Try Wonders for freeHow the literature is organised
Three questions organise the field: how to detect postpartum depression, who gets it, and what it does to families.
Detection is dominated by the Edinburgh Postnatal Depression Scale. Cox, Holden and Sagovsky (1987) validated a ten-item self-report scale on 84 mothers against the Research Diagnostic Criteria, reported satisfactory sensitivity and specificity, and noted it takes about five minutes to complete. That scale is why the paper has more than 14,000 citations in the snapshot behind this page.
On prevalence and risk, the landmarks are meta-analyses. O'Hara and Swain (1996) put the average prevalence of non-psychotic postpartum depression at 13% and found the strongest predictors to be past psychopathology, psychological disturbance during pregnancy, a poor marital relationship, low social support and stressful life events. Beck (2001) re-ran the exercise over 84 studies from the 1990s, confirmed the earlier predictors and added four: self-esteem, marital status, socioeconomic status and unplanned or unwanted pregnancy. Robertson et al. (2004) synthesised antenatal risk factors. Gavin et al. (2005) reviewed 109 articles, included 28, and found combined point prevalence of major and minor depression between 6.5% and 12.9% across trimesters and postpartum months, with as many as 19.2% having an episode in the first three months after delivery. O'Hara and McCabe (2013) summarise the state of knowledge.
On consequences, Field (2010) reviews effects on early interactions, parenting and safety practices, and Slomian et al. (2019) synthesise 122 studies of maternal and infant outcomes, organised into maternal consequences, infant consequences and mother-child interaction. Saharoy et al. (2023) is a recent narrative account of the same ground.
Main debates
The first thing to understand is why prevalence figures disagree, because they do, widely. O'Hara and Swain (1996) identified the two causes: the assessment method (self-report measures give larger estimates) and the length of the postpartum period assessed (longer windows give higher prevalence). Al-abri et al. (2023), reviewing 128 systematic reviews, put numbers on the first: mean prevalence was 27.4% in studies using self-reported measures against 17.0% in those using structured interviews. So a page of sources reporting 13%, 24.7% and 34% is not necessarily contradicting itself; check the instrument, the cut-off and the window before you compare.
The second argument is about screening. O'Hara and McCabe (2013) describe a growing movement to build mental health screening into routine primary care for pregnant and postpartum women, and to follow it with treatment or referral — the follow-up being the part that decides whether screening helps. Gastaldon et al. (2022) supply what an evidence-based risk algorithm would need: in their umbrella review of 185 observational studies, the associations they rated highly suggestive were premenstrual syndrome, violent experiences and unintended pregnancy, and no association reached "convincing evidence".
Where recent work is heading
Four directions stand out. Fathers: Paulson and Bazemore (2010) meta-analysed 43 studies and 28,004 participants, estimating paternal depression at 10.4% between the first trimester and the first year, rising to 25.6% at three to six months postpartum, and moderately correlated with maternal depression (r = 0.31); Wang et al. (2021) review what influences it. Global evidence: Roddy Mitchell et al. (2023) pooled 589 studies covering 616,708 women in 51 low- and middle-income countries at 24.7%, with the highest subgroup figure, 38.9%, among women who had experienced intimate partner violence. Alshikh Ahmad, Alkhatib and Luo (2021) report 27% across 15 Middle East studies, and Chen et al. (2022) 34% across eight studies conducted during the COVID-19 pandemic. Treatment: the two zuranolone phase 3 trials (Deligiannidis et al., 2021, with 150 women at 30 mg; 2023, with 196 women with severe postpartum depression at 50 mg) both found significant improvement over placebo on the 17-item Hamilton scale at day 15, with somnolence, dizziness and sedation the most common adverse events. Sun et al. (2021) tested smartphone-delivered mindfulness training in pregnancy. Synthesis of risk: Gastaldon et al. (2022), plus meta-analyses by Liu, Wang and Wang (2022), who pool prevalence at 14% across 133,313 women, and Yang, Wu and Chen (2022), across 79,043.