Assertive Community Treatment (ACT) was developed in the early 1970s as a response to the closing down of psychiatric hospitals. ACT is a team-based approach aiming at keeping ill people in contact with services, reducing hospital admissions and improving outcome, especially social functioning and quality of life. To determine the effectiveness of Assertive Community Treatment (ACT) as an alternative to i. standard community care, ii. traditional hospital-based rehabilitation, and iii. case management. For each of the three comparisons the main outcome indices were i. remaining in contact with the psychiatric services, ii. extent of psychiatric hospital admissions, iii. clinical and social outcome and iv. costs. Electronic searches of CINAHL (1982-1997), the Cochrane Schizophrenia Group's Register of trials (1997), EMBASE (1980-1997), MEDLINE (1966-1997), PsycLIT (1974-1997) and SCISEARCH (1997) were undertaken. References of all identified studies were searched for further trial citations. The inclusion criteria were that studies should i. be randomised controlled trials, ii. have compared ACT to standard community care, hospital-based rehabilitation, or case management and iii. have been carried out on people with severe mental disorder the majority of whom were aged from 18 to 65. Studies of ACT were defined as those in which the investigators described the intervention as "Assertive Community Treatment" or one of its synonyms. Studies of ACT as an alternative to hospital admission, hospital diversion programmes, for those in crisis, were excluded. The reliability of the inclusion criteria were evaluated. Three types of outcome data were available: i. categorical data, ii. numerical data based on counts of real life events (count data) and iii. numerical data collected by standardised instruments (scale data). Categorical data were extracted twice and then cross-checked. Peto Odds Ratios and the number needed to treat (NNT) were calculated. Numerical count data were extracted twice and cross-checked. Count data could not be combined across studies for technical reasons (the data were skewed) but all relevant observations based on count data were reported in the review. Numerical scale data were subject to a quality assessment. The validity of the quality assessment was itself assessed. Numerical scale data of suitable quality were combined using the standardised mean difference statistic where possible, otherwise the data were reported in the text or 'Other data tables' of the review. ACT versus standard community care Those receiving ACT were more likely to remain in contact with services than people receiving standard community care (OR 0.51, 99%CI 0.37-0.70). People allocated to ACT were less likely to be admitted to hospital than those receiving standard community care (OR 0.59, 99%CI 0.41-0.85) and spent less time in hospital. In terms of clinical and social outcome, significant and robust differences between ACT and standard community care were found on i. accommodation status, ii. employment and iii. patient satisfaction. There were no differences between ACT and control treatments on mental state or social functioning. ACT invariably reduced the cost of hospital care, but did not have a clear cut advantage over standard care when other costs were taken into account. ACT versus hospital-based rehabilitation services Those receiving ACT were no more likely to remain in contact with services than those receiving hospital-based rehabilitation, but confidence intervals for the odds ratio were wide. People getting ACT were significantly less likely to be admitted to hospital than those receiving hospital-based rehabilitation (OR 0.2, 99%CI 0.09-0.46) and spent less time in hospital. Those allocated to ACT were significantly more likely to be living independently (OR (for not living independently) 0.19, 99%CI 0.06-0. (A