Decrease of Anticholinergic Drug Use in Nursing Home Residents in the United States, 2009 to 2017

Abstract

Objective: To investigate the time course of anticholinergic drug use in nursing home residents and assess if any temporal change in anticholinergic use varied by nursing home quality rating.

Design: Retrospective repeated cross-sectional analysis of Medicare enrollment, Parts A, B, and D, claims data linked to the Minimum Data Set from 2009 to 2017.

Setting: Medicare-certified nursing homes.

Participants: Long-term residents 65 years or older with nursing home stay of at least 100 consecutive days within a given calendar year.

Measurements: Estimates of anticholinergic drug prescription rates between 2009 and 2017 were based on a binary variable indicating whether a resident received a drug with high anticholinergic activity, as defined by the Anticholinergic Cognitive Burden scale, for at least 1 day during the initial 100 consecutive days of nursing home stay in a given calendar year. We used mixed effects logistic regression models to determine adjusted rates of anticholinergic use each year and test the interaction between nursing home quality rating and year, while adjusting for patient and nursing home characteristics.

Results: The cohort included 786,858 100-day nursing home stays (299,354 unique residents) in 6,703 nursing homes for the years 2009 to 2017. Prescription rates were stable at approximately 34% to 35% between 2009 and 2011, then gradually decreased to 24.3% in 2017 (P < .0001), with the decline being more pronounced in nursing homes having high quality ratings (P < .0001). Rates for anticholinergic drugs in nursing homes with 4 to 5 star quality rating (33.7% in 2011 to 23.3% in 2017) showed a steeper decline over time relative to nursing homes with 1 to 2 star quality rating (34.2% in 2011 to 25.2% in 2017) (P < .0001).

Conclusions: The use of drugs with high anticholinergic activity has declined from 2009 to 2017, with a greater decline in higher-quality nursing homes.

Keywords: anticholinergic drug use; antipsychotics; nursing homes.

Figure 1.

Adjusted Prevalence of Anticholinergic Prescriptions in Nursing Homes by Year and Joinpoint Analysis for the Years 2009 to 2017. The adjusted rates for the calendar year were estimated using Mixed Effects Logistic Regression modeling, while adjusting for race, sex, age, quality rating, location, bed number, ownership type and geographic region. Error bars are 95% confidence intervals. The model-derived rates were used in subsequent joinpoint analysis, which indicated a statistically significant change of slope with Joinpoint estimate for the year 2011 (95% Confidence Interval 2011 – 2013). For the period 2009 to 2011, the slope was not statistically different from zero (p ≈ 0.6), and for the years 2011 to 2017 the estimated rate of decline in the prevalence of anticholinergic prescriptions was 1.7% per year (p < 0.0001).

Figure 2.

Adjusted Prevalence of Anticholinergic Prescriptions per Year by Nursing Home Quality Rating. The moderating effect of nursing home quality rating on the decline in anticholinergic prescription rate over time was investigated using Mixed Effects Logistic Regression modeling, controlling for the nesting of observations in nursing homes while adjusting for the fixed effects race, sex, age, location, bed number, ownership type and geographic region. Error bars are 95% confidence intervals. Prescribing rate in nursing homes with a quality rating “4-5” declined from 33.7% in 2011 to 23.3% in 2017, whereas it declined from 34.2% in 2011 to 25.2% in 2017 in nursing homes with a quality rating “1-2”. Prescribing rate in nursing homes with a quality rating “3” declined from 34.3% in 2011 to 24.5% in 2017.

Figure 3.

Prescription Rates of Specific Anticholinergic Drug Classes Over the Period 2009 - 2017. Please note that the y-axes denoting rates of drug use are different in each plot. The choice of y-axis' scale on each plot was made based on are intention to focus on the relative changes of rates among years. The absolute and relative changes between 2009 and 2017 are reported on the upper-right corner of each plot. The adjusted rates for each drug class were estimated using Mixed Effects Logistic Regression modeling, controlling for the nesting of observations in nursing homes while adjusting for the fixed effects race, sex, age, quality rating, location, bed number, ownership type and geographic region. In all drug classes the effect of year was statistically significant (p < .0001). Error bars are 95% confidence intervals.