Maternal anaemia remains a pressing global health challenge, with a notable burden in low- and middle-income countries. Existing studies in sub-Saharan Africa have largely relied on average associations, thereby concealing key variation among women and failing to account for heterogeneity.

To assess the association between completing at least eight antenatal care (ANC) contacts and maternal anaemia in Ghana and to explore heterogeneity in this association using causal machine learning.

An institution-based cross-sectional study was conducted using a retrospective review of medical records and causal machine learning analysis.

Juaben Government Hospital.

Of 2326 women who delivered at the hospital, 2114 with complete data on the main exposure and outcome variables were included in the analysis.

Completion of at least eight ANC contacts. ANC contact was defined as the in-person visit to the clinic with a healthcare professional for routine ANC services and follow-up.

Maternal anaemia, defined as a haemoglobin level below 11 g/dL in the last ANC before delivery.

A causal forest model was used to estimate the association between completing at least eight ANC contacts and maternal anaemia. Conditional average treatment effects were used to explore individual-level variation in these associations, providing policy-relevant insights.

Completing ≥8 ANC contacts was associated with a 6 percentage-point lower probability of maternal anaemia compared with having fewer visits (average treatment effect: -0.06, 95% CI –0.11 to –0.02). Predicted individual-level effects ranged from –0.21 to 0.09. Participants’ age, malaria prophylaxis, marital status, parity and educational level were the five most important contributors to the observed heterogeneity.

This study demonstrated that completing ≥8 ANC contacts is associated with a lower probability of maternal anaemia, with heterogeneity across subgroups. We recommend differentiated, context-specific ANC interventions that focus on high-impact subgroups while strengthening the effectiveness and quality of care delivered at each visit.

To translate and culturally adapt six self-report measures for depression, anxiety, post-traumatic stress disorder (PTSD) and somatic symptom disorder into Hindi and determine their diagnostic accuracy against a diagnostic clinical interview.

Cross-sectional validation study.

Rural Kangra, Himachal Pradesh, northern India.

480 perinatal (pregnant or within 12 months postpartum) and non-perinatal (not currently pregnant and not given birth within 12 months) women at one tertiary hospital and district-level Anganwadi (community health) centres.

Symptom endorsement; and discriminant validity, sensitivity, specificity, positive and negative predictive values and area under the receiver operating characteristic curve (AUROC) of the Kessler Scale of Psychological Distress (K10), Patient Health Questionnaire (PHQ9), Edinburgh Postnatal Depression Scale (EPDS), Generalised Anxiety Disorder Scale (GAD7), Perinatal Anxiety Screening Scale (PASS), PTSD Checklist (PCL-5) and Scale for the Assessment of Somatic Symptoms (SASS).

Complete data were available for 443 participants. Tiredness and body weakness were the most commonly endorsed symptoms among participants with common mental disorders. Among perinatal participants, the AUROC was highest for the GAD7 (0.88, 95% CI 0.79 to 0.96) and SASS (0.84, 95% CI 0.71 to 0.96). Among non-perinatal participants, the AUROC was highest for the SASS (0.92, 95% CI 0.88 to 0.97) and PHQ9 (0.91, 95% CI 0.86 to 0.96).

Measures which assess for fatigue, tiredness and somatic symptoms may help to identify women experiencing common mental disorders in this setting. Small numbers of participants with clinically diagnosed mental disorders in our sample mean results must be interpreted cautiously.

NCT05485701.