Prediction models for postpartum post-traumatic stress disorder: a systematic review and meta-analysis.
Huijing Shi, Jinyang Liu, Chaochao Yang et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis assessed multivariable models predicting post-traumatic stress disorder within 12 months after childbirth. It included 17 studies with 17,236 participants and pooled performance from validated models. The pooled area under the curve was 0.84 (95% CI: 0.80-0.88), with sensitivity of 0.77 and specificity of 0.82. Common predictors included mode of delivery, social support, pregnancy complications, and fear of childbirth. However, performance varied substantially across studies, and reporting of calibration and external validation was limited.
Key findings
- Across validated models, pooled AUC was 0.84 (95% CI: 0.80-0.88), with sensitivity of 0.77 and specificity of 0.82.
- The review included 17 studies involving 17,236 participants.
- Mode of delivery, social support, pregnancy complications, and fear of childbirth were frequently identified predictors.
- Performance estimates showed considerable heterogeneity (I² = 89.7%); calibration and external validation were insufficiently reported.
Why it matters
Risk prediction could support earlier identification of people at risk of postpartum PTSD, complementing screening approaches with limitations. The pooled results suggest potential, but substantial heterogeneity and gaps in validation and calibration mean that reported accuracy may not translate reliably across settings. Establishing clinical utility and performance in external populations is important before these models can inform care.
What to keep in mind
Considerable heterogeneity (I² = 89.7%) limits confidence in pooled performance and generalizability; calibration and external validation were insufficiently reported.