Twenty randomised controlled trials (14,416 participants) were included. Fifteen trials used cluster randomisation. One trial was judged as having a low risk of bias, five as high risk and 14 had an unclear risk of bias. The most common problems were unclear randomisation and/or allocation concealment, and incomplete follow-up. None of the cluster randomised trials reported blinded recruitment and two had unit of analysis errors. Follow-up periods ranged from one to 12 months.
Educational outreach initiatives in nursing homes (two trials, 1,538 residents): One trial found statistically significant reductions for intervention compared with usual care in antipsychotic drug use. The other trial found no significant results.
Educational meetings in the work setting (five trials, 811 residents): There were conflicting results. Two trials found significant reductions in neuroleptic and NSAID use, one after training and support for health personnel and the other after a staff educational programme. No other significant results were reported.
Educational meetings with at least one additional intervention (three trials, 8,559 residents): Interventions varied. One trial found a significant reduction in the number of antibiotic prescriptions for suspected urinary tract infection at 12 months. One trial found a significant reduction in antipsychotics consumption. No other significant differences were reported.
Medication review with a pharmacist (seven trials, 3,212 residents): Results were variable. One trial found a statistically significant difference in the mean number of drug changes at six months. Two trials reported significant differences in medication appropriateness index scores at three and eight months. One trial reported a significant reduction in psychoactive drug use at 12 months. Four trials found no differences in falling incidences. All other comparisons were non-significant.
Other interventions (three trials, 256 residents): One trial found significantly fewer drugs were prescribed with medical care by a geriatric assessment team compared with standard care at three months. One trial of a programme of dementia care found significantly less use of physical restraints for intervention groups during activities arranged outside the nursing unit. There were no other significant differences.