Use of electronic health records to develop an actionable taxonomy of persistent hypertension

ABSTRACT Background The digital transformation of medical data presents opportunities for novel approaches to manage patients with persistent hypertension (defined as multiple measurements of elevated BP over 6 months). We sought to develop an actionable taxonomy of patients with persistent hypertension based on clinical data from the electronic health records (EHR). Methods This qualitative study was a content analysis of clinician notes in the EHR of patients in the Yale New Haven Health System. Eligible patients were 18 to 85 years and had blood pressure ≥160/100 mmHg at five or more consecutive outpatient visits between January 1 st 2013 to October 31 st 2018. A total of 4,828 patients met criteria, of which 200 records were randomly selected for chart review. Through a systematic, inductive approach, we developed a rubric to abstract data from the EHR and then analyzed the abstracted data qualitatively using conventional content analysis until saturation was reached. Results We reached saturation with 115 patients, who had a mean age of 68.1 (SD, 11.6) years; 54.8% were female; 52.2%, 30.4%, and 13.9% were White, Black, and Hispanic people. We identified three content domains related to persistence of hypertension: (1) non-intensification of pharmacological treatment (defined as absence of antihypertensive treatment intensification in response to persistent severely elevated blood pressure) with four subcategories, including provider purview, competing medical priorities, patient preference, and de-emphasis of the office measurement; (2) non-implementation of prescribed treatment (defined as a documentation of provider recommending a specified treatment plan to address hypertension but treatment plan not being implemented) with four subcategories, including obstacles to obtaining medications, psychosocial barriers, patient misunderstanding, and negative medication experience; and (3) non-response to prescribed treatment (defined as clinician-acknowledged persistent hypertension despite documented effort to escalate existing pharmacologic agents and addition of additional pharmacologic agents with presumption of adherence) with two subcategories, including resistant hypertension and secondary hypertension. Conclusions This study presents a novel actionable taxonomy for classifying patients with persistent hypertension by their contributing causes based on EHR data. These categories can be automated and linked to specific types of actions to address them. Clinical Perspective What is Known? This study developed a novel actionable taxonomy for classifying patients with persistent hypertension by their contributing causes, using qualitatively content analysis of clinician notes in the EHR. We identified three main content domains and a variety of subcategories contributing to persistent hypertension (non-intensification of pharmacological treatment, non-implementation of prescribed treatment, non-response to prescribed treatment), providing actionable information to inform solutions. This taxonomy is based on real-world data in the EHR, so it is pragmatic for use in the clinical setting. What the Study Adds? This actionable taxonomy lays the foundation for developing effective tools for health systems to rapidly identify and classify people with persistent hypertension and connect them with targeted, personalized interventions at scale. Personalized medicine depends on distinguishing patients with persistent hypertension by their contributing factors; such knowledge is essential for tailoring care to individuals with appropriate evidence-based interventions.