes symptoms, including difficulty speaking, feeling confused, trouble remembering, weakness, and drowsiness/sleepiness. Some of these are rare and difficult to detect by traditional patient-reported outcomes studies. To the best of our knowledge, this is the first study to use a nationwide EHR database to characterize type 2 diabetes-related symptoms and their temporal patterns. Fifty-nine symptoms, including both over-represented and rare diabetes-related symptoms, were identified.

Assessment of Risk Prediction Models Applied to Longitudinal Electronic Health Record Data for the Prediction of Major Cardiovascular Events in the Presence of Data Shifts

Assessment of Risk Prediction Models Applied to Longitudinal Electronic Health Record Data for the Prediction of Major Cardiovascular Events in the Presence of Data Shifts