Process evaluation should be an integral part of randomised controlled trials (RCTs), and this is important in examining complex interventions with a number of active ingredients.

To analyse effectiveness and effect moderators in a multi-centre RCT of needs-oriented discharge planning vs. treatment as usual (TAU) for high utilisers of mental health care.

From April 2006 to July 2007, 491 adults with high utilisation of mental health care for psychotic disorders gave informed consent to participate in a multi-site RCT at five German psychiatric inpatient services. Subjects allocated to the intervention group were offered a manualised needs-led discharge planning intervention of two 45 minute sessions delivered by a trained study worker. Outcomes (service use, needs, psychopathology, depression, quality of life, direct costs) were assessed at four measurement points during 18-month follow-up. Latent class modelling was used to scrutinise differential effectiveness. Characteristics of subgroups of participants showing variable treatment response will be described.

There was improvement both in the intervention and (TAU) control group but no differential effectiveness of the discharge planning intervention. Only few variables significantly moderated effects. Effect moderators pertained to quality of delivery and intervention dose.

Study results could help to tailor interventions to local conditions and guide future implementation of discharge planning in routine care. Further research should focus on scrutinising study process, ensuring ecological suitability of interventions and on head-to-head comparison studies.