Deprescribing Medications Among Older Adults From End of Hospitalization Through Postacute Care: A Shed-MEDS Randomized Clinical Trial

Key Points

Question  Does a pharmacist- or nurse practitioner–led patient-centered deprescribing intervention reduce or stop medications for older adults at hospital discharge, postacute care (PAC) facility discharge, and 90-day follow-up?

Findings  In this randomized clinical trial that included 372 older adults with polypharmacy who were transitioning from hospitalization to PAC, those who received a patient-centered deprescribing intervention had significantly fewer medications compared with the control group who received usual care at PAC facility discharge and at the 90-day follow-up.

Meaning  The findings suggest the safety and effectiveness of the deprescribing intervention in reducing the total medication burden at PAC facility discharge, which was sustained 90 days after this discharge.

Importance  Deprescribing is a promising approach to addressing the burden of polypharmacy. Few studies have initiated comprehensive deprescribing in the hospital setting among older patients requiring ongoing care in a postacute care (PAC) facility.

Objective  To evaluate the efficacy of a patient-centered deprescribing intervention among hospitalized older adults transitioning or being discharged to a PAC facility.

Design, Setting, and Participants  This randomized clinical trial of the Shed-MEDS (Best Possible Medication History, Evaluate, Deprescribing Recommendations, and Synthesis) deprescribing intervention was conducted between March 2016 and October 2020. Patients who were admitted to an academic medical center and discharged to 1 of 22 PAC facilities affiliated with the medical center were recruited. Patients who were 50 years or older and had 5 or more prehospital medications were enrolled and randomized 1:1 to the intervention group or control group. Patients who were non–English speaking, were unhoused, were long-stay residents of nursing homes, or had less than 6 months of life expectancy were excluded. An intention-to-treat approach was used.

Interventions  The intervention group received the Shed-MEDS intervention, which consisted of a pharmacist- or nurse practitioner–led comprehensive medication review, patient or surrogate-approved deprescribing recommendations, and deprescribing actions that were initiated in the hospital and continued throughout the PAC facility stay. The control group received usual care at the hospital and PAC facility.

Main Outcomes and Measures  The primary outcome was the total medication count at hospital discharge and PAC facility discharge, with follow-up assessments during the 90-day period after PAC facility discharge. Secondary outcomes included the total number of potentially inappropriate medications at each time point, the Drug Burden Index, and adverse events.

Results  A total of 372 participants (mean [SD] age, 76.2 [10.7] years; 229 females [62%]) were randomized to the intervention or control groups. Of these participants, 284 were included in the intention-to-treat analysis (142 in the intervention group and 142 in the control group). Overall, there was a statistically significant treatment effect, with patients in the intervention group taking a mean of 14% fewer medications at PAC facility discharge (mean ratio, 0.86; 95% CI, 0.80-0.93; P < .001) and 15% fewer medications at the 90-day follow-up (mean ratio, 0.85; 95% CI, 0.78-0.92; P < .001) compared with the control group. The intervention additionally reduced patient exposure to potentially inappropriate medications and Drug Burden Index. Adverse drug event rates were similar between the intervention and control groups (hazard ratio, 0.83; 95% CI, 0.52-1.30).

Conclusions and Relevance  Results of this trial showed that the Shed-MEDS patient-centered deprescribing intervention was safe and effective in reducing the total medication burden at PAC facility discharge and 90 days after discharge. Future studies are needed to examine the effect of this intervention on patient-reported and long-term clinical outcomes.

Trial Registration  ClinicalTrials.gov Identifier: NCT02979353

Polypharmacy is prevalent among older hospitalized patients and is associated with adverse postdischarge outcomes.^1-4 Patients and caregivers, overwhelmingly, are willing to deprescribe (ie, stop or reduce) 1 or more of their medications if their physicians agree.^5-7 Although deprescribing is effective, substantial evidence gaps remain. For example, the majority of deprescribing interventions have been limited to specific drug classes or medical conditions.^8-10 Few deprescribing interventions have considered the entirety of the medication list across the breadth of medical conditions. Furthermore, fewer trials have initiated deprescribing in the hospital, and, to our knowledge, none of these trials have included patients needing continuing care in a postacute care (PAC) facility.¹¹

There may be advantages to deprescribing in both the acute care and PAC settings for older adults. First, the high prevalence of polypharmacy and multimorbidity suggests a likelihood of benefit.⁴ Second, older patients may be admitted for conditions associated with specific medications (eg, syncope), which may enhance the receptiveness of patients and clinicians to reducing medications.¹² Third, both care settings provide an opportunity to supervise and monitor potential deprescribing-related adverse drug withdrawal events (ADWEs).¹³

The purpose of this randomized clinical trial was to evaluate the efficacy of a patient-centered deprescribing intervention (Shed-MEDS [Best Possible Medication History, Evaluate, Deprescribing Recommendations, and Synthesis]) among hospitalized older adults transitioning or being discharged to a PAC facility. The primary objective of the intervention was to reduce overall medication count at hospital discharge and PAC facility discharge and to assess the maintenance of intervention effects 90 days after PAC facility discharge. The secondary objective was to identify the intervention effects on the number of potentially inappropriate medications (PIMs)^14-16 and the anticholinergic and sedative drug burden.¹⁷

The Shed-MEDS randomized clinical trial ^18,19 was approved by the institutional review board at Vanderbilt University Medical Center (VUMC) and a National Institute on Aging–appointed Data Safety Monitoring Board. The trial protocol is provided in Supplement 1. All patients or surrogates provided written informed consent for their participation. We followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.

Between March 2016 and October 2020, we recruited patients 50 years or older who were admitted to VUMC and had 5 or more prehospital medications. These patients were referred to 1 of 22 PAC facilities within a 9-county area surrounding VUMC in Nashville, Tennessee. The geographic inclusion area enabled a home visit by the research team at the 90-day follow-up. We excluded patients who were non–English speaking, were unhoused, were long-stay residents of nursing homes, or had less than 6 months of life expectancy. Participants were identified through a daily physical therapy report of patients who had been recommended for PAC placement and subsequently confirmed in the case management system as having a referral to a partner PAC facility.

Demographic characteristics, comorbidities, and medication measures of participants (Table 1) were extracted from their medical record or assessed by study personnel via a standardized interview. Demographic characteristics were collected to characterize the trial population and included age, sex, race and ethnicity (including Asian; Black or African American; Hispanic or non-Hispanic; Native American or Alaska Native; Pacific Islander or Native Hawaiian; White; and other, unknown, or declined/refused to answer), and educational level (≤high school or > high school diploma). Medical diagnoses (using International Classification of Diseases, Ninth Revision, and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, codes) were used to calculate the Charlson Comorbidity Index, which ranged from 0 to 37, with higher scores indicating greater number of comorbidities.²⁰ We assessed cognitive status using the Brief Interview for Mental Status, which had scores ranging from 0 to 15, with scores below 13 indicating cognitive impairment.²¹ We conducted a Best Possible Medication History (BPMH) process, supplementing the initial electronic medical record medication list with patient or surrogate interview, pharmacy refill history, review of outside medical records (eg, Medication Admission Record from a transferring facility), and the Tennessee Controlled Substance Monitoring Database to determine the total number of medications for each patient.^18,22 We also collected the number of outpatient prescribers and pharmacies attributed to the medications present at hospital admission. Finally, we asked patients if they received assistance with medication management at home (yes or no), such as picking up refills from the pharmacy, organizing pills, and/or being reminded to take their medications.

We randomized participants in a 1:1 ratio using permuted blocks to the Shed-MEDS intervention (intervention group) or to BPMH plus usual care at the hospital and PAC facility (control group). Block size was either 2 or 4 and was selected uniformly at random. A total of 9283 hospitalized patients were screened for eligibility. Figure 1 indicates the frequency of exclusions, with living outside the geographic area being the most common reason.

A description of the intervention protocol has been previously published.^19,22 In brief, the hospital intervention phase included 4 steps that were conducted by trained research clinicians (ie, pharmacists or nurse practitioners). First, a research clinician reviewed all BPMH-listed medications, including over-the-counter (OTC) medications and supplements, for deprescribing (defined as stopping or reducing medications) indications. Second, research clinicians led a deprescribing conversation with the patient and/or surrogate, including medications that were identified in the first step along with medications that the patient or surrogate had an interest in deprescribing. This semistructured interview included medication-specific questions about medication knowledge, adherence, adverse effects, and efficacy.²³

Third, we discussed potential deprescribing actions with the outpatient prescribers, including dose reductions or titrations. We synthesized deprescribing targets into final actions and discussed these targets with the inpatient treatment team, who implemented them at their own discretion. Fourth, within 48 hours of the transfer to the PAC facility, the research clinician contacted via telephone the admission or charge nurse at the receiving PAC facility to review the medication list and agreed-on deprescribing recommendations. During the PAC facility stay, research clinicians called weekly to review the medication administration record and monitor deprescribing actions with the facility’s prescribing authority, including documentation of recommended dose reductions or titrations and related symptom monitoring. At the time of PAC facility discharge, research clinicians sent the reconciled PAC discharge list and ongoing deprescribing recommendations to relevant outpatient prescribers.

The Shed-MEDS deprescribing intervention ended at PAC facility discharge. However, we followed up with participants for up to 90 days after discharge, including performing an in-person home visit at the final follow-up time point.

The primary outcome was the total medication count at hospital discharge and PAC facility discharge, with follow-up assessments during the 90 days after PAC facility discharge. Medications included all prescribed or OTC drugs that were scheduled or as needed with oral, intravenous, subcutaneous, rectal, transdermal, and ophthalmic administrations, while topicals (eg, lotions) and otic medications were excluded. Enrollment (baseline) medication counts were based on the prehospital medication count (per the BPMH) and any newly prescribed in-hospital medications.

The secondary outcomes were the total number of PIMs at each time point, the Drug Burden Index (DBI), and adverse events. A PIM was defined as any medication on 1 of 3 previously published lists, including the Beers Criteria,¹⁴ the STOPP (Screening Tool of Older Persons' Prescriptions) Criteria,¹⁵ and the RASP (Rationalization of Home Medication by an Adjusted STOPP in Older Patients) List.¹⁶ The PIM classification assessed only for presence or absence on any list regardless of the clinical context or indication. Additionally, we measured the DBI, a validated continuous measure of sedative and anticholinergic burden.¹⁷ The DBI is the sum of each medication’s daily dose divided by the minimum effective dose (as estimated by the US Food and Drug Administration [FDA] minimum recommended dose) and the patient’s daily dose. The DBI captures reductions in dose, even when the total number of medications is not reduced. For both the intervention and control groups, we examined the frequency, by FDA general drug categories, of dose-reduced or stopped medications at PAC facility discharge.²⁴

We measured adverse events throughout the trial (from enrollment to 90 days after PAC facility discharge) in both groups. An adverse event was defined as any unplanned emergency department visit, an intensive care unit transfer among currently hospitalized patients, a new hospitalization, or death, with the latter 3 events being categorized as serious. All adverse events were assessed by a trained physician reviewer using the 10-point Naranjo Scale to determine whether events were possibly related (score >1) to an adverse drug event (ADE) or ADWE.^25,26 Physician reviewers were blinded to the group assignment. Further details on adverse event determination are provided in the eMethods in Supplement 2.

Due to the in-person nature of the Shed-MEDS intervention, the clinical research staff conducting the intervention and collecting data at the follow-up time point could not be blinded to the intervention status. The primary investigators and trained physician reviewers (including A.S.M. and S.K.) for safety measures (eg, ADEs) were blinded to the group assignment.

The effect of the intervention on the total number of medications and PIMs at hospital discharge, PAC facility discharge, and 90 days after PAC facility discharge was quantified using mixed-effects Poisson regression, adjusting for measurement time point (as a categorical covariate), the enrollment medication count for each outcome, and the interaction of intervention and time point. The within-participant association among repeated measurements was modeled using a random intercept indexed by participant. The overall statistical significance of the treatment effect was evaluated using a Wald-type multiple degree of freedom, which tested the null hypothesis that the intervention had no effect at any time point (hospital discharge, PAC facility discharge, and 90-day follow-up) after randomization. Two-sided P < .05 was considered statistically significant. No adjustments were made to control the familywise type-I error probability for multiple testing. The intervention effect at each time point was summarized using the estimated relative effect on the mean, or mean ratio, with 95% CIs.

In a sensitivity analysis, we repeated the main analysis excluding OTC medications. The effect of intervention on the DBI was assessed in a similar fashion using linear mixed-effects regression rather than Poisson regression. Data for all randomized patients who were discharged from the hospital to a partner PAC facility were included in the analyses. All participant measurements, regardless of subsequent attrition, were used in the analyses. The incidence of missing data due to attrition was examined by group assignment. Unadjusted comparisons across groups were made using the Wilcoxon test or Pearson χ² test, and summaries were generated using the sample proportion, median, and IQR as appropriate. The rates of adverse events, including those that reoccurred, were compared across groups by computing a hazard ratio (HR) with 95% CI using a shared frailty model adjusted for the assigned group (intervention vs control), with frailty term indexed by participant. This type of model is an extension of the Cox proportional hazards regression model, in which the hazard function has a random effect that accounts for heterogeneity among participants. The frailty model is used when the event of interest may occur multiple times for any one participant (eg, rehospitalization).²⁷

At the time of trial initiation, we expected to enroll approximately 576 participants across 4 years of enrollment with 27.5% attrition, or 420 total participants. The pilot intervention was associated with an approximately 50% reduction in the count of total medications from enrollment to hospital discharge, whereas an approximately 25% reduction was observed in the control group under usual hospital care.¹⁹ Using the mixed-effects Poisson regression method, we implemented a simulation-based power analysis assuming that these effects would be attenuated by 20% at PAC facility discharge and again by 20% at the 90-day follow-up. A planned sample of 420 participants provided greater than 95% power to detect this magnitude of effect.

An intention-to-treat method was used for all statistical analyses: all participants were analyzed according to the group assignment at the time of randomization, regardless of subsequent compliance with the study protocol or follow-up completeness. Data were managed using the REDCap platform.²⁸ All statistical analyses were performed with R, version 4.0.3 (R Foundation for Statistical Computing).

Of the 1353 eligible patients, 372 (28%) provided consent and were randomized to either the intervention group (n = 186) or control group (n = 186) (Figure 1). A total of 284 participants (142 in the intervention group and 142 in the control group) were included in the intention-to-treat analysis. Table 1 shows participant demographic and clinical characteristics for the total sample and by group. Overall, participants had a mean (SD) age of 76.2 (10.7) years and included 229 females (62%) and 143 males (38%), with most being of White race and ethnicity (312 [84%]) and admitted from home (303 [81%]). Participants had a median (IQR) Charlson Comorbidity Index of 6.0 (5.0-9.0), and 78 (21%) exhibited cognitive impairment. The median (IQR) hospital length of stay from enrollment to hospital discharge was 2.0 (1.0-4.0) days, and the median (IQR) PAC length of stay was 22 (15.0-34.0) days. Patients received prescriptions from a median (IQR) of 3 (2-3) outpatient prescribers and 1 (1-2) pharmacy in the previous 3 months. A total of 212 patients (57%) reported receiving help with medication management prior to hospital admission, and 123 patients (33%) reported not receiving such help.

The median (IQR) number of prehospital medications was 16 (12.0-20.0) per patient. At enrollment, the median (IQR) total number of prehospital medications in addition to new hospital medications was 23.0 (19.0-29.0) per patient (Table 2).

The median (IQR) total number of medications stopped at PAC facility discharge (from the time of enrollment) was 14.0 (11.0-18.0) in the intervention group and 12.0 (9.0-16.0) in the control group. The median (IQR) total number of medications started at PAC facility discharge was 3.0 (1.0-5.0) in the intervention group and 3.0 (1.2-4.8) in the control group. Adjusting for the total medication count at enrollment, the mean total medication count was similar between intervention and control groups at hospital discharge, but patients in the intervention group had a mean of 14% fewer medications at PAC facility discharge (mean ratio, 0.86; 95% CI, 0.80-0.93; P < .001) when the intervention ended and a mean of 15% fewer medications 90 days after PAC facility discharge (mean ratio, 0.85; 95% CI, 0.78-0.92; P < .001) (Table 2).

The magnitude of the overall treatment effect was high across the entire period of observation (P < .001, testing the null hypothesis that the intervention had no effect at any time point), with the intervention group having significantly fewer medications compared with the control group (hospital discharge mean ratio, 0.96 [95% CI, 0.90-1.02]; PAC discharge mean ratio, 0.86 [95% CI, 0.80-0.93]; 90-day follow-up mean ratio, 0.85 [95% CI, 0.78-0.92]) (Figure 2). Similar treatment effects were observed in sensitivity analyses that excluded all OTC medications (hospital discharge mean ratio, 0.95 [95% CI, 0.89-1.02]; PAC discharge mean ratio, 0.87 [95% CI, 0.80-0.95]; 90-day follow-up mean ratio, 0.87 [95% CI, 0.80-0.96]) (eTable 1 in Supplement 2).

The mean (SD) number of PIMs was similar between the intervention and control groups at hospital discharge (9.3 [3.5] vs 8.8 [3.3]). The intervention group, however, was prescribed significantly fewer PIMs at PAC facility discharge (7.7 [3.0]) and during the 90 days after PAC facility discharge (8.9 [3.5]). The DBI was significantly lower for the intervention group at each time point (hospital discharge mean difference, –0.28 [95% CI, –0.51 to –0.04]; PAC discharge mean difference, –0.60 [95% CI, –0.86 to –0.34]; 90-day follow-up mean difference, –0.35 [95% CI, –0.63 to –0.07]). For both PIMs and DBI, the overall treatment effects had a great magnitude (P < .001). The PIM counts and DBI values across time points are shown in eFigures 1 and 2 in Supplement 2, respectively.

The Shed-MEDS intervention deprescribed medications across numerous drug classes according to FDA general drug categories. eFigure 3 in Supplement 2 displays the most frequently deprescribed medication classes (at least 40 deprescribing events), with vitamins or supplements, laxatives, and antihypertensives being the most frequently deprescribed medications at PAC facility discharge. eTable 2 in Supplement 2 provides further description of the top 3 medications deprescribed by drug categories.

The rates of overall adverse events, ADEs, and ADWEs were comparable between intervention and control groups (HR, 0.83; 95% CI, 0.52-1.30). There was a consistent pattern toward lower rates in the intervention group (Table 3). For example, the most frequent adverse event was hospitalization, with 84 events in the intervention group and 107 in the control group (HR, 0.76; 95% CI, 0.53-1.09).

After the patient-centered Shed-MEDS deprescribing intervention, participants at PAC facility discharge experienced a significant reduction in total medications across a wide range of drug categories, which was sustained up to 90 days after the end of active deprescribing, when compared with patients who received usual care. Additionally, the intervention decreased PIMs and lowered the DBI across the intervention period and during the 90-day follow-up. The intervention was not associated with increased rates of overall adverse events or ADWEs. These findings demonstrated that a hospital-initiated, pharmacist- or nurse practitioner–led, patient-centered deprescribing intervention can effectively and safely reduce the medication burden among hospitalized older patients needing PAC.

Although this trial was the first, to our knowledge, to implement deprescribing among hospitalized older patients transitioning to PAC, this trial was preceded by other inpatient deprescribing studies.^29-35 The overall effect size of deprescribing in this trial was larger at the end of the intervention (PAC facility discharge) in comparison to the effect size in previous reports. Compared with other inpatient trials, in the present work, we were able to suggest deprescribing actions throughout the PAC facility stay, thus potentially enhancing the efficacy of the Shed-MEDS intervention. In addition, most previous trials used explicit tools to target PIMs for deprescribing, most often defined by the Beers Criteria. A systematic review identified 9 such trials, 3 of which demonstrated statistically significant reductions in PIMs between hospital admission and discharge.¹¹ The number of PIMs at enrollment in those studies was lower than the population of the present trial, a difference that may be explained by the more restrictive approach taken by previous studies to classify PIMs in comparison to this study, which included any medication on 3 PIM lists, regardless of prescribing indication or patient condition. Additionally, the BPMH process we followed was exhaustive and identified patient medications that may be missing from standard admission medications lists.¹⁸ In a meta-analysis of 25 deprescribing trials, the population effect was a mean difference of 0.4 drugs between the intervention and control groups.³⁶ Recently, a large multicenter deprescribing cluster randomized trial (OPERAM [Optimizing Therapy to Prevent Avoidable Hospital Admissions in Multimorbid Older Adults]) did not result in significant differences in PIMs between the intervention and control groups.³⁷ The larger absolute amount of deprescribing achieved in the present trial (eg, 2 fewer PIMs at PAC facility discharge) may be attributed to a consideration of a broader range of medications and the continued deprescribing throughout the PAC facility stay.

The intervention group in this trial experienced significant reductions in total medications at PAC facility discharge, and those reductions were maintained 90 days after discharge. Medication reductions, however, were modest at the earliest postenrollment time point (hospital discharge). The smaller effect size at hospital discharge may be explained by 2 factors. First, the median time between enrollment and hospital discharge was only 2 days. This period allowed for a limited time to fully implement all deprescribing steps. The disadvantage of delayed enrollment was counterbalanced by the ability to continue the intervention in the PAC setting, where the median length of stay was 22 days. This time allowed for weekly opportunities to perform medication reviews, respond to clinical concerns, and reinforce deprescribing recommendations with the PAC team. Without the ongoing involvement of a clinician facilitating deprescribing after PAC facility discharge, the intervention effect appeared durable but did not increase in magnitude. Hospitalized older patients commonly experience recurrent acute care episodes (ie, emergency department visits or rehospitalizations), which is a risk factor for initiation of new PIMs.³⁸ Future interventions may incorporate ongoing surveillance to address potential new deprescribing indications.

The PIMs and DBI similarly decreased among participants in the intervention group. This finding suggests that medication reductions include those with high clinical relevance, such as those associated with geriatric syndromes (eg, falls). Clearly identifying PIMs (including sedating and anticholinergic medications) according to multiple validated lists and the DBI may aid clinicians in prioritizing medications for deprescribing, especially when the starting list is substantial. There are multiple drug classes that may not be included on specific PIM lists or the DBI but may cause unintended harm. For example, guidelines for diabetes control change among older patients; however, patients may continue to be on higher medication doses for which the risk of harms may outweigh the benefits.³⁹ In addition, although there are benefits to lower blood pressure targets for older patients, particularly among those who are frail and continuing care at a PAC facility, there continue to be opportunities to reduce the medication burden of patients with hypertension in response to either ongoing hypotension; ongoing symptoms (eg, lightheadedness); or admissions with falls, fractures, or syncope.^40-44 In this trial, the most frequently deprescribed drug classes were supplements and laxatives. Commonly deprescribed medications for chronic disease included those for hypertension and diabetes. This finding suggests that in addition to focusing on PIMs, assessing the appropriateness of supplements and blood pressure and hemoglobin A_1c targets is important in developing deprescribing strategies to reduce the complexity and potential harm of medication regimens.

A key concern of initiating deprescribing, from both patient and clinician perspectives, is developing serious adverse events, including hospitalization and death.⁴⁵ We found no increases in emergency department visits, hospitalizations, or mortality. The number of ADWEs was modest in the intervention group and did not differ significantly from that in the control group. This similarity in adverse event outcomes supports the safety of a patient-centered deprescribing strategy when initiated in the hospital and continued in the PAC facility. These results can support the conversations about the safety of deprescribing. Another barrier of deprescribing, specifically in the PAC setting, is the availability and training of clinicians. The efficacy of deprescribing in PAC would suggest the benefits of engaging clinical pharmacists or nurse practitioners with specific abilities to perform comprehensive medication reviews, identify opportunities to deprescribe, monitor symptoms, and provide patient and clinician education at the time of discharge.

These findings must be interpreted in consideration of trial limitations. First, patient enrollment at a single academic hospital may limit generalizability, although this was balanced by the inclusion of 22 PAC facilities. Second, enrolled patients may be more willing to deprescribe. This potential for enrollment bias could increase intervention effects; however, it also could increase deprescribing behavior in the control group. Third, the research clinicians provided deprescribing recommendations but did not implement the actual deprescribing actions, which remained the responsibility of the primary prescribers. Although this approach may have weakened the intensity of the deprescribing effect, we believe it was a pragmatic and acceptable approach for both clinicians and patients and may help explain the sustainability of the intervention effects. Fourth, due to the large number of PAC sites and small numbers of enrolled participants at some of these sites, we did not adjust for site-level effects. Fifth, inpatient clinicians were not blinded to the intervention. Thus, it is possible that new deprescribing skills and behaviors may have been applied to patients in the control group, which would reduce the observed degree of differences between the intervention and control groups.

In this randomized clinical trial, older adults requiring PAC after hospitalization experienced high levels of polypharmacy. The patient-centered Shed-MEDS deprescribing intervention was found to be safe and effective in reducing the total medication burden at PAC facility discharge, with a sustained effect of up to 90 days after discharge. Future research is needed to examine the effects of deprescribing on patient-reported and additional long-term clinical outcomes.

Accepted for Publication: December 4, 2022.

Published Online: February 6, 2023. doi:10.1001/jamainternmed.2022.6545

Corresponding Author: Eduard E. Vasilevskis, MD, MPH, Vanderbilt University, 2525 West End Ave, Ste 450, Nashville, TN 37203 (ed.vasilevskis@vumc.org).

Author Contributions: Drs Vasilevskis and Simmons had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Vasilevskis, Shah, Shotwell, Kripalani, Simmons.

Acquisition, analysis, or interpretation of data: Vasilevskis, Shah, Hollingsworth, Shotwell, Mixon, Simmons.

Drafting of the manuscript: Vasilevskis, Shah, Hollingsworth, Shotwell, Simmons.

Critical revision of the manuscript for important intellectual content: Vasilevskis, Shah, Shotwell, Kripalani, Mixon, Simmons.

Statistical analysis: Vasilevskis, Shah, Shotwell, Simmons.

Obtained funding: Vasilevskis, Simmons.

Administrative, technical, or material support: Shah, Hollingsworth, Mixon.

Supervision: Vasilevskis, Shah, Kripalani, Simmons.

Other - I'm the principal investigator of this study: Simmons.

Conflict of Interest Disclosures: Dr Kripalani reported receiving grants from Bristol Myers Squibb/Sanofi and grants from IBM Corporation outside the submitted work. Dr Mixon reported receiving a Health Services Research and Development Service grant from the US Department of Veterans Affairs outside the submitted work. No other disclosures were reported.

Funding/Support: This study was funded by grant R01AG053264 from the National Institute on Aging of the National Institutes of Health (co–principal investigators: Drs Vasilevskis and Simmons). The use of institutional data management system REDCap was supported by a Clinical and Translational Science Awards grant UL1TR000445 from the National Center for Advancing Translational Sciences.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Additional Contributions: We thank the members of the Shed-MEDS team, including the following personnel who are not listed as coauthors: Carole Bartoo, GNP; Jennifer Kim, GNP; Kanah Lewallen, GNP; Whitney Narramore, PharmD; Robin Parker, PharmD; Susan Lincoln, BS; Joanna Gupta, MEd; and Jessica Lovell, PharmD. They received no additional compensation, beyond their usual salary, for their contribution to this work.