Natural language processing for scalable feature engineering and ultra-high-dimensional confounding adjustment in healthcare database studies.

Background: To improve confounding control in healthcare database studies, data-driven algorithms may empirically identify and adjust for large numbers of pre-exposure variables that indirectly capture information on unmeasured confounding factors ('proxy' confounders). Current approaches for high-dimensional proxy adjustment do not leverage free-text notes from EHRs. Unsupervised natural language processing (NLP) technology can scale to generate large numbers of structured features from unstructured notes.

Comparative ranking of marginal confounding impact of natural language processing-derived versus structured features in pharmacoepidemiology.

Objective: To explore the ability of natural language processing (NLP) methods to identify confounder information beyond what can be identified using claims codes alone for pharmacoepidemiology.

Methods: We developed a retrospective cohort for high vs low dose proton pump inhibitors from linked Medicare claims (2008-2017) and electronic health record data for patients with a history of peptic ulcer disease or osteoarthritis. Clinical notes authored one year before first dispensing date were processed by off-the-shelf tools: bag-of-n-grams, latent Dirichlet allocation, a linguistics-focused tool, BERT sentence embeddings, BioBERT word embeddings, and GloVe word embeddings.

Patient Characteristics Associated With Using Transcatheter Left Atrial Appendage Occlusion Versus Oral Anticoagulants for Atrial Fibrillation.

Background: Transcatheter left atrial appendage occlusion (LAAO) is an alternative to oral anticoagulants (OACs) for stroke prevention in patients with atrial fibrillation, but the predictors of LAAO use in routine care are unclear. We aimed to assess the utilization trends of LAAO and compare the change in characteristics of LAAO users versus OACs since its marketing.

Methods: Using the US Medicare claims database (March 15, 2015, to December 31, 2020), we identified patients with atrial fibrillation, ≥65 years, and CHADS-VASc score ≥2 (men) or ≥3 (women), with either first implantation of an LAAO device or initiation of OACs, including apixaban, dabigatran, rivaroxaban, edoxaban, or warfarin.

Development and Validation of a Claims-Based Model to Predict Categories of Obesity.

We developed and validated a claims-based algorithm that classifies patients into obesity categories. Using Medicare (2007-2017) and Medicaid (2000-2014) claims data linked to 2 electronic health record (EHR) systems in Boston, Massachusetts, we identified a cohort of patients with an EHR-based body mass index (BMI) measurement (calculated as weight (kg)/height (m)2). We used regularized regression to select from 137 variables and built generalized linear models to classify patients with BMIs of ≥25, ≥30, and ≥40.

Risk of Cataract Surgery and Age-Related Macular Degeneration After Initiation of Denosumab vs Zoledronic Acid for Osteoporosis: A Multi-Database Cohort Study.

Background And Objective: There is a relative lack of head-to-head comparisons of denosumab against other osteoporosis drugs on safety. We aimed to explore ocular outcomes in patients with osteoporosis initiating denosumab vs zoledronic acid.

Methods: We conducted a cohort study using claims data (2010-15) from two large US commercial insurance databases including patients with osteoporosis who were aged 50 years or older and initiators of denosumab or zoledronic acid.

Comparative Safety and Effectiveness of Denosumab Versus Zoledronic Acid in Patients With Osteoporosis: A Cohort Study.

Limited head-to-head comparative safety and effectiveness data exist between denosumab and zoledronic acid in real-world healthcare. We aimed to examine the safety and effectiveness of denosumab compared to zoledronic acid with regard to risk of serious infection and cardiovascular disease (CVD) and osteoporotic fracture. We conducted a cohort study using claims data (2009-2013) from a US commercial insurance plan database.