Potentially Harmful Medication Prescribing by the Degree of Physician Specialization in Nursing Home Practice: An Observational Study

Abstract

Objectives: The use of anticholinergics, antipsychotics, benzodiazepines, and other potentially harmful medications (PHMs) is associated with particularly poor outcomes in nursing home (NH) residents with Alzheimer's disease and related dementias (ADRD). Our objective was to compare PHM prescribing by NH physicians and advanced practitioners who focus their practice on NH residents (NH specialists) vs non-NH specialists.

Design: Retrospective cohort study.

Setting and participants: We included a 20% random sample of Medicare beneficiaries with ADRD who resided in 12,278 US NHs in 2017. Long-stay NH residents with ADRD were identified using MDS, Medicare Parts A and B claims. Residents <65 years old or without continuous Part D coverage were excluded.

Methods: Physicians in generalist specialties and advanced practitioners with ≥90% of Part B claims for NH care were considered NH specialists. Residents were assigned to NH specialists vs non-NH specialists based on plurality of Part D claims submitted for that resident. Any PHM use (defined using the Beers Criteria) and the proportion of NH days on a PHM were modeled using generalized estimating equations. Models included resident demographics, clinical characteristics, cognitive and functional status, behavioral assessments, and facility characteristics.

Results: Of the 54,713 residents in the sample, 27.9% were managed by an NH specialist and 72.1% by a non-NH specialist. There was no statistically significant difference in any PHM use [odds ratio (OR) 0.97, 95% CI 0.93-1.02, P = .23]. There were lower odds of prolonged PHM use (OR 0.87, 95% CI 0.81-0.94, P < .001, for PHM use on >75% vs >0%-<25% of NH days) for NH specialists vs non-NH specialists.

Conclusions and implications: Although the use of PHMs among NH residents with ADRD managed by NH specialists was not lower, they were less likely to receive PHMs over longer periods of time. Future work should evaluate the underlying causes of these differences to inform interventions to improve prescribing for NH residents.

Keywords: Potentially harmful medications; antipsychotics; nursing home; physicians; specialization.

Figure 1.

Sample Selection The flow diagram depicts inclusion and exclusion criteria for the study sample.

Figure 2.

Forest Plot of Odds of Any PHM Use and of Prolonged PHMs Use by Patients of NH-specialists vs. Non-NH-Specialists This forest plot shows the adjusted odds for each category of outcome overall and by drug class. To accommodate the distribution of PHM prescribing rates, two sets of models were estimated for each outcome: an any vs. none ‘zero component’ model and a multinomial model with three categories of PHM use (>0 to <25% of nursing home days, 25–75% of days, and >75% of days). * All odds ratios adjusted for resident demographics (age, gender, race, dual Medicare and Medicaid coverage), clinical characteristics (Cognitive Function Scale, Charlson comorbidity index, CHESS score, ARBS score, MDS-ADL score), the presence of six mental health conditions (anxiety, depression, bipolar disorder, schizophrenia, other psychotic disorder, and PTSD), and nursing home characteristics (profit status, size, whether the facility was part of a chain, urban vs. rural location, percent of residents insured by Medicare, Medicaid, percent of residents who were Black, patient acuity index, direct care staffing in hours per resident day, and the number of deficiencies). ^† Two models were estimated for each outcome: an any vs. none ‘zero component’ model and a multinomial model with three categories of PHM use (>0 to <25% of nursing home days, 25–75% of days, and >75% of days).