Retrospective Evaluation of a Pragmatic Deprescribing Initiative in a Skilled Nursing Facility System

Context: Polypharmacy increases risk of adverse drug effects.

Objective: To retrospectively determine whether a pragmatic deprescribing protocol reduced 8 common classes of medications in two skilled nursing facilities (SNFs) in a single system.

Study Design and Analysis: Retrospective, longitudinal pre/post evaluation. A preliminary analysis was published using data from 2017-first half of 2021. This follow up examines whether initial improvements were maintained over a longer evaluation interval, comparing the pre (2017-2019) with post-intervention years (2020-2021).

Setting or Dataset: Long-term resident data reported through annual comprehensive minimum data set (MDS) reviews conducted at two skilled nursing facilities in a single system.

Population Studied: Long-term residents at two skilled nursing facilities.

Intervention/Instrument: Interdisciplinary deprescribing effort to reduce medication in SNF residents including clinician education, guideline development, and chart reviews.

Outcome Measures: Odds of being administered each of 8 classes of medication (Diuretic, Opioid, Antipsychotic, Anticoagulant, Antianxiety, Antibiotic, Hypnotic, Antidepressant, and average total classes of medications (0-8) per record. Odds Ratios for each class of medication, and mean number of medication classes, were compared between the first and updated analyses. Each analysis controlled for race, female gender, and age.

Results: There were 15,117 resident data points available for analysis. The mean total medications per resident slightly decreased in the post-intervention period (mean=1.85 classes of medication per resident at both facilities from 2020-2021 vs 1.88 from 2017-2019). Significant decreases in ORs were maintained for Diuretic (.846, p<.001 vs .82 p<.001), Opioid (.740, p<.001 vs .79 p<.001), and Antipsychotic (.848, p<.001 vs .80 p=.006) administration; a decrease in odds of being administered an antibiotic became significant (.888, p<.004 vs .98 p= .711). A nonsignificant decrease in the odds of Antianxiety drug administration was also maintained (.932, p=.187 vs .89 p=.138).

Conclusions: Reductions were maintained for medication classes with serious side effects). These results are convergently supported by a separate cost-effectiveness analysis.