PurposeEvidence on the effectiveness of Independent Supported Housing (ISH) for non-homeless people with mental illness primarily comes from observational cohort studies, which have high risk of bias due to confounding by time-invariant sample characteristics. The present study proposes an alternative study design known from pharmacology to overcome this bias and strengthen evidence. MethodsWe conducted a retrospective mirror-image analysis with medical records of 144 ISH service users to assess the effectiveness of ISH in reducing the number and duration of hospitalisations. Outcomes occurring in equal periods before and during ISH utilisation were compared for every ISH user. Differences between the periods were tested with incidence rate ratios (IRR). ResultsIncluded service users were on average 38.2 years old, female (54%) and predominately had an affective (28.5%) or a schizophrenic or psychotic (22.9%) disorder with ISH utilisation days ranging from 36-960. Fewer admissions (IRR=0.41, 95%-CI 0.27-0.64) and fewer person-days hospitalised (IRR=0.38, 95%-CI 0.35-0.41) were observed during utilisation period compared to prior to their ISH utilisation. While the reduction in psychiatric admissions may be somewhat confounded by time-variant characteristics, the substantial reduction in hospitalised bed-bays represents at least partially an intervention effect. ConclusionsThe mirror-image study design allowed for a cost-effective investigation of ISH effectiveness in reducing hospitalisation without confounding by time-invariant sample characteristics. We provide recommendations for the design’s application and suggest further research with larger samples.