Electronic Medical Records Dataset Research Articles

An Effective DNA Methylation Biomarker Screening Mechanism for Amyotrophic Lateral Sclerosis (ALS) Based on Comorbidities and Gene Function Analysis.

This study used epigenomic methylation differential expression analysis to identify primary biomarkers in patients with amyotrophic lateral sclerosis (ALS). We combined electronic medical record datasets from MIMIC-IV (United States) and NHIRD (Taiwan) to explore ALS comorbidities in depth and discover any comorbidity-related biomarkers. We also applied word2vec to these two clinical diagnostic medical databases to measure similarities between ALS and other similar diseases and evaluated the statistical assessment of the odds ratio to discover significant comorbidities for ALS subjects. Important and representative DNA methylation biomarker candidates could be effectively selected by cross-comparing similar diseases to ALS, comorbidity-related genes, and differentially expressed methylation loci for ALS subjects. The screened epigenomic and comorbidity-related biomarkers were clustered based on their genetic functions. The candidate DNA methylation biomarkers associated with ALS were comprehensively discovered. Gene ontology annotations were then applied to analyze and cluster the candidate biomarkers into three different groups based on gene function annotations. The results showed that a potential testing kit for ALS detection can be composed of SOD3, CACNA1H, and ERBB4 for effective early screening of ALS using blood samples. By developing an effective DNA methylation biomarker screening mechanism, early detection and prophylactic treatment of high-risk ALS patients can be achieved.

TSOANet: Time-Sensitive Orthogonal Attention Network for medical event prediction

Medical Event Prediction (MEP) based on Electronic Medical Records (EMR) is an essential and valuable task for healthcare. For a patient, information in the EMR can be organized into a structured sequence, consisting of multiple visits each with details about visit time and various types of medical events. As the time intervals between neighboring visits are irregular and the medical events at different visits can vary significantly, MEP based on EMR is still challenging. Many studies have been proposed to model the irregular time intervals, relations among different types of medical events within each visit and relations among medical events across visits, and reported exciting results. However, most of these studies focus on two out of the three aspects mentioned above, with only a few addressing all the three aspects simultaneously. In this study, we propose a novel network, the Time-Sensitive Orthogonal Attention Network (TSOANet), which can fully utilize the irregular time intervals, relations among different types of medical events within and across visits. In particular, we design two key components: (1) Time-Sensitive Block, used to model the time intervals at both local and global levels to determine the impact of each visit in EMR; (2) Orthogonal Attention Block, used to model relations among different types of medical events within each visit and across visits in two axes, that is, event axis and time axis. Extensive experiments on two public real-world EMR datasets demonstrate that TSOANet outperforms the state-of-the-art models for various prediction tasks, thereby verifying the effectiveness of our approach. The source code of TSOANet is released at https://github.com/chh13502/TSOANet.

Reducing unwarranted chest x-rays in bronchiolitis: Importance of a robust analysis.

Bronchiolitis is the commonest reason for hospitalisation amongst infants and is often a target for low-value care (LVC) reduction. We aimed to assess the impact of a multifaceted intervention (clinician education, parent engagement, audit-feedback) on rates of chest x-rays (CXR) in bronchiolitis. Longitudinal study of CXRs ordered in infants (1-12 months) diagnosed with bronchiolitis in the Emergency Department (ED) of an Australian paediatric hospital between May 2016 and February 2023. We used logistic regression to measure the impact of the intervention on unwarranted CXR orders, controlling for other potential impacting variables such as time, patient characteristics (age/sex), clinical variables (fever, hypoxia, tachypnoea), seasonal factors (month, day of the week, business hours) and time passed since intervention. Ten thousand one hundred and nine infants were diagnosed with bronchiolitis in the ED over the study period, with 939 (9.3%) receiving a CXR, of which 69% (n = 651) were considered unwarranted. Rates of unwarranted CXRs reduced from 7.9% to 5.4% post-intervention (P < 0.0001). Logistic regression showed the intervention had no significant effect (OR 0.89, 95% CI 0.65-1.23) once other variables and underlying time-based trends were accounted for. Although pre-post rates appeared significantly improved, a robust analysis demonstrated that our multi-faceted intervention was not effective in reducing CXRs in bronchiolitis. The decision to order CXR was associated with clinical features that overlap with pneumonia suggesting ongoing misconceptions regarding the role of CXR for this indication. Our study highlights the value of large electronic medical record datasets and robust methodology to avoid falsely attributing underlying trends to the LVC intervention.

Use of Sequential Hot-Deck Imputation for Missing Health Care Systems Data for Population Health Research.

Electronic medical record (EMR) data present many opportunities for population health research. The use of EMR data for population risk models can be impeded by the high proportion of missingness in key patient variables. Common approaches like complete case analysis and multiple imputation may not be appropriate for some population health initiatives that require a single, complete analytic data set. In this study, we demonstrate a sequential hot-deck imputation (HDI) procedure to address missingness in a set of cardiometabolic measures in an EMR data set. We assessed the performance of sequential HDI within the individual variables and a commonly used composite risk score. A data set of cardiometabolic measures based on EMR data from 2 large urban hospitals was used to create a benchmark data set with simulated missingness. Sequential HDI was applied, and the resulting data were used to calculate atherosclerotic cardiovascular disease risk scores. The performance of the imputation approach was assessed using a set of metrics to evaluate the distribution and validity of the imputed data. Of the 567,841 patients, 65% had at least 1 missing cardiometabolic measure. Sequential HDI resulted in the distribution of variables and risk scores that reflected those in the simulated data while retaining correlation. When stratified by age and sex, risk scores were plausible and captured patterns expected in the general population. The use of sequential HDI was shown to be a suitable approach to multivariate missingness in EMR data. Sequential HDI could benefit population health research by providing a straightforward, computationally nonintensive approach to missing EMR data that results in a single analytic data set.

Agreement between Patient-reported Pain Medication Use and Electronic Medical Record Data in Surgical Amputation Patients.

Opioid misuse after surgery remains a public health crisis in the United States. Recent efforts have focused on tracking pain medication use in surgical populations. However, accurate interpretations of medication use remain quite challenging given inconsistent usage of different datasets. The purpose of this study was to investigate the agreement between electronic medical records (EMR) versus patient self-reported use of pain medications in a surgical amputation population. Patients undergoing major lower extremity amputation or amputation-related procedures were included in this study. Both self-reported and EMR data for pain medication intake were obtained for each patient at three time points (preoperatively, 4 months postoperatively, and 12 months postoperatively). Percentage agreement and the kappa statistic were calculated for both usage (yes/no) and dose categories. Forty-five patients were included in this study, resulting in 108 pairs of self-reported and EMR datasets. Substantial levels of agreement (>70% agreement, kappa >0.61) for opioid use was seen at preoperative and 12 months postoperative. However, agreement dropped at 4 months postoperatively. Anticonvulsant medication showed high levels, whereas acetaminophen showed lower levels of agreements at all time points. Either self-reported or EMR data may be used in research and clinical settings for preoperative or 12-month postoperative patients with little concern for discrepancies. However, at time points immediately following the expected end of acute surgical pain, self-reported data may be needed for more accurate medication reporting. With these findings in mind, usage of datasets should be driven by study objectives and the dataset's strength (eg, accuracy, ease, lack of bias).

Use of EMR data to validate performance of machine learning prognostic models for patients with cancer.

512 Background: Machine learning models that predict survival time for patients with cancer can be useful in the clinic. Validating their performance in deployment is important but challenging, because the only timely source of follow-up/death data is the electronic medical record (EMR), which is known to under-capture deaths resulting in informative censoring. We examined whether validation using EMR data can distinguish between low- and high-quality models, by using gold-standard Cancer Registry data to calculate the true model performance level. Methods: This was a retrospective study of 3417 patients diagnosed with metastatic cancer from 2008-2014, who were either diagnosed with cancer or received their initial treatment at our institution. We used regularized logistic regression on 525 demographic and diagnostic features to predict whether each patient survived for &gt;1 year following diagnosis. We trained the model on high-quality Cancer Registry data from 2197 patients diagnosed between 2008 and 2012 and tested it on two sets of patients. The first set consisted of 1184 patients who had Cancer Registry data evaluable for the 1 year survival endpoint (not lost to follow-up prior to 1 year). Of these, 335 were marked as lost to follow-up prior to 1 year in the EMR dataset, leaving 849 patients evaluable for the endpoint using the EMR data. Results: The model using all available features yielded an AUC of 0.760 when validating with Cancer Registry data (n=1184), and an AUC of 0.771 with EMR data (n=849). When using fewer features, observed AUC dropped for both validation data sources. When using Cancer Registry validation data the model was found to be well calibrated, but with EMR validation data the model incorrectly appeared to systematically underpredict survival. Conclusions: EMR data was useful for validation of model discrimination, but not calibration.

Integration of natural and deep artificial cognitive models in medical images: BERT-based NER and relation extraction for electronic medical records.

Medical images and signals are important data sources in the medical field, and they contain key information such as patients' physiology, pathology, and genetics. However, due to the complexity and diversity of medical images and signals, resulting in difficulties in medical knowledge acquisition and decision support. In order to solve this problem, this paper proposes an end-to-end framework based on BERT for NER and RE tasks in electronic medical records. Our framework first integrates NER and RE tasks into a unified model, adopting an end-to-end processing manner, which removes the limitation and error propagation of multiple independent steps in traditional methods. Second, by pre-training and fine-tuning the BERT model on large-scale electronic medical record data, we enable the model to obtain rich semantic representation capabilities that adapt to the needs of medical fields and tasks. Finally, through multi-task learning, we enable the model to make full use of the correlation and complementarity between NER and RE tasks, and improve the generalization ability and effect of the model on different data sets. We conduct experimental evaluation on four electronic medical record datasets, and the model significantly out performs other methods on different datasets in the NER task. In the RE task, the EMLB model also achieved advantages on different data sets, especially in the multi-task learning mode, its performance has been significantly improved, and the ETE and MTL modules performed well in terms of comprehensive precision and recall. Our research provides an innovative solution for medical image and signal data.

Empowering Precision Medicine: Unlocking Revolutionary Insights through Blockchain-Enabled Federated Learning and Electronic Medical Records.

Precision medicine has emerged as a transformative approach to healthcare, aiming to deliver personalized treatments and therapies tailored to individual patients. However, the realization of precision medicine relies heavily on the availability of comprehensive and diverse medical data. In this context, blockchain-enabled federated learning, coupled with electronic medical records (EMRs), presents a groundbreaking solution to unlock revolutionary insights in precision medicine. This abstract explores the potential of blockchain technology to empower precision medicine by enabling secure and decentralized data sharing and analysis. By leveraging blockchain's immutability, transparency, and cryptographic protocols, federated learning can be conducted on distributed EMR datasets without compromising patient privacy. The integration of blockchain technology ensures data integrity, traceability, and consent management, thereby addressing critical concerns associated with data privacy and security. Through the federated learning paradigm, healthcare institutions and research organizations can collaboratively train machine learning models on locally stored EMR data, without the need for data centralization. The blockchain acts as a decentralized ledger, securely recording the training process and aggregating model updates while preserving data privacy at its source. This approach allows the discovery of patterns, correlations, and novel insights across a wide range of medical conditions and patient populations. By unlocking revolutionary insights through blockchain-enabled federated learning and EMRs, precision medicine can revolutionize healthcare delivery. This paradigm shift has the potential to improve diagnosis accuracy, optimize treatment plans, identify subpopulations for clinical trials, and expedite the development of novel therapies. Furthermore, the transparent and auditable nature of blockchain technology enhances trust among stakeholders, enabling greater collaboration, data sharing, and collective intelligence in the pursuit of advancing precision medicine. In conclusion, this abstract highlights the transformative potential of blockchain-enabled federated learning in empowering precision medicine. By unlocking revolutionary insights from diverse and distributed EMR datasets, this approach paves the way for a future where healthcare is personalized, efficient, and tailored to the unique needs of each patient.

Real-world outcomes with adjuvant nonsteroidal aromatase inhibitors (NSAIs) vs tamoxifen (TAM) in patients with hormone receptor−positive/human epidermal growth factor receptor 2−negative (HR+/HER2−) early breast cancer (EBC): A US database analysis.

542 Background: Randomized controlled trial (RCT) meta-analysis data have shown a significantly greater disease-free survival (DFS) benefit with aromatase inhibitors than TAM in patients with HR+/HER2− EBC. Here we compare real-world outcomes in patients with HR+/HER2− EBC who received adjuvant endocrine therapy (ET) with NSAIs or TAM. Methods: This was a retrospective analysis of ConcertAI’s deidentified electronic medical records data set of patients treated at US academic and community oncology clinics from January 1, 1995 to April 30, 2021. The cohort included patients with stage II-III (if IIIB or IIIC, confirmation was required on residual tumor status) HR+/HER2− EBC who underwent surgery and initiated adjuvant ET (excluding those who switched classes of ET); ovarian function suppression (OFS) was permitted. Menopausal status was determined using age (≥ 50 years) as a proxy for undocumented status. Invasive DFS (IDFS; risk of disease recurrence, death, or second primary tumor), distant DFS (DDFS; risk of distant recurrence, death, or second primary tumor), and overall survival (OS; risk of death) were assessed and defined as the time interval between start of ET and first event. Multivariate Cox regression analyses were performed, adjusting for age, stage, Charlson Comorbidity Index, and prior adjuvant or neoadjuvant chemotherapy for the ET cohort. Results: Of the total 3133 patients analyzed, 2507 were included in this analysis (NSAI ± OFS, n = 1854; TAM ± OFS, n = 653). Among these patients, the median (range) age was 59.5 (22.8-86.6) years, 25.4% were premenopausal, and 56.2% had tumor node involvement. The mean (SD) time to ET discontinuation was 46.2 (35.9) months overall and was similar between the NSAI and TAM cohorts. Reasons for ET discontinuation were undocumented or missing from the database for 76.8% of patients. Multivariate Cox regression analyses suggested a significant risk reduction with NSAIs vs TAM, with a 17% reduction in risk of an IDFS event (hazard ratio [HR], 0.83 [95% CI, 0.69-0.98]) and an 18% reduction in risk of a DDFS event (HR, 0.82 [95% CI, 0.69-0.98]). Although there were few events, OS analysis suggested a numerical trend favoring NSAIs over TAM, with a 12% reduction in risk of death (HR, 0.88 [95% CI, 0.69-1.13]). Conclusions: This real-world data analysis revealed a reduction in risk of IDFS and DDFS events with adjuvant NSAIs vs TAM in patients with HR+/HER2− EBC. Collectively, RCT and real-world data demonstrate greater benefit with adjuvant NSAIs vs TAM with respect to risk of recurrence in this patient population.

Treatment patterns and outcomes among locally advanced cervical cancer patients receiving concurrent chemoradiotherapy.

e17511 Background: Concurrent chemoradiotherapy (CCRT) is standard treatment for patients with locally advanced cervical cancer (LACC). However, little is known on the real-world treatment patterns and outcomes among LACC patients. This study evaluated patient characteristics and treatment patterns of LACC patients, and real-world outcomes among patients receiving CCRT as the first treatment after diagnosis (CCRT-first) in US academic and community settings. Methods: Data were drawn from the ConcertAI Oncology Dataset, a US-based electronic medical record dataset. We included adult patients diagnosed between 2010-2018 with LACC, defined as stage IB2-IIB with node involvement or stage III-IVA cervical cancer, with squamous cell carcinoma (excl. verrucous), adenocarcinoma (excl. clear cell and endometroid), or adenosquamous carcinoma. Patients with prior immunotherapy or a second primary cancer (other than in situ or non-melanoma skin cancer) were excluded. Patients were followed from initial diagnosis through end of the second regimen of systemic anti-cancer therapy, end of record, or death, whichever occurred first. Patient characteristics and treatments were reported overall and in CCRT-first patients. Among CCRT-first patients, real-world time on CCRT treatment (rwTOT), recurrence-free survival (rwRFS), and progression-free survival (rwPFS) among patients with persistent disease were estimated using Kaplan-Meier methods. Results: Overall, 300 patients with LACC were included. At LACC diagnosis, median age of patients was 51 years, 53.7% were White, 30.0% were Black, 48.0% were peri/postmenopausal, 89.3% had squamous cell histology, 75.3% had stage III disease, 92.7% had no evidence of performance status impairment, 50.3% were treated in community settings, and 21.7% had only public insurance (11.0% Medicaid, 10.7% Medicare; 56.3% had no documentation of insurance). Distributions were similar among CCRT-first patients. First treatment after diagnosis included CCRT (N=229), surgery (N=28), systemic therapy (N=11), and radiation therapy alone (N=5). 27 were untreated, and 29 patients received CCRT after another therapy. Of the 229 CCRT-first patients, median (95% CI) rwTOT was 1.6 (1.4-1.7) months; 78.2% received cisplatin within CCRT, and median duration of cisplatin treatment was 35 days; 28.4% received a systemic therapy after CCRT, and 11.8% further initiated a second systemic therapy. 27 patients had recurrent disease after complete response (median rwRFS not reached). 179 patients had persistent disease after CCRT, among whom median (95% CI) rwPFS was 29.7 (16.9-59.3) months from CCRT start. Conclusions: In US clinical practice during 2010-2018, most LACC patients received CCRT as the first treatment after diagnosis. The high proportion of patients who develop persistent disease after CCRT indicates a need for improved first treatment options.

Intelligent Diagnosis of Obstetric Diseases Using HGS-AOA Based Extreme Learning Machine

This paper aimed to realize intelligent diagnosis of obstetric diseases using electronic medical records (EMRs). The Optimized Kernel Extreme Machine Learning (OKEML) technique was proposed to rebalance data. The hybrid approach of the Hunger Games Search (HGS) and the Arithmetic Optimization Algorithm (AOA) was adopted. This paper tested the effectiveness of the OKEML-HGS-AOA on Chinese Obstetric EMR (COEMR) datasets. Compared with other models, the proposed model outperformed the state-of-the-art experimental results on the COEMR, Arxiv Academic Paper Dataset (AAPD), and the Reuters Corpus Volume 1 (RCV1) datasets, with an accuracy of 88%, 90%, and 91%, respectively.

Abstract P3-03-12: Risk of recurrence with adjuvant endocrine therapy in real world patients with hormone receptor positive/human epidermal growth factor receptor-negative early breast cancer: a US database analysis

Abstract Background The recommended 5-10 years of adjuvant endocrine therapy (ET) as standard of care has improved outcomes in patients with HR+/HER2− early breast cancer (EBC); however, risk of recurrence persists. The monarchE study demonstrated improvement in invasive disease-free survival (iDFS) with abemaciclib in node-positive, high-risk EBC, the majority of which were stage III (Johnston SRD, et al. J Clin Oncol. 2021;38:3987); however, monarchE’s relatively short follow-up time to date only allowed for observation of early recurrences—longer follow-up is needed to evaluate late recurrences. Risk of recurrence is a concern for patients with all stages of EBC (Pan H, et al. N Engl J Med. 2017;377:1836). We assessed the risk of recurrence in a real-world setting among patients with stage II-III HR+/HER2− EBC after initiation of adjuvant ET. Methods This was a retrospective analysis of ConcertAI’s deidentified electronic medical records data set among patients treated from January 1, 1995, to April 30, 2021. The cohort included stage II-III patients with HR+/HER2− EBC who underwent surgery and initiated adjuvant ET (if IIIB or IIIC, confirmation was required on residual tumor status). Patients were from academic and community oncology clinics across the US. Data were missing from the database for some variables; thus, percentages may not add up to 100%. iDFS was assessed to determine the risk of disease recurrence, death, or second primary tumor and was defined as the time interval between start of ET and first iDFS event. Kaplan-Meier methods were used to estimate cumulative probabilities of experiencing an iDFS event at 5 and 10 years from the start of ET. Results A total of 3133 patients (98.8% female) with stage II-III HR+/HER2− EBC were included in the analysis. The median follow-up time was 68.1 months. The median age was 59 years; 22.2% of female patients were pre/perimenopausal and 44.7% were postmenopausal; 80.9% of all patients had stage II disease, while 19.1% had stage III disease. In total, 42.1% and 3.7% of those with stage II and III disease had no nodal involvement, respectively. Overall, 51.2% of patients did not receive radiotherapy, and 59.5% of patients did not receive (neo)adjuvant chemotherapy. In the total sample, the 5- and 10-year risk of an iDFS event was 26.1% and 45.0%, respectively. In patients with stage II disease, the 5- and 10-year risks were 22.7% and 40.5%, respectively, and 40.4% and 62.9% among stage III. Conclusions These real-world data demonstrate that the risk of recurrence with adjuvant ET, particularly in those with stage II EBC, is higher than reported in many randomized controlled trials. Importantly, our data show that, in the real-world setting, the stage II population represents a significant proportion of patients (nearly 4 times more than stage III). These findings confirm and underscore the clear unmet need in this population and highlight the need for improved treatment options in this broader EBC population. The NATALEE study will investigate ribociclib + ET in a broad population of patients including stage II and stage III HR+/HER2− EBC with high-risk features, regardless of nodal status. Citation Format: Joyce O’Shaughnessy, Denise Yardley, Lowell Hart, Pedram Razavi, Stephanie L. Graff, Jenifer Wogen, Courtney McDermott, Pierre-Alexandre Dionne, Sina Haftchenary, Purnima Pathak, Sara Tolaney. Risk of recurrence with adjuvant endocrine therapy in real world patients with hormone receptor positive/human epidermal growth factor receptor-negative early breast cancer: a US database analysis [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P3-03-12.

Social Determinants of Health and Traumatic Brain Injury: Implications for Rehabilitation Service Delivery and Outcomes.

Social Determinants of Health and Traumatic Brain Injury: Implications for Rehabilitation Service Delivery and Outcomes.

Entity relationship extraction from Chinese electronic medical records based on feature augmentation and cascade binary tagging framework.

Extracting entity relations from unstructured Chinese electronic medical records is an important task in medical information extraction. However, Chinese electronic medical records mostly have document-level volumes, and existing models are either unable to handle long text sequences or exhibit poor performance. This paper proposes a neural network based on feature augmentation and cascade binary tagging framework. First, we utilize a pre-trained model to tokenize the original text and obtain word embedding vectors. Second, the word vectors are fed into the feature augmentation network and fused with the original features and position features. Finally, the cascade binary tagging decoder generates the results. In the current work, we built a Chinese document-level electronic medical record dataset named VSCMeD, which contains 595 real electronic medical records from vascular surgery patients. The experimental results show that the model achieves a precision of 87.82% and recall of 88.47%. It is also verified on another Chinese medical dataset CMeIE-V2 that the model achieves a precision of 54.51% and recall of 48.63%.

Chronic disease diagnosis model based on convolutional neural network and ensemble learning method.

Chronic diseases have become one of the main causes of premature death all around the world in recent years. The diagnosis of chronic diseases is time-consuming and costly. Therefore, timely diagnosis and prediction of chronic diseases are very necessary. In this paper, a new method for chronic disease diagnosis is proposed by combining convolutional neural network (CNN) and ensemble learning. This method utilizes random forest (RF) as the base classifier to improve classification performance and diagnostic accuracy, and then combines AdaBoost to successfully replace the Softmax layer of CNN to generate multiple accurate base classifiers while determining their optimal attributes, achieving high-quality classification and prediction of chronic diseases. To verify the effectiveness of the proposed method, real-world Electronic Medical Records dataset (C-EMRs) was used for experimental analysis. The results show that compared with other traditional machine learning methods such as CNN, K-Nearest Neighbor, and RF, the proposed method can effectively improve the accuracy of diagnosis and reduce the occurrence of missed diagnosis and misdiagnosis. This study will provide effective information for the diagnosis of chronic diseases, assist doctors in making clinical decisions, develop targeted intervention measures, and reduce the probability of misdiagnosis.

Risk of Erectile Dysfunction in Male Patients with Gout Treated with Febuxostat or Allopurinol: A Propensity Score-Matched Cohort Study.

To evaluate and compare the risk of erectile dysfunction (ED) associated with the use of allopurinol and febuxostat in adult male gout patients. We conducted a cohort study using TriNetX (Cambridge, MA, USA), a global federated health research network that provides real-time electronic medical record datasets. We analyzed and compared the associated risk of ED in gout patients who started taking allopurinol or febuxostat within 12 months. Propensity score matching was performed to adjust for demographic variables, comorbidities, and medication use. Kaplan-Meier analysis was used to estimate the probability of the outcome of interest. The hazard ratio (HR) and associated confidence intervals were calculated along with the proportionality test using R's Survival Package v3.2-3. We identified 679,862 patients with gout among 107,517,445 patients in the database. Of these patients, 24,000 were treated with febuxostat and 299,726 with allopurinol. After propensity matching, 9075 patients receiving febuxostat without allopurinol (febuxostat group) and 9075 corresponding patients receiving allopurinol without febuxostat (allopurinol group) were analyzed for comparison. Among all male patients over 19 years of age, febuxostat was associated with a significantly higher risk of ED versus allopurinol (HR 1.354; 95% confidence interval (CI) 1.003-1.829; log rank test, p = 0.047). After subgroup analysis, in gout patients aged 19-64 years, a significantly higher incidence of ED was observed in the febuxostat group than in the allopurinol group (HR 2.002, 95% CI 1.282-3.126). The risk of ED did not differ significantly between the allopurinol and febuxostat groups in gout patients older than 65 years. Febuxostat may be associated with a higher risk of ED than allopurinol in adult male patients with gout. Future large-scale prospective studies are warranted to confirm our results.

The prediction capability of a cataract surgery risk stratification model based on a large electronic medical record dataset.

The aim of this study was to develop a risk stratification system that predicts visual outcomes (uncorrected corrected visual acuity at one week and five weeks postoperative) in patients undergoing cataract surgery. This was a retrospective analysis in a multitier ophthalmology network. Data from all patients who underwent phacoemulsification or manual small-incision cataract surgery between January 2018 and December 2019 were retrieved from an electronic medical record system. There were 122,911 records; 114,172 (92.9%) had complete data included. Logistic regression analyzed unsatisfactory postoperative outcomes using a main effects model only. The final model was cross-checked using forward stepwise selection. The Hosmer-Lemeshow goodness of fit test, the Bayesian information criterion, and Nagelkerke's R2 assessed model fit. Dispersion was calculated from deviance and degrees of freedom and C-stat from receiving operating characteristics analysis. The final phacoemulsification model (n = 48,169) had a dispersion of 1.08 with a Hosmer-Lemeshow goodness of fit of 0.20, a Nagelkerke R2 of 0.19, and a C-stat of 0.72. The final manual small-incision cataract surgery model (n = 66,003) had a dispersion of 1.05 with a Hosmer-Lemeshow goodness of fit of 0.00015, a Nagelkerke R2 of 0.14, and a C-stat of 0.68. The phacoemulsification model had reasonable model fit; the manual small-incision cataract surgery model had poor fit and was likely missing variables. The predictive capability of these models based on a large, real-world cataract surgical dataset was suboptimal to determine which patients could benefit most from sight-restoring surgery. Appropriate patient selection for cataract surgery in developing settings should still rely on clinician thought processes, intuition, and experience, with more complex cases allocated to more experienced surgeons.

Prediction of Venous Thrombosis Chinese Electronic Medical Records Based on Deep Learning and Rule Reasoning

Aiming at the problems of heavy workload of medical staff in the process of venous thrombosis prevention and treatment, error evaluation, missed evaluation, and inconsistent evaluation, we propose a joint extraction model of Chinese electronic medical records based on deep learning. The approach was to first construct the handshake annotation, then use bidirectional encoder representations from transformers (BERT) as the word vector embedding, then use the bidirectional long short-term memory network (BiLSTM) to extract the contextual features, and then integrate the contextual information into the process of normalizing the word vector. Experiments show that our proposed method achieves 93.3% and 94.3% of entity and relation F1 on the constructed electronic medical record dataset, which effectively improves the effect of medical information extraction. At the same time, the venous thromboembolism (VTE) risk factors extracted from the electronic medical record were used to judge the risk factors of venous thrombosis by means of rule reasoning. Compared with the assessment of clinicians on the Wells and Geneva scales, the accuracy rates of 84.7% and 86.1% were obtained.

Borrowing strength from clinical trials in analysing longitudinal data from a treated cohort: investigating the effectiveness of acetylcholinesterase inhibitors in the management of dementia.

Health care professionals seek information about effectiveness of treatments in patients who would be offered them in routine clinical practice. Electronic medical records (EMRs) and randomized controlled trials (RCTs) can both provide data on treatment effects; however, each data source has limitations when considered in isolation. A novel modelling methodology which incorporates RCT estimates in the analysis of EMR data via informative prior distributions is proposed. A Bayesian mixed modelling approach is used to model outcome trajectories among patients in the EMR dataset receiving the treatment of interest. This model incorporates an estimate of treatment effect based on a meta-analysis of RCTs as an informative prior distribution. This provides a combined estimate of treatment effect based on both data sources. The superior performance of the novel combined estimator is demonstrated via a simulation study. The new approach is applied to estimate the effectiveness at 12 months after treatment initiation of acetylcholinesterase inhibitors in the management of the cognitive symptoms of dementia in terms of Mini-Mental State Examination scores. This demonstrated that estimates based on either trials data only (1.10, SE = 0.316) or cohort data only (1.56, SE = 0.240) overestimated this compared with the estimate using data from both sources (0.86, SE = 0.327). It is possible to combine data from EMRs and RCTs in order to provide better estimates of treatment effectiveness.

Data on Patient Record Trajectory for Linkage (DataPRinT Linkage).

The linkage of Electronic Medical Records, Administrative and other data sources is highly valuable for research and health system monitoring. Once linked, combined resources can be analyzed to provide the answers to a variety of health questions that otherwise could not be answered. However, legislative and administrative barriers, including lengthy processes for data sharing agreements, may preclude timely linkage which is a key requirement during pandemics. ObjectiveTo develop a method using a patient’s health trajectory to probabilistically link primary care Electronic Medical Record (EMR) data with administrative and other data, without the need to transfer large datasets or identifiable information. To determine the legislative feasibility, accuracy and validity of this linkage process. Study DesignIdentify data strings that do not directly identify patients and could be used as unique linkage variables. The data strings, which we are calling dataprints, are sufficiently similar over time in different databases. One example in Ontario, Canada, is the pattern of submitted health claims. For every patient seen by a family physician, there exists a unique pattern of dates/billing codes/diagnoses over time. These unique patterns are reasonably similar in EMR and administrative datasets. We will apply an algorithm which turns the string in the selected dataprints to an irreversibly hashed code for each person. The hashed code and no additional information will be provided by both data controllers to a trusted-third party who will determine which records match and send a mapping table to both. This enables analyses to be run in parallel, without divulging any direct person identifiers. DatasetIndividuals contained in the University of Toronto Practice Based Research Network (UTOPIAN). Outcome MeasuresLinkage quality will be assessed by the number of true matches and represented by sensitivity, specificity and positive and negative predictive values. ResultsThe method will be evaluated against a validated, deterministically linked reference standard at North York General Hospital using de-identified EMR and hospital data. Results will inform processes to enable analyses across datasets while adhering to privacy legislation.