Impact of on-site pharmacists in residential aged care facilities on the quality of medicines use: a cluster randomised controlled trial (PiRACF study)

Residents of residential aged care facilities (RACFs) have a high prevalence of use of potentially inappropriate medications (PIMs) and resultant medicines-related harm. This study investigated the effect of an on-site pharmacist model on PIMs use and other medication outcomes for residents in RACFs. A multi-facility, non-blind, cluster randomised controlled trial, with randomisation at the facility level, was conducted. Fifteen facilities enrolled and participated in the study, 7 facilities (560 residents) were allocated to the intervention arm and 8 facilities (737 residents) were allocated to the control arm. Each facility in the intervention arm employed an on-site pharmacist for 12 months to perform medication management activities as part of an interdisciplinary care team. The primary outcome was the proportion of residents taking at least one PIM according to the 2019 Beers® Criteria. Using generalised linear mixed-effects models, accounting for confounders and clustering, there was a significant reduction in the proportion of residents prescribed at least one PIM (odds ratio 0.50, 95% confidence interval, 0.335–0.750; p = 0.001) in the intervention arm. There were also significant decreases in the Anticholinergic Cognitive Burden scale and chlorpromazine equivalent daily dose of antipsychotics. The on-site pharmacist intervention significantly improved the appropriateness of medicines use in RACFs.


Randomisation and recruitment
Randomisation was performed at the facility level.RACFs were stratified by size of facility and randomised into either the intervention or control group through computer-generated allocation by an independent person external to the research team.All RACFs in the ACT, Australia, that met the inclusion criteria were invited to participate in this study.Due to the nature of the intervention, participants were not blinded.Permanent residents of participating RACFs were included unless they requested to opt out.New residents entering RACFs after the baseline data collection that remained at the RACFs until the end-point data collection, were also included.Respite (non-permanent) residents were excluded.Only residents who were aged 65 years or over were included in the data analysis.

Intervention
Each participating intervention RACF directly employed a pharmacist as a facility staff member reporting to the RACF manager, to work on-site for 2 to 2.5 days per week for 12 months.Salaries for pharmacists were funded by the research grant.The intervention (model of care) was informed by the findings of a pilot study and discussion with RACF managers, general practitioners (GPs), pharmacists, and a consumer representative who participated in the pilot [30][31][32] .Training and support were provided to pharmacists prior to starting their role and during the period of the intervention 28 .RMMR and QUM services were conducted by RACF usual contractors in both study arms and did not impact the services provided by the on-site pharmacists.On-site pharmacists performed activities including medication reviews and medication reconciliation at transitions of care; clinical audits to identify those at highest risk of medication-related problems; group and individual education with staff, residents, and families; and improving facility's medication management policies.A list of pharmacist activities in the model of care was detailed in the protocol 28 .On-site pharmacists self-reported their daily activities through an online diary.

Outcomes measures
The primary outcome measure was the proportion of residents who were taking at least one PIM according to the American Geriatrics Society Beers® 2019 criteria 33 .The Beers® criteria were modified in this study to fit the Australian setting to include medications that are available in Australia 8 .The use of PIMs has often been used a marker for QUM in RACFs 6,[34][35][36] .The first author (I.H.) applied the Beers® Criteria to residents' medications with ongoing review by the chief investigators.The unit of analysis for medication-related outcomes was at the resident-level.
Secondary outcomes included the proportion of residents who were prescribed at least one psychotropic medicine (defined as antipsychotics and benzodiazepines in the absence of major psychiatric disease or epilepsy), residents' daily dose of psychotropic medicines (measured as chlorpromazine and diazepam equivalent daily dose per resident) 37,38 , residents' mean Anticholinergic Cognitive Burden (ACB) scale 39 , the number of regular medications, and the proportion of residents who had complete documentation of drug allergies or adverse drug reactions in their RACF records.

Statistical analyses
Descriptive statistics included mean and standard deviation for numeric variables and proportion for categorical variables.Mann-Whitney-U tests and Chi-square were used for unadjusted comparison of (continuous and categorical, respectively) baseline characteristics between control and intervention arms.
This study used generalised linear mixed-effects models (GLMMs) to compare outcome variables between intervention and control groups at baseline and endpoint.Logistic GLMM was used for binary outcome variables (e.g., taking at least one PIM: yes/no), Poisson was used for discrete count variables (e.g., number of regular medications), and gamma distributions were used for continuous positively skewed variables (e.g., daily dose of psychotropic medicines).The models adjusted for potential confounders (age, gender, Charlson Comorbidity Index (CCI), presence of other educational intervention 40 , diagnosis of dementia and number of regular medicines), and accounted for clustering of residents within RACFs and repeated observation within residents.The intra-cluster correlation coefficient (ICC) at baseline was calculated for the primary outcome 41 .For each outcome variable, the model estimated main effects of group (control or intervention), time (baseline and endpoint) and the interaction group by time effect (group*time).The estimated coefficient associated to the interaction term reflects the difference in changes over time by study arms, also known as the difference-in-difference approach between group effect (intervention or control) and time effect (baseline and intervention) 42 .Results are reported with 95% confidence intervals (CI).Missing values were uncommon (less than 1%) and a listwise deletion was applied.Data analysis was conducted using the Statistical Package for Social Sciences (SPSS, version 27.0;IBM Corp. Armonk, USA).Any result with an associated probability value less than 5% (p < 0.05) was considered statistically significant.

Study population
Fifteen RACFs enrolled and participated in the study.After randomisation, 7 RACFs (560 residents) were allocated to the intervention arm and 8 RACFs (737 residents) were allocated to the control arm.Twenty-two residents were lost to follow-up and 16 residents were excluded because they were under 65 years of age, where the Beers® Criteria are not applicable.Overall, 1275 residents from the baseline and 1301 residents from the endpoint were included in the final analysis.The flow diagram of the study population is presented in Fig. 1.
The baseline characteristics of participating residents were similar between groups, although the control group had slightly more females and had higher percentage of residents diagnosed with dementia (Table 1).The demographics of participating residents were compared to Australian national data for RACF residents and is presented in Table 2.

Impact of intervention on the primary outcome
Table 3 shows descriptive results at baseline and endpoint for each of the study outcomes.The proportion of residents who were prescribed at least one regular PIM in the intervention group reduced from 70.6% [interquartile interval 66.6-74.4%]at baseline to 60.8% [interquartile interval 56.8-64.7%]at endpoint, with little change in the control group over the study period (Table 3).Table 4 presents a summary of the results for the primary and secondary study outcomes.This is presented as the two-way interaction effect of intervention at the endpoint (i.e., intervention arm * Endpoint) for both unadjusted models and after adjusting for clustering and potential confounders.Supplementary Tables 1-7 present the full details of each model with the main effects, interactions effects, and confounders considered in adjusted models.In the GLMM logistic model for the primary outcome when after adjusting for confounders, residents in the intervention group had halved the odds of having at least one prescribed PIM over the period of the study, when compared to the control group (odds ratio 0.50, 95% CI: 0.335-0.750;P = 0.001).Clustering of the primary outcome was lower than anticipated at baseline (ICC = 3.17%, 95% CI 1.0-9.64%).

Impact of intervention on secondary outcomes
Resident's mean ACB scale decreased from 1.21 at baseline to 0.94 at endpoint in the intervention group, and from 1.21 at baseline to 1.14 at endpoint in the control group.Similarly, models estimating resident's ACB scale showed a lowered risk of ACB scale with a rate ratio of 0.800 (95% CI: 0.678-0.0944) in the intervention group when compared to the control group over the period of the study.Residents in the intervention group had lower odd of taking one or more benzodiazepines or antipsychotics (odds ratio 0.68, 95% CI 0.439-1.042)over the study period but was not statistically significant (P = 0.076).When estimating the effect of the intervention on the doses of residents' antipsychotics, the chlorpromazine equivalent daily dose showed a statistically significant decrease of 0.250 mg (95% CI − 0.456 to − 0.043) (P = 0.018) over the study period.The diazepam equivalent daily dose of benzodiazepines per resident was reduced by 0.129 mg (95% CI − 0.428 to 0.170) but was not statistically significant (P = 0.397).Models estimating the effect on the number of regular medications and the proportion of residents with complete ADR documentation showed no statistically significant differences between control and intervention groups (see Supplementary Tables 1-7 for details).

Discussion
This is the first randomised controlled trial conducted to assess the effects of an on-site pharmacist in residential aged care model on medication appropriateness.The study found a statistically significant reduction in the number of residents using PIMs, residents' mean anticholinergic burden scale, and chlorpromazine equivalent dose of antipsychotic medications per resident in the intervention group compared to the control.There were no significant changes identified in other outcomes, including number of regular medications, ADR documentation status, the proportion of residents taking one or more benzodiazepine or antipsychotic, and the dose of benzodiazepines between control and intervention groups.
In this study, pharmacists were included as part of each facility's multi-disciplinary care team to perform resident-level clinical activities, including comprehensive medication reviews, and facility-level activities such www.nature.com/scientificreports/as clinical audits and contributing to policies and procedures.This holistic approach to medication management allowed pharmacists to prioritise and resolve medication-related problems as they occurred and mitigate risks through improving organisational systems within the RACF [43][44][45] .An advantage of integrating pharmacists within the healthcare team of RACF is the potential for improved collaboration and communication, as pharmacists work in proximity with others healthcare team members.In this model pharmacists are part of the team of RACFs as compared to external contractors who visit the RACFs.On-site pharmacists can regularly communicate and collaborate with RACF team members, residents, families, GPs and other RACF health professionals and contribute to the multi-disciplinary care of residents.A study of the activities conducted by OSPs in the PiRACF study showed the wide range of activities conducted by OSPs including clinical medication reviews, education, clinical audits, and quality improvement activities.The OSPs spent a large proportion of their time communicating and collaborating with the GPs, RACFs healthcare team, residents, and their families.Of all recommendations made by OSPs, the rate of prescriber agreement was 51.5% 46 .
The study found a significant decrease in the proportion of residents taking at least one PIM in the intervention group compared to the control group over the study period.According to the adjusted model presented, the odds of having at least one prescribed PIM in the intervention group had halved over the period of the study, when compared to the control group.The use of PIMs has been considered as an indicator for the QUM in healthcare settings 5 , and the presence of PIMs has been linked with significant adverse drug events and hospitalisations among older people 6,8 .In this trial, the control sites followed the usual care where an accredited pharmacist could perform medication reviews (referred to as RMMR), on a visitational basis when they are deemed clinically necessary by prescribers.Visitational services may have limitations, especially during periods of transitions of care where residents are 3 times more likely to experience a medication-related problem 47 .Although the current RMMR service in Australia has demonstrated effectiveness in identifying medication-related issues 23 , a recent  48 .Additionally, a recent study showed that only 1 in 5 residents receive an RMMR in the 3 months after being admitted to an RACF in Australia 49 .Australian QUM services aim to improve medicine-related practices in RACFs but limited evidence is available for their effectiveness 27 .The integrated pharmacist approach in RACFs may enhance medication management through improved interdisciplinary collaboration and goes beyond the existing RMMR and QUM services.
The study observed a statistically significant reduction of residents' anticholinergic drug load in the intervention group.Older residents are particularly vulnerable from the detrimental effects of anticholinergic medications, such as cognitive impairment, constipation, and behavioural disturbances 50 .The use of medications with anticholinergic activity has been linked with an increased risk of falls in older adults 51 , and the Beers® Criteria recommend avoiding the use of highly anticholinergic medications in the elderly 33 .According to a systematic review of pharmacist interventions to improve anticholinergic prescribing practice in older people, medication reviews and education were the most commonly used interventions 52 .A systematic review found that pharmacistled interventions that involved collaboration with medical practitioners reduced anticholinergic drug burden more than other interventions 45 .A qualitative systematic review found working in silos resulted in low motivation   Mean ACB scale Baseline www.nature.com/scientificreports/ to reduce anticholinergic use, while enablers of successful anticholinergic burden reduction involved good communication with patients, carers, and other healthcare professionals 40 .The nature of the on-site pharmacist intervention enables improved collaboration with other healthcare professionals as well as residents which may facilitate the reduction of residents' ACB.The proportion of residents taking psychotropics, defined as one or more benzodiazepines or antipsychotics, had a reduction in the intervention arm when compared to the control but the difference was not statistically significant.While the dose of benzodiazepines didn't show a reduction, notably, the dose of antipsychotics after adjusting for potential confounders, showed a statistically significant reduction in the intervention arm.This finding suggests the pharmacists were able to initiate reduction in dose of antipsychotics but may have not been able to achieve cessation over the period of the intervention, as the process requires dose tapering and monitoring 53 .
Polypharmacy is prevalent amongst residents in residential age care.In the RACF setting polypharmacy is often defined as the use of nine or more regular medications 54,55 .The average number of regular medications for each resident in this cohort ranged between 9 to 10 regular medications.This study showed no significant effects on the number of regular medications when compared between intervention and control over the period of study.While high levels of polypharmacy have been associated with medication-related problems, hospital admission and falls 56 , there may be a clinically appropriate need for polypharmacy in some residents with high level of multimorbidity.Balancing the risk of undertreatment and reducing polypharmacy can be complex, and further research may be needed to explore barriers in addressing polypharmacy by on-site pharmacists.The proportion of residents who had their allergy and ADR status documented was recorded and was found to be high at baseline with little scope for improvement.Allergies and ADRs have the potential to cause significant harm and hospital admissions 3 , and on-site pharmacists need to pay attention to complete ADR documentation in their roles.
The study had the following limitations.Due to COVID-19 we could not assess the effect of the on-site pharmacist in RACF model on medication round time and inappropriate administration of dosage forms, as was initially planned.The Beers criteria tool was used as the primary outcome of the study to identify the use of potentially inappropriate medications amongst residents in RACFs.While the Beers Criteria tool is a useful proxy for QUM, it does not assess the underuse of helpful medications.All participating facilities in the study were in the Australian Capital Territory (ACT) Australia which is a metropolitan area.A careful analysis of the contextual differences should be considered when generalising to other regions in Australia and beyond, such as

Outcomes
Unadjusted model (1) p value Adjusted model (2) p value Primary outcome: Proportion of residents who were prescribed 1 + regular PIMs Control versus intervention at baseline (3)  www.nature.com/scientificreports/regional and rural areas.Future studies could assess the longer-term impact of OSP in RACF model on resident health outcomes and include economical analysis for the wider roll out of this model.

Conclusion
The on-site pharmacist intervention was successful in reducing PIMs, anticholinergic cognitive burden, and the dose of antipsychotics in RACF residents, suggesting that the on-site pharmacist in RACF model improves medication management.The findings strengthen the case for wider implementation and integration of pharmacists into multi-disciplinary care teams in RACFs to help improve residents' quality use of medicines. https://doi.org/10.1038/s41598-023-42894-5

Table 4 .
Summary of results from generalized linear mixed-models of medication-related outcomes in the PiRACF study (reference category: controls at baseline) (1) the unadjusted model controlled for the clustering structure of the data only.(2) The adjusted model controlled for the clustering structure of the data and adjusted for potential confounders.(3) Main effect of intervention group in comparison to control as reference at baseline.(4) Interaction term (intervention × endpoint).Significant values are in bold.OR = odds ratio, RR = rate risk, Coeff = coefficient, ACB = Anticholinergic Burden, ADR = Adverse Drug reaction, PIM = Potentially Inappropriate Medication.Adjusted model includes age, sex, presence/absence of a dementia diagnosis, CCI, number of regular medications, and presence/absence of concurrent NPS MedicineWise intervention.Baseline Control n = 734, Intervention n = 541, Endpoint Control n = 681, Intervention n = 620.

Table 2 .
Comparison between PiRACF study sample characteristics (at baseline) and national data.*Residents under 65 years were excluded from the national data.Statistically significant differences between PiRACF study and nationally representative data shown in bold text.

Table 3 .
Descriptive data for medication-related outcomes at baseline and endpoint.*Proportions presented as % and continuous variables as means.ACB Anticholinergic Cognitive Burden, ADR Adverse Drug Reaction, mg milligram, PIM Potentially Inappropriate Medication.