Record Data Research Articles

Risk Factors for Lung Cancer in Non-smoking Women, North Sumatera, Indonesia

Basic Health Research Data in 2018 reported that lung cancer in women in Indonesia ranked third after breast cancer and cervical cancer. A person's risk of developing lung cancer can be evaluated by looking at several things, namely age and gender, genetic predisposition to cancer, tobacco use, and exposure to toxic agents at home or work. The purpose of this research is to obtain data on several risk factors with the incidence of lung cancer in non-smoking women at Adam Malik General Hospital and Santa Elisabeth General Hospital. This research is a case-control matching study involving 224 research subjects through medical record data collection from January 2019 to December 2020. All research subjects were recorded from medical record data confirmed by questionnaires related to risk factors, namely ages, exposure to cigarette smoke (passive smokers), genetics, mosquito coils, limes, pesticides, garbage burning, exposure to firewood, and air pollution. Data were analyzed by logistic regression test. The average age in the case group was 57.86 years, while in the control group, the average age was 47.4 years. The largest age group in both study groups was 40 years. In this research, five risk factors of lung cancer influenced the incidence of lung cancer in non-smoking women, namely age (p = 0.008), passive smokers (p = 0.043), genetic factors (p < 0.001), exposure to firewood (p = 0.007), and air pollution factors (p = 0.041). It was estimated that risk factors that significantly influence the incidence of lung cancer in women are not the age of 40 years, genetic factors, passive smokers, and exposure to firewood.

Pharmacogenetics: Opportunities for the All of Us Research Program and Other Large Data Sets to Advance the Field.

Pharmacogenetic variation is common and an established driver of response for many drugs. There has been tremendous progress in pharmacogenetics knowledge over the last 30 years and in clinical implementation of that knowledge over the last 15 years. But there have also been many examples where translation has stalled because of the lack of available data sets for discovery or validation research. The recent availability of data from very large cohorts with linked genetic, electronic health record, and other data promises new opportunities to advance pharmacogenetics research. This review presents the stages from pharmacogenetics discovery to widespread clinical adoption using prominent gene-drug pairs that have been implemented into clinical practice as examples. We discuss the opportunities that the All of Us Research Program and other large biorepositories with genomic and linked electronic health record data present in advancing and accelerating the translation of pharmacogenetics into clinical practice.

Knowledge, attitudes, and practices regarding urinary tract infections among women in the United Arab Emirates.

Urinary tract infections (UTIs), which are infections of the kidneys, ureters, bladder, or urethra, are a worldwide public health concern. As compared to men, women are more prone to UTIs. There have been several studies that explore the knowledge, attitudes, and practices of women regarding UTIs in different countries, but no such study has been conducted in the UAE; therefore, we conducted this study in the UAE setting. This study was conducted using an online survey created on Microsoft Forms. The minimum sample size required for our study was 385. This study was conducted after obtaining ethical approval from the Research Ethics Committee at the University of Sharjah. A personally designed questionnaire consisting of 21 items, derived from previous research was used to record data. The data was analyzed using SPSS. A total of 475 females were included in the study. Most respondents were aged 18-30 years (47.4%). Our study found that a majority of the participants (69.7%) correctly identified bacteria as the most common cause of UTIs. With regards to practices undertaken during UTIs, among the participants with a history of UTI episodes, 32.6% waited 24-48 hours before seeking medical attention at a hospital or clinic, while 10% did not visit a hospital at all. Distinct trends were found when comparing demographic factors with knowledge levels. Most notably, the age group of 18-30 years showed the highest percentage of high-knowledge individuals (49%) compared to other age groups (p < 0.05). Education level was significantly (p = 0.003) associated with UTI knowledge. Going to the hospital/clinic was reported by 41% with high knowledge but only by 20% of those with poor knowledge. Moreover, a higher proportion of individuals with high knowledge sought medical attention immediately within 24 hours (47%). Most of the participants possessed adequate knowledge regarding UTIs. Higher knowledge levels were associated with more proactive and appropriate health behaviors, such as seeking medical attention promptly and drinking more water.

PAIN AND MOBILITY OF OLDER ADULTS IN MEDICAL AND SURGICAL UNITS: AN EHR-BASED DATA ANALYSIS

Abstract Pain accelerates age-related decline in mobility. One in five hospitalized older adults have pain. Pain experience varies by the type of hospitalization. The purpose of this study was to describe pain and mobility in hospitalized older adults. We hypothesized that pain would be higher in patients with lower mobility. We analyzed electronic health record data from an academic health center in Florida between 2012 and 2021. We used the mean pain score and mean score on the mobility subscale of the Braden Scale for each encounter. We used the Wilcoxon Rank-Sum Test and Pearson’s correlation coefficient to measure the differences and relationship between pain and mobility. The cohort included 19,854 older adults (13,878 medical and 5,976 surgical patients). The mean mobility score was 3.05 ± 0.57 in medical and 2.88 ± 0.51 in surgical patients (W = 48697407, p &amp;lt; 0.001). The mean pain score was 1.93 ± 2.72 in medical and 2.25 ± 2.61 in surgical patients (W = 36893487, p &amp;lt; 0.001). There was no correlation between pain and mobility in either medical (r = 0.01, p = 0.427) or surgical patients (r = 0.02, p = 0.085). Mean pain intensity may not be associated with mobility in hospitalized older adults. Instead, the highest level of pain or uncontrolled pain may be associated with mobility decline. Our findings provide motivation for a future study to account for confounders of the relationship between pain and mobility, such as analgesics, and temporality of pain and mobility assessments in the hospital.

Correlation Between Serum CA-125 Levels and Age in Advanced-Stage Epithelial Ovarian Cancer Patients at Dr. Soetomo General Academic Hospital, Surabaya, Indonesia (2021-2022)

Ovarian cancer is a gynecological malignancy that originates in the ovaries. Symptoms of ovarian cancer often remain unnoticed in its early stages, leading to a diagnosis at an advanced stage. CA-125 is a widely recognized biomarker used in ovarian cancer for diagnosis, monitoring treatment response, and predicting prognosis. Research has shown that CA-125 levels can vary with age and are influenced by multiple biological factors, including hormonal changes and the progression of ovarian cancer. Age-related factors can influence CA-125 levels, which may affect the interpretation of such examination results. This study aims to determine the correlation between serum CA-125 levels and ages in advanced-stage epithelial ovarian cancer patients. This research is a retrospective study using the comparative analytic research method by taking the medical record data of epithelial ovarian cancer with advanced-stage patients treated in oncology clinic of Dr. Soetomo General Hospital Surabaya for January 1, 2021 – December 2022. Based on the analysis of 61 patient subjects categorized into three age groups, the median serum CA-125 levels for patients aged &lt;40 years were 172.85 U/mL, for those aged 40-50 years the median was 261 U/mL, and for patients aged &gt;50 years the median was 116.2 U/mL. The Kruskal-Wallis test (p-value) yielded a result of 0.518, indicating that there was no statistically significant relationship between serum CA-125 levels of patients and patients’ age (p = 0.518). In addition, there is no correlation between serum CA-125 levels and age in advanced-stage epithelial ovarian cancer patients at Dr. Soetomo General Hospital, Surabaya.

Profile and resistance pattern of Stenotrophomonas maltophilia in Dr. Soetomo General Academic Hospital During 2020-2022 Period

Introduction: Stenotrophomonas maltophilia is a challenging respiratory tract pathogen with increasing number of cases. S. maltophilia also shows a high resistance towards various antibiotics. The aim for this study is to know the antibiotic resistance pattern and profile of patient infected with S. maltophilia in order to find out effective antibiotic therapy in treating future cases. Methods: This cross-sectional study uses descriptive analytic method using medical record data of 36 patients infected with S. maltophilia in Dr. Soetomo General Academic Hospital during 2020-2022 period. The measured variables included antibiotic resistance pattern, age, sex, clinical diagnosis, comorbidity, signs and symptoms, medical device usage, total leukocyte count, and mortality rate. Result: S. maltophilia shows highest sensitivity index towards Trimetophrim-sulfomethoxazole (56%). Cases of S. maltophilia is most often found in 0-17 years age group (53%) and male (63%). It may cause infection whether alone or alongside other bacteria. The most common clinical diagnosis that occurs in S. maltophilia infections is pneumonia (81%). Dyspnea is the most occurred symptoms. Hypertension is a common comorbidity. IV infusion is the most used medical device. Might cause infection alongside other bacteria. Leukocytosis is often present. Mortality rate among subjects is 50%. Conclusion: Trimetophrim-sulfomethoxazole is the best treatment for S. maltophilia infections. Cases most often occurs in age 0-17 years, male, diagnosed with pneumonia, and has hypertension as comorbidity. In the future, antibiotics must be used wisely to prevent bacterial resistance and regular changing of medical device is encouraged to prevent nosocomial infections.

Therapy Pattern and Potential for Drug Interactions In COPD Patients at S.K. Lerik Regional Hospital Of Kupang City

Chronic Obstructive Pulmonary Disease (COPD) is a respiratory disease characterized by persistent airflow limitation and is associated with an increased chronic inflammatory response to certain irritants, gases, or particles. COPD management uses a combination of drugs to treat the disease's symptoms, which varies depending on the patient's condition. The combination of these drugs has the potential to cause drug interactions that result in Drug Related Problems (DRP). A DRP is defined as a change in the effect of a drug due to being given together with other drugs. This study aims to describe the pharmacological therapy given and the potential for drug interactions in COPD therapy at the S.K Lerik Regional General Hospital. This research is descriptive, non-experimental, and observational, using retrospective data. The sample in this study was the medical record data of COPD patients hospitalized at the S.K Lerik Regional General Hospital, with a total sampling technique. The study’s results revealed that the COPD patients’ treatment at the S.K Lerik Hospital mainly uses a combination of corticosteroid therapy (methylprednisolone), bronchodilators (salbutamol), and antibiotics (ceftriaxone). The highest potential for drug interactions was at the mild severity level with the interaction between methylprednisolone and salbutamol, which was given to all patients (12 patients). Due to the high potential for drug interactions of these two drugs, further monitoring is needed in COPD patients to determine the manifestations of these drug interactions in clinical conditions

Rebuilding the clinical and operational acute care pharmacist productivity model: Leveraging electronic health record data and analytics at an academic medical center.

In an effort to expedite the publication of articles, AJHP is posting manuscripts online as soon as possible after acceptance. Accepted manuscripts have been peer-reviewed and copyedited, but are posted online before technical formatting and author proofing. These manuscripts are not the final version of record and will be replaced with the final article (formatted per AJHP style and proofed by the authors) at a later time. Acute care pharmacy practice continues to evolve, and useful pharmacist productivity models should reflect contemporary practice. The purpose of this project was to rebuild an acute care pharmacist productivity model to accurately capture and categorize the variable workload of acute care generalists and specialists using electronic health record (EHR) data reports and analytic capabilities. The acute care pharmacist productivity model was rebuilt with 5 variable workload drivers, including order verification, medication preparation verification (product check), clinical scoring, pharmacist-documented progress notes, and pharmacy intensity score-weighted patient admissions and patient days. Reports from the EHR database were used to capture all workload drivers. For each variable workload driver, more granular categories included in the reports were assigned time standards based on the work's complexity. Work output data, represented as units of service, were mapped by the user's job title (ie, generalist or specialist) and by the inpatient department. The rebuilt pharmacist productivity model leverages EHR data and standard documentation of patient care activities to capture workload and reflect clinical practice. The EHR-generated reports enabled construction of a comprehensive and sustainable productivity model for acute care pharmacists without the need for additional manual documentation for productivity purposes.

Physician-Visit Frequency and Its Impact on Glycemic Control in People With Type 2 Diabetes: Quantifying Care Acceptance Parameters From Retrospective Electronic Health Record Data.

Type 2 diabetes is a metabolic disorder characterized by insulin resistance and hyperglycemia affecting many individuals worldwide. For effective management, adherence to recommended physician visits is important, along with lifestyle modification and pharmacological interventions. Regular doctor visits can improve adherence and help prevent complications. This study examined how doctor-visit frequency impacts blood glucose (BG) control in people with type 2 diabetes. This retrospective, single-center study included adults with type 2 diabetes who had at least two visits and two hemoglobin A1c (HbA1c) values recorded at least 90 days apart between March 2019 and November 2023. Visit per quarter ratio (VQR) and the standardized delta HbA1c (ΔHbA1c) were defined. Data were analyzed for descriptive statistics and compared for significance between the groups. Statistical differences among the groups were determined using the Kruskal-Wallis test and Dwass-Steel-Critchlow-Fligner pairwise comparisons. Five hundred seventy-seven participants with type 2 diabetes were analyzed, with a mean age of 53.9 years (±12.7). The HbA1c outcome was significantly lower in the good compliance group (7.3%) than in the poor (7.7%). The high VQR group had a significantly lower HbA1c outcome (7.4%) than the low VQR group (7.7%); 50.98% (n = 104) in the low VQR group, 65.54% (n = 97) in the medium VQR group, and 60.44% (n = 136) in the high VQR group achieved HbA1c below the target. The mean ΔHbA1c was significantly lower in the good compliance group compared to the poor compliance group. The average follow-up durations were 11.22 quarters (±4.11) for low VQR, 9.75 quarters (±4.69) for medium VQR, and 5.50 quarters (±4.84) for high VQR. Ahigher frequency of follow-up may be needed to encourage people with type 2 diabetes mellitus to visit their doctor regularly. The frequency of doctor visits has a positive impact on BG control. Regular visits enable timely adjustment of therapy and ensure high compliance with prescribed treatment. These findings have significant implications for mitigating nonadherence in chronic conditions like type 2 diabetes mellitus and warrant further investigation.

Satisfaction and Preferences for Infusion Therapies in Advanced Parkinson's Disease-Patient Perspective.

Background and Objectives: The rapid growth of the number of advanced Parkinson's disease (PD) patients has caused a significant increase in the use of device-aided therapies (DATs), including levodopa-carbidopa intestinal gel (LCIG) and continuous subcutaneous apomorphine infusion (CSAI). The objective of this study was to evaluate patients' satisfaction and the factors influencing preferences for CSAI and LCIG. Materials and Methods: The research focused on individuals diagnosed with advanced PD undergoing DAT at the Neurology Department of the University Hospital in Katowice. A telephone survey conducted between June and July 2024 evaluated the experiences of patients with LCIG and CSAI. The Parkinson's Disease Questionnaire (PDQ-8) and the Stress Scale for Family Caregivers (BSFC-s) were applied. Based on medical record data comprising reasons for the exclusion of individuals, disease-related and treatment data were collected. Results: Among the original cohort of 64 patients, 50 completed the survey, including 31 who might choose between infusion therapies. The average patient ages were 70.6 ± 4.7 (CSAI) and 71.2 ± 7.2 years (LCIG), with disease durations of 15 (IQR: 12-19) and 18 (IQR: 13-19) years, respectively. LCIG patients presented higher PDQ-8 scores (20 (IQR: 13-27) vs. 13 (IQR: 6-19), p = 0.008), and higher BSFC-s scores (19 (IQR: 12-21) vs. 9 (IQR: 2.5-13), p = 0.011). Furthermore, significant factors influencing patient preferences included fear of surgery (75% vs. 36.8%, p = 0.043) and concerns about DAT safety (83.3% vs. 47.4%, p = 0.049). Conclusions: LCIG and CSAI therapies offer benefits and disadvantages, with safety concerns and fear of surgery seeming to be decisive in the decision-making process.

Reusing routine electronic health record data for nationwide COVID-19 surveillance in nursing homes: barriers, facilitators, and lessons learned

BackgroundAt the beginning of the COVID-19 pandemic in 2020, little was known about the spread of COVID-19 in Dutch nursing homes while older people were particularly at risk of severe symptoms. Therefore, attempts were made to develop a nationwide COVID-19 repository based on routinely recorded data in the electronic health records (EHRs) of nursing home residents. This study aims to describe the facilitators and barriers encountered during the development of the repository and the lessons learned regarding the reuse of EHR data for surveillance and research purposes.MethodsUsing inductive content analysis, we reviewed 325 documents written and saved during the development of the COVID-19 repository. This included meeting minutes, e-mails, notes made after phone calls with stakeholders, and documents developed to inform stakeholders. We also assessed the fitness for purpose of the data by evaluating the completeness, plausibility, conformity, and timeliness of the data.ResultsKey facilitators found in this study were: 1) inter-organizational collaboration to create support; 2) early and close involvement of EHR software vendors; and 3) coordination and communication between partners. Key barriers that hampered the fitness of EHR data for surveillance were: 1) changes over time in national SARS-CoV-2 testing policy; 2) differences between EHR systems; 3) increased workload in nursing homes and lack of perceived urgency; 4) uncertainty regarding the legal requirements for extracting EHR data; 5) the short notice at which complete and understandable information about the repository had to be developed; and 6) lack of clarity about the differences between various COVID-19 monitors.ConclusionsDespite the urgent need for information on the spread of SARS‐CoV‐2 among nursing home residents, setting up a repository based on EHR data proved challenging. The facilitators and barriers found in this study affected the extent to which the data could be used. We formulated nine lessons learned for developing future repositories based on EHR data for surveillance and research purposes. These lessons were in three main areas: legal framework, contextual circumstances, and quality of the data. Currently, these lessons are being applied in setting up a new registry in the nursing home sector.

Practice and Match Workload Characteristics of Elite Men’s Ice Hockey

The purpose of this study was to record and present data related to the external workload demand of elite male ice hockey players, in both matches and practices, including the differences between positions, as well as time on ice values for the population. External workload monitoring was performed using a CatapultTMinertial measurement unit system and data was collected for the duration of one full season in the Kontinental Hockey League (Russia). The study found forwards had significantly higher (p ≤ 0.05) absolute and relative values in both games and practices for most metrics, when compared to defensemen, and regardless of position, games had significantly higher workload demand than practices. Furthermore, defensemen had significantly higher (p ≤ 0.05) time on ice values than forwards. Ice time was not spread evenly between players of the same position, but was consistent for each individual athlete. The findings can be used by coaches to assist in the planning of on-ice training to optimize player performance and readiness.

Development and validation of computable social phenotypes for health-related social needs.

Measurement of health-related social needs (HRSNs) is complex. We sought to develop and validate computable phenotypes (CPs) using structured electronic health record (EHR) data for food insecurity, housing instability, financial insecurity, transportation barriers, and a composite-type measure of these, using human-defined rule-based and machine learning (ML) classifier approaches. We collected HRSN surveys as the reference standard and obtained EHR data from 1550 patients in 3 health systems from 2 states. We followed a Delphi-like approach to develop the human-defined rule-based CP. For the ML classifier approach, we trained supervised ML (XGBoost) models using 78 features. Using surveys as the reference standard, we calculated sensitivity, specificity, positive predictive values, and area under the curve (AUC). We compared AUCs using the Delong test and other performance measures using McNemar's test, and checked for differential performance. Most patients (63%) reported at least one HRSN on the reference standard survey. Human-defined rule-based CPs exhibited poor performance (AUCs=.52 to .68). ML classifier CPs performed significantly better, but still poor-to-fair (AUCs = .68 to .75). Significant differences for race/ethnicity were found for ML classifier CPs (higher AUCs for White non-Hispanic patients). Important features included number of encounters and Medicaid insurance. Using a supervised ML classifier approach, HRSN CPs approached thresholds of fair performance, but exhibited differential performance by race/ethnicity. CPs may help to identify patients who may benefit from additional social needs screening. Future work should explore the use of area-level features via geospatial data and natural language processing to improve model performance.

Interpersonal Continuity of Care May Help Delay Progression to Type 2 Diabetes.

The association between interpersonal continuity of care (CoC) and progression from the prediabetic state to Type 2 Diabetes (T2D) remains unknown. To evaluate the association between interpersonal CoC and the progression to T2D among persons with prediabetes. A retrospective cohort study using electronic health record (EHR) data from 6620 patients at Geisinger, a large rural health care system in Danville, PA. Cox regression methods were used to estimate the hazard ratio associated with progression to T2D within 3-years of being diagnosed with prediabetes. One additional visit with the primary care provider most frequently seen by the patient is associated with 14% decreased risk (HR = 0.86; 95% CI = 0.85, 0.87; P < .001) of transitioning to type 2 diabetes within 3 years of being diagnosed with prediabetes. This study demonstrates an association between increased interpersonal CoC after a person is diagnosed with prediabetes and a reduced risk of progressing to T2D within 3 years.

Accuracy of Electronic Health Record Phenotypes to Detect Recognition of Hypertension in Pediatric Primary Care.

To evaluate the accuracy of extractable electronic health record (EHR) data to define clinician recognition of hypertension in pediatric primary care. We used EHR data to perform a cross-sectional study of children aged 3-18 years at well-visits in Connecticut from 2018-2023 (n=50,290) that had either: (1) incident hypertension (hypertensive BP at the well-visit and ≥2 prior hypertensive BPs without prior diagnosis of hypertension); or (2) isolated hypertensive BP at the well-visit without necessarily having prior hypertensive BPs. We tested the accuracy of EHR phenotypes to detect recognition of incident hypertension or hypertensive BP using structured elements, including diagnosis codes, problem list entries, number of BP measurements, orders, and follow-up information. The primary outcome of hypertension recognition was determined by chart review. Among 239 children with incident hypertension and a random sample of 220 children with hypertensive BP, 13% in each sample had clinician recognition of hypertension and hypertensive BP, respectively. An algorithm using ICD-10 encounter diagnosis code, ICD-10 problem list, or multiple BPs during the visit had the highest AUC for attention to incident hypertension (AUC, 0.84; sensitivity, 71.9%; specificity, 95.7%). Adding follow-up BP information to this algorithm had the highest AUC for attention to hypertensive BP (AUC, 0.85; sensitivity, 75.9%; specificity, 93.2%). For patients with hypertension recognition by chart review, ~20% had only free text documentation of hypertension without any structured elements. EHR phenotypes for hypertension recognition have high specificity and moderate sensitivity and may be used in clinician decision support to improve guideline-recommended care.

Every Day I Write the Book: Transforming EHR Data from Daily Practice into New Knowledge.

Despite producing mountains of data in the daily course of care, the documentation labors of frontline clinicians currently return very little value to them or to the health system. The potential of these painstakingly collected data are enormous and clinical registries can extract the extraordinary capacity of these data and transform them into research-ready datasets while protecting the confidentiality of the patients and clinicians. Clinical registries represent transformative tools for primary care research, bringing together the dimensions of clinical practice, research, quality improvement, and policy impact from a large, nationally reflective, diverse sample of practices and patients. The PRIME Registry is one such clinical registry that extracts electronic health record data from more than 600 primary care practices across the United States that is helping advance research, improve quality, and shape the policies required to achieve high-performing primary care for all. Other examples of primary care registries exist, but most of the painstakingly captured data from frontline care remains fallow. Enabling use of these data are important for research, to prevent harm from mis-trained machine learning algorithms, and for monitoring the health of the public.

Caring for Patients Seeking Asylum: Early Data From the Safety Net System in New York City.

Over 200000 people seeking asylum have arrived in New York City from 2022 to 2024. As the safety net hospital system for our city, New York City (NYC) Health+Hospitals (H+H) has taken the lead in caring for newly arrived asylum seekers. We used electronic medical record data to gain early insights into utilization and needs among these patients. We developed a hybrid definition to identify patients who are asylum seekers and examined their demographics, insurance, utilization, primary billing diagnoses, and immunizations from the Epic EMR. We included data on other patients as a point of comparison and analyzed data separately for adult and pediatric patients. In 2023, 15233 or 1.4% of H +H patients were asylum seekers. Asylum seekers had fewer visits than other patients in 2023, and utilization was particularly driven by pregnancy and childbirth. Children seeking asylum were engaged in primary care at high rates. Documented childhood vaccinations were higher among asylum-seeking children engaged in primary care compared to those who were not. The majority of adults and children were insured. Adult primary care engagement was low, as were visits for those seeking care for behavioral health issues. For adult asylum seekers, pregnancy and childbirth are key needs for adults. Increasing vaccination rates among children is critical and promoting engagement in primary care is a promising way to do so. Coordinated efforts are required to meet the varied needs of migrant arrivals, and safety net hospitals can play a key role in addressing their health care needs.

Real-World Effectiveness of Bebtelovimab Versus Nirmatrelvir/Ritonavir in Outpatients with COVID-19.

This real-world study assessed the effectiveness of bebtelovimab (BEB) versus nirmatrelvir/ritonavir (NR) among outpatients with COVID-19 during the Omicron variant era. We conducted a cohort study evaluating patients treated with BEB or NR from February to August 2022 (study period). Follow-up began the day after treatment and continued for 30days. Cohorts were constructed using de-identified electronic health record data from TriNetX Dataworks USA. The study assessed 30-day all-cause hospitalization or death (composite) using the risk difference (RD) and 95% confidence interval (95%CI). Unmatched cohorts included 12,920 BEB- and 70,741 NR-treated patients. After exact matching on key baseline covariates (age > 65years, immunocompromised, recent emergency department [ED] visit, and COVID-19 vaccination) and high-dimensional propensity score matching (1:1) on a broader set of covariates, 5827 patients were included in each cohort. BEB-treated patients were older and had more comorbidities compared to NR-treated patients prior to matching. After matching, baseline characteristics were well balanced. The cumulative incidence of the primary outcome (hospitalization or death) was 2.0% and 1.8% for BEB and NR, respectively (RD 0.2%; 95%CI -0.3%, 0.7%). The upper bound of the RD 95%CI (0.7%) excluded the noninferiority margin (1.795%), demonstrating that BEB was not inferior to NR. The RDs of the secondary outcomes were (BEB vs NR): hospitalization (RD 0.1%; 95%CI -0.4%, 0.6%); ED visit (RD 0.5%; 95%CI -0.3%, 1.3%); and death (RD 0.09%; 95%CI -0.003%, 0.2%). Results from subgroup, sensitivity, and linked analyses (EHR + claims + mortality data) were consistent with the main results. Treatment with BEB was not inferior to NR with respect to 30-day all-cause hospitalization or death. The risk of secondary outcomes was not different for patients treated with BEB compared to NR.

Variation in Emergency Department Physician Admitting Practices and Subsequent Mortality

An emergency department (ED) physician's decision to admit a patient to the hospital plays a pivotal role in determining the type and intensity of care that patient will receive. ED physicians vary widely in their propensity to admit patients to the hospital, but it is unknown whether higher admission propensities result in lower subsequent mortality rates. To measure the variation in ED physicians' admission propensities and estimate their association with patients' subsequent mortality rates. This cross-sectional study used nationwide Veterans Affairs electronic health record data from January 2011 to December 2019, comparing physicians practicing within the same ED. The study population was composed of patients visiting the ED with 1 of the 3 most frequent chief complaints in US EDs (chest pain, shortness of breath, and abdominal pain). The data analyses were performed from May 2022 to October 2024. The main outcomes were variation in physicians' adjusted admission rates, short inpatient stays (<24 hours), and 30-day mortality. The study population included 2098 physicians seeing 2 137 681 patient visits across 105 EDs. The mean (SD) patient age was 63 (15) years, and 9.8% of patients were female. The mean admission rate was 41.2%, and the mean 30-day mortality rate was 2.5%. Physicians' adjusted admission rates varied greatly within the same ED (eg, for chest pain: 90th percentile of physicians, 56.6% admitted vs 10th percentile, 32.6% admitted; difference, 24.0 percentage points), despite finding no association between these adjusted admission rates and patients' prior health status as measured by their Elixhauser Comorbidity Index score before the ED visit. However, patients admitted by physicians with higher admission rates were more likely to be discharged within 24 hours (eg, 31.0% vs 24.8%, respectively), while patients of physicians with higher admission rates had subsequent mortality rates that were no less than those of patients of physicians with lower admission rates. This cross-sectional study demonstrated that ED physicians vary widely in their admission propensity, despite seeing patients with similar prior health status. The results suggest that patients treated by physicians with higher admission propensities are more likely to be discharged after only a short inpatient stay and experience no reduction in subsequent mortality rates.

Lessons learned on information retrieval in electronic health records: a comparison of embedding models and pooling strategies.

Applying large language models (LLMs) to the clinical domain is challenging due to the context-heavy nature of processing medical records. Retrieval-augmented generation (RAG) offers a solution by facilitating reasoning over large text sources. However, there are many parameters to optimize in just the retrieval system alone. This paper presents an ablation study exploring how different embedding models and pooling methods affect information retrieval for the clinical domain. Evaluating on 3 retrieval tasks on 2 electronic health record (EHR) data sources, we compared 7 models, including medical- and general-domain models, specialized encoder embedding models, and off-the-shelf decoder LLMs. We also examine the choice of embedding pooling strategy for each model, independently on the query and the text to retrieve. We found that the choice of embedding model significantly impacts retrieval performance, with BGE, a comparatively small general-domain model, consistently outperforming all others, including medical-specific models. However, our findings also revealed substantial variability across datasets and query text phrasings. We also determined the best pooling methods for each of these models to guide future design of retrieval systems. The choice of embedding model, pooling strategy, and query formulation can significantly impact retrieval performance and the performance of these models on other public benchmarks does not necessarily transfer to new domains. The high variability in performance across different query phrasings suggests that the choice of query may need to be tuned and validated for each task, or even for each institution's EHR. This study provides empirical evidence to guide the selection of models and pooling strategies for RAG frameworks in healthcare applications. Further studies such as this one are vital for guiding empirically-grounded development of retrieval frameworks, such as in the context of RAG, for the clinical domain.