BackgroundHomelessness and associated mortality and multimorbidity rates are increasing. Systematic reviews have failed to identify complex interventions that decrease unscheduled emergency health services utilisation or increase scheduled care. Better evidence is needed to inform policy responses. We examined the feasibility of a complex intervention (PHOENIx: Pharmacist led Homeless Outreach Engagement Nonmedical Independent prescribing (Rx)) to inform a subsequent pilot randomised controlled trial (RCT).MethodsNon-randomised trial with comparator group set in Greater Glasgow and Clyde Health Board, Scotland. Participants were adult inpatients experiencing homelessness in a city centre Glasgow hospital, at the point of hospital discharge. All patients were offered and agreed to receive serial consultations with the PHOENIx team (NHS Pharmacist prescriber working with Simon Community Scotland (third sector homeless charity worker). Patients who could not be reached by the PHOENIx team were allocated to the usual care (UC) group. The PHOENIx intervention included assessment of physical/mental health, addictions, housing, benefits and social activities followed by pharmacist prescribing with referral to other health service specialities as necessary. All participants received primary (including specialist homelessness health service general practitioner care, mental health and addictions services) and secondary care. Main outcome measures were rates of: recruitment; retention; uptake of the intervention; and completeness of collected data, from recruitment to censor date.Results Twenty four patients were offered and agreed to participate; 12 were reached and received the intervention as planned with a median 7.5 consultations (IQR3.0-14.2) per patient. The pharmacist prescribed a median of 2 new (IQR0.3-3.8) and 2 repeat (1.3-7.0) prescriptions per patient; 10(83%) received support for benefits, housing or advocacy. Twelve patients were not subsequently contactable after leaving hospital, despite agreeing to participate, and were assigned to UC. Two patients in the UC group died of drug/alcohol overdose during follow up; no patients in the Intervention group died. All 24 patients were retained in the intervention or UC group until death or censor date and all patient records were accessible at follow up: 11(92%) visited ED in both groups, with 11(92%) hospitalisations in intervention group, 9(75%) comparator group. Eight (67%) intervention group patients and 3(25%) comparator group patients attended scheduled out patient appointments. Conclusions Feasibility testing of the PHOENIx intervention suggests merit in a subsequent pilot RCT.