The Care Home Independent Pharmacist Prescriber Study (CHIPPS): development and implementation of an RCT to estimate safety, effectiveness and cost-effectiveness

Programme delivery

There have been a few changes in the programme to improve procedures or optimise effectiveness.

Qualitative evaluation

A focus group and an online survey were planned to capture the PIP’s experience at the end of the feasibility study; this was changed to a face-to-face workshop to facilitate a fuller understanding of the intervention delivery and research procedures.

Recruitment of triads

The unit of recruitment was refined to triads consisting of a PIP/general practitioner (GP)/care home(s). To manage workload, triad recruitment was divided into three time phases to ensure that peak activity points did not overlap.

Recruitment of resident participants

The target of residents recruited in each triad was changed from ‘20’ to a ‘mean of 20’.

Health economic model beyond trial

This was not undertaken because the intervention was estimated to increase costs and reduce quality-adjusted life-years (QALYs), such that the extrapolation would not change the outcome.

Validation of methods for capturing quality of life

This was not undertaken as EuroQol Five Dimensions (EQ-5D) was the only utility measure used in the final trial. Previously, a validation work for use in care homes (by proxy) has already been undertaken.

Coronavirus

The trial was delayed in Scotland for the triads recruited at the end of Phase III. Final data collection was undertaken remotely immediately after the first lockdown was eased and with appropriate approval. The dissemination events were conducted on Zoom.

Approvals

Research ethics approved by the NRES Committee Yorkshire and the Humber Sheffield REC Ref.: 15/YH/0172.

NHS Research and Development approved by NHS Grampian R&D Ref.: IRAS: 173232.

Aim

To qualitatively explore stakeholders’ expectations and understanding of introducing a new service to inform ways to introduce an acceptable service, anticipating and mitigating potential barriers.

Research question

What components stakeholders would specify in a feasible and acceptable PIP service and what did they consider were barriers and enablers to implementing it?

Objectives

To access stakeholders in four demographically and culturally diverse study sites [England (two), Scotland and Northern Ireland (one in each)] to collect their views through focus group discussions and interviews and validate analytic themes through consensus discussions.

Method

A purposive sampling approach gained stakeholders’ views with experience of living or working in, or with, care homes, to maximise the range of data relevant to the research goals. 8 Stakeholders were GPs, pharmacists (Ps), care home managers (CHMs), care home staff (CHS), care home residents and residents’ relatives.

Our semi-structured topic guide was informed by the Theoretical Domains Framework (TDF) for behaviour change9 to identify expected characteristics, barriers for and benefits of the proposed PIP role (see Appendix 1, Table 1).

We drew on the interdisciplinary research team’s broad experience to identify theory- and practice-related data collection topics relevant to

• working in current medication management

• stakeholders’ knowledge of scientific, procedural or environmental factors shaping care home practices

• how a PIP service and GP-PIP partnership might work

• potential benefits and risks

• social and professional roles and identity

• topics for PIP training (reported in WP4).

We contacted GPs, pharmacists and care homes – through our own and local professional networks. Additional care homes were recruited through national and site-specific regulatory bodies, local primary care and care home networks. Our maximum variation sample targeted: differently managed care homes, varying resident needs, nursing and care residents and diverse funding arrangements.

Transcripts were iteratively analysed using the TDF to identify key components, initial codes and themes for our study design.

Results

Thirteen stakeholder group-specific focus groups (n = 72 participants) and interviews (n = 13 participants) were held with GPs, pharmacists (P), one pharmacy technician, care home managers, care home staff, residents and residents’ relatives in four study sites (see Appendix 1, Table 2). Fifteen pharmacists described themselves as being employed in primary care, 11 within community pharmacy and one as split across the two (see Appendix 1, Table 3).

All focus group, interview and workshop stakeholders largely welcomed introducing the PIP service as offering benefits for residents, care homes and doctors. Reasons included viewing this new role as relevant for improving medication management, benefiting residents, overcoming communication lapses between care home, GP practice and pharmacy, residents and their relatives. Nonetheless, stakeholders identified specific potential contextual and implementation barriers and facilitators specifically in relation to

• chronic disease management (contextual)

• knowledge of older people’s medication and care homes (contextual)

• clarity of PIPs’ role and responsibilities (implementation)

• integrated social and professional team-working (implementation).

GPs’ working patterns, together with multiple care home staff involved in medication, were seen to constrain effective chronic disease management and effective communication around residents’ medication needs.

All stakeholders prioritised regular, responsive medication reviews by PIPs to address gaps in managing chronic disease and enhance the safety of residents living with comorbidities.

GPs’ onerous workload limited their capacity for ‘time-consuming’ procedures and ‘complexities’ in reviewing and managing medication (GP8-FG), and pharmacists argued that PIPs could do ‘more proactive work’ (P10-FG).

Discussion

Work package 1 explored stakeholders’ expectations of the components and context for a feasible PIP service, with a TDF-informed approach identifying components, stakeholders and contextual practices as relevant to mitigate implementation barriers. This identified care home environments as complex, with diverse participants, organisational processes, systems and resources providing care to frail older people, posing chronic disease management and review challenges.

A PIP model was envisaged to offer means to address this, but only if well informed about older people’s medication and care homes. Stakeholders believed that an acceptable service required the PIP to offer means to strengthen mechanisms to ensure efficient, effective ‘whole team’ approach to prescribing in care homes. 1 All stakeholders gave priority to a PIP conducting medication reviews to save GPs’ time/work. They welcomed PIP-related relationships as enhancing trustful communication around medication issues, mutually recognising remits and competencies.

We gained comprehensive understanding from stakeholders about processes to maximise chances of the impact of the PIP intervention on practice. 10 Central here was for everyone involved to ‘understand each other’s systems’ by recognising established organisational and cultural practices in care homes and primary care during implementation and improving communication around related changes. Care home managers, residents and relatives saw PIPs spending time with residents to explain and reassure around medication, to be more fully partners in their own care. 11

Integrated team-working was another key component: stakeholders were more likely to consider the PIP’s role as acceptable and viable if effective team-working were embedded in implementing such healthcare change. 12 GPs and pharmacists strongly preferred PIPs to be integrated into their practice teams.

Clearly defined PIP roles were considered crucial at micro-level experiences of individual actions and at macro-level experiences in care home and GP practice organisation. Residents and relatives were more likely to accept the new service if its purpose was carefully explained to them. GPs and pharmacists required explicit agreement on areas for PIP prescribing.

Contextual factors framed how stakeholders envisaged implementation issues. For example, effective team-working with the PIP, in GP practices and care homes (an implementation concern), depended upon the PIP acquiring appropriate knowledge of older people’s medication and frailties (a contextual issue). Guidelines and existing research underscore that specifically addressing both context and implementation barriers can better guarantee improved outcomes for older people in residential settings. 13 Multiple stakeholders shared the belief that for any proposed PIP innovation to be acceptable and viable, it would be dependent on all stakeholders understanding each other’s systems. 14

Aim

This WP aimed to identify the most appropriate outcomes to measure the impact of the CHIPPS intervention through the development of a core outcome set (COS).

Objectives

The purpose of determining the effectiveness of the CHIPPS service was to

• identify potentially appropriate outcome measures

• rate and select outcome measures based on validity, reliability, utility and proximity to the intervention.

Method

In WP2, we followed methodology for the development of a COS (standardised set of outcomes that should be measured and reported at minimum in all clinical trials)15 for CHIPPS. The full study has been published. 16

The methodology used was informed by published guidance17 and conducted in two phases. Phase I encompassed the first three steps recommended by Williamson et al. : (1) to explicitly describe the scope of the COS, (2) to identify existing outcomes, and (3) to identify outcomes that are important to key stakeholders. This resulted in a long list of outcomes that fed into Phase II, corresponding to step (4) of the aforementioned guidance – prioritising the most important outcomes using a consensus method.

Key findings

An overview of the COS development process is provided in Appendix 1, Figure 2.

Discussion

This WP was informed, in part, by the stakeholder engagement activities organised in WP1. Stakeholder engagement is a crucial part of COS development methodology to ensure that outcomes represent those that are considered of importance to service users and/or their representatives.

The final COS provided the basis for planning the next phases of the trial, notably the feasibility study (WP5), internal pilot and main trial (WP6).

Background

Medication administration errors are common within care homes. 1 The PIP intervention therefore has the potential to improve outcomes. However, not all treatments can be provided. We therefore undertook an economic evaluation of the PIP intervention to assess whether it represented a good use of scarce resources. This was deemed feasible for the full trial (outlined in WP6) based on, amongst other things, the high response rate achieved for proxy EQ-5D scores in the feasibility study (see WP5). The methods of resource use data collection were also informed in a previous work in the care home setting. 23

Objective

To estimate the cost-effectiveness of the PIP intervention.

Methods

Outcomes

Quality of life was measured using the EuroQol Five Dimensions and Five Level Rating Scale (EQ-5D-5L). 30 The EQ-5D-5L (proxy version) was chosen as it is deemed the preferred measure of utility in the National Institute for Health and Clinical Excellence (NICE) methods guide31 and in the light of work undertaken in the CHIPPS feasibility study32 (see WP5). The latter included an assessment of outcome measures used in the CHIPPS feasibility study where the EQ-5D-5L was chosen to be used in the full RCT, as it had been previously validated,33 and had a relatively good completion rate and low time taken to collect. It is recommended that QALYs are used to measure and value health effects, as QALY is a generic measure of health benefit that considers both mortality and health-related quality of life. 34

For all participants, proxy respondents were asked to report participants’ level of problems (none to extreme/unable) in five dimensions (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) at baseline, 3- and 6-month follow-up. In line with the NICE position statement,33 the crosswalk mapping function was used to convert these responses into utility scores. 35 Those who were known to have died by the particular follow-up points were assigned a utility score of zero. QALY scores were subsequently estimated based on the total area under the curve method and the assumption of linear interpolation. 36

Analyses

Base-case results were based on those with complete cost data at both baseline and 6-month follow-up and those with complete EQ-5D data at baseline, 3- and 6-month follow-up. Bivariate regression37 was used to analyse the cost and QALY data, based on the intention-to-treat approach, enabling the mean incremental cost (mean difference in cost between the two arms) and the mean incremental effect (the mean difference in QALYs) to be estimated. The cost-effectiveness acceptability curve (CEAC)38 was also used to estimate the probability of the PIP intervention being cost-effective at the λ value of £20,000/QALY compared with standard care. In addition, three key sensitivity analyses were undertaken: using multiple imputation to account for missing values, removing training costs (as they are one-off cost) and undertaking a threshold analysis to establish at what threshold the intervention would be effective (in terms of cost or effect).

Results

Outcomes

EQ-5D scores are shown in Appendix 1, Table 25, where it can be seen that both arms had lower mean scores at the 3- and 6-month follow-up points. Although the mean QALY score was higher for the intervention arm, the mean baseline EQ-5D score was also higher for these participants. This means we cannot infer that the intervention was more effective based on these results and that there is a need to adjust for, amongst other things, the baseline difference in EQ-5D scores between arms.

Analyses

A total of 609 participants (70%) had complete cost and EQ-5D data. Bivariate regression estimated that the mean (95% confidence interval) incremental cost of the intervention, compared with standard care, after adjusting for baseline costs, age and gender was £279.86 (£19.39–540.33). The estimated mean incremental effect was −0.004 (−0.016 to 0.009) QALYs. The PIP intervention was therefore estimated to be dominated, as it was associated with higher costs and lower effect, and the CEAC estimated that it had a 3.8% probability of being cost-effective at the £20,000/QALY value. These results suggest that there would be no added benefit of a long run model as (given lower effect and higher cost) as the results of a long-term model would be unchanged/in line with the within trial analysis, that is, the PIP intervention would not be estimated to be cost-effective. All sensitivity analyses were not found to change conclusions from the above findings.

Discussion

In terms of overall costs, the estimated mean incremental cost of £280 was lower than the estimated cost of the PIP intervention, suggesting that intervention costs were partially offset by certain lower costs, for example, medication. However, as the PIP intervention had higher mean costs and was not estimated to be associated with an improved QALY score, it is not estimated to be cost-effective. These conclusions are based on the 6-month within trial analysis for those with complete data, but the conclusions were the same when multiple imputations or other sensitivity analyses were carried out (see Appendix 1, Table 26).

Approvals

Research ethics approved by the Faculty of Medicine and Health Sciences Research Ethics Committee Ref.: 20142015-77.

NHS Research and Development approved by NHS Grampian Research and Development Ref.: NRS15/PH18.

Aim

Develop a training package to ensure that pharmacist independent prescribers are appropriately prepared to deliver the service.

Method

WP4 consisted of six phases:

• systematic review and narrative synthesis

• initial stakeholder engagement

• training specific interviews and focus groups

• stakeholder engagement and consensus

• feasibility testing

• validation.

Analytical approach

Data were themed and collated to inform the development of a care home pharmacist training package. All training methods outlined within selected papers were extracted.

Systematic review

Stakeholder engagement

Thirteen interviews and 13 focus groups with 72 participants were undertaken. The main results have been reported elsewhere. 44 The different types of knowledge identified as important through the interviews and focus groups are summarised in Appendix 1, Box 1. Appendix 1, Figure 3: Draft 1 provides a copy of the first draft of the PDF used to underpin the training package.

A variety of activities were added to enable the development of identified cultural knowledge requirements, for example, spending time with different medical practice and care home staff to identify expectations and preferences and learning to use local systems.

The practical knowledge identified as important was how to provide pharmaceutical care for older people with frailty.

The EAP identified the need for ‘context’ to be included as a domain within the PDF and to change the ‘chronic disease management’ domain to ‘managing complexity in late life’ (see Appendix 1, Figure 4: Draft 2).

Training-specific focus groups and interviews

Six primary care pharmacists, six GPs, five community pharmacists, six care home staff and four local experts participated. Additional changes to the training package were identified as being required, including the addition of activities to enable the PIP to understand local cultures and to integrate into the teams.

Expert consensus

Four consensus panels were held (Grampian n = 12, Yorkshire and Humber n = 12, Norfolk n = 15, Belfast n = 14). The content of the face-to-face training days was agreed along with the expectation that a geriatrician should be involved in delivery.

The consensus panel emphasised the need for a significant amount of time to be allocated to the development of relationships and that training care home staff was an important element within this.

A number of further changes to competencies and behaviours were recommended.

A large number of topics about which PIPs should be knowledgeable were identified (see Appendix 1, Box 2). It was agreed to provide this information through a knowledge pack, consisting of relevant links.

It was proposed that the PIPs would use the PDF for self-assessment purposes and they should be allocated a mentor (senior care home pharmacist) to support them through the process.

Completion of these competencies would be signed off both by their mentor and by a medical practitioner with expertise in providing medical care to care home residents (independent assessor). Sign-off by both parties would provide the ‘accreditation’ for the PIPs.

A copy of the training package and PDF created following this exercise and before feasibility testing is provided in Appendix 1 and Figure 5.

Feasibility testing

The PIPs reported that the process of personal development planning, being supported by a mentor and assessed on their final competence through oral viva, was appropriate and effective. Greater guidance on evidence collection for assessment purposes was requested at the outset.

The elements of the face-to-face training that were perceived as particularly effective were the case studies surrounding the management of complexity, legal issues and covert administration and the session on the management of psychotropic medication.

The training was viewed positively and reported to be motivational, helping to enhance confidence. This resulted in no major changes to the training package.

Evaluation

All PIPs found the training useful, specifically the input from the geriatric specialists who produced case studies on medication management in older people: ‘probably about the best CPD I’ve done for a long time’. Overwhelmingly, the PIPs strongly agreed that the training content was appropriate.

• Comprehensive training, mentorship, competency assessment and explanation of research processes amply equipped these experienced PIPs to carry out their role.

• Input from older people psychiatric specialists and specific training on antipsychotics was reported to be of huge value to the majority of PIPs, increasing their confidence and understanding around this area of medication management.

Post-training mentorship by a qualified pharmacist with extensive care home experience was reported as being extremely helpful in guiding their CPD and in preparing a portfolio of evidence for the assessment of competence by a GP. All PIPs achieved competency at first attempt and within expected timelines. Only one PIP stated that mentorship was not very useful at this stage. The PIPs reported that they had less contact with the mentor as the intervention progressed.

The training days and mentoring appeared to increase confidence, especially in PIPs with limited experience in older people medicine. Even for those with extensive experience in older people medicine, the training provided a chance to consolidate knowledge. Alongside consolidating clinical knowledge, the PIPs needed to be able to successfully develop relationships with care home staff and primary care staff if they were joining a new practice. PIPs reported arranging peer support through the cloud-based messaging app ‘Telegram’. Mentorship and peer support was reported as useful in the early stages of the intervention but use tailed off as individuals’ confidence grew.

The need to demonstrate competency through completion of the competency framework and professional discussion with a GP was appreciated by the PIPs. Evidence from the PCP reviews suggested all were competent in the role. PIPs’ comments on their practice during the trial, evidence the increased competency and confidence they had due to the training programme:

There were a few people that we’d got off a medication, antipsychotics particularly, because that’s something I probably wouldn’t have touched, but after having the training session, and the group discussions, and more of an awareness, I felt more comfortable

Pharmacist independent prescribers suggested that individualised country-specific sessions could run in tandem during main training days without affecting the length of the training programme. An additional refresher event mid-intervention was recommended as was advice on building relationships in CHs where work cultures may be new to pharmacists.

Discussion

The results from the feasibility study suggest that the iteratively designed training package is likely to ensure that PIPs are competent to undertake their envisaged role. Evaluation within WP6 will enable the researchers to determine whether it is generalisable to a broader range of PIPs.

Approvals

Research ethics approved by NRES East of England – Essex REC Ref: 16/EE/0284 and Scotland A REC Ref.: 16/SS/0125.

Health Research Authority approved Ref: IRAS 206970.

Aims

To test and refine the service specification and proposed study processes to inform the cRCT (WP6).

Objectives

To

• test processes for participant identification, recruitment and consent and assess retention rates

• determine suitability of outcome measures and data collection processes from care homes and GP practices

• assess service and research acceptability

• test and refine the service specification.

Method

Estimating the participating proportion of the eligible population

The proportion of general practices, pharmacists, care homes and residents approached and consented was recorded along with the proportion of residents followed up at 3 months to assess recruitment and retention.

Suitability of outcome measures

Potential outcome measures identified in WP2 (see Appendix 1, Box 3) were collected at baseline and 3-month follow-up for each participant. To determine their suitability for inclusion in the RCT, each outcome measure was assessed against the following criteria: availability of the data source, potential for bias, potential for missing data, resident centeredness, sensitivity to the intervention, reliability, whether validated, potential for third party completion, ability to blind and time taken to collect per patient and quantity of data. At minimum, the measure had to be judged objective, discriminating and efficient to collect.

Assessment of service acceptability and trial feasibility

Data analysis

Descriptive statistics were used for the quantitative outcome measures at baseline and follow-up. No statistical comparisons were conducted. Interview and focus group transcripts were thematically analysed.

Approvals and registration information

A favourable ethical opinion was received from East of England – Essex Research Ethics Committee (5 September 2016: 16/EE/0284) and Scotland A Research Ethics Committee (8 September 2016:rec ref. number 206970) with subsequent approval from the Health Research Authority/NHS Research and Development.

The trial was registered on the ISRCTN registry Registration number ISRCTN10663852.

Results

Discussion

The acceptability of the service and feasibility of proceeding to the main trial were confirmed. The processes to identify and recruit trial participants (GPs, PIPs, care homes and residents) were successful and scalable, including those for participants without capacity. Participants were retained for 3 months. The primary outcome measure for the main trial was confirmed as the fall rate per person. In contrast to other potential outcomes, a clinically important difference in falls would be detectable with acceptable power within a realistic sample size. Minor areas of refinement to the service specification and the research process were identified. For example, using the professional judgement of a study team member to assess causality between reported adverse event(s) (AEs) and the intervention was subject to bias, and an alternative approach was developed, that is, using independent GPs for WP6. Similarly, a review of the PCPs, conducted subjectively by the study team members, could have been biased, and WP6 includes details of the standardised protocol and reporting templates used in the main trial.

The participant demographics were similar to those of the UK care home population. 1 Participation rates were high, suggesting there had not been selective recruitment. The PIPs participating in this study included pharmacists employed by either primary care or the GP practice, providing evidence that this service specification is adaptable to either model; PIPs with a pre-existing relationship with the GP found it easier to arrange meetings.

The level of EOI confirmed that there would be sufficient participants for the main trial, and the consent rate of residents informed the target number of care home patients required to be registered with the participating GP.

Approvals

Research ethics approved by NRES East of England Central Cambridge REC Ref.: 17/EE/0360 and Scotland A REC Ref.: 17/SS/0118.

Health Research Authority approved, Ref.: IRAS 233964.

Aim

To estimate the effectiveness, cost-effectiveness and safety of a PIP assuming responsibility for providing pharmaceutical care to residents in care homes.

Methods

This was a cRCT conducted in primary care involving triads of a GP practice, PIP and a sufficient number of care homes to provide 20 care home residents per triad. 50

Recruitment

Participant identification and recruitment

Pharmacist Independent Prescribers were identified using local networks and related GPs (if links already established) were recruited concurrently, followed by their relevant care home(s) who were approached by participating GPs. Where one single linked home had too few potential resident participants, up to two further homes were recruited.

Randomisation and blinding

Randomisation was at triad level to minimise contamination occurring between two homes in the same practice where one received the intervention but the other did not. Residents were not individually randomised, as part of the intervention was for PIPs to help care homes improve their overall medication management processes. Randomisation was stratified by the four geographical areas, using a web-based electronic system integrated into the study database, in an allocation ratio of one intervention triad to one control triad. Because of the nature of the intervention, PIPs, care homes and GPs were not blinded to their allocation arm once the intervention began. Researchers, however, were blinded to arm allocation during the recruitment phase.

Six weeks were provided for PIP training to be undertaken and completed. Consequently, time zero for data collection purposes was standardised at 6 weeks after randomisation.

Intervention

This was delivered by trained PIPs for a period of 6 months and involved the PIP, in collaboration with the care home resident’s GP and care home staff, assuming responsibility for managing the medication of each resident, including

• reviewing the resident’s medication and developing and implementing a PCP

• assuming prescribing/deprescribing responsibilities

• supporting systematic ordering, prescribing and administration processes with each care home, general practice and supplying pharmacy where needed

• providing care home staff training

• liaising with GP practice, care home and supplying community pharmacy.

We anticipated each PIP providing approximately 4 hours of intervention per week per 20 residents for 6 months.

Control: Triads allocated to the control arm received usual GP-led care, which could include pharmacist review/services to care homes where routinely provided, excluding those of an intensity equivalent to the study intervention.

Outcome measures

The primary outcome was fall rate per person over 6 months after time zero, as documented in care home falls record for those residents recruited into the trial only.

Secondary outcomes, all at 6 months after time zero unless stated otherwise:

• Resident (i.e. self-report) or proxy resident quality of life (EQ-5D-5L) at 3 months and 6 months: utility scale where a score of 0 indicates equivalent to death and that of 1 indicates full health. 30

• Proxy modified physical functioning score (Barthel): a score of 0 is most dependent on 20, which indicates least dependent. 47

• DBI: a measure of anti-cholinergic and sedative drug exposure, which was collected through GP-recorded medication data: higher scores indicate greater anticholinergic potential and increased risk of drug-related morbidity. 46

• Health service utilisation (and associated costs); notably unplanned hospital admissions in the past 6 months, at 6-month follow-up, collected from care home and GP records (see WP3 for health economics results of WP6).

• Mortality.

Sample size

A sample size calculation indicated that 880 participants (440 in each arm) would detect a 21% decrease in fall rate from 1.50 per resident over 6 months with 80% power, at the 5% significance level and an assumed intraclass correlation coefficient not more than 0.05. With approximately 20 patients per triad, this equated to 44 triads. The relative reduction of 21% was half of that detected within a UK-based, pharmacist-led medication review service provided to care homes. 51 Furthermore, we assumed a loss to follow-up of 20% based on mortality and loss observed in the CAREMED study. 23

Safety

Processes were developed for recording sudden unexpected serious adverse reactions (SUSARs), SAEs and AEs. SAEs were defined as inpatient hospitalisation and death requiring immediate reporting. If SAEs were believed to be related to the study intervention, then they were reported as SUSARs. We used a mixture of prospective and retrospective SUSAR notification by asking GPs to report SUSARs immediately, and retrospectively, and by our trial manager proactively contacting care homes each month to ask about SAEs. The resident’s GP was then assessed for causality of the SAE, and whether it was linked, or not, to the PIP intervention.

Concerns could also be confidentially raised by any member of care home staff using a dedicated email address.

A random (computerised number generation) 20% sample of the PCPs and associated resident documents were assessed by a study geriatrician, to ensure clinical appropriateness and safety.

Process evaluation

Complementary to the main RCT, a process evaluation was conducted,52 following MRC guidance. 7 This evaluation used a mixed methods approach to inform interpretation of trial findings and subsequent implementation, should the intervention be effective. Objectives were as follows: to provide description of the intervention in terms of quality, quantity and variability in delivery; to explore the effect of individual components on the primary outcome; to investigate the mechanisms of action; to describe views of the intervention (including training of PIPs and care home staff) from GP, care home, PIP, resident and relative perspectives; to describe the characteristics of each arm and to estimate how ‘normalised’ the intervention became.

A mix of quantitative (surveys of care home staff, GPs and PIPs, PIP activity logs, PCP review and the trial outcomes) and qualitative (interviews with care home staff, residents, GPs and PIPs) approaches were used. Data were collected relating to delivery of detailed tasks required to implement the new service, to collect data to confirm the mechanism of action as hypothesised in the logic model (see Appendix 6), to collect explanatory process data and data on contextual factors that could have facilitated/hindered effective and efficient delivery of the service. Detailed analysis of PCPs additionally involved determining which medication-related changes would be associated with the risk of falls guided by recent comprehensive systematic reviews. 53–55

All data were collected, from intervention arm participants only, after the study period for each home had finished. The tasks, aims, data and data sources are summarised in Appendix 1, Tables 18–21.

Interviews schedules were informed by normalisation process theory (NPT),56,57 and questionnaires included a set of NoMaD questions that translates NPT domains for survey use. 58 Interviews were conducted face-to-face or by telephone; all were audio recorded and transcribed verbatim. Thematic analysis of qualitative data was based on the NPT framework, but a complementary inductive approach enabled recognition of unexpected emergent themes. All data sets (qualitative and quantitative) were integrated to identify relationships, explain findings and identify optimal intervention contexts. 52 Full details are published elsewhere. 50

Ethics

Ethics approval was provided by East of England Central Cambridge Research Ethics Committee (for England and Northen Ireland) – Ref.: 17/EE/0360; and by Scotland A REC – Ref.: 17/SS/0118.

Results

As shown in Appendix 1, Figure 8 (CONSORT diagram), we recruited 49 triads (49 general practices, 49 PIPs and 72 care homes) between December 2017 and May 2019. Of these 49 triads, 25 were randomly allocated to the intervention arm and 24 to the control arm. There were 454 residents in the intervention arm and 428 residents in the control arm. Almost all losses were due to resident deaths (134/166 losses = 81%). Three intervention PIPs (12%) failed to deliver the service due to personal reasons.

Baseline comparison between arms is provided in Appendix 1 and Table 12. Whilst most variables including age, medications, admissions, DBI, Charlson comorbidity, and EQ-5D-5L proxy scores were similar between arms, the control arm had a slightly greater proportion of male residents (33% vs. 28%), and a considerably greater proportion in nursing home care (59% vs. 42%). In line with those findings, the intervention arm had a higher Barthel score (8.34 vs. 7.07 of 20, where higher scores imply greater independence) and those self-reporting EQ-5D (11% of participants) reported better health 0.50 versus 0.35 (scale 0–1, 1 = perfect health). Residents in the intervention arm had a mean number of falls of 0.78 in the previous 90 days compared with that of 0.57 in residents of the control arm. Follow-up data collection commenced in September 2018 and concluded in July 2020 (this was delayed by 4 months for the final triads due to COVID-19).

Discussion

This large, rigorously conducted, cRCT did not demonstrate that PIPs who took responsibility for medication management for older patients in care homes could reduce falls significantly. Moreover, health economic analysis showed an overall cost per resident of £280, with no effect on quality of life.

However, the PIPs did contribute to a decrease in the residents’ mean DBI compared with that of the controls, suggesting the occurrence of effective deprescribing, and the data suggest there were savings in medication costs. In the longer term, a reduced DBI would have theoretical health benefits. All outcomes’ point estimates, after adjustment for baseline differences, including Barthel index, mortality, hospital admission, proxy quality of life and falls favoured the intervention arm although not reaching statistical significance.

Drug burden is associated with increased mortality,59 falls,60 hip fractures,61 frailty62 and reduced quality of life. 62 Consequently, the significant reduction in DBI is predictive of improved resident outcomes. With data on DBI and risk based on a minimum 12 months of observation,60–63 the 6-month follow-up in this study may again have been unlikely to fully realise clinical improvements.

Whilst we should be cautious in overinterpreting such findings, the data combined with the generally very positive views from the stakeholders and confirmed safety suggest that this intervention merits further refinement and investigation. In particular, the facilitating contextual factors need to be further explored in light of the ongoing NHS roll out of this and similar pharmacist-led medication-related services.

Work package 1

Minimal representation of residents and relatives in intervention-development workshops, held on university premises outside their usual residence. Participant self-selection may bias them toward favourable expectations for the PIP role.

Work package 2

Delphi panel participants, drawn from the research team, were located in the UK, which limited generalisability. We included just two PPI representatives in our panel where other COS developers have recruited PPI representatives and professionals to their Delphi panels in a 2:1 ratio. 67,68 Outcomes that had not received consensus following the second round were excluded from the COS. Further rounds may have resulted in consensus for more outcomes. For pragmatic reasons, a final face-to-face ‘consensus meeting’ did not occur, potentially limiting further consideration of some outcomes. This approach has been used in other COS research. 67–69

Work package 3

We were unable to collect data in relation to all items of resource use. Assumptions were made about the GP time saved due to no longer approving prescriptions, but there could have been other differences, for example, shorter GP medication reviews and length of home visits. The analysis also did not include any time implications for care home staff. Estimated PIP intervention costs were also based on data reported by PIPs in the activity logs, which suggested under-reporting as mean reported time tended to be lower than the time remunerated for.

Work package 4

The training package was tested for final validation purposes only on 25 PIPs.

Work package 5

Research processes and service specification tested in one triad only in each location, that is, four PIPs, four general practices and six care homes in total.

Work package 6

Whilst GP selection of care homes may have introduced an element of selection bias, the number of homes available to them was frequently limited and there was no evidence of obvious selection bias in the final cohort of randomised homes.

Three PIPs (12%) failed to deliver the intervention.

Follow-up was carried out for 6 months only, and this may have been an insufficient time period. Some residents’ PCPs had agreed only 3 months into the intervention period. Furthermore, where medication changes reduced the likelihood of falls, this is rarely an immediate effect.

The primary outcome of falls was proximal to only a third of the PIP interventions, and therefore, more focused inclusion criteria based on the risk of falls due to medication may have been more appropriate.

The concurrent national roll out of MOCH in England may have reduced intervention opportunities.

WP1 service specification development

To effectively embed this new pharmaceutical care service in care homes, stakeholders highlighted that securing PIP role acceptability and viability would require time and steps to ensure that all involved could ‘understand each other’s systems’. This should address contextual and implementation barriers and identify what practices were feasible for addressing residents’ medication-related safety issues. GP and care homes required reassurance that the model was likely safe.

WP2 outcome identification and selection

Using the evidence-based and stakeholder views, we conducted a Delphi consensus exercise to identify outcomes that should be measured to evaluate the service and published this as a COS for trials testing prescribing interventions in care homes. This was registered on the COMET database. The outcomes meeting our criteria were subsequently feasibility tested in WP5.

WP3 health economics

The methods used to estimate cost-effectiveness in WP6 were developed through WP1 to WP4 and tested in WP5. The mean incremental cost of the PIP intervention, compared with that in control participants, was estimated to be £280. This suggests that the cost of the PIP intervention might be partially offset by, for example, lower medication costs. However, no improvement in QALYs was identified, and consequently, based on the results presented, the PIP intervention was not estimated to be cost-effective.

WP4 training package

The CHIPPS training package consisted of 2-day face-to-face training, a PDF, personal development planning with supported mentorship, time to ‘understand each other’s systems’ and independent formative assessment by oral viva, which enabled the PIPs to practice safely.

WP5 feasibility study

Feasibility testing demonstrated that the PIP service was acceptable and practical. Trial processes for recruitment, retention, data collection and choice of outcome measures were confirmed. A robust process for monitoring safety of the PIP medication changes was developed, which involved independent assessment of all deaths and hospitalisations, independent review of PIP-created PCPs through sampling and provision of an independent email address to enable reporting of safety concerns.

WP6 definitive trial with internal pilot

No differences were found in the primary outcome of falls. This can be explained by the relatively small proportion of interventions that were identified as potentially affecting this outcome. The intervention did significantly reduce residents’ drug burden. No differences in the other secondary outcomes were identified, although all outcomes favoured the intervention arm. Proactive monitoring found the model of care to be safe.

Frequent discontinuation of medication used within the central nervous system explained the reduction in drug burden seen. Most, but not all, PIPs delivered the intervention as intended. The view of stakeholders was generally very positive. Care home staff valued the pharmacist’s advice and prompt resolution of queries. GPs reported reduced workload, improved patient safety and less inappropriate prescribing. To be successful in integrating the new systems of working, PIPs need to become fully integrated into GP and care home teams.

Recommendations for future research

Future research should

• determine the optimal model, processes and time for integrating pharmacists effectively into care home and GP practice teams

• determine the optimum follow-up time for interventions focused on fall reduction

• identify mechanisms of action leading to proactive deprescribing to enhance current practice

• develop an outcome measure that is meaningful to patients and practical for researchers, which accurately captures the impact of pharmacist interventions.

Pharmacist independent prescriber

Care home

Residents

Recruitment and employment of the pharmacist independent prescriber

Initial identification and recruitment of the PIP will be conducted by the CHIPPS management committee. The PIP will require

• excellent interpersonal, communication and IT skills

• familiarity with relevant GP software systems

• experience of providing prescribing and medicine management advice and support

• previous experience of working in a GP practice environment

• be able to travel to site locations

• a mobile phone to be contactable for the purposes of delivering this service

• appropriate indemnity insurance for prescribing.

The PIP will be employed according to local arrangements and seconded to the relevant GP practice for the study duration and during training and competency assessments (see Training and competency assessment of PIP).

Training and competency assessment of the PIP

See text in main paper.

PIP roles and responsibilities (NB: categorised as essential or not)

The PIP will, where appropriate:

Safe and effective service provision

• The PIP will be contactable and respond to messages within 24 hours (Monday–Friday).

• The GP practice will triage all medicine-related contacts for CHIPPS participants following a locally agreed protocol for referral to the PIP [see Training provision (undertaken at PIP’s discretion)].

• PIP will have full (read/write) access to the GP record system to issues prescriptions and update records.

• Where possible, the PIP will use remote access to update records when changes are made to GP-held records.

• Where remote access is not feasible, the PIP must update records within 24 hours of making a change.

• PIP will have full (read/write) access to care home records to update records during all visits using appropriate local reporting systems.

• The PIP will visit/contact the care home at least once a week.

• The PIP will visit/contact the GP practice at least once a week.

• All annual leaves must be agreed at least 4 weeks before the leave and clear system for the transfer of responsibility communicated to GP, care home and supplying pharmacy.

• The PIP will work within the local prescribing formularies of GP practice and primary care organisation.

• The PIP will report and document all significant clinical events or near misses using local reporting procedures and study documentation.

• The PIP will ensure that all records are aligned.

Costs

Costs were estimated in Great British pounds (£) at 2017/2018 financial year levels from the perspective of the NHS and personal social services (PSS). No discounting was undertaken due to the time frame of the study (costs were estimated over a 3-month period before the baseline/intervention and a 6-month follow-up period).

Estimating the cost of the PIP intervention

To estimate the PIP costs, each PIP was asked to complete two particular items over the 6-month post-randomisation period: a training log and an activity log. The training log enabled the PIP to record the time spent with others/courses undertaken as part of their training/professional development, for example, in-person training, contacts with a mentor, GP, community pharmacist, as well as any self-directed learning. The PIPs were also asked to record any activities associated with the PIP role within the PIP activity log, including the duration of (1) any contacts with others, for example, care home residents (participants), care home staff, GP, or any other professionals (e.g. geriatricians, community pharmacists, and district nurses), and (2) any other non-contact activity, for example, resident prescription management, care home medication storage, staff training and travel.

For both training and activity, log estimates of the cost per hour of employment were assigned to estimated times for the PIP and other professionals (including care home staff), where it was assumed that the PIP would have band 7 costs and their mentors/trainers band 8a costs. 24 Overheads were included as part of these unit costs, and these were deemed to cover any additional costs associated with the intervention.

As in the intervention arm, the PIP’s role enabled the undertaking of prescription management for the individual care home residents, and it was envisaged that this would enable GPs to spend less time undertaking prescription management for intervention arm participants. We estimated the GP time saving that would have occurred, for each individual participant, because of no longer undertaking prescription time management, would have been as follows. It has been estimated that the mean time spent per prescription-related event is 56.0 seconds (28). Based on this figure, if each participant were to have two prescriptions issued each month (each of which contain up to four items), which no longer required GP approval, then this would equate to a total GP time saving of 11.2 minutes (per resident across the 6-month study period). At a cost of £110 per hour for a GP24, this equates to a total cost saving, for GP time, of £20.60 per resident/participant.

Subsequently, the total PIP intervention cost was estimated by summing the per-participant PIP training and PIP activity costs and deducting the estimated GP time/cost saving.

Other costs

As it was considered that there was the potential for the introduction of the PIP role to improve outcomes/impact the use of other NHS and PSS resource-items, the following items were extracted from medical records.

Medication details were extracted from primary care records for both the 3-month period before baseline and the 6-month study follow-up period. In terms of unit cost, when extracting the medication data (directly into an electronic format), researchers were able to select medication details from the 2015 version of the Prescription Cost Analysis (PCA),28,29 This was undertaken with a view to reduce the number of misspelt medications to which it would not be possible to attach cost at the end of the study (it was still possible to enter a medication name if the drug was not listed in the 2015 PCA). Unit costs were automatically selected from the 2015 PCA as part of the data collection process, and no adjustment was made to these costs, as the average cost per prescription item has not increased in recent years. 28,29 Medications that were not automatically assigned a 2015 PCA price at the time of data entry were assigned a unit cost from the 2018 PCA. 28,29 The above enabled overall medication costs to be estimated.

Details of outpatient attendances, inpatient stays, tests and investigations and GP and practice nurse visits/phone calls were also extracted from primary care records. Conversely, informed by previous research,23 the details of all other health professional contacts were extracted from care home records, where professional data of the number of visits (and location: care home/community) and number of phone calls were extracted. For all items, data in relation to the previous 3 months were collected at baseline and those for the previous 6 months were collected at the 6-month follow-up. Unit costs were subsequently assigned to each of these resource items (see Appendix 1, Table 1) and totalled to estimate the total other (non-medication) costs.

Finally, the total PIP intervention cost, overall medication costs and other (non-medication) costs were summed to estimate the total (NHS and PSS) cost.

Cost of the PIP intervention

The training logs were returned by 23/25 PIPs, and the activity logs by 22/25 PIPs. Certain assumptions were therefore necessary to deal with this and other types of missing data, in order that mean PIP costs be estimated for all PIPs/participants. By way of example, in cases where a particular training/activity was reported, but the associated time was not, the mean time listed for that particular training/activity by other PIPs was estimated and assigned to the described activity. This meant, for example, that the mean cost for the 23 PIPs who returned the training logs was applied to the two PIPs who did not as it was known that these two PIPs had completed their training.

The mean training time per PIP was 4.3 days (see Table 2), where this included a 2-day course, approximately a further day of contact time with other professionals, for example, mentor/GP along and a further day of self-directed learning. The mean total training cost (including the costs for contacts with other professionals) was estimated to be £61,916.07, which was equivalent to £2476.64 per PIP or £137.90 per study participant. Though these costs are based on actual study data, they may overestimate future training costs as in practice training could be delivered to more individuals at once, and the costs spread across more residents, reducing the cost per resident (this is assessed as part of the aforementioned second sensitivity analysis).

No PIP activity costs were assigned to intervention participants in the care homes where the three PIPs did not return their activity logs (these PIPs did not deliver the intervention and zero PIP time/costs were therefore assigned to the residents under their care). Conversely, one PIP reported activities that amounted to a time of 19.4 days (assuming a 7.5-hour working day) per resident. As such, there was wide variation around the mean PIP reported activity time of 7.6 days (this equates to 191 minutes per participant, see Appendix 1, Table 23).

In terms of costs, the mean PIP activity cost was £205.29 (range: £0–1168.67) per participant, where the mean PIP cost was £169.68 per participant, compared to £35.61 per participant for other professionals’ time. When added to the aforementioned training cost, and the assumed GP cost saving due to them undertaking fewer prescriptions, this gives an estimated mean PIP intervention cost of £322.59 (see Appendix 1, Table 24).

Other costs

In terms of the other NHS and PSS resource-items that it was considered could change due to the introduction of the PIP role (medication, out-patient attendances, in-patient stays, tests and investigations, GP and practice nurse visits/phone calls and other health professional visits/phone calls), the associated mean estimated levels of resource use and costs are reported in Tables 2 and 3, respectively, where it can be seen that overall medication costs and other (non-medication) costs are broadly similar between groups. Accordingly, the mean total (NHS and PSS) cost was estimated to be higher in the PIP intervention group (see Appendix 1, Table 24), where this difference between groups was largely due to the costs associated with the PIP intervention.