Abstract P355: Random-effects Modeling Of Implicit Differences Between Patients, Clinicians, And Institutions And Antihypertensive Therapeutic Intensification

Abstract

Background: Evidence points to treatment intensification (TI) as the most important factor impacting BP control. In our modeling of TI from EHR data, 13.2% of variance in TI could be explained by care-process and clinical factors. TI is usually framed as a clinician trait, but evidence has yet to show a significant contribution from clinician-related variables to TI in hypertension management. Random effects modeling can quantify the impact of implicit differences between patients, clinicians, and institutions on antihypertensive (AH) TI. Methods: Data for defining and modeling AH TI comes from ambulatory primary care clinics at five health care organizations. Multivariable mixed-effects logistic regression was used to model AH TI and several clinical and encounter related variables. Random intercepts were added to the model for each individual patient, clinician, site, and health care organization (HCO). Results: Comparing stepwise R 2 for a multivariable model of AH TI with random intercepts for patient, clinician, site, and HCO indicates implicit variation between clinicians adds virtually nothing to variation in TI (0.001%) vs. 0.5% for patient, 0.3% for site, and 0.2% for HCO-related differences. Conclusion: Among the variables examined, the primary source of variation in TI is explained by patient differences. The collective contribution to TI by site and HCO differences were comparable to patient, but clinician differences were essentially non-contributory. The data suggest the patient-clinician relationship (patient-centered care) and HCO impact TI. Greater understanding of these dynamics could potentially improve the low rate of TI for uncontrolled BP.