Holistic health and social care outreach for people experiencing homelessness with recent non-fatal overdose in Glasgow, Scotland: the Pharmacist and third sector Homeless charity worker Outreach Engagement Non-medical Independent prescriber Rx (PHOENIx) pilot randomised controlled trial

Study design and setting

We conducted a pilot RCT in 20 venues in Glasgow, which is Scotland’s largest city (circa 600 000 inhabitants). Venues included different locations where people experiencing homelessness live, for example, hostels, or visit, for example, third sector homelessness charity drop-in centres or congregate, for example, begging pitches. In Scotland, the NHS provides free healthcare including prescriptions. We followed the Consolidated Standards of Reporting Trials guidelines31 and the trial was preregistered. A detailed protocol has been previously published.28

Participants

Inclusion criteria were as follows14 28:

Homeless (living in temporary homeless accommodation, no fixed abode or rough sleeping);

and

Aged 18 years or over;

and

At least one self-reported, non-prescribed drug overdose (blackout/lack of response and slow/irregular breathing that was thought to progress to complete cessation of respiratory effort unless treated) in the past 6 months confirmed by:

• A witness; or

• An ambulance call out; or

• An emergency department (ED) visit; or

• An administration of naloxone.

Exclusion criteria were as follows: living in a residential or community-based rehabilitation facility which had direct access to in-house medical and nursing care; or unable to give written informed consent. If necessary, researchers confirmed overdose occurrence(s) by contacting witnesses. Witnesses were usually known to the participant, for example, workers in the temporary accommodation, or other residents in the temporary accommodation. Researchers also checked different records for confirmation of overdose(s): accommodation workers’ records (a diary, kept at the accommodation); social care or addictions team records; secondary care records.

In recruitment venues, researchers collected a wide range of information including demographics, diagnoses, prescribing, street drug use, laboratory tests, subjective and objective health measures, validated questionnaires and healthcare utilisation data. These data were collected during in-person consultations or from clinical records after the in-person consultation (online supplemental web appendix 1).28 Participants received a £10 shopping voucher (redeemable at a shop not selling tobacco or alcohol) on completion of baseline, 6- and 9-month data collection.

Randomisation and blinding

After baseline data collection, researchers phoned a study administrator on a dedicated randomisation number. On receiving the call, the administrator randomly picked a concealed, sealed envelope from a bundle. The concealed, sealed envelopes were produced in advance of recruitment, by an independent member of the research team who had no other study involvement. The administrator opened the envelope (while the researcher remained on the phone line) and divulged the participant’s allocation (PHOENIx plus usual care (UC) or UC). Due to the nature of the intervention, blinding of participants and staff to allocation was not possible. Outcome assessments and analysis were conducted by blinded researchers.

Researchers aimed to locate participants and collect follow-up data at 6 and 9 months postrandomisation. Reasons, why follow-up assessments could not be conducted, included participant dead; participant lacking the mental capacity to make decisions (confirmed in writing in the participant’s medical case records); participant withdrawn consent; participant under the influence of alcohol and/or street drugs and unable to complete the assessment; or could not be located by researchers. For all participants (except those who had died, lost capacity or had consent withdrawn), researchers accessed health and social care records at 6- and 9-month follow-ups to extract service utilisation data and outcomes. Details of parallel economic and qualitative evaluations will be reported elsewhere.

Intervention

The PHOENIx intervention has been described previously14 16–18 28 and full details are given in online supplemental web appendix 2. Briefly, generalist pharmacists with an independent prescribing qualification paired with third sector homelessness outreach workers and visited participants weekly for approximately 7 months. During visits, PHOENIx collaborated with primary health, social care and third sector teams to assess and address a wide range of health and social care problems prioritised by the participant.

Outcomes

The coprimary outcomes of interest were the following progression criteria28:

1. Recruitment: at least 100 participants within 4 months of trial start date;

2. Data collection: at least 80% of participants with data collected (baseline, 6- and 9-month follow-ups);

3. Intervention adherence: at least 60% of participants in the PHOENIx group receiving the intervention;

4. Retention: at least 60% of participants remaining in the study (receiving in-person or telephone follow-up assessment at 6 and 9 months postrandomisation); and

   • any improvement in the rate of presentation to EDs, and overdoses, at 6- or 9-month follow-up.

Secondary outcomes included the number of participants with, and time to first: overdose; hospitalisation. Also, the number of participants receiving prescribed treatment for physical or mental health and problem drug use; number of treatments per participant; health-related quality of life (QoL) (Euro Qol-Visual Analogue Scale which forms part of the self-rated EuroQol 5 Dimension-3 Level (EQ5D5L). Participants were asked to measure their own health from 0 (worst health state imaginable) to 100 (best health state imaginable).32 Other measures included: Patient Experience with Treatment and Self-management measure (PETS: a measure of the work and impact of self-management on functioning and well-being)33 (the research team worked with the developer (Dr David Eton) to adapt PETS version 2.0. The PETS including adapted versions are protected by copyright, ©2020 Mayo Foundation for Medical Education and Research. All rights reserved. Permission to use the PETS from Dr Eton.) number of primary healthcare contacts; number of missed and attended outpatient appointments; a measure of frailty (Fried’s adapted frailty phenotype)34; anxiety/ depression score (Patient Health Questionnaire 4 (PHQ4)35; modified Medical Research Council breathlessness scale36; peak flow rate; participant reported injecting drug use; attempted suicide and self-harm. Welfare entitlements, social prescribing and tenancy type were also collected. Adverse events were not formally assessed although events such as overdoses and ED attendances were collected by researchers and recorded onto data collection forms at 6- and 9-month follow-ups (online supplemental web appendix 1). PHOENIx provided care in accordance with established local and national clinical guidance.

Procedures

Baseline characteristics for each participant were documented at inclusion. Primary and secondary outcomes were assessed as close as possible to the planned 6 and 9 months after randomisation. Independent researchers collected outcome data during in-person/phone interviews and from participants’ health and social care records. Overdose occurrences at follow-up were collected from primary and secondary care clinical records.

Participants who died or lost capacity were excluded from numerators and denominators of calculations of retention in the trial at 6 and 9 months.

Sample size and statistical methods

We aimed to recruit at least 100 participants to inform the sample size of a subsequent definitive RCT28 37 and based on an earlier feasibility study,18 we invited 160 to account for loss to follow-up. No interim analyses or stopping guidelines were planned.28

Participant characteristics are summarised using counts and percentages for categorical data, means and SD, or medians (IQR) 25th and 75th percentiles for continuous data depending on the distribution. Primary and secondary outcome measures are reported using the same summaries as noted above at each of the planned 6- and 9-month follow-ups. Number and percentage of participants experiencing outcomes of interest, as well as the mean and SD/median (IQR) for the number of events per participant are reported as required.

This approach was used for exploring the pilot data because this is a hypothesis-generating study, the sample size was small, outcomes were anticipated to have low prevalence and the study was not powered to detect differences between groups.28 37 As noted previously,28 we followed recommended advice for the analysis of pilot RCTs by reporting descriptive statistics rather than reporting formal tests of significance.37 The aim of a pilot trial is not to assess effectiveness (or efficacy) and it will usually be underpowered to do this.31 Formal hypothesis testing for effectiveness (or efficacy) is not recommended.31 37 Our aims were to assess participant recruitment, retention, data collection and intervention delivery, with exploration of any signal of effect in clinical outcomes, for example, ED attendance or overdose occurrence. The information gathered will inform the design of a planned, future full-scale RCT. All summaries were conducted using MINITAB 21.38

Patient and public involvement

People experiencing homelessness were involved at each stage of the study including identifying the need for the study, recruitment and providing feedback on the proposed intervention. This was done independently by Johnsen and colleagues.39 Participants (n=7) and people with lived experience of homelessness and street drug use provided comments on the research materials (consent form, participant information and baseline assessments) when they were consulted during protocol development. People (n=4) with lived experience from participating third sector homelessness organisations were part of the trial management group, which met monthly throughout the trial, and contributed to the final manuscript.

Process and economic evaluation

Semistructured interviews will explore future implementation of the PHOENIx intervention. Participants’ reasons for overdose and protective factors are described.40 An economic evaluation will assess the feasibility of conducting a cost-effectiveness analysis in a subsequent definitive trial.28

Results (primary outcome)

Recruitment

Glasgow has over 30 venues accommodating or providing support to people experiencing homelessness, most of which are located in the city centre. The research team started recruiting in city centre venues for convenience and then travelled to three outlying congregate dwelling places for people experiencing homelessness. Target recruitment was reached after visiting the 20th venue. Researchers visited homeless venues and passed study information to staff. Staff then approached eligible participants, explaining the nature of the study, and passed eligible participants the study information leaflet. Participants who were interested in participating made themselves known to the researchers. A median of 3.5 participants were recruited from each venue (IQR 2–9). 130 eligible people experiencing homelessness with a non-fatal overdose in the preceding 6 months were approached over 2.9 months (65 working days) between May and September 2021 (figure 1). Two participants declined and 128 (98%) provided informed consent thereby satisfying the first progression criterion.

Request permission

Data collection (baseline)

Baseline interviews were conducted by researchers asking participants a mixture of open and closed questions (online supplemental appendix 1). Researchers were not viewed as being part of social welfare and they did not describe themselves as such to participants. They did not affect care directly. Instead, they obtained informed, written consent and collected baseline data (tables 1 and 2) before randomisation. Participants’ responses were written directly onto the paper data collection form, by the researcher during the interview. Following in-person assessments, researchers accessed participants’ health (primary care and secondary care), social care and third sector records to confirm relevant variables, for example, overdose, diagnoses and prescribed medicines.

Tables 1 and 2 give participant characteristics at baseline prior to randomisation. None of the characteristics at baseline relate to services given or received after baseline. The tables describe summary characteristics by randomised group. There were few missing data; where these did occur, they are quantified in table footnotes. One exception was the collection of data from participants, on duration of homelessness, where 36 intervention and 44 UC participants had missing data. The reason for this was that the question on duration of homelessness was omitted in error from earlier versions of the data collection form. While we have a lot of missing data for this measure, we have no reason to believe the findings would be different if all participants had provided data in relation to this.

Baseline data were collected on all recruited participants, satisfying the second part of the progression criteria, as reported previously.14

Approximately 25% of participants were not registered with a GP.14 Of those who were registered, contacts were infrequent (one-third of participants contacted their GP in the previous 6 months). 90% of participants were registered with Alcohol and Drug Recovery Services (ADRS: the health and social care service with responsibility for managing alcohol and problem substance use), of whom, 60% had at least one contact in the past 6 months. One-third of participants were registered with specialist mental health services although 75% had not had any contact in the past 6 months.

Randomisation achieved comparable characteristics for most variables, with some exceptions including patterns of street drug use: 45 (73%) in the PHOENIx arm used heroin compared with 32 (48%) in the UC arm; 41 (66%) in the PHOENIx arm used cocaine versus 35 (53%) in UC and 32 (76%) used cannabis compared with 21 (65%) in UC. Four (6%) used gabapentinoids in PHOENIx group compared with 9 (16%) in UC. There appeared to be fewer PHOENIx group participants prescribed diazepam maintenance by ADRS: as part of a wider harm reduction intervention (5 (9%) PHOENIx versus 8 (12%) UC.

In table 1, diazepam means ‘prescribed diazepam’. In Scotland, because of the extent of the street benzodiazepine problem,41 diazepam is prescribed by ADRS staff, as a substitute (although unlicensed) for some people who have problem street benzodiazepine use.

All participants had severe and multiple disadvantages. Quality-of-life scores were low, and experience of burden of treatment scores was high (indicating more difficulty for participants) in both groups at baseline. There appeared to be fewer participants in the PHOENIx arm with hospitalisations and ED visits in the previous 6 months and contacts with ADRS prescribers (medical, nursing and pharmacy staff all prescribed; medical staff given as an example in table 2). Approximately one-third of participants in each arm received at least one contact from a GP in the previous 6 months, and 11%–20% had received care from a practice nurse.

Data collection (6- and 9-month follow-ups) and participant retention

Figure 1 shows losses and exclusions after randomisation with reasons, by treatment allocation. No safety concerns were raised by trial staff, participants or the Trial Management Group at any point during the study. The mean duration of 6 months in-person follow-up was 5.8 (SD 1.6) months: PHOENIx 5.5 months (SD 1.6) and UC 6.0 months (SD 1.5). Due to delays in locating and interviewing five participants (two in the intervention arm and three in the UC arm) who were in prison at 6 months, the mean duration of planned 9-month (in-person) follow-up was 10.3 months (SD 1.5): 10.5 (SD 1.5) months in the intervention group and 10.2 (SD 1.6) months in UC. Figure 1 shows data collected at baseline (128/128; 100%), 6 (94/120; 78%) and 9 (91/116; 78%) months give 86% combined, while satisfying the minimum threshold of 60% for proportions retained in the study at 6- and 9-month follow-ups. At both follow-up time points and in both study arms, all participants other than those who had died, lost capacity or were withdrawn, had secondary outcome data collected from health and social care records.

PHOENIx intervention adherence

It was planned that at least 60% of participants would receive the PHOENIx intervention at least weekly. 60 (97%) of PHOENIx participants received at least one visit or phone call from the PHOENIx team. Every weekly visit/phone call counted as part of intervention delivery was a consultation rather than an attempt at outreach. One participant moved into permanent accommodation on the day of recruitment and another voluntarily withdrew (giving no reason). Figure 1 shows the PHOENIx intervention was offered to 60 participants from the date of the first participant being randomised, until 3 March 2022 (the end of funded intervention time). 46 (74%) received at least weekly visits (or phone calls) from the PHOENIx team until study close, death, loss of capacity or withdrawal of consent. Table 3 (PHOENIx consultations and interventions) summarises the types of contact, and health and social care actions taken by the PHOENIx team. On average, participants received 14.9 in-person consultations, 13.8 phone consultations and 2.5 text messages (combined average of 28 contacts per participant) over a mean period of 201.8 (SD 58.2) days (28.8 weeks; 7.2 months). The minimum number of consultations per patient was 2 and the maximum was 86. The minimum duration of PHOENIx support was 39 days and the maximum was 290 days (10.4 months). Eight out of 10 attempted consultations resulted in face-to-face or phone consultations. Participants received almost 20 hours of in-person or phone support from the PHOENIx team over an average of 7.2 months (approximately 40 min/week). Those participants who received the intervention for shorter periods had either died during the trial, were lost to follow-up, or lost mental capacity to continue in the study. At the end of the study intervention period, the PHOENIx team stopped seeing participants. They stepped back and UC (without the PHOENIx intervention) resumed. If there were any outstanding actions for participant care, these were recorded in medical and social care records to ensure continuity of care. The intervention was individualised to each participant. This was necessary because participants had a large range of different, longstanding health and social care problems as described in tables 1 and 2. The wide range of support (duration of intervention and range of problems addressed) did not come as a surprise to the staff delivering the intervention, or the research team, because we had conducted extensive feasibility testing prior to the pilot RCT.16–18

Many participants had not had any previous therapeutic relationships with health or other workers and were wary of entering into any kind of alliance–outcome relationship due to past traumas. This meant that some time was needed to build trust before beginning to tackle the participants’ problems. When the team addressed health and social care problems, these took some time because there were normally multiple problems. For example, the participant may have prioritised their dental pain, necessitating the team to take the participant to a dentist for registration and booking serial appointments for assessment and treatment. Following this, or while waiting for the next dental appointment, the participant may have asked for help with their leg ulcer, or respiratory problems, or help to apply for welfare benefits. All of these actions took time and persistent follow-up on outreach.

Changes to care resulting directly from PHOENIx

49 (82%) participants received at least one physical health medicine prescription from the pharmacist. Online supplemental web appendix 3 describes the range of medicine groups prescribed for physical health problems including pain/epilepsy; anaemia and nutritional problems; chronic obstructive pulmonary disease and wound care dressings. Medicines prescribed for mental health conditions included antidepressants and in a minority of instances, antipsychotics which were prescribed for people with established, relevant diagnoses who had recently discontinued due to difficulty obtaining prescriptions. Approximately half (32/60 (53%)) received at least one prescription for an untreated mental health problem. A PHOENIx Pharmacist-initiated prescription to treat problem drug use was rare (3% of participants) because PHOENIx Pharmacists worked under Homeless Health Service GP Service governance, which did not include governance to prescribe medication-assisted treatment for problem drug use. Instead, patients were prescribed treatments for problem drug use, by specialists in the ADRS. Requests for help with problem diazepam use constituted most PHOENIx team referrals to ADRS. These occurred after several consultations, the team having built trust to the point where participants began to think about addressing their problem street benzodiazepine use, then accepted the offer of specialist assessment.

At baseline, participants had a range of conditions, some of which were treated and others untreated (table 1). PHOENIx Pharmacists prescribed medicines for conditions whether they were treated or not. For treated conditions, the pharmacists tended to stop existing treatments (because they were not achieving the desired effect, for example, antidepressant without any evidence of benefit) and start a new type of medicine for the same condition, or change the dose of an existing medicine, for example, increase the dose of a steroid inhaler for chronic obstructive pulmonary disease. Pharmacists made the new diagnoses described in table 3.

In all cases when medicines were prescribed, or participants received treatment or referral, PHOENIx teams followed up at the next contact, to confirm the patient was responding as planned.

Third sector homelessness worker interventions received by most participants: advice and social support, and accompanied appointment attendance, for example, to job centres and outpatient appointments, were common. Half of participants received items to help with daily living, for example, kettles and televisions, for use in emergency accommodation. Social workers do not tend to support people experiencing homelessness in this way. Usual social support does not include working with people experiencing homelessness (or the wider population in recovery from drug use) to encourage participation in work training programmes, voluntary or paid work. Through social prescribing networks established by third sector organisations, half of participants began participation in purposeful activities, for example, fixing bicycles or other voluntary work. These types of activities were welcomed by participants and viewed as a step towards recovery and re-integration into mainstream society.

Rate of presentation to ED at 6- and 9-month follow-up

Three measures of ED attendance were collected at follow-up: ED attendances per participant; participants with at least one ED attendance; the number of days between randomisation and first overdose after randomisation.

PHOENIx supported some participants to attend ED (table 3). Table 4 shows what appears to be more participants with at least one ED visit but delayed time to first ED visit. From Table 4, there appeared to be a delay in the time to first overdose in the intervention group compared with UC. Similarly, there appeared to be more intervention group participants with at least one hospitalisation compared with UC but delayed time to hospitalisation.

Participants in the PHOENIx arm received increases in the prescribing of medicines for physical and mental health and problem opiate use and more ADRS contacts at follow-up. Baseline differences in diazepam (fewer in the intervention arm) prescribing persisted. Quality-of-life scores (on the EQ5D5L Visual Analogue Scale) in PHOENIx appeared to be higher at follow-up but any apparent difference waned at 9 months. Of note is the improvement in EQ5D5L scores in both groups from baseline (table 1) to follow-up (table 4).

Online supplemental web appendix 4 shows reduction in treatment burden workload and impact of self-management in PHOENIx participants was more pronounced at 6 months (while the intervention was being delivered) than at 9 months. Online supplemental web appendix 4 also shows eight (15%) PHOENIx group participants lived in temporary furnished flats at 6 months, signalling a step up in terms of independence, as compared with supported or unsupported emergency accommodation. Seven (16%) PHOENIx group lived in their own tenancies at 9 months compared with UC (4(9%)) and increases in welfare benefits appeared higher in participants receiving PHOENIx support. High levels of psychological distress persisted throughout the study for most participants (online supplemental appendix 4).

Strengths and limitations

As far as we know, our approach involving collaboration between health and third sector workers is novel.25 Third sector and voluntary organisations exist across the world, aiming to support people experiencing homelessness. In addition to the UK, pharmacists can legally prescribe medicines in Canada, New Zealand, Nigeria, Argentina, Israel and the USA.21–23 Our intervention, if shown to be effective in a subsequent RCT, may be generalisable worldwide to other healthcare settings where clinical pharmacists and third sector homeless teams work collaboratively with GPs/family physicians. However, in healthcare systems other than the UK National Health Service, transferability of the pharmacist intervention may be limited by other impediments faced by people experiencing homelessness, for example, a lack of health insurance which complicates access to medicines. It is possible that the intervention could be delivered by independent prescriber nurses or other healthcare professionals, alongside third sector homelessness workers. Adherence with the intervention was comparable to rates in other studies targeting homeless participants.14 25 We included a diverse range of outcomes that are important to patients including, for the first time, a measure of treatment burden (PETS) and explored the possibility of time to overdose being a primary outcome in subsequent work.

Nonetheless, the study has several limitations. Pilot studies are not designed to detect intervention effectiveness. The study setting and availability of other services may have impacted outcomes. For example, uptake of PHOENIx interventions may have been conditional on a relative lack of alternative health and social care outreach or, for example, tailored mental health services such as cognitive–behavioural therapy—trauma-focused care. Generalist pharmacists worked closely with specialist homelessness/inclusion health GPs, obtaining advice on outreach when needed. Specialist homelessness GP services exist because of the unique needs of this group of patients, and the evidence of effectiveness for people experiencing homelessness.60 The UK also has mainstream (regular) GP practices catering for people who are not homeless. People are free to register with one or the other. The specialist homelessness GP service offered psychologically informed unconditional care to people experiencing homelessness. The PHOENIx Pharmacists, acting as the outreach arm of the specialist homeless GP practice, also provided trauma-informed care. Continuity of care was offered through repeated outreach visits, even if the participant moved from one temporary accommodation to another. The specialist homelessness GPs and pharmacists never ‘closed’ participants to their services and did not move people who were registered with the homelessness GP service onto a mainstream GP practice, until they were ready.60

The extent of access to permanent supported housing and rehabilitation in Glasgow also has a bearing on the PHOENIx intervention and outcomes. Other weaknesses include some baseline imbalances between groups which may have arisen following 1:1 randomisation without stratification, for example, baseline diazepam prescribing.