Use Of Electronic Health Record Data Research Articles

Development and validation of an open-source pipeline for automatic population of case report forms from electronic health records: a pediatric multi-center prospective study

Clinical trials and registry studies are essential for advancing research and developing novel treatments. However, these studies rely on manual entry of thousands of variables for each patient. Repurposing real-world data can significantly simplify the data collection, reduce transcription errors, and make the data entry process more efficient, consistent, and cost-effective. We developed an open-source computational pipeline to collect laboratory and medication information from the electronic health record (EHR) data and populate case report forms. The pipeline was developed and validated with data from two independent pediatric hospitals in the US as part of the Long-terM OUtcomes after Multisystem Inflammatory Syndrome In Children (MUSIC) study. Our pipeline allowed the completion of two of the most time-consuming forms. We compared automatically extracted results with manually entered values in one hospital and applied the pipeline to a second hospital, where the output served as the primary data source for case report forms. We extracted and populated 51,845 laboratory and 4913 medication values for 159 patients in two hospitals participating in a prospective pediatric study. We evaluated pipeline performance against data for 104 patients manually entered by clinicians in one of the hospitals. The highest concordance was found during patient hospitalization, with 91.59% of the automatically extracted laboratory and medication values corresponding with the manually entered values. In addition to the successfully populated values, we identified an additional 13,396 laboratory and 567 medication values of interest for the study. The automatic data entry of laboratory and medication values during admission is feasible and has a high concordance with the manually entered data. By implementing this proof of concept, we demonstrate the quality of automatic data extraction and highlight the potential of secondary use of EHR data to advance medical science by improving data entry efficiency and expediting clinical research. NIH Grant 1OT3HL147154-01, U24HL135691, UG1HL135685.

Optimizing EHR Data

Electronic health records (EHRs) contribute to quality-of-care delivery directly. With a comprehensive and effective use of EHR data, healthcare providers can create precise, personalized treatment plans and seamlessly communicate across departments. By optimizing the data in a health system, clinical engineers can improve these processes and, subsequently, patient flow. This article combines case studies and existing research to explore the challenges and opportunities in current EHR data systems, how evolving artificial intelligence tools can be used to improve data utilization, and the importance of interoperability.

Gap Analysis of Glaucoma Examination Concept Representations within Standard Systemized Nomenclature of Medicine – Clinical Terms

PurposeStandardization of eye care data is important for clinical interoperability and research . We aimed to address gaps in the representations of glaucoma examination concepts within Systemized Nomenclature of Medicine – Clinical Terms (SNOMED-CT), the preferred terminology of the American Academy of Ophthalmology. DesignStudy of data elements. MethodsStructured eye exam data fields from two electronic health records (EHR) systems (Epic Systems and Medisoft) were compared against existing SNOMED-CT codes for concepts representing glaucoma examination findings3. Glaucoma specialists from multiple institutions were surveyed to identify high-priority gaps in representation, which were discussed among the SNOMED International Eye Care Clinical Reference Group. Proposals for new codes to address the gaps were formulated and submitted for inclusion in SNOMED-CT. Main Outcome MeasuresGaps in SNOMED-CT glaucoma examination concept representations ResultsWe identified several gaps in SNOMED-CT regarding glaucoma examination concepts. A survey of glaucoma specialists identified high-priority data elements within the categories of tonometry and gonioscopy. For tonometry, there was consensus that we need to define new codes related to maximum intraocular pressure (IOP) and target IOP, and to delineate all methods of measuring IOP. These new codes were proposed and successfully added to SNOMED-CT for future use. Regarding gonioscopy, the current terminology did not include the ability to denote the gonioscopic grading system used (e.g., Shaffer or Spaeth), degree of angle pigmentation, iris configuration (except for plateau iris), and iris approach. There was also no ability to specify eye laterality or angle quadrant for gonioscopic findings. We proposed a framework for representing gonioscopic findings as observable entities in SNOMED-CT. DiscussionThere are existing gaps in the standardized representation of findings related to tonometry and gonioscopy within SNOMED-CT. These are important areas for evaluating clinical outcomes and enabling secondary use of EHR data for glaucoma research. This international, multi-institutional collaborative process enabled identification of gaps, prioritization, and development of data standards to address these gaps. ConclusionAddressing these gaps and augmenting SNOMED-CT coverage of glaucoma examination findings could enhance clinical documentation and future research efforts related to glaucoma.

Regulatory Issues in Electronic Health Records for Adolescent HIV Research: Strategies and Lessons Learned.

Electronic health records (EHRs) are a cost-effective approach to provide the necessary foundations for clinical trial research. The ability to use EHRs in real-world clinical settings allows for pragmatic approaches to intervention studies with the emerging adult HIV population within these settings; however, the regulatory components related to the use of EHR data in multisite clinical trials poses unique challenges that researchers may find themselves unprepared to address, which may result in delays in study implementation and adversely impact study timelines, and risk noncompliance with established guidance. As part of the larger Adolescent Trials Network (ATN) for HIV/AIDS Interventions Protocol 162b (ATN 162b) study that evaluated clinical-level outcomes of an intervention including HIV treatment and pre-exposure prophylaxis services to improve retention within the emerging adult HIV population, the objective of this study is to highlight the regulatory process and challenges in the implementation of a multisite pragmatic trial using EHRs to assist future researchers conducting similar studies in navigating the often time-consuming regulatory process and ensure compliance with adherence to study timelines and compliance with institutional and sponsor guidelines. Eight sites were engaged in research activities, with 4 sites selected from participant recruitment venues as part of the ATN, who participated in the intervention and data extraction activities, and an additional 4 sites were engaged in data management and analysis. The ATN 162b protocol team worked with site personnel to establish the necessary regulatory infrastructure to collect EHR data to evaluate retention in care and viral suppression, as well as para-data on the intervention component to assess the feasibility and acceptability of the mobile health intervention. Methods to develop this infrastructure included site-specific training activities and the development of both institutional reliance and data use agreements. Due to variations in site-specific activities, and the associated regulatory implications, the study team used a phased approach with the data extraction sites as phase 1 and intervention sites as phase 2. This phased approach was intended to address the unique regulatory needs of all participating sites to ensure that all sites were properly onboarded and all regulatory components were in place. Across all sites, the regulatory process spanned 6 months for the 4 data extraction and intervention sites, and up to 10 months for the data management and analysis sites. The process for engaging in multisite clinical trial studies using EHR data is a multistep, collaborative effort that requires proper advanced planning from the proposal stage to adequately implement the necessary training and infrastructure. Planning, training, and understanding the various regulatory aspects, including the necessity of data use agreements, reliance agreements, external institutional review board review, and engagement with clinical sites, are foremost considerations to ensure successful implementation and adherence to pragmatic trial timelines and outcomes.

Patterns of multimorbidity in primary care electronic health records: A systematic review.

Multimorbidity, the coexistence of multiple chronic conditions in an individual, is a complex phenomenon that is highly prevalent in primary care settings, particularly in older individuals. This systematic review summarises the current evidence on multimorbidity patterns identified in primary care electronic health record (EHR) data. Three databases were searched from inception to April 2022 to identify studies that derived original multimorbidity patterns from primary care EHR data. The quality of the included studies was assessed using a modified version of the Newcastle-Ottawa Quality Assessment Scale. Sixteen studies were included in this systematic review, none of which was of low quality. Most studies were conducted in Spain, and only one study was conducted outside of Europe. The prevalence of multimorbidity (i.e. two or more conditions) ranged from 14.0% to 93.9%. The most common stratification variable in disease clustering models was sex, followed by age and calendar year. Despite significant heterogeneity in clustering methods and disease classification tools, consistent patterns of multimorbidity emerged. Mental health and cardiovascular patterns were identified in all studies, often in combination with diseases of other organ systems (e.g. neurological, endocrine). These findings emphasise the frequent coexistence of physical and mental health conditions in primary care, and provide useful information for the development of targeted preventive and management strategies. Future research should explore mechanisms underlying multimorbidity patterns, prioritise methodological harmonisation to facilitate the comparability of findings, and promote the use of EHR data globally to enhance our understanding of multimorbidity in more diverse populations.

Predicting adverse outcomes in adults with a community-acquired lower respiratory tract infection: a protocol for the development and validation of two prediction models for (i) all-cause hospitalisation and mortality and (ii) cardiovascular outcomes

BackgroundCommunity-acquired lower respiratory tract infections (LRTI) are common in primary care and patients at particular risk of adverse outcomes, e.g., hospitalisation and mortality, are challenging to identify. LRTIs are also linked to an increased incidence of cardiovascular diseases (CVD) following the initial infection, whereas concurrent CVD might negatively impact overall prognosis in LRTI patients. Accurate risk prediction of adverse outcomes in LRTI patients, while considering the interplay with CVD, can aid general practitioners (GP) in the clinical decision-making process, and may allow for early detection of deterioration. This paper therefore presents the design of the development and external validation of two models for predicting individual risk of all-cause hospitalisation or mortality (model 1) and short-term incidence of CVD (model 2) in adults presenting to primary care with LRTI.MethodsBoth models will be developed using linked routine electronic health records (EHR) data from Dutch primary and secondary care, and the mortality registry. Adults aged ≥ 40 years with a GP-diagnosis of LRTI between 2016 and 2019 are eligible for inclusion. Relevant patient demographics, medical history, medication use, presenting signs and symptoms, and vital and laboratory measurements will be considered as candidate predictors. Outcomes of interest include 30-day all-cause hospitalisation or mortality (model 1) and 90-day CVD (model 2). Multivariable elastic net regression techniques will be used for model development. During the modelling process, the incremental predictive value of CVD for hospitalisation or all-cause mortality (model 1) will also be assessed. The models will be validated through internal-external cross-validation and external validation in an equivalent cohort of primary care LRTI patients.DiscussionImplementation of currently available prediction models for primary care LRTI patients is hampered by limited assessment of model performance. While considering the role of CVD in LRTI prognosis, we aim to develop and externally validate two models that predict clinically relevant outcomes to aid GPs in clinical decision-making. Challenges that we anticipate include the possibility of low event rates and common problems related to the use of EHR data, such as candidate predictor measurement and missingness, how best to retrieve information from free text fields, and potential misclassification of outcome events.

Challenges in and Opportunities for Electronic Health Record-Based Data Analysis and Interpretation.

Electronic health records (EHRs) have been increasingly adopted in clinical practices across the United States, providing a primary source of data for clinical research, particularly observational cohort studies. EHRs are a high-yield, low-maintenance source of longitudinal real-world data for large patient populations and provide a wealth of information and clinical contexts that are useful for clinical research and translation into practice. Despite these strengths, it is important to recognize the multiple limitations and challenges related to the use of EHR data in clinical research. Missing data are a major source of error and biases and can affect the representativeness of the cohort of interest, as well as the accuracy of the outcomes and exposures. Here, we aim to provide a critical understanding of the types of data available in EHRs and describe the impact of data heterogeneity, quality, and generalizability, which should be evaluated prior to and during the analysis of EHR data. We also identify challenges pertaining to data quality, including errors and biases, and examine potential sources of such biases and errors. Finally, we discuss approaches to mitigate and remediate these limitations. A proactive approach to addressing these issues can help ensure the integrity and quality of EHR data and the appropriateness of their use in clinical studies.

ICD-10 diagnosis codes in electronic health records do not adequately capture fracture complexity for proximal humerus fractures

The ability to do comparative effectiveness research (CER) for proximal humerus fractures (PHF) using data in Electronic Health Record (EHR) systems and administrative claims databases was enhanced by the tenth revision of the International Classification of Diseases (ICD-10) which expanded the diagnosis codes for PHF to describe fracture complexity including displacement and the number of fracture parts. However, these expanded codes only enhance secondary use of data for research if the codes selected and recorded correctly reflect the fracture complexity. The objective of this project was to assess the accuracy of ICD-10 diagnosis codes documented during routine clinical practice for secondary use of EHR data. A sample of patients with PHFs treated by orthopedic providers across a large, regional health care system between January 1, 2016 and December 31, 2018 were retrospectively identified from the EHR. Four fellowship-trained orthopedic surgeons reviewed patient X-rays and recorded the Neer Classification characteristics of displacement, number of parts, and fracture location(s). The fracture characteristics were then reviewed by a trained coder and the most clinically appropriate ICD-10 diagnosis code based on the number of fracture parts was assigned. We assessed congruence between ICD-10 codes documented in the EHR and X-ray validated codes, and assessed sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for EHR-documented ICD-10 codes. There were 761 patients with unilateral, closed PHF who met study inclusion criteria. On average, patients were 67 years of age and 77% were female. Based on X-ray review, 37% were 1-part fractures, 42% were 2-part, 11% were 3-part and 10% were 4-part fractures. Of the EHR diagnosis codes recorded during clinical practice, 59% were "unspecified" fracture diagnosis codes that did not identify the number of fracture parts. Examination of fracture codes revealed PPV was highest for 1-part (PPV = 0.66, 95% CI 0.60-0.72) and 4-part fractures (PPV = 0.67, 95% CI 0.13-1.00). Current diagnosis coding practices do not adequately capture the fracture complexity needed to conduct subgroup analysis for PHF. Conclusions drawn from population studies or large databases using ICD-10 codes for PHF classification should be interpreted within this limitation. Future studies are warranted to improve diagnostic coding to support large observational studies using EHR and administrative claims data.

Diagnosing major depressive disorder and substance use disorder using the electronic health record: A preliminary validation study

BackgroundOne mechanism to examine if major depressive disorder (MDD) is related to the development of substance use disorder (SUD) is by leveraging naturalistic data available in the electronic health record (EHR). Rules for data extraction and variable construction linked to psychometrics validating their use are needed to extract data accurately. ObjectiveWe propose and validate a methodologic framework for using EHR variables to identify patients with MDD and non-nicotine SUD. MethodsProxy diagnoses and index dates of MDD and/or SUD were established using billing codes, problem lists, patient-reported outcome measures, and prescriptions. Manual chart reviews were conducted for the 1-year period surrounding each index date to determine (1) if proxy diagnoses were supported by chart notes and (2) if the index dates accurately captured disorder onset. ResultsThe results demonstrated 100% positive predictive value for proxy diagnoses of MDD. The proxy diagnoses for SUD exhibited strong agreement (Cohen's kappa of 0.84) compared to manual chart review and 92% sensitivity, specificity, positive predictive value, and negative predictive value. Sixteen percent of patients showed inaccurate SUD index dates generated by EHR extraction with discrepancies of over 6 months compared to SUD onset identified through chart review. ConclusionsOur methodology was very effective in identifying patients with MDD with or without SUD and moderately effective in identifying SUD onset date. These findings support the use of EHR data to make proxy diagnoses of MDD with or without SUD.

Equity and bias in electronic health records data

Embedded pragmatic clinical trials (ePCTs) are conducted during routine clinical care and have the potential to increase knowledge about the effectiveness of interventions under real world conditions. However, many pragmatic trials rely on data from the electronic health record (EHR) data, which are subject to bias from incomplete data, poor data quality, lack of representation from people who are medically underserved, and implicit bias in EHR design. This commentary examines how the use of EHR data might exacerbate bias and potentially increase health inequities. We offer recommendations for how to increase generalizability of ePCT results and begin to mitigate bias to promote health equity.

What does it take to create a European Health Data Space? International commitments and national realities

Public health concerns in Europe demonstrate the necessity of building a health policy that could contribute to the long-term sustainable development of the European Union (EU), as stated in the European Health Union (EHU) manifesto. The main desire to create an EHU is embodied in the launch of the European Health Data Space (EHDS). The EHDS seeks to foster a genuine single market for digital health services and products by, among other things, accelerating the uptake and implementation of harmonised and interoperable electronic health record (EHR) systems across the EU. In the context of primary and secondary use of EHR data, developments in Europe have thus far resulted in patchy and, in some places, non-interoperable solutions. Taking the gap between international ambitions and national realities as a starting point, this paper contends that both EU level and Member State level circumstances should be considered to make the EHDS a reality.

P013 A three-year audit of dose tapering biologic medications in a rheumatology service

Abstract Background/Aims Dose tapering of biologic therapies is supported by both ACR and EULAR guidelines. Identification of appropriate patients may deliver cost savings whilst minimising the risk of disease flare. To audit dose tapering activity for patients treated with bDMARD for rheumatoid (RA), psoriatic arthritis (PsA) and ankylosing spondylosis (AS) meeting locally agreed criteria over three years. Methods All patients treated with a bDMARD for RA, PsA or AS fulfilling criteria for dose tapering between 1st October 2019 - 1st October 2021 were identified from electronic health records (EHR). Rituximab patients were excluded due to reactive treatment responding to flare. Patients were required to have been on their bDMARD for = > one year, not prescribed oral steroids / treated with parenteral steroids in the past six months, no swollen joints in the past six months, and not failed dose tapering in the previous year. RA patients were ineligible if DAS ≥3.2 in the previous year; PsA if their biologic was prescribed by dermatology /active psoriasis; AS patients if BASDAI score of > 4 in past year. Records were flagged to prompt a shared decision-making discussion between the treating clinician and patient during routine appointments; additional patients were identified through clinical assessment. Failing dose tapering was defined as a post tapering flare requiring dose increase within one year. Patients were tapered by increasing the interval of a bDMARD. Results The number of patients that dose tapered increased from 24 in 2019 to 51 in 2021, making a total estimated NHS list saving of up to £170,000 per annum. (Table 1). Tapering was unsuccessful for 7/116 (6%) due to post tapering flares. Patients flaring after tapering were reassessed urgently and had their bDMARD dose reviewed. More patients were eligible than identified from EHR search due to missing data in coding; 787 (61%) across all years. Conclusion Dose tapering was well tolerated in eligible patients making this a safe and acceptable initiative. Use of EHR data and behavioural prompts for clinicians supported this in practice. Cost savings were reinvested to increase pharmacy support for homecare prescription management. Disclosure H. Light: None. J. Firth: None. J. Bluett: Grants/research support; JB has received a research grant from Pfizer. Other; JB has received travel/conference fees from UCB, Pfizer and Eli Lilly. K. Kinsey: None.

Leveraging Electronic Health Record Data for Timely Chronic Disease Surveillance: The Multi-State EHR-Based Network for Disease Surveillance.

Electronic health record (EHR) data can potentially make chronic disease surveillance more timely, actionable, and sustainable. Although use of EHR data can address numerous limitations of traditional surveillance methods, timely surveillance data with broad population coverage require scalable systems. This report describes implementation, challenges, and lessons learned from the Multi-State EHR-Based Network for Disease Surveillance (MENDS) to help inform how others work with EHR data to develop distributed networks for surveillance. Funded by the Centers for Disease Control and Prevention (CDC), MENDS is a data modernization demonstration project that aims to develop a timely national chronic disease sentinel surveillance system using EHR data. It facilitates partnerships between data contributors (health information exchanges, other data aggregators) and data users (state and local health departments). MENDS uses query and visualization software to track local emerging trends. The program also uses statistical and geospatial methods to generate prevalence estimates of chronic disease risk measures at the national and local levels. Resulting data products are designed to inform public health practice and improve the health of the population. MENDS includes 5 partner sites that leverage EHR data from 91 health system and clinic partners and represents approximately 10 million patients across the United States. Key areas of implementation include governance, partnerships, technical infrastructure and support, chronic disease algorithms and validation, weighting and modeling, and workforce education for public health data users. MENDS presents a scalable distributed network model for implementing national chronic disease surveillance that leverages EHR data. Priorities as MENDS matures include producing prevalence estimates at various geographic and subpopulation levels, developing enhanced data sharing and interoperability capacity using international data standards, scaling the network to improve coverage nationally and among underrepresented geographic areas and subpopulations, and expanding surveillance of additional chronic disease measures and social determinants of health.

Electronic health record data for assessing risk of hospitalization for COVID-19: Methodological considerations applied to multiple sclerosis.

During the COVID-19 pandemic, electronic health record (EHR) data has been used to investigate disease severity and risk factors for severe COVID-19 in people with multiple sclerosis (pwMS). Methodological challenges including sampling bias, and residual confounding should be considered when conducting EHR-based studies. We aimed to address these limitations related to the use of EHR data in order to identify risk factors, including the use of disease modifying therapies (DMTs), associated with hospitalization for COVID-19 amongst pwMS. We performed a retrospective cohort study including a sample of 47,051 pwMS using a large US-based EHR and claims linked database. Follow-up started at the beginning of the pandemic, February 20th 2020, and continued until September 30th 2020. COVID-19 diagnosis was determined by the presence of ICD-10 diagnostic code for COVID-19, or a positive diagnostic laboratory test, or an ICD-10 diagnostic code for coronaviruses. We used Cox regression modeling to assess the impact of baseline demographics, MS disease history and pre-existing comorbidities on the risk of hospitalization for COVID-19. Then, we identified 5,169 pwMS using ocrelizumab (OCR) and 3,351 pwMS using dimethyl fumarate (DMF) at baseline, and evaluated the distribution of the identified COVID-19 risk factors between the two groups. Finally, we used Cox regression models, adjusted for the identified confounders, to estimate the risk of hospitalization for COVID-19 in pwMS treated with OCR compared to DMF. Among the pwMS cohort, we identified 799 COVID-19 cases (1.7%) which resulted in 182 hospitalizations for COVID-19 (0.4%). Population differences between the pwMS and COVID-19 cohorts were observed. Statistical modeling identified older age, male gender, African-American race, walking with assistance, non-ambulatory status, severe relapse requiring hospitalization in year prior to baseline, and specific comorbidities to be associated with a higher risk of COVID-19 related-hospitalization. Comparing the COVID-19 risk factors between OCR users and DMF users, MS characteristics including ambulatory status and MS subtype were highly imbalanced, likely arising from key differences in the labelled indications for these therapies. Compared to DMF use, in unadjusted (HR 1.58, 95% CI 0.73 - 3.44), adjusted (HR 1.28, 95% CI 0.58 - 2.83), propensity score weighted (HR 1.25, 95% CI 0.56 - 2.80), and doubly robust models (HR 1.29, 95% CI 0.57 - 2.89), no significantly increased risk of hospitalization for COVID-19 was associated with OCR use. We observed significant population differences when comparing all pwMS to COVID-19 cases, as well as significant differences in key confounders between OCR and DMF treated patients. In unadjusted analyses we did not observe a statistically significant higher risk of COVID-19 hospitalization in pwMS treated with OCR compared to DMF, with further attenuation of risk when adjusting for the key confounders. This study re-emphasises the importance to appropriately consider both sampling and confounding bias in EHR-based MS research.

Trends in Electronic Health Record Capabilities for Tracking Documentation Time

To understand hospitals' access to and use of data from electronic health record (EHR) developers that quantify the amount of time clinicians spend documenting clinical care in EHRs. Descriptive analysis of 4 waves of a nationally representative survey of US nonfederal acute care hospitals from 2017 to 2019 and 2021 (N = 10,662 across years). We identified the share of hospitals that had access to EHR documentation time measures between 2017 and 2021 and how access varied by hospital and EHR characteristics. We then described how EHR data were used among hospitals with access and whether use varied by developer. The share of hospitals with access to EHR documentation time measures increased significantly each year between 2017 and 2021, when more than two-thirds of hospitals reported having access to these measures. Despite hospitals' increased access to measures that track EHR time, lower-resourced hospitals, nonteaching hospitals, and hospitals with non-market-leading EHR developers were less likely to report having access than their counterparts. In 2021, the 2 most common uses of EHR data were "identifying providers in need of training and support" and "identifying areas to improve clinical workflow." The share of hospitals indicating use of EHR data increased between 2019 and 2021 for all studied uses. A higher proportion of hospitals with access to EHR documentation time measures used them for more purposes over time, suggesting their increased value. Although hospitals' access to and use of EHR documentation measures increased significantly in the last 5 years, future research efforts should investigate whether the use of these measures translates into reduced burden for providers.

Case study - Feature engineering inspired by domain experts on real world medical data

To perform data mining projects for knowledge discovery based on health data produced in a daily health care stored in electronic health records (EHR) can be time consuming. This study exemplifies that the involvement of a data scientist improves classification performances. We have performed a case study that comprises two real world medical research projects, comparing feature engineering and knowledge discovery based on classification performance. Project (P1) comprised 82,742 patients with the research question “Can we predict patient falls by use of EHR data” and the second project (P2) included 23,396 patients with the focus on “Negative side effects of antiepileptic drug consumption on bone structure”.The results concluded three salient results. (i) It is valuable for medical researchers to involve a data scientist when medical research based on real world medical data is performed. The findings were justified with an analysis of classification metrics when iteratively engineered features were used. The features were generated from domain experts and computer scientists in collaboration with medical researchers. We gave this process the name domain knowledge-driven feature engineering (KDFE).To evaluate the classification performance the metric area under the receiver operating characteristic curve (AUROC) was used. (ii) Domain experts are benefited in quantitative terms by KDFE. When KDFE was compared to baseline, the average classification performance measured by AUROC for the engineered features rose for P1 from 0.62 to 0.82 and for P2 from 0.61 to 0.89 (p-values << 0.001). (iii) The engineered features were represented in a systematic structure, which is the foundation of a theoretical model for automated KDFE (aKDFE).To our knowledge, this is the first study that proves that via quantitative measures KDFE adds value to real-world. However, the method is not limited to the medical domain. Other areas with similar data properties should also benefit from KDFE.

A semi-supervised adaptive Markov Gaussian embedding process (SAMGEP) for prediction of phenotype event times using the electronic health record

While there exist numerous methods to identify binary phenotypes (i.e. COPD) using electronic health record (EHR) data, few exist to ascertain the timings of phenotype events (i.e. COPD onset or exacerbations). Estimating event times could enable more powerful use of EHR data for longitudinal risk modeling, including survival analysis. Here we introduce Semi-supervised Adaptive Markov Gaussian Embedding Process (SAMGEP), a semi-supervised machine learning algorithm to estimate phenotype event times using EHR data with limited observed labels, which require resource-intensive chart review to obtain. SAMGEP models latent phenotype states as a binary Markov process, and it employs an adaptive weighting strategy to map timestamped EHR features to an embedding function that it models as a state-dependent Gaussian process. SAMGEP’s feature weighting achieves meaningful feature selection, and its predictions significantly improve AUCs and F1 scores over existing approaches in diverse simulations and real-world settings. It is particularly adept at predicting cumulative risk and event counting process functions, and is robust to diverse generative model parameters. Moreover, it achieves high accuracy with few (50–100) labels, efficiently leveraging unlabeled EHR data to maximize information gain from costly-to-obtain event time labels. SAMGEP can be used to estimate accurate phenotype state functions for risk modeling research.

Computable phenotypes for cohort identification: core content for a new class of FAIR Digital Objects

Computable phenotypes for cohort identification: core content for a new class of FAIR Digital Objects

Testing an Automated Approach to Identify Variation in Outcomes among Children with Type 1 Diabetes across Multiple Sites.

Efficient methods to obtain and benchmark national data are needed to improve comparative quality assessment for children with type 1 diabetes (T1D). PCORnet is a network of clinical data research networks whose infrastructure includes standardization to a Common Data Model (CDM) incorporating electronic health record (EHR)-derived data across multiple clinical institutions. The study aimed to determine the feasibility of the automated use of EHR data to assess comparative quality for T1D. In two PCORnet networks, PEDSnet and OneFlorida, the study assessed measures of glycemic control, diabetic ketoacidosis admissions, and clinic visits in 2016-2018 among youth 0-20 years of age. The study team developed measure EHR-based specifications, identified institution-specific rates using data stored in the CDM, and assessed agreement with manual chart review. Among 9,740 youth with T1D across 12 institutions, one quarter (26%) had two or more measures of A1c greater than 9% annually (min 5%, max 47%). The median A1c was 8.5% (min site 7.9, max site 10.2). Overall, 4% were hospitalized for diabetic ketoacidosis (min 2%, max 8%). The predictive value of the PCORnet CDM was >75% for all measures and >90% for three measures. Using EHR-derived data to assess comparative quality for T1D is a valid, efficient, and reliable data collection tool for measuring T1D care and outcomes. Wide variations across institutions were observed, and even the best-performing institutions often failed to achieve the American Diabetes Association HbA1C goals (<7.5%).

MixEHR-Guided: A guided multi-modal topic modeling approach for large-scale automatic phenotyping using the electronic health record

Electronic Health Records (EHRs) contain rich clinical data collected at the point of the care, and their increasing adoption offers exciting opportunities for clinical informatics, disease risk prediction, and personalized treatment recommendation. However, effective use of EHR data for research and clinical decision support is often hampered by a lack of reliable disease labels. To compile gold-standard labels, researchers often rely on clinical experts to develop rule-based phenotyping algorithms from billing codes and other surrogate features. This process is tedious and error-prone due to recall and observer biases in how codes and measures are selected, and some phenotypes are incompletely captured by a handful of surrogate features. To address this challenge, we present a novel automatic phenotyping model called MixEHR-Guided (MixEHR-G), a multimodal hierarchical Bayesian topic model that efficiently models the EHR generative process by identifying latent phenotype structure in the data. Unlike existing topic modeling algorithms wherein the inferred topics are not identifiable, MixEHR-G uses prior information from informative surrogate features to align topics with known phenotypes. We applied MixEHR-G to an openly-available EHR dataset of 38,597 intensive care patients (MIMIC-III) in Boston, USA and to administrative claims data for a population-based cohort (PopHR) of 1.3 million people in Quebec, Canada. Qualitatively, we demonstrate that MixEHR-G learns interpretable phenotypes and yields meaningful insights about phenotype similarities, comorbidities, and epidemiological associations. Quantitatively, MixEHR-G outperforms existing unsupervised phenotyping methods on a phenotype label annotation task, and it can accurately estimate relative phenotype prevalence functions without gold-standard phenotype information. Altogether, MixEHR-G is an important step towards building an interpretable and automated phenotyping system using EHR data.