OBJECTIVE: Patients who receive most care within a single healthcare system (colloquially called a "loyalty cohort" since they typically return to the same providers) have mostly complete data within that organization's electronic health record (EHR). Loyalty cohorts have low data missingness, which can unintentionally bias research results. Using proxies of routine care and healthcare utilization metrics, we compute a per-patient score that identifies a loyalty cohort. MATERIALS AND METHODS: We implemented a computable program for the widely adopted i2b2 platform that identifies loyalty cohorts in EHRs based on a machine-learning model, which was previously validated using linked claims data. We developed a novel validation approach, which tests, using only EHR data, whether patients returned to the same healthcare system after the training period. We evaluated these tools at 3 institutions using data from 2017 to 2019. RESULTS: Loyalty cohort calculations to identify patients who returned during a 1-year follow-up yielded a mean area under the receiver operating characteristic curve of 0.77 using the original model and 0.80 after calibrating the model at individual sites. Factors such as multiple medications or visits contributed significantly at all sites. Screening tests' contributions (eg, colonoscopy) varied across sites, likely due to coding and population differences. DISCUSSION: This open-source implementation of a "loyalty score" algorithm had good predictive power. Enriching research cohorts by utilizing these low-missingness patients is a way to obtain the data completeness necessary for accurate causal analysis. CONCLUSION: i2b2 sites can use this approach to select cohorts with mostly complete EHR data.

OBJECTIVE: The increasing translation of artificial intelligence (AI)/machine learning (ML) models into clinical practice brings an increased risk of direct harm from modeling bias; however, bias remains incompletely measured in many medical AI applications. This article aims to provide a framework for objective evaluation of medical AI from multiple aspects, focusing on binary classification models. MATERIALS AND METHODS: Using data from over 56 000 Mass General Brigham (MGB) patients with confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), we evaluate unrecognized bias in 4 AI models developed during the early months of the pandemic in Boston, Massachusetts that predict risks of hospital admission, ICU admission, mechanical ventilation, and death after a SARS-CoV-2 infection purely based on their pre-infection longitudinal medical records. Models were evaluated both retrospectively and prospectively using model-level metrics of discrimination, accuracy, and reliability, and a novel individual-level metric for error. RESULTS: We found inconsistent instances of model-level bias in the prediction models. From an individual-level aspect, however, we found most all models performing with slightly higher error rates for older patients. DISCUSSION: While a model can be biased against certain protected groups (ie, perform worse) in certain tasks, it can be at the same time biased towards another protected group (ie, perform better). As such, current bias evaluation studies may lack a full depiction of the variable effects of a model on its subpopulations. CONCLUSION: Only a holistic evaluation, a diligent search for unrecognized bias, can provide enough information for an unbiased judgment of AI bias that can invigorate follow-up investigations on identifying the underlying roots of bias and ultimately make a change.

OBJECTIVE: High-throughput electronic phenotyping algorithms can accelerate translational research using data from electronic health record (EHR) systems. The temporal information buried in EHRs is often underutilized in developing computational phenotypic definitions. This study aims to develop a high-throughput phenotyping method, leveraging temporal sequential patterns from EHRs. MATERIALS AND METHODS: We develop a representation mining algorithm to extract 5 classes of representations from EHR diagnosis and medication records: the aggregated vector of the records (aggregated vector representation), the standard sequential patterns (sequential pattern mining), the transitive sequential patterns (transitive sequential pattern mining), and 2 hybrid classes. Using EHR data on 10 phenotypes from the Mass General Brigham Biobank, we train and validate phenotyping algorithms. RESULTS: Phenotyping with temporal sequences resulted in a superior classification performance across all 10 phenotypes compared with the standard representations in electronic phenotyping. The high-throughput algorithm's classification performance was superior or similar to the performance of previously published electronic phenotyping algorithms. We characterize and evaluate the top transitive sequences of diagnosis records paired with the records of risk factors, symptoms, complications, medications, or vaccinations. DISCUSSION: The proposed high-throughput phenotyping approach enables seamless discovery of sequential record combinations that may be difficult to assume from raw EHR data. Transitive sequences offer more accurate characterization of the phenotype, compared with its individual components, and reflect the actual lived experiences of the patients with that particular disease. CONCLUSION: Sequential data representations provide a precise mechanism for incorporating raw EHR records into downstream machine learning. Our approach starts with user interpretability and works backward to the technology.

OBJECTIVE: Due to a complex set of processes involved with the recording of health information in the Electronic Health Records (EHRs), the truthfulness of EHR diagnosis records is questionable. We present a computational approach to estimate the probability that a single diagnosis record in the EHR reflects the true disease. MATERIALS AND METHODS: Using EHR data on 18 diseases from the Mass General Brigham (MGB) Biobank, we develop generative classifiers on a small set of disease-agnostic features from EHRs that aim to represent Patients, pRoviders, and their Interactions within the healthcare SysteM (PRISM features). RESULTS: We demonstrate that PRISM features and the generative PRISM classifiers are potent for estimating disease probabilities and exhibit generalizable and transferable distributional characteristics across diseases and patient populations. The joint probabilities we learn about diseases through the PRISM features via PRISM generative models are transferable and generalizable to multiple diseases. DISCUSSION: The Generative Transfer Learning (GTL) approach with PRISM classifiers enables the scalable validation of computable phenotypes in EHRs without the need for domain-specific knowledge about specific disease processes. CONCLUSION: Probabilities computed from the generative PRISM classifier can enhance and accelerate applied Machine Learning research and discoveries with EHR data.

OBJECTIVE: The Consortium for Clinical Characterization of COVID-19 by EHR (4CE) is an international collaboration addressing coronavirus disease 2019 (COVID-19) with federated analyses of electronic health record (EHR) data. We sought to develop and validate a computable phenotype for COVID-19 severity. MATERIALS AND METHODS: Twelve 4CE sites participated. First, we developed an EHR-based severity phenotype consisting of 6 code classes, and we validated it on patient hospitalization data from the 12 4CE clinical sites against the outcomes of intensive care unit (ICU) admission and/or death. We also piloted an alternative machine learning approach and compared selected predictors of severity with the 4CE phenotype at 1 site. RESULTS: The full 4CE severity phenotype had pooled sensitivity of 0.73 and specificity 0.83 for the combined outcome of ICU admission and/or death. The sensitivity of individual code categories for acuity had high variability-up to 0.65 across sites. At one pilot site, the expert-derived phenotype had mean area under the curve of 0.903 (95% confidence interval, 0.886-0.921), compared with an area under the curve of 0.956 (95% confidence interval, 0.952-0.959) for the machine learning approach. Billing codes were poor proxies of ICU admission, with as low as 49% precision and recall compared with chart review. DISCUSSION: We developed a severity phenotype using 6 code classes that proved resilient to coding variability across international institutions. In contrast, machine learning approaches may overfit hospital-specific orders. Manual chart review revealed discrepancies even in the gold-standard outcomes, possibly owing to heterogeneous pandemic conditions. CONCLUSIONS: We developed an EHR-based severity phenotype for COVID-19 in hospitalized patients and validated it at 12 international sites.

MOTIVATION: Expert-labeled data are essential to train phenotyping algorithms for cohort identification. However expert labeling is time and labor intensive, and the costs remain prohibitive for scaling phenotyping to wider use-cases. RESULTS: We present an approach referred to as polar labeling (PL), to create silver standard for training machine learning (ML) for disease classification. We test the hypothesis that ML models trained on the silver standard created by applying PL on unlabeled patient records, are comparable in performance to the ML models trained on gold standard, created by clinical experts through manual review of patient records. We perform experimental validation using health records of 38 023 patients spanning six diseases. Our results demonstrate the superior performance of the proposed approach. AVAILABILITY AND IMPLEMENTATION: We provide a Python implementation of the algorithm and the Python code developed for this study on Github. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

OBJECTIVE: The study sought to design, pilot, and evaluate a federated data completeness tracking system (CTX) for assessing completeness in research data extracted from electronic health record data across the Accessible Research Commons for Health (ARCH) Clinical Data Research Network. MATERIALS AND METHODS: The CTX applies a systems-based approach to design workflow and technology for assessing completeness across distributed electronic health record data repositories participating in a queryable, federated network. The CTX invokes 2 positive feedback loops that utilize open source tools (DQe-c and Vue) to integrate technology and human actors in a system geared for increasing capacity and taking action. A pilot implementation of the system involved 6 ARCH partner sites between January 2017 and May 2018. RESULTS: The ARCH CTX has enabled the network to monitor and, if needed, adjust its data management processes to maintain complete datasets for secondary use. The system allows the network and its partner sites to profile data completeness both at the network and partner site levels. Interactive visualizations presenting the current state of completeness in the context of the entire network as well as changes in completeness across time were valued among the CTX user base. DISCUSSION: Distributed clinical data networks are complex systems. Top-down approaches that solely rely on technology to report data completeness may be necessary but not sufficient for improving completeness (and quality) of data in large-scale clinical data networks. Improving and maintaining complete (high-quality) data in such complex environments entails sociotechnical systems that exploit technology and empower human actors to engage in the process of high-quality data curating. CONCLUSIONS: The CTX has increased the network's capacity to rapidly identify data completeness issues and empowered ARCH partner sites to get involved in improving the completeness of respective data in their repositories.

Objective: To provide an open source, interoperable, and scalable data quality assessment tool for evaluation and visualization of completeness and conformance in electronic health record (EHR) data repositories. Materials and Methods: This article describes the tool's design and architecture and gives an overview of its outputs using a sample dataset of 200 000 randomly selected patient records with an encounter since January 1, 2010, extracted from the Research Patient Data Registry (RPDR) at Partners HealthCare. All the code and instructions to run the tool and interpret its results are provided in the Supplementary Appendix. Results: DQe-c produces a web-based report that summarizes data completeness and conformance in a given EHR data repository through descriptive graphics and tables. Results from running the tool on the sample RPDR data are organized into 4 sections: load and test details, completeness test, data model conformance test, and test of missingness in key clinical indicators. Discussion: Open science, interoperability across major clinical informatics platforms, and scalability to large databases are key design considerations for DQe-c. Iterative implementation of the tool across different institutions directed us to improve the scalability and interoperability of the tool and find ways to facilitate local setup. Conclusion: EHR data quality assessment has been hampered by implementation of ad hoc processes. The architecture and implementation of DQe-c offer valuable insights for developing reproducible and scalable data science tools to assess, manage, and process data in clinical data repositories.