Reporting of outcomes.

Impact studies were excluded from this table because these studies had outcomes of a different type (e.g., costs). Hence, the total number of studies is 68.aNot applicable in 11/68 studies, because all cause death was the outcome.bTypes of outcomes and how they were analysed (unclear for five studies). The sum 6+23+8+30 is higher than 63 because some outcomes were analysed in more than one way (e.g., a time-to-event outcome that was analysed as time to event and as a binary outcome neglecting time). If a study analysed two binary outcomes, it was here counted as one binary outcome.cAfter dichotomization of a continuous outcome.dOne study used the Cochran–Mantel–Haenszel procedure, another calculated odds ratios.CART, classification and regression tree.