Antihypertensive deprescribing in frail long-term care residents (OptimizeBP): protocol for a prospective, randomised, open-label pragmatic trial

Alberta Health Services Databases

Alberta Health Services is a steward of the linked databases which track encounters with the healthcare system for all Albertans. This includes community physician services and diagnoses, prescriptions dispensed (PIN), reasons for hospitalisation and emergency visits (Discharge Abstract Database (DAD), National Ambulatory Care Reporting System (NACRS), laboratory results (Lab), vital statistics (eg, mortality) and the Alberta Continuing Care Information System which contains LTC admission and discharge dates, mortality, and clinical data from Resident Assessment Instrument–Minimum Dataset 2.0 (RAI-MDS).

Resident Assessment Instrument–Minimum Dataset

RAI-MDS is a comprehensive, standardised assessment tool used in LTC throughout Canada. In Alberta, RAI-MDS assessments are administered by LTC staff on admission to LTC and every 3 months thereafter, with an additional assessment potentially performed if there is a significant change in health status. Data from the assessment tool are used to inform and guide clinical practice and policy decisions. Alberta Health Services has an RAI-MDS provincial team and a zone team for each of its five geographically defined zones to support LTC operators and provide quality assurance of the assessment data.

Although it is not directly part of RAI-MDS, we will be using elements of RAI-MDS data to calculate a frailty index for each participant. The frailty index is an internationally validated frailty measure that accommodates a variety of datasets, with diverse data elements, by simply determining the proportion of detectable deficits/limitations that are present.14 The scores can be categorised as follows: fit (0 to 0.12), mild frailty (>0.12 to 0.24), moderate frailty (>0.24 to 0.36) and severe frailty (>0.36 to 1.0).15 Hubbard and Peel’s validated frailty index coding16 17 will be used for the trial.

BP data

PointClickCare is healthcare management software used by many Alberta LTC facilities that tracks, among other things, BP measurements. The vast majority of facilities participating in OptimizeBP record BP collected as part of routine care, including BP collected to support deprescribing for this trial, directly into PointClickCare. PointClickCare transfers these BP measurements to Alberta Health Services for the purpose of determining eligibility and process outcomes for this study. If participating LTC facilities instead record BP onto paper charts, facility staff are supporting the trial by entering BP measurements into either a REDCap database,18 or an electronic spreadsheet that is transferred securely to Alberta Health Services.

To optimise generalisability of our findings, during the trial, each facility continues to collect BP measurements as per their prior routine (table 1), except for frequency of assessment—which increases to biweekly for the intervention group until deprescribing is completed (ie, until 2 weeks after the last dose reduction/discontinuation).

Inclusion criteria

• Resident of a participating Alberta LTC facility for at least 30 days prior to randomisation.

• ≥70 years old.

• ≥2 health system encounters with a diagnosis of hypertension (ICD-9 and ICD-10 codes) since 2002.

• Taking ≥1 daily antihypertensive medication (requires a dispensation in the preceding 15 days).

• Average recorded systolic BP<135 mm Hg. The average BP is calculated from the three most recent BP readings taken on separate days. If a resident only has one or two BP measurements, the average of these is taken. If a resident has >1 BP measurement in a single day, then the average BP for this day is calculated and used as a single data point to calculate the overall average (ie, each day can contribute only one value to the overall average BP calculation). As these were usual care BP assessments, no set time period for collection was required (with periods for collection potentially as long as 4 months).

Exclusion criteria

• History of CHF

  • Either ≥2 community physician encounters or ≥1 hospitalisation/ED encounter with a diagnosis of CHF (ICD-9 and ICD-10 codes) since 2002.

  • History of CHF recorded in RAI-MDS .

  • Dispensation of furosemide in the preceding 15 days (a loop diuretic commonly prescribed for CHF in Canada and uncommonly used for hypertension management).

• Only one antihypertensive and it is likely not being used to manage hypertension.

  • The only antihypertensive is a betablocker (common treatment for angina, atrial fibrillation, tremor and migraine).

  • The only antihypertensive is a calcium channel blocker and the resident has ≥1 health system encounter with a diagnosis of angina, tachyarrhythmia, or atrial fibrillation (ICD-9 and ICD-10 codes) since 2002.

  • The only antihypertensive is an alpha-blocker (common treatment for prostatism/nocturia).

• Admitted to hospital at the time of randomisation.

• Opted out of the trial by the resident, their family, their physician or the facility.

Treatment

Continually reducing antihypertensives provided an upper systolic threshold of 145 mm Hg is not exceeded (achieved using a trial-specific antihypertensive deprescribing algorithm).

Control

Usual care.

Antihypertensive deprescribing algorithm

The deprescribing algorithm was developed through an iterative and informal process. First, two family physicians (RYK and SRG), two Care of the Elderly physicians (MM and PK), one pharmacist (CAS) and the co-ordinating editor of the Cochrane Hypertension Review Group provided input. The algorithm was then further revised with pharmacy/nursing feedback from the first few sites enrolling participants. We selected an upper systolic BP threshold of 145 mm Hg for our trial based on the most applicable guidelines for BP targets in frail older adults.8 The algorithm appears in figure 2. Essentially, BP medications are continually targeted for reduction or discontinuation every 2 weeks so long as the systolic BP is less than 145 mm Hg. If the systolic BP rises and stays above 145 mm Hg, sufficient BP medication is reintroduced to place the systolic BP at or below 145 mm Hg. Pharmacists/NPs are given flexibility in determining the order of medications to deprescribe when more than one antihypertensive is in use, and the amount of each dose decreases (ie, 25%–50%, until the smallest dose has been reached, at which point it can be discontinued).

Figure 2

OptimizeBP antihypertensive deprescribing algorithm. BP, blood pressure.

Pilot

CapitalCare Lynnwood was our pilot site, which we used to test and refine our methods. As the only major alteration to our methods was the expansion of our exclusion criteria to better pick out those with CHF, or those using antihypertensives for reasons other than hypertension, our pilot participants were rolled into the overall trial. While fully blinded to allocation, and prior to our interim analysis being conducted, we additionally added some secondary and tertiary outcomes we believed would more comprehensively capture process, cost and hypoperfusion-related outcomes—in particular the addition of ulcers and renal insufficiency. All protocol changes are fully disclosed on ClinicalTrials.gov.

Implementation

Facility pharmacists (n=19) and NPs (n=3) perform the deprescribing. They are asked to first attend an orientation session led by RYK and CAS. This session provides background in support of less aggressive BP targets in frail older adults, resources and evidence for deprescribing, as well as the trial timeline and protocols. Pharmacists and NPs are provided with trial materials including the deprescribing algorithm, trial selection criteria and a tracking sheet used to help them keep track of their deprescribing. Facilities complete a survey prior to their initial randomisation to determine the method for measuring BP at the facility (table 1).

During the agreed on start week, the Alberta Health Services data analyst sends an encrypted email to the pharmacist/NP with the names and personal health numbers of residents randomised to the intervention group. Within a week, the research team emails the pharmacist/NP to ensure they have received this list and will start deprescribing. Subsequently, the research team emails the pharmacist/NP monthly to troubleshoot any issues with implementation. Most pharmacists or NPs use the tracking sheet and forward it (after they have removed all identifiers) to the study team. This further aids in identifying process hurdles.

BP measurements of residents in the intervention group are taken biweekly, using the facility’s usual method, during the period that deprescribing is actively being carried out. BP measurements of residents in the control group continue to be taken at their usual frequency (table 1).

Pharmacists/NPs deprescribe per the trial algorithm, typically every 2 weeks, basing their decision on at least 4 BP readings during that 2-week interval. Depending on the procedures at the facility, pharmacists/NPs either sign the deprescribing order themselves or require approval of the primary care physician. The resident’s care team (often a family physician, pharmacist, NP and/or care manager) remain responsible for the resident’s care and are free to alter/reintroduce antihypertensives if clinically indicated.

Intention-to-treat assumptions

Loss to follow-up: We will follow all randomised residents given complete data capture within the province, and we will continue to follow residents even if they change facilities. We will not censor residents that move outside the province given (1) few LTC residents are expected to migrate and (2) there is a delay in accurately detecting this (those who leave Alberta are not quick to cancel their provincial health insurance and the registry database is infrequently updated).

Withdrawal: In some instances, family/residents/physicians do not initially opt out of the trial but ultimately decline or are unable to participate when notified about an intended reduction in antihypertensive medication. As this happens only in the intervention group, we will continue to follow and analyse outcomes in all such individuals to prevent attrition from being imbalanced, even though antihypertensive deprescribing did not take place.

Lag in electronic health data: Alberta Health Services databases and PointClickCare BP data access can be up to several weeks behind for informing our inclusion/exclusion criteria. If, when these data are updated, it is clear that residents were not actually eligible on the date of randomisation (eg, were no longer taking an antihypertensive medication, had died or were in the hospital), those individuals will be retroactively excluded from the trial. This applies equally to both groups of the trial.

Missing RAI-MDS data: If RAI-MDS baseline or postrandomisation reports are not available for a resident, we will exclude the resident from the relevant analysis.

Inability to deprescribe: All individuals will be included in each analysis as allocated irrespective of whether deprescribing was possible (eg, antihypertensives are in use for reasons other than BP lowering).

Selecting regression covariates

For continuous outcomes, we will use a maximum of 1 covariate per 20 randomised subjects. For dichotomous outcomes, we will use a maximum of 1 covariate per 10 outcomes. Table 3 predefines the covariate list and all are measured at baseline (date of randomisation). We will always use the maximum number of covariates an analysis can support and employ them in the order listed. If a given covariate has too many levels to be accommodated by the remaining number of outcomes or subjects (eg, facility), we will skip it and go to the next covariate in the list.

Day-to-day management

OptimizeBP is managed collaboratively by RYK, SRG and EY, with input from coinvestigators as needed (CAS, JAB, DF and CSK).

Independent data safety monitoring board

Dr James M. Wright, Co-ordinating Editor of the Cochrane Hypertension Review Group, will chair a 5-member committee set to review all clinical outcomes on 50% of primary outcome events being reached (124). This is expected in the first half of 2024. Stopping rules: If p value is ≤0.001 for primary outcome benefit (the Haybittle-Peto boundary),21 or if p value is ≤0.05 for harm, the board will apply clinical judgement and make recommendations as to whether or not the trial should stop early.

End-point validation committee

A panel of three physicians will examine the spectrum of most responsible diagnostic codes for each hospitalisation and each emergency visit routinely captured in Alberta Health Services databases to ensure they match the outcome definition. Other outcomes will be accepted as they appear in the electronic health claims records, given there are no additional clinical records to review.