The aim of the study was to examine the feasibility of recruitment, retention, the extent of intervention delivery and collection of outcome data, prior to a pilot randomised controlled trial. Twenty four people experiencing homelessness were offered, and 24 agreed to participate in the study. Half were subsequently engaged and received serial health checks leading to prescribing, and onward referral. Six of the 12 who were un-contactable, could be traced but their circumstances contra-indicated a visit from the PHOENIx team or researchers suggesting overall recruitment in a subsequent RCT with the need for face to face baseline assessment, would be in the region of 50%, which is similar to one previous study.30 The sample size of any planned pilot study would therefore require to be doubled. Engagement and retention with the intervention was excellent (all participants in the intervention arm engaged with the team throughout the duration of the study). There was no loss to follow up in terms of accessing participants’ hospital clinical records, providing a low cost approach to collecting follow up acute health service utilisation data in a subsequent pilot study.
Fewer patients had missed appointments in the intervention group with fewer missed appointments per patient, possibly due to the PHOENIx team arranging out patient appointments and supporting travel arrangements. The team also supported patients to attend their GP, although primary health care utilisation data were not sought during this study. The team encouraged patients to go directly to ED where this was deemed necessary, particularly out of hours when GP surgeries were closed. This may account for the slightly higher number of patients with hospitalisations in the intervention group, particularly for mental health problems.
Patients interviewed in a recent qualitative study suggested the PHOENIx team may directly improve health.24 By design, the current study lacks the power to determine if the intervention does objectively improve health, however our findings (recruitment, retention, uptake of the intervention, extent of data collection) warrant further examination in a randomised controlled pilot study, with parallel health economic and qualitative process evaluation to assess implementation potential.
Strengths and limitations
Recruitment and attrition difficulties in trials involving people experiencing homelessness is described previously.32,33 Our recruitment method (AHLT approaching patients in hospital) achieved 100% recruitment of targeted patients, but 50% subsequently unavailable for baseline assessment suggests the need for baseline assessment at the point of recruitment in hospital, or participant incentives e.g. vouchers, or additional approaches e.g. peer recruiters.
Participants’ characteristics were similar to those described in previous studies; 13,15,18,34 increasing the chances of generalisability in different settings and healthcare systems. Findings can be compared to the characteristics of patients experiencing homelessness in Glasgow (Lowrie F et al)22 and Glasgow/Edinburgh (Zeitler et al).13 A comparison of the study sample Table 2 (patient characteristics) with the samples described by Lowrie22 and Zeitler13 show similarity in terms of age and gender, but our current sample had a higher prevalence of drug use (92% compared with 62%22, and 73%13), hostel dwelling (60% vs 40%22) and rough sleeping (21% vs 13%).22 These differences are possibly explained by previous studies sampling exclusively from patients registered with the Homeless Health services in Glasgow and Edinburgh, whereas our inclusion criteria were homelessness and discharged from hospital, 75% of whom had and 25% had no registered GP. Lack of any type of accommodation and not having a registered GP together suggests a lower level of priority attached to, or difficulty associated with GP registration, by patients in our sample.
The complex PHOENIx intervention included components seen as important from the perspective of a person experiencing homelessness e.g. the immediacy of response. 24, 35 It was delivered as planned, with patients receiving a median of seven consultations each following 12 attempts by the PHOENIx team, and no refusals, suggesting acceptability in our sample.
The PHOENIx response to the patient’s health check was tailored, and the patient’s shared primary and secondary care clinical record was updated at each consultation. A wide range of problems were identified and resolved during the consultation. Detailed actions taken as a result of patient consultations, have been described previously. 20-22 Many of the changes to prescribing are evidence based e.g. initiation of an antidepressant for a major depressive episode, and would be expected to reduce the risk of depression causing or contributing to ED attendance. In this feasibility study, by design, limited inferences can be made about the link between prescribing, adherence and health service utilisation outcomes. In a subsequent pilot or definitive study, with a larger sample, if there is a difference in pre-specified outcomes between groups, we would be able to consider how, by prescribing and on occasion, going to a pharmacy and having the medicines dispensed then taken back to patients, adherence would likely improve. In turn, if the prescription was a direct treatment for a specified condition e.g. an infection, or essential to prevent a medical emergency e.g. insulin for a type 1 diabetic, it is likely that subsequent ED visits would be averted in a short time period. The team delivering the intervention were adept at gaining patients’ trust which is known to be important for improving health outcomes36 and enlisting the support of a wide range of individuals from diverse organisations, to prioritise and co-ordinate care for people with complex needs. The intervention team may have helped to reduce patients’ treatment burden.37 Treatment burden refers to the demands made of patients by healthcare systems. It includes the work patients have to do to gain an understanding of their health problems, the engagement work they do to access services and treatments (including prescriptions), the work of attending appointments, undergoing investigations and taking medication, as well as self monitoring work they may have to undertake. An example of the treatment burden experienced by people experiencing homelessness, who have a median of 6 or 7 long term conditions, is adhering to complex treatment regimens, or organising already challenging life circumstances to order and re-order their medicines, or attend different, un-co-ordinated services for their mental, physical and addictions long term conditions. 37 Partnership working with GPs is a pre-requisite to co-ordination of primary care: clinical governance for the PHOENIx team rested with the Homelessness Health Service GP service.
Independent Prescriber Pharmacists and Nurses, and SCS street outreach workers are available to provide this intervention beyond the life of the feasibility study, as part of routine service, underscoring opportunities for rollout into routine primary care should the intervention prove beneficial, and cost effective. Outcome measures in this study were relevant, built on previous work, inexpensive to collect, objectively assessed, and reproducible. Surrogate end points could be assessed in subsequent trials, however unlike patients recruited in previous trials of patients recruited because of a single condition, 38-40 people experiencing homelessness have multiple complex health and social care needs meaning outcomes are more diverse. Subsequent pilot work on the PHOENIx intervention could evaluate patient reported outcomes, or effectiveness using health state utilities (for a future cost utility analysis) e.g. through the use of EQ-5D-5L to generate Quality Adjusted Life Years which could then be used alongside cost data to give an indicative picture of cost effectiveness. Two previous UK based studies of patients experiencing homelessness have included economic analyses.41,42
It is possible that patients allocated to the usual care group had greater levels of unmet health needs than those who could be reached. Because of the additional difficulty associated with engaging with these patients, they are likely to have had fewer contacts with services e.g. health and social care, addictions, more chaotic lifestyles and perhaps lower prioritisation of health needs and worse health. Randomisation in any subsequent controlled pilot study is needed to reduce selection bias however those who are hard to reach may remain so even within the context of a RCT. In some settings, where the majority of people experiencing homelessness are migrants and low paid workers, and in healthcare systems that are not free at the point of use, our findings may not be generalizable.
Through our pragmatic trial we aim to demonstrate improvements in the quality and reductions in the cost of care. Quality could be measured by patients receiving more appropriate care through being able to access first contact, co-ordinated, continuous care by the PHOENIx team who act as a bridge back into GPs and other specialists in homelessness health.24 In order to decrease ED visits through an intervention, we need a sufficient number of ED visits for any reduction to be visible. To maximise the chances of an intervention showing a difference between intervention and control groups in a subsequent trial, targeting patients with higher rates of baseline ED presentations may be more appropriate, or those with higher ED presentations due to physical health problems, given the tendency for the team to assess, diagnose and treat physical health problems.
A small number of patients were recruited, but recommendations for sample sizes in feasibility studies are sparse, because the aim is to examine recruitment, retention, intervention fidelity and outcomes.25,26,43,44 A recent systematic review of RCTs, non RCTs and controlled before-after studies of interventions to improve care of people who are homeless, identified only one feasibility study, with nine participants (six intervention; three control).19 More definitive RCTs are needed, including assessment of a wider range of outcomes e.g. quality of life measures, primary care health utilisation, and duration of hospitalisation.
Comparison with existing literature
In accordance with the recommended stages of complex intervention testing, 25 the PHOENIx intervention was developed and optimised previously.24, 20-22 The pairing of third sector worker and pharmacist independent prescriber on outreach, as far as we are aware, is a novel approach to improving engagement and uptake of health and social care interventions in people who are homeless.19 The only other pharmacist led intervention study in this area did not include assessment of wider health needs, prescribing or referral, or offer serial encounters. 45
Systematic and other reviews have described health interventions to improve health in people who are homeless. However there are no known effective and cost effective ‘off the shelf’ interventions involving healthcare professionals. 19, 46, 47
Our target population are people who pay the ultimate price of extreme inequity: they experience severe and multiple disadvantage across their lives and often die sooner because they are worse off. Our findings suggest a pilot study of the intervention would be justified, and thereafter, a definitive randomised controlled trial. If the definitive study shows reduction in emergency health service utilisation, or length of hospitalisation, or fewer deaths in the intervention group with causality attributable to the PHOENIx intervention, then we will have created an intervention to address one aspect of inequity in health for people experiencing homelessness.
The lack of published feasibility and pilot studies preceding definitive RCTs, suggests previous investigators may not have worked through the phases of developing their complex interventions, introducing the possibility of weaknesses in the reporting and conduct of their definitive trials.48,49 In terms of generalisability of the study population, observed ED visit rates were lower than those reported in studies from Canada 30, 49 and North America. 49 Multiple factors interact to influence healthcare utilisation in people who are homeless, however the lower rates observed in our study are surprising given the high rates of mental health and substance misuse in our target group,13,18 which are associated with higher rates of ED use. Differences in access and payment in different health care systems may also explain differential ED rates.50
Interventions involving tailoring primary care to people who are homeless have been tried previously to decrease ED use 51 or complex interventions which include respite and social care/housing support 52 although not in the context of adequately powered RCTs. The PHOENIx intervention is tailored to individuals over time, combining assessment of health, with housing and opportunities to involve patients in social activities, because housing is an integral component of disease management53-54 and having a structure and purpose to daily life is rated as important for patients to remain healthy.55 The diverse range of outcomes described in this feasibility study suggest the PHOENIx intervention assessed and addressed factors that take precedence over health care in addition to the predisposing needs that drive health seeking behaviour. 56 The social prescribing component of the PHOENIx intervention offered patients an opportunity to move into a place of safety by day and night. We offered patients opportunities to occupy their time with productive, meaningful activities. These activities took patients away from the city centre streets where they ordinarily would spend their time and where drug use and violence was likely. Assessment and intervention to address homelessness led to some patients obtaining accommodation, or better accommodation, giving them a place of relative safety by night, as compared with rough sleeping for example. We suspect most people would regard these components of the intervention as important priorities, in a hierarchy of needs, which underpin good health and may be a pre-requisite to seeking healthcare, at least healthcare that the patient may view as non urgent.