Electronic Health Record Phenotyping Research Articles

Developing Phenotypes from Electronic Health Records for Chronic Disease Surveillance

ObjectiveTo utilize clinical data in Electronic Health Records (EHRs) to develop chronic disease phenotypes appropriate for conducting population health surveillance.IntroductionChronic diseases, including hypertension, type 2 diabetes mellitus (diabetes), obesity, and hyperlipidemia, are some of the leading causes of morbidity and mortality in the United States. Monitoring disease prevalence guides public health programs and policies that help prevent this burden. EHRs can supplement traditional sources of chronic disease surveillance, such as health surveys and administrative claims datasets, by offering near real-time data, large sample sizes, and a rich source of clinical data. However, few studies have provided clear, consistent EHR phenotypes that were developed to inform population health surveillance.MethodsRetrospective EHR data were obtained for patients seen at New York University Langone Health in 2017 (n=1,397,446). To better estimate chronic disease burden among New York City (NYC) adults, the patient population was limited to NYC residents aged 20 or older, who were seen in the ambulatory primary care setting (n=153,653). Rule-based algorithms for identifying patients with hypertension, statin-eligibility, diabetes, and obesity were developed based on a combination of diagnostic codes, lab results or vitals, and relevant prescriptions. We compared the performance of our metric definitions to selected phenotypes from the literature using percent agreement and Cohen’s kappa. Patients with discordant disease classifications between the two sets of definitions were analyzed through natural language processing (NLP) on the patients’ 2017 medical notes using a support vector machine model. Statin-eligibility is a novel phenotype and therefore did not have a comparable definition in the literature. Sensitivity analyses were conducted to determine how disease burden changed under alternative rules for each metric.ResultsOf 153,653 adult ambulatory care patients in 2017, an estimated 53.7% had hypertension, 12.4% had diabetes, 27.8% were obese, and 30.0% were statin-eligible under our proposed definitions. The estimated prevalence of hypertension increased from 28.1% to 53.7% when diagnostic codes were supplemented with blood pressure measurements and anti-hypertensive medications, while the estimated prevalence of diabetes increased less than one percentage point with inclusion of diabetes-related medications and elevated A1C measurements. There was high agreement between our obesity (94.5% agreement, k=0.86) and diabetes (96.2% agreement, k=0.81) definitions and selected definitions from the literature and moderate agreement between the hypertension definitions (74.8% agreement, k=0.41). NLP classification of discordant cases had greater alignment with the classification results of our definitions for both hypertension (78.0% agreement) and diabetes (71.2% agreement) but did not show strong agreement with either obesity algorithm. Sensitivity analyses did not have large impacts on prevalence estimates for any of the indicators, with all estimates within two percentage points of the final algorithms.ConclusionsOur proposed rule-based phenotypes using prescriptions, labs, and vitals improved ascertainment of conditions beyond diagnostic codes and were robust to modifications per sensitivity analyses. Results from our algorithms were highly consistent with standard phenotypes from the literature and may improve case capture for surveillance purposes. These algorithms can be replicated across diverse EHR networks and can be weighted to generate population prevalence estimates.

Machine Learning, Natural Language Processing, and the Electronic Health Record: Innovations in Mental Health Services Research.

An unprecedented amount of clinical information is now available via electronic health records (EHRs). These massive data sets have stimulated opportunities to adapt computational approaches to track and identify target areas for quality improvement in mental health care. In this column, three key areas of EHR data science are described: EHR phenotyping, natural language processing, and predictive modeling. For each of these computational approaches, case examples are provided to illustrate their role in mental health services research. Together, adaptation of these methods underscores the need for standardization and transparency while recognizing the opportunities and challenges ahead.

Feature extraction for phenotyping from semantic and knowledge resources.

Phenotyping algorithms can efficiently and accurately identify patients with a specific disease phenotype and construct electronic health records (EHR)-based cohorts for subsequent clinical or genomic studies. Previous studies have introduced unsupervised EHR-based feature selection methods that yielded algorithms with high accuracy. However, those selection methods still require expert intervention to tweak the parameter settings according to the EHR data distribution for each phenotype. To further accelerate the development of phenotyping algorithms, we propose a fully automated and robust unsupervised feature selection method that leverages only publicly available medical knowledge sources, instead of EHR data. SEmantics-Driven Feature Extraction (SEDFE) collects medical concepts from online knowledge sources as candidate features and gives them vector-form distributional semantic representations derived with neural word embedding and the Unified Medical Language System Metathesaurus. A number of features that are semantically closest and that sufficiently characterize the target phenotype are determined by a linear decomposition criterion and are selected for the final classification algorithm. SEDFE was compared with the EHR-based SAFE algorithm and domain experts on feature selection for the classification of five phenotypes including coronary artery disease, rheumatoid arthritis, Crohn's disease, ulcerative colitis, and pediatric pulmonary arterial hypertension using both supervised and unsupervised approaches. Algorithms yielded by SEDFE achieved comparable accuracy to those yielded by SAFE and expert-curated features. SEDFE is also robust to the input semantic vectors. SEDFE attains satisfying performance in unsupervised feature selection for EHR phenotyping. Both fully automated and EHR-independent, this method promises efficiency and accuracy in developing algorithms for high-throughput phenotyping.

Electronic Health Record Phenotypes for Identifying Patients with Late-Stage Disease: a Method for Research and Clinical Application

Electronic Health Record Phenotypes for Identifying Patients with Late-Stage Disease: a Method for Research and Clinical Application

EHR phenotyping via jointly embedding medical concepts and words into a unified vector space

BackgroundThere has been an increasing interest in learning low-dimensional vector representations of medical concepts from Electronic Health Records (EHRs). Vector representations of medical concepts facilitate exploratory analysis and predictive modeling of EHR data to gain insights about the patterns of care and health outcomes. EHRs contain structured data such as diagnostic codes and laboratory tests, as well as unstructured free text data in form of clinical notes, which provide more detail about condition and treatment of patients.MethodsIn this work, we propose a method that jointly learns vector representations of medical concepts and words. This is achieved by a novel learning scheme based on the word2vec model. Our model learns those relationships by integrating clinical notes and sets of accompanying medical codes and by defining joint contexts for each observed word and medical code.ResultsIn our experiments, we learned joint representations using MIMIC-III data. Using the learned representations of words and medical codes, we evaluated phenotypes for 6 diseases discovered by our and baseline method. The experimental results show that for each of the 6 diseases our method finds highly relevant words. We also show that our representations can be very useful when predicting the reason for the next visit.ConclusionsThe jointly learned representations of medical concepts and words capture not only similarity between codes or words themselves, but also similarity between codes and words. They can be used to extract phenotypes of different diseases. The representations learned by the joint model are also useful for construction of patient features.

Automated feature selection of predictors in electronic medical records data.

The use of Electronic Health Records (EHR) for translational research can be challenging due to difficulty in extracting accurate disease phenotype data. Historically, EHR algorithms for annotating phenotypes have been either rule-based or trained with billing codes and gold standard labels curated via labor intensive medical chart review. These simplistic algorithms tend to have unpredictable portability across institutions and low accuracy for many disease phenotypes due to imprecise billing codes. Recently, more sophisticated machine learning algorithms have been developed to improve the robustness and accuracy of EHR phenotyping algorithms. These algorithms are typically trained via supervised learning, relating gold standard labels to a wide range of candidate features including billing codes, procedure codes, medication prescriptions and relevant clinical concepts extracted from narrative notes via Natural Language Processing (NLP). However, due to the time intensiveness of gold standard labeling, the size of the training set is often insufficient to build a generalizable algorithm with the large number of candidate features extracted from EHR. To reduce the number of candidate predictors and in turn improve model performance, we present an automated feature selection method based entirely on unlabeled observations. The proposed method generates a comprehensive surrogate for the underlying phenotype with an unsupervised clustering of disease status based on several highly predictive features such as diagnosis codes and mentions of the disease in text fields available in the entire set of EHR data. A sparse regression model is then built with the estimated outcomes and remaining covariates to identify those features most informative of the phenotype of interest. Relying on the results of Li and Duan (1989), we demonstrate that variable selection for the underlying phenotype model can be achieved by fitting the surrogate-based model. We explore the performance of our methods in numerical simulations and present the results of a prediction model for Rheumatoid Arthritis (RA) built on a large EHR data mart from the Partners Health System consisting of billing codes and NLP terms. Empirical results suggest that our procedure reduces the number of gold-standard labels necessary for phenotyping thereby harnessing the automated power of EHR data and improving efficiency.

Using whole genome scores to compare three clinical phenotyping methods in complex diseases

Genome-wide association studies depend on accurate ascertainment of patient phenotype. However, phenotyping is difficult, and it is often treated as an afterthought in these studies because of the expense involved. Electronic health records (EHRs) may provide higher fidelity phenotypes for genomic research than other sources such as administrative data. We used whole genome association models to evaluate different EHR and administrative data-based phenotyping methods in a cohort of 16,858 Caucasian subjects for type 1 diabetes mellitus, type 2 diabetes mellitus, coronary artery disease and breast cancer. For each disease, we trained and evaluated polygenic models using three different phenotype definitions: phenotypes derived from billing data, the clinical problem list, or a curated phenotyping algorithm. We observed that for these diseases, the curated phenotype outperformed the problem list, and the problem list outperformed administrative billing data. This suggests that using advanced EHR-derived phenotypes can further increase the power of genome-wide association studies.

Research use of electronic health records: patients' perspectives on contact by researchers.

ObjectiveThe use of electronic health records (EHRs) for research has the potential to improve the diagnosis and treatment of disease, yet contact with patients based on results of EHR phenotyping has received little attention. Researchers will almost certainly discover discrepancies in EHRs that call for resolution and, in some cases, raise the ethical dilemma of whether to contact patients about a potentially undiagnosed or untreated health concern. The objective of this study was to explore patients’ attitudes and opinions about potential contact by researchers who have had access to their EHRs.Materials and methodsWe conducted 15 focus groups in four diverse counties in the southeastern United States. We designed vignettes to describe different situations in which researchers conducting a hypothetical study might have reason to consider contact with patients.ResultsMany patients believed it was important for researchers to take action if they discovered information suggesting a current serious health concern. Relaying the information through patients’ physicians was considered the most appropriate course of action. Across vignettes, there were significant differences between urban and rural sites.Discussion and conclusionsResearchers may increasingly encounter situations involving contact with patients following EHR phenotyping. They should carefully consider the possibility of such contact when planning their studies, including the time and expertise needed to adjudicate potentially serious discrepancies. Our focus group results are one source of input for the development of ethical approaches to the research use of EHRs.

Using Electronic Health Records for Public Health Hypertension Surveillance

ObjectiveTo assess the equivalence of hypertension prevalence estimates between longitudinal electronic health record (EHR) data from a community-based health information exchange (HIE) and the Behavioral Risk Factor Surveillance System (BRFSS).IntroductionHypertension (HTN) is a highly prevalent chronic condition and strongly associated with morbidity and mortality. HTN is amenable to prevention and control through public and population health programs and policies. Therefore, public and population health programs require accurate, stable estimates of disease prevalence, and estimating HTN prevalence at the community-level is acutely important for timely detection, intervention, and effective evaluation. Current surveillance methods for HTN rely upon community-based surveys, such as the BRFSS. While BRFSS is the standard at the state- and national-level, they are expensive to collect, released once per year, and their confidence intervals are too wide for precise estimates at the local level. More timely, frequently updated, and locally precise prevalence estimates could greatly improve the timeliness and precision of public health interventions. The current study evaluated EHR data from a large, mature HIE as an alternative to community-based surveys for timely, accurate, and precise HTN prevalence estimation.MethodsTwo years (2014-2015) of EHR data were obtained from the Indiana Network for Patient Care for two major health systems in Marion County, Indiana, representing approximately 75% of the total county population (n=530,244). These data were linked and evaluated for prevalent HTN. Six HTN phenotypes were defined using structured data variables including clinical diagnoses (ICD9/10 codes), blood pressure (BP) measurements (HTN = ≥140mmhg systolic or ≥90mmHg diastolic), and dispensed HTN medications (Table 1). Phenotypes were validated using a random sample of 600 records, comparing EHR phenotype HTN to HTN as determined through manual chart review by a Registered Nurse. Each phenotype was further evaluated against BRFSS estimates for Marion County, and stratified by sex, race, and age to compare EHR-generated HTN prevalence measures to those known and in current use for chronic disease surveillance. Comparisons were made using the two one-sided statistical test (TOST) of equivalence, wherein the null hypothesis is the BRFSS and EHR prevalence estimates are different by +/-5% and the alternative is estimates differ by less than +/-5%. Rejection of the null resulted in the conclusion of equivalence of the estimates for use in population/public health.ResultsIn general, the performance of the EHR phenotypes was characterized by high specificity (>87%) and low to moderate sensitivity (range 25.4%-95.3%). The false positive rate was lowest among the phenotype defining HTN by both clinical diagnosis and BP measurements (0.3%), and sensitivity was greatest for the phenotype combining all three structured data elements (95.2%). The prevalence of HTN in Marion County, Indiana (2014-2015) for the EHR sample (n=530,244) ranged between 13.7% and 36.2%, compared to 28.4% in the BRFSS sample (Table 1). Only one EHR phenotype (≥1 HTN BP measurement) demonstrated equivalence with BRFSS prevalence at the county level (difference 0.9%, 90% CI for difference -2.3%-4.0%). HTN prevalence by sex, race, age, sex and age, and sex and race (n=120 comparisons) failed to demonstrate equivalence between EHR and BRFSS measures in all but two comparisons, both among females aged 18-39 years. Differences between EHR and BRFSS HTN prevalence at the subgroup level varied but were particularly pronounced among older adults. As suspected, HTN prevalence precision was improved in the EHR sample with the largest subgroup 95% CI width of 0.7% for male African Americans compared to the BRFSS sample 95% CI width of 29.6%.ConclusionsThe applicability of the tested HTN phenotypes will vary based upon which EHR structured data elements are available to public health (i.e., ICD10, vitals, medications). We found that HTN surveillance using a community-based HIE was not a valid replacement for the BRFSS, although the HIE-based estimates could be readily generated and had much narrower confidence intervals.ReferencesMozaffarian D, et al. Heart Disease and Stroke Statistics — 2016 Update. Circulation. 2016; 133: e38-e360.Yoon S, Fryar C, Carroll M. HTN Prevalence and Control Among Adults: United States, 2011–2014. NCHS Data Brief No. 220. 2015; Hyattsville, MD: National Center for Health Statistics, Centers for Disease Control and Prevention, US Dept of Health and Human Services.

A National Initiative in Data Science for Health: An Evaluation of the UK Farr Institute

Objective: To evaluate the extent to which the inter-institutional, inter-disciplinary mobilisation of data and skills in the Farr Institute contributed to establishing the emerging field of data science for health in the UK. Design and Outcome measures: We evaluated evidence of six domains characterising a new field of science: defining central scientific challenges, demonstrating how the central challenges might be solved, creating novel interactions among groups of scientists, training new types of experts, re-organising universities, demonstrating impacts in society. We carried out citation, network and time trend analyses of publications, and a narrative review of infrastructure, methods and tools. Setting Four UK centres in London, North England, Scotland and Wales (23 university partners), 2013-2018. Population Subsets of the UK’s 65 million population with health records accessible for research. The Farr Institute published a research corpus around a central scientific challenge, demonstrating insights from electronic health record (EHR) and administrative data at each stage of the translational cycle in 593 papers with at least one Farr Institute author affiliation on PubMed. Sample sizes showed some evidence of increase but remained less than 10% of the UK population in primary care-hospital care linked studies. The Farr Institute established the first four ISO27001 certified trusted research environments in the UK, and approved more than 1000 approved research users, published on 102 unique EHR and administrative data sources and established open platforms for the scalable re-use of EHR phenotyping (>70 diseases, CALIBER). The co-author publication network expanded from 944 unique co-authors (based on 67 publications in the first 30 months) to 3839 unique co-authors (545 papers in the final 30 months). 4/5 Centres established 27 new faculty (tenured) positions, 3 new university institutes, and 3 new master’s courses, training >400 people at master’s, short-course and leadership level and 48 PhD students. There were over 2300 citations for the 10 most cited papers and Farr research informed eight practice-changing clinical guidelines and policies relevant to the health of millions of UK citizens. Conclusion: The Farr Institute played a major role in establishing and growing the field of data science for health in the UK, with some initial evidence of benefits for health and healthcare. The Farr Institute has now expanded into Health Data Research UK but a key challenge is to network such activities internationally. What is already known: • National research initiatives in data science for health are under way in several countries seeking to harness insights from electronic health record (EHR) and administrative data at regional and national scale for patient and public benefit. • One approach to grow this emerging field, adopted by the UK, is to establish a dedicated national research institute, the Farr Institute. • We do not know how effective such initiatives are: multi-centre, inter-disciplinary research initiatives are common, but there is a lack of research evaluating such initiatives (in general) and national research institutes (in particular). • The Farr Institute ran from 2013 until 2018 when its larger-scale successor, Health Data Research UK, was established. What this study adds: • We provide a framework of six domains relevant for evaluating new inter-institutional, inter-disciplinary initiatives seeking to establish and grow an emerging field of science: defining central scientific challenges, demonstrating how the central challenges might be solved, creating novel interactions among groups of scientists, training new types of experts, re-organising universities, demonstrating impacts in society. • We show that the Farr Institute created new activities in and across each of the six domains for developing a distinctive research field. • The Farr Institute demonstrated the ability of multiple UK health research funders and multiple universities to partner in mobilising data, methodology and expertise across disciplines, organisations and information governance domains, resulting in a larger scale of research and improved methodology. • We have demonstrated globally relevant challenges and opportunities for developing data science for health across institutional and disciplinary barriers, consistent with the need for big investigation not simply big data.

Electronic Health Record Phenotypes for Precision Medicine: Perspectives and Caveats From Treatment of Breast Cancer at a Single Institution.

Precision medicine is at the forefront of biomedical research. Cancer registries provide rich perspectives and electronic health records (EHRs) are commonly utilized to gather additional clinical data elements needed for translational research. However, manual annotation is resource‐intense and not readily scalable. Informatics‐based phenotyping presents an ideal solution, but perspectives obtained can be impacted by both data source and algorithm selection. We derived breast cancer (BC) receptor status phenotypes from structured and unstructured EHR data using rule‐based algorithms, including natural language processing (NLP). Overall, the use of NLP increased BC receptor status coverage by 39.2% from 69.1% with structured medication information alone. Using all available EHR data, estrogen receptor‐positive BC cases were ascertained with high precision (P = 0.976) and recall (R = 0.987) compared with gold standard chart‐reviewed patients. However, status negation (R = 0.591) decreased 40.2% when relying on structured medications alone. Using multiple EHR data types (and thorough understanding of the perspectives offered) are necessary to derive robust EHR‐based precision medicine phenotypes.

A Novel Screening Method to Identify Late-Stage Dementia Patients for Palliative Care Research and Practice

ContextInvestigators need novel methods for timely identification of patients with serious illness to test or implement new palliative care models. ObjectivesThe study's aim was to develop an electronic health record (EHR) phenotype to identify patients with late-stage dementia for a clinical trial of palliative care consultation. MethodsWe developed a computerized method to identify patients with dementia on hospital admission. Within a data warehouse derived from the hospital's EHR, we used search terms of age, admission date, and ICD-9 and ICD-10 diagnosis codes to create an EHR dementia phenotype, followed by brief medical record review to confirm late-stage dementia. We calculated positive predictive value, false discovery rate, and false negative rate of this novel screening method. ResultsThe EHR phenotype screening method had a positive predictive value of 76.3% for dementia patients and 24.5% for late-stage dementia patients; a false discovery rate of 23.7% for dementia patients and 75.5% for late-stage dementia patients compared to physician assessment. The sensitivity of this screening method was 59.7% to identify hospitalized patients with dementia. Daily screening—including confirmatory chart reviews—averaged 20 minutes and was more feasible, efficient, and more complete than manual screening. ConclusionA novel method using an EHR phenotype plus brief medical record review is effective to identify hospitalized patients with late-stage dementia. In health care systems with similar clinical data warehouses, this method may be applied to serious illness populations to improve enrollment in clinical trials of palliative care or to facilitate access to palliative care services.

The use of electronic health records for psychiatric phenotyping and genomics.

The widespread adoption of electronic health record (EHRs) in healthcare systems has created a vast and continuously growing resource of clinical data and provides new opportunities for population-based research. In particular, the linking of EHRs to biospecimens and genomic data in biobanks may help address what has become a rate-limiting study for genetic research: the need for large sample sizes. The principal roadblock to capitalizing on these resources is the need to establish the validity of phenotypes extracted from the EHR. For psychiatric genetic research, this represents a particular challenge given that diagnosis is based on patient reports and clinician observations that may not be well-captured in billing codes or narrative records. This review addresses the opportunities and pitfalls in EHR-based phenotyping with a focus on their application to psychiatric genetic research. A growing number of studies have demonstrated that diagnostic algorithms with high positive predictive value can be derived from EHRs, especially when structured data are supplemented by text mining approaches. Such algorithms enable semi-automated phenotyping for large-scale case-control studies. In addition, the scale and scope of EHR databases have been used successfully to identify phenotypic subgroups and derive algorithms for longitudinal risk prediction. EHR-based genomics are particularly well-suited to rapid look-up replication of putative risk genes, studies of pleiotropy (phenomewide association studies or PheWAS), investigations of genetic networks and overlap across the phenome, and pharmacogenomic research. EHR phenotyping has been relatively under-utilized in psychiatric genomic research but may become a key component of efforts to advance precision psychiatry.

Electronic health record phenotype in subjects with genetic variants associated with arrhythmogenic right ventricular cardiomyopathy: a study of 30,716 subjects with exome sequencing.

PurposeArrhythmogenic right ventricular cardiomyopathy (ARVC) is an inherited heart disease. Clinical follow-up of incidental findings in ARVC-associated genes is recommended. We aimed to determine the prevalence of disease thus ascertained.Methods30,716 individuals underwent exome sequencing. Variants in PKP2, DSG2, DSC2, DSP, JUP, TMEM43, or TGFβ3 that were database-listed as pathogenic or likely pathogenic were identified and evidence-reviewed. For subjects with putative loss-of-function (pLOF) variants or variants of uncertain significance (VUS), electronic health records (EHR) were reviewed for ARVC diagnosis, diagnostic criteria, and International Classification of Diseases (ICD-9) codes.Results18 subjects had pLOF variants; none had an EHR diagnosis of ARVC. Of 14 patients with an electrocardiogram (ECG), one had a minor diagnostic criterion, 13 were normal. 184 subjects had VUSs; none had an ARVC diagnosis. In subjects with VUSs, there was no difference in the proportion with major (4%) or minor (13%) ECG diagnostic criteria compared to variant-negative controls. ICD-9 codes showed no difference in defibrillator utilization, electrophysiologic abnormalities or non-ischemic cardiomyopathies in patients with pLOF or VUSs compared to controls.ConclusionpLOF variants in an unselected cohort were not associated with ARVC phenotypes based on EHR review. The negative predictive value of EHR review remains uncertain.

Abstract WP315: An Electronic Health Record Phenotype of Ischemic Stroke Using Non-Claims Clinical Data and Machine Learning

Introduction: Real-time identification of patients with acute ischemic stroke (AIS) in the electronic health record (EHR) can enhance care delivery systems, clinical decision support, and research subject recruitment. EHR data that is accessible during a patient’s admission may be used to identify patients with AIS, but established methods for characterizing which data to use have not yet been determined. Hypotheses: 1. An EHR “phenotype” of AIS can be identified using clinical EHR data. 2. Machine learning can identify the AIS phenotype using similar inputs with greater accuracy than clinician-specified identification algorithms. Methods: Two stroke neurologists selected generalizable AIS-related clinical data points from the Columbia University Medical Center EHR (clinical laboratory results and medication, imaging, and stroke service list orders) to identify the AIS phenotype, and determined pre-specified priority logic based on institutional practice patterns. Separately, a regularized logistic regression (RLR) model was applied to all available neurology-related order sets and clinical laboratory inputs. The classification accuracy of the two algorithms was compared using a “gold standard” data set, consisting of our institution’s ischemic stroke registry from January 1 st , 2015 to March 31 st , 2016. Negative controls were selected from all patients admitted to the neurology service at our institution during the same time period. Results: Our data contained 482 patients with AIS and 3,628 negative controls. The clinician-specified identification algorithm identified the AIS phenotype with sensitivity of 90.6%, specificity of 50.4%, and positive predictive value (PPV) of 93.5%. In comparison, the RLR-based algorithm had a sensitivity of 96.3%, specificity of 52.2%, and PPV of 93.8%. Conclusions: We determined an AIS phenotype that could be identified using clinical, non-claims data that is available during a patient’s admission, and used machine learning to optimize the classifying ability. While specificity is low, the high sensitivity may allow use for screening and clinical decision support. Further studies are needed to externally validate these findings and optimize algorithm specificity.

"Pheno"menal value for human health.

Rare genetic variants are linked to electronic health record phenotypes at a population scale

Semi-supervised learning of the electronic health record for phenotype stratification

Patient interactions with health care providers result in entries to electronic health records (EHRs). EHRs were built for clinical and billing purposes but contain many data points about an individual. Mining these records provides opportunities to extract electronic phenotypes, which can be paired with genetic data to identify genes underlying common human diseases. This task remains challenging: high quality phenotyping is costly and requires physician review; many fields in the records are sparsely filled; and our definitions of diseases are continuing to improve over time. Here we develop and evaluate a semi-supervised learning method for EHR phenotype extraction using denoising autoencoders for phenotype stratification. By combining denoising autoencoders with random forests we find classification improvements across multiple simulation models and improved survival prediction in ALS clinical trial data. This is particularly evident in cases where only a small number of patients have high quality phenotypes, a common scenario in EHR-based research. Denoising autoencoders perform dimensionality reduction enabling visualization and clustering for the discovery of new subtypes of disease. This method represents a promising approach to clarify disease subtypes and improve genotype-phenotype association studies that leverage EHRs.

Assessing electronic health record phenotypes against gold-standard diagnostic criteria for diabetes mellitus.

We assessed the sensitivity and specificity of 8 electronic health record (EHR)-based phenotypes for diabetes mellitus against gold-standard American Diabetes Association (ADA) diagnostic criteria via chart review by clinical experts. We identified EHR-based diabetes phenotype definitions that were developed for various purposes by a variety of users, including academic medical centers, Medicare, the New York City Health Department, and pharmacy benefit managers. We applied these definitions to a sample of 173 503 patients with records in the Duke Health System Enterprise Data Warehouse and at least 1 visit over a 5-year period (2007-2011). Of these patients, 22 679 (13%) met the criteria of 1 or more of the selected diabetes phenotype definitions. A statistically balanced sample of these patients was selected for chart review by clinical experts to determine the presence or absence of type 2 diabetes in the sample. The sensitivity (62-94%) and specificity (95-99%) of EHR-based type 2 diabetes phenotypes (compared with the gold standard ADA criteria via chart review) varied depending on the component criteria and timing of observations and measurements. Researchers using EHR-based phenotype definitions should clearly specify the characteristics that comprise the definition, variations of ADA criteria, and how different phenotype definitions and components impact the patient populations retrieved and the intended application. Careful attention to phenotype definitions is critical if the promise of leveraging EHR data to improve individual and population health is to be fulfilled.

A Standards-based Semantic Metadata Repository to Support EHR-driven Phenotype Authoring and Execution.

This study describes our efforts in developing a standards-based semantic metadata repository for supporting electronic health record (EHR)-driven phenotype authoring and execution. Our system comprises three layers: 1) a semantic data element repository layer; 2) a semantic services layer; and 3) a phenotype application layer. In a prototype implementation, we developed the repository and services through integrating the data elements from both Quality Data Model (QDM) and HL7 Fast Healthcare Inteoroperability Resources (FHIR) models. We discuss the modeling challenges and the potential of our system to support EHR phenotype authoring and execution applications.

Electronic health record analysis via deep Poisson factor models

Electronic Health Record (EHR) phenotyping utilizes patient data captured through normal medical practice, to identify features that may represent computational medical phenotypes. These features may be used to identify at-risk patients and improve prediction of patient morbidity and mortality. We present a novel deep multi-modality architecture for EHR analysis (applicable to joint analysis of multiple forms of EHR data), based on Poisson Factor Analysis (PFA) modules. Each modality, composed of observed counts, is represented as a Poisson distribution, parameterized in terms of hidden binary units. Information from different modalities is shared via a deep hierarchy of common hidden units. Activation of these binary units occurs with probability characterized as Bernoulli-Poisson link functions, instead of more traditional logistic link functions. In addition, we demonstrate that PFA modules can be adapted to discriminative modalities. To compute model parameters, we derive efficient Markov Chain Monte Carlo (MCMC) inference that scales efficiently, with significant computational gains when compared to related models based on logistic link functions. To explore the utility of these models, we apply them to a subset of patients from the Duke-Durham patient cohort. We identified a cohort of over 16,000 patients with Type 2 Diabetes Mellitus (T2DM) based on diagnosis codes and laboratory tests out of our patient population of over 240,000. Examining the common hidden units uniting the PFA modules, we identify patient features that represent medical concepts. Experiments indicate that our learned features are better able to predict mortality and morbidity than clinical features identified previously in a large-scale clinical trial.