A maximum likelihood approach to electronic health record phenotyping using positive and unlabeled patients

Abstract

Objective Phenotyping patients using electronic health record (EHR) data conventionally requires labeled cases and controls. Assigning labels requires manual medical chart review and therefore is labor intensive. For some phenotypes, identifying gold-standard controls is prohibitive. We developed an accurate EHR phenotyping approach that does not require labeled controls. Materials and Methods Our framework relies on a random subset of cases, which can be specified using an anchor variable that has excellent positive predictive value and sensitivity independent of predictors. We proposed a maximum likelihood approach that efficiently leverages data from the specified cases and unlabeled patients to develop logistic regression phenotyping models, and compare model performance with existing algorithms. Results Our method outperformed the existing algorithms on predictive accuracy in Monte Carlo simulation studies, application to identify hypertension patients with hypokalemia requiring oral supplementation using a simulated anchor, and application to identify primary aldosteronism patients using real-world cases and anchor variables. Our method additionally generated consistent estimates of 2 important parameters, phenotype prevalence and the proportion of true cases that are labeled. Discussion Upon identification of an anchor variable that is scalable and transferable to different practices, our approach should facilitate development of scalable, transferable, and practice-specific phenotyping models. Conclusions Our proposed approach enables accurate semiautomated EHR phenotyping with minimal manual labeling and therefore should greatly facilitate EHR clinical decision support and research.

Application of concise machine learning to construct accurate and interpretable EHR computable phenotypes

“…There are several possible directions for further improving FEAT. For one, the ability of FEAT to recapitulate expert-curated heuristics suggests that simpler expert heuristics, such as anchor variables 49 , may be leveraged as teachers in a semi-supervised approach. This could be implemented with multi-stage learning, first to predict heuristics and then to predict chart-review.…”

Phenotyping coronavirus disease 2019 during a global health pandemic: Lessons learned from the characterization of an early cohort

“… Assign probability of known disease; [30] evaluate data driven selection of cases or controls such as a maximum likelihood approach. [31] Defined Logic Data use requires knowledge of data cleaning processes. Build a data dictionary documenting representation of data elements (e.g., Boolean, temporal) as well as cleaning methods.…”

Desiderata for a Synthetic Clinical Data Generator

“…The first approaches transfer real patient data into synthetic data by learning the patterns of disease and care in real background EHRs. The EMERGE framework [7] and machine learning-based approaches such as medGAN [8] or statistical approaches [9] fall into this category. While these algorithms require a comparatively minor amount of manual input from subject matter experts, they will inevitably make severe errors when capturing the disease progression and careflow models, compromising the realism of the output data [7,8].…”