Background: This study evaluated a multifaceted psychiatric intervention targeted at the complex medically ill identified by means of the INTERMED, an instrument to assess case complexity. Methods: Of 885 rheumatology inpatients and diabetes outpatients who were assessed for eligibility, 247 were identified as complex (INTERMED score >20) and randomized to the intervention (n = 125, 84 rheumatology and 41 diabetes patients) or care as usual (n = 122, 78 rheumatology and 44 diabetes patients). For the majority of the cases the multifaceted intervention consisted of an intervention conducted by a psychiatric liaison nurse and/or of referral to a liaison psychiatrist, followed by advice to the treating physician or organization of a multidisciplinary case conference. Baseline and follow-up at months 3, 6, 9 and 12 measured prevalence of major depression (Mini-International Neuropsychiatric Interview), depressive symptoms (Center for Epidemiological Studies Depression Rating Scale), physical and mental health (SF-36), quality of life (EuroQol), health care utilization and HbA1c levels (diabetic patients). Results: Prevalence of major depression was reduced from 61% (T0) to 28% (T4) in the intervention group and remained stable in care as usual (57% at T0 to 50% at T4). Compared to care as usual, significant improvement over time was observed in the intervention group with regard to depressive symptoms (F = 11.9; p = 0.001), perception of physical (F = 5.7; p = 0.018) and mental health (F = 3.9; p = 0.047) and quality of life (F = 21.8; p < 0.001). Effects tended to be stronger in diabetes patients, in patients with baseline major depression and in patients with moderate INTERMED scores. Finally, hospital admissions occurred less often in the intervention group, reaching statistical significance for the period between 6 and 9 months of follow-up (p = 0.02). Conclusions: The results suggest that a psychiatric intervention targeted for complex medical patients can improve health outcomes.