Background: It is unclear how best to identify and treat women with mental disorders in pregnancy and the year after birth (i.e. the perinatal period). Objectives: (1) To investigate how best to identify depression at antenatal booking [work package (WP) 1]. (2) To estimate the prevalence of mental disorders in early pregnancy (WP1). (3) To develop and examine the efficacy of a guided self-help intervention for mild to moderate antenatal depression delivered by psychological well-being practitioners (WP1). (4) To examine the psychometric properties of the perinatal VOICE (Views On Inpatient CarE) measure of service satisfaction (WP3). (5) To examine the clinical effectiveness and cost-effectiveness of services for women with acute severe postnatal mental disorders (WPs 1–3). (6) To investigate women’s and partners’/significant others’ experiences of different types of care (WP2). Design: Objectives 1 and 2 – a cross-sectional survey stratified by response to Whooley depression screening questions. Objective 3 – an exploratory randomised controlled trial. Objective 4 – an exploratory factor analysis, including test–retest reliability and validity assessed by association with the Client Satisfaction Questionnaire contemporaneous satisfaction scores. Objective 5 – an observational cohort study using propensity scores for the main analysis and instrumental variable analysis using geographical distance to mother and baby unit. Objective 6 – a qualitative study. Setting: English maternity services and generic and specialist mental health services for pregnant and postnatal women. Participants: Staff and users of mental health and maternity services. Interventions: Guided self-help, mother and baby units and generic care. Main outcome measures: The following measures were evaluated in WP1(i) – specificity, sensitivity, positive predictive value, likelihood ratio, acceptability and population prevalence estimates. The following measures were evaluated in WP1(ii) – participant recruitment rate, attrition and adverse events. The following measure was evaluated in WP2 – experiences of care. The following measures were evaluated in WP3 – psychometric indices for perinatal VOICE and the proportion of participants readmitted to acute care in the year after discharge. Results: WP1(i) – the population prevalence estimate was 11% (95% confidence interval 8% to 14%) for depression and 27% (95% confidence interval 22% to 32%) for any mental disorder in early pregnancy. The diagnostic accuracy of two depression screening questions was as follows: a weighted sensitivity of 0.41, a specificity of 0.95, a positive predictive value of 0.45, a negative predictive value of 0.93 and a likelihood ratio (positive) of 8.2. For the Edinburgh Postnatal Depression Scale, the diagnostic accuracy was as follows: a weighted sensitivity of 0.59, a specificity of 0.94, a positive predictive value of 0.52, a negative predictive value of 0.95 and a likelihood ratio (positive) of 9.8. Most women reported that asking about depression at the antenatal booking appointment was acceptable, although this was reported as being less acceptable for women with mental disorders and/or experiences of abuse. Cost-effectiveness analysis suggested that both the Whooley depression screening questions and the Edinburgh Postnatal Depression Scale were more cost-effective than with the Whooley depression screening questions followed by the Edinburgh Postnatal Depression Scale or no-screen option. WP1(ii) – 53 women with depression in pregnancy were randomised. Twenty-six women received modified guided self-help [with 18 (69%) women attending four or more sessions] and 27 women received usual care. Three women were lost to follow-up (follow-up for primary outcome: 92%). At 14 weeks post randomisation, women receiving guided self-help reported fewer depressive symptoms than women receiving usual care (adjusted effect size −0.64, 95% confidence interval −1.30 to 0.06). Costs and quality-adjusted life-years were similar, resulting in a 50% probability of guided self-help being cost-effective compared with usual care at National Institute for Health and Care Excellence cost per quality-adjusted life-year thresholds. The slow recruitment rate means that a future definitive larger trial is not feasible. WP2 – qualitative findings indicate that women valued clinicians with specialist perinatal expertise across all services, but for some women generic services were able to provide better continuity of care. Involvement of family members and care post discharge from acute services were perceived as poor across services, but there was also ambivalence among some women about increasing family involvement because of a complex range of factors. WP3(i) – for the perinatal VOICE, measures from exploratory factor analysis suggested that two factors gave an adequate fit (comparative fit index = 0.97). Items loading on these two dimensions were (1) those concerning aspects of the service relating to the care of the mother and (2) those relating to care of the baby. The factors were positively correlated (0.49; p < 0.0001). Total scores were strongly associated with service (with higher satisfaction for mother and baby units, 2 degrees of freedom; p < 0.0001) and with the ‘gold standard’ Client Service Questionnaire total score (test–retest intraclass correlation coefficient 0.784, 95% confidence interval 0.643 to 0.924; p < 0.0001). WP3(ii) – 263 of 279 women could be included in the primary analysis, which shows that the odds of being readmitted to acute care was 0.95 times higher for women who were admitted to a mother and baby unit than for those not admitted to a mother and baby unit (0.95, 95% confidence interval 0.86 to 1.04; p = 0.29). Sensitivity analysis using an instrumental variable found a markedly more significant effect of admission to mother and baby units (p < 0.001) than the primary analysis. Mother and baby units were not found to be cost-effective at 1 month post discharge because of the costs of care in a mother and baby unit. Cost-effectiveness advantages may exist if the cost of mother and baby units is offset by savings from reduced readmissions in the longer term. Limitations: Policy and service changes had an impact on recruitment. In observational studies, residual confounding is likely. Conclusions: Services adapted for the perinatal period are highly valued by women and may be more effective than generic services. Mother and baby units have a low probability of being cost-effective in the short term, although this may vary in the longer term.