Abstract
Blood pressure measurement plays a critical role in the screening of atherosclerotic cardiovascular disease and in the diagnosis and management of a broad range of medical conditions (Arnett et al., 20219; Tsao et al., 2023; Whelton et al., 2017). However, measuring and recording blood pressures in the setting of remotely-delivered care remains an understudied phenomenon. Since the rapid healthcare transformations in the wake of the COVID-19 pandemic, approximately 20-30% of primary care services are conducted through remote means (Graetz et al., 2024; Lee et al,, 2023); how this shift in care modality impacts hypertension screening and blood pressure-informed clinical decision-making remains unknown. Applying traditional and machine learning methodologies to a large regional health system dataset combining electronic health record, administrative, and billing data for over 1.4 million encounters between 2020- 2024, blood pressure measurement was noted as a nearly universal phenomenon in office-based encounters (99.9%, n = 1,029,289) but rare in video- (20%, n = 91,128) and audio-based encounters (19%, n = 63), p<0.001. Machine learning algorithms including logistic regression with and without regularization, CART, Boosted Trees, and Random Forests were applied to identify factors predictive of blood pressure measurement. Across all models, modality of care overwhelmingly outperformed all other variables as the leading predictor for blood pressure measurement, with AUCs ranging between 0.968 to 0.990. This research highlights urgent implications for policy, practice, research, and health systems to ensure that quality, safety, and the integrity of medical diagnosis and preventive objectives remain robust in the setting of hybridized primary care.