BACKGROUND: Inequitable access to health care increases morbidity and mortality among people experiencing homelessness. Peer advocates ('peers') with lived experience may help others to access health care.

OBJECTIVES: To evaluate the impact and cost-consequence of Groundswell's Homeless Health Peer Advocacy programme on healthcare access, the processes through which it operates and the impact for peer advocates themselves.

ETHICS AND DESIGN: A participatory mixed-method design with three components: qualitative study (A), prospective cohort (B), and cost-consequence analysis (C) using cohort and programmatic data. Ethical approval: Dulwich Research Ethics Committee (Integrated Research Application System 271312).

SETTING: London, United Kingdom (2019-23) coinciding with COVID-19 and disruptions to the National Health Service, Homeless Health Peer Advocacy and housing services.

PARTICIPANTS: Homeless Health Peer Advocacy clients and non-clients (A-C); Homeless Health Peer Advocacy staff, volunteers and homelessness-sector stakeholders (A).

INTERVENTION: Peer advocates accompany clients to healthcare appointments and provide support to address barriers to access.

MAIN OUTCOME MEASURES: Primary: probability of 'did not attend' at a scheduled outpatient appointment within 12 months of cohort enrolment. Secondary: number of inpatient admissions and accident and emergency visits.

DATA SOURCES: (A) In-depth interviews and focus groups; (B) Structured questionnaires and National Health Service Hospital Episode Statistics; (C) Groundswell programme data and cohort findings.

RESULTS: Qualitative (A): Peer advocacy empowered clients by building cultural health capitals (skills and communication that support healthcare interactions) and strengthening social and economic resources. Advocates themselves gained social, cultural, human and physical resources, though benefits were greatest for those with some pre-existing stability. Cohort (B): Compared with non-clients, Homeless Health Peer Advocacy clients showed no difference in did not attend rates (rate ratio 0.97, 95% confidence interval 0.67 to 1.42) or accident and emergency visits (mean difference 0.86, 95% confidence interval -0.06 to 1.79) for the other pre-specified outcomes. Clients had 1.14 more inpatient admissions (95% confidence interval 0.52 to 1.75). Sensitivity analyses with imputed data suggested higher numbers of outpatient attendances, outpatient 'did not attends', accident and emergency visits and admissions among clients. Secondary analyses suggested differences by levels of anxiety and depression. Cost-consequence (C): Median annual cost per client was £353 (£176 per scheduled engagement). Evidence of National Health Service cost saving was inconclusive.

LIMITATIONS: The COVID-19 disrupted both Homeless Health Peer Advocacy delivery and National Health Service services. Non-randomised design may have introduced bias.

CONCLUSIONS: Homeless Health Peer Advocacy enhances clients' cultural health capital and helps peer advocates achieve their goals. We cannot state whether peer advocacy reduces 'did not attends' or demonstrate cost savings, but it was associated with more inpatient admissions and, in sensitivity analyses, more outpatient appointments.