In this study, five outcome scenarios are possible: 1) completion of the trial; 2) entry into aged residential care (ARC) prior to completion (this would be linked to a change in GP and medication control); 3) prolonged hospitalisation; 4) death; and 5) withdrawal or other loss to follow up. The first two outcome scenarios would involve the post medication review being undertaken either with the participant or via the medical records of a care facility. The third outcome scenario could lead to delays in recording the post intervention review or may lead to any of the four other scenarios.
The primary outcome will be the change in a participant’s DBI (ΔDBI) between the time of the baseline interRAI assessment (T1) and time of 6 months follow-up (T2; ΔDBI = DBIT1 – DBIT2). Data for the calculation will be collected by comparing medication use pre- and post- intervention. We will determine if there is a greater reduction in the DBI of participants in the experimental arm of the trial compared with participants with the same level of frailty in the control arm. Subgroup analysis will determine if deprescribing is more pronounced for those with more severe frailty.
Secondary outcomes measures will be compared between the two arms of the trial after 6 months and are outlined in Table 3. These include number of hospitalisations, entry into aged residential care, all-cause mortality, and cost utility from the funder’s perspective. We will measure the number of emergency department visits and unplanned hospital admissions. Data on entry into or change in level of care in aged residential care will be extracted from relevant national databases by the analytical services of the Ministry of Health. Patient mortality data will be matched using Participants’ national health index (NHI) number and added to the dataset and analysed using competing risk regression. In a separate analysis we will conduct a health cost utility assessment.
Power and Sample size
For this RCT we define the clinically significant change in DBI as 0.5, the equivalent of one medication contributing to DBI given at the minimal efficacious dose (14). Approximately 4% of recent interRAI assessments show a ΔDBI ≥ 0.5 over a six-month period. A meaningful outcome from deprescribing in this study would be to increase the percentage, in the intervention cohort, with a DBI change ≥ 0.5 by 10% points. This would bring the percentage ΔDBI ≥ 0.5 over a six-month period to 14%.
The null hypothesis is that there is no difference in proportions with a ΔDBI of ≥ 0.5 over a six-month period between the control and intervention groups given that 4% of participants in the control group have a ΔDBI of ≥ 0.5. To disprove the null hypothesis, we need to detect a change in number of participants of 10% or more with a ΔDBI of ≥ 0.5 with a power of 90% at an α = 0.05. This requires 167 participants in each arm of the study, 334 in total.
For each frailty stratum under the null hypothesis there will be no difference in change in DBI between the control and intervention groups, and assume 4% of participants in the control group have a reduction in DBI of ≥ 0.5, then to detect a change in number of participants of 20% or more with a reduction in DBI of ≥ 0.5 with a power of 80%, and at α = 0.017 (0.05/3), requires 56 participants in each arm of the study (112 in each strata, 336 in all).
It is estimated that over 12 months within the target areas of the district health boards’ approximately 650 participants will meet the basic inclusion criteria including the minimum DBI. Previous data suggests that approximately 50% of the study’s cohort will take the target medicines and therefore 325 people would be eligible to take part in the study per annum.
The data analysis will use the intention-to-treat principle (where all available data from participants will be included in the analysis).
We will compare the proportions of the outcome measures using a chi-squared test and present results with a 95% confidence interval.
We will present Kaplan-Meier survival curves for subgroup analysis. Subgroups will include control low frailty, control medium frailty, control high frailty, intervention low frailty, intervention medium frailty, intervention high frailty. This analysis will be controlled for age and sex, with death as a censored event. We will then conduct a competing risk analysis using cumulative incidence functions (CIF). For example, for entry into residential care, entry is the primary event of interest and death is the competing event and the CIFs are the probability of observing these events before the end of the 6-month follow-up period.
Cost benefit analysis
As a separate analysis, a health economist will oversee an analysis of financial costs of routine screening for frailty compared to the expected benefits from more appropriate prescribing, reduced pharmaceutical costs, and avoided hospital admissions and entry into residential care. Given that we will stratify our cohort by frailty, we anticipate identifying a group of participants, with a degree of frailty, who will benefit most from targeted medication reviews.
The cost-effectiveness analysis will include costs and benefits both for the participants and for the health care system. Standard robustness checks will be performed.
A desirable feature is that the benefits of implementing the intervention are likely to be realized soon after implementing, and this will contribute to a favourable cost-effectiveness ratio. (33)
We will report according to the CONSORT reporting guidelines (34).
Participants will be blinded to their study arm. The pharmacists conducting the first medication reviews will be made aware of which arm of the trial participants are allocated to when they open the envelopes at the first home visit. In the follow-up home visits, pharmacists conducting the post intervention medication reviews will be blinded to the participants’ allocation. Patient unblinding is permitted in case of health concerns requiring immediate attention.
The study administrator and the data manager will have access to all participant data. The primary investigator and the study statistician will be blinded to participant data and allocation during the trial.
Ethics and Protocol changes
Ethics approval has been obtained from the Health and Disability Ethics Committee (HDEC) based on this study protocol in revision 8 under amendment AM06. Minor changes to consent forms and participant information sheet were made under amendments AM07 – AM10. Changes to allow conducting the second medication review over the phone during covid-19 lockdown were approved under amendment AM12. Any further changes made to protocol revision 8 will be communicated to the Health and Disability Ethics Committee and approval will be sought to implement the amended protocol. Incremental changes to the Participant Information Sheet (PIS) and Consent to Contact form will be made under subsequent amendments.
Any protocol modifications will be internally reviewed by the DMC, by the funder HRC and if required approved by HDEC.