Psychotropic deprescribing across different prescribing professions in New Mexico and Louisiana

This study aimed to assess differences in the rates of deprescribing between prescribing psychologists, psychiatrists, and primary care physicians. MarketScan private insurance claims were used to develop a longitudinal active-comparator, prevalent-user cohort of patients who were treated with a psychotropic medication from psychologists, psychiatrists, or primary care physicians for at least 90 days in New Mexico or Louisiana (states where psychologists can prescribe) between 2005-2021. The type of provider (psychologist, psychiatrist, or primary care physicians) who prescribed the psychotropic medication was the exposure of interest. Three measures of deprescribing were used as outcomes: deprescribing without replacement, complete discontinuation of prescribing, or a sustained reduction in the prescribed days' supply. Patient demographic and clinical characteristics during the six months prior to their initial prescription were measured as covariates. Inverse propensity of treatment weighting was used to adjust for baseline differences between provider groups, creating two weighted analytic cohorts with covariate balance: psychologists versus psychiatrists and psychologists versus primary care physicians. We estimated doubly-robust Cox Proportional Hazards models for each deprescribing measure in both cohorts. Prescribing psychologists deprescribed without replacement more than psychiatrists (Hazard ratio [95% CI] = 1.13 [1.06, 1.20]) and less than primary care physicians (0.73 [0.69, 0.78]). Conversely, they reduced the days' supply less often than psychiatrists (0.79 [0.69, 0.91]) and more than primary care physicians (1.64 [1.42, 1.90]). There were no differences in complete discontinuation between provider types. Prescribing psychologists deprescribe at a rate between psychiatrists and primary care physicians. Findings varied depending on the deprescribing measure used, suggesting psychotropic-specific deprescribing measures are needed.