“…Overall, nine of the studies were at high risk of selection bias [2,1,14,[20][21][22]27,30,29,31,36] due to uncontrolled confounders or, for before and after studies, due to having no control group for comparison. All studies were also unclear [20,30,29,31,40] or at high risk [2,1,14,16,19,21,22,26,27,[36][37][38] of performance bias due to an imbalance in treatment intensity or the existence of additional co-occurring interventions -such as one treatment group receiving extended 24-hour-care, assertive engagement, or home treatment whilst the other does not. Design specific issues, such as having only one site per intervention group in cluster-level studies or a lack of an appropriate concurrent control altogether, means all the studies were also at high risk of other potential sources of bias.…”