Internationally, an increasing proportion of emergency department visits are mental health related. Concurrently, psychiatric wards are often occupied above capacity. Healthcare providers have introduced short-stay, hospital-based crisis units offering a therapeutic space for stabilisation, assessment and appropriate referral. Research lags behind roll-out, and a review of the evidence is urgently needed to inform policy and further introduction of similar units. This systematic review aims to evaluate the effectiveness of short-stay, hospital-based mental health crisis units. We searched EMBASE, Medline, CINAHL and PsycINFO up to March 2021. All designs incorporating a control or comparison group were eligible for inclusion, and all effect estimates with a comparison group were extracted and combined meta-analytically where appropriate. We assessed study risk of bias with Risk of Bias in Non-Randomized Studies - of Interventions and Risk of Bias in Randomized Trials. Data from twelve studies across six countries (Australia, Belgium, Canada, The Netherlands, UK and USA) and 67 505 participants were included. Data indicated that units delivered benefits on many outcomes. Units could reduce psychiatric holds (42% after intervention compared with 49.8% before intervention; difference = 7.8%; P < 0.0001) and increase out-patient follow-up care (χ2 = 37.42, d.f. = 1; P < 0.001). Meta-analysis indicated a significant reduction in length of emergency department stay (by 164.24 min; 95% CI -261.24 to -67.23 min; P < 0.001) and number of in-patient admissions (odds ratio 0.55, 95% CI 0.43-0.68; P < 0.001). Short-stay mental health crisis units are effective for reducing emergency department wait times and in-patient admissions. Further research should investigate the impact of units on patient experience, and clinical and social outcomes.