Daily PubMed evidence board
Prediction models for postpartum post-traumatic stress disorder: a systematic review and meta-analysis.
This systematic review and meta-analysis looked at studies that developed or tested models to predict which people might develop post-traumatic stress dis…
Signal score64Research triage score
CertaintyLow to moderate (limited by high risk of bias, sparse external validation, and heterogeneity)Verify in full text
PMID42218472Source identifier
Research triage, not medical advice
Do not use this summary, score, or benefit-cost estimate to diagnose, treat, prescribe, or change care without reviewing the full study and consulting qualified professionals.
Verify before acting
Check full-text methods, eligibility, outcomes, risk of bias, harms, conflicts, funding, replication, and applicability.
Plain-English signal
This systematic review and meta-analysis looked at studies that developed or tested models to predict which people might develop post-traumatic stress disorder after childbirth (PP-PTSD). The authors found 16 studies with 29 models (many from China; many using machine learning). All studies had high risk of bias, and only seven did external validation. For the externally validated logistic models, the pooled discrimination (AUC) was 0.86, suggesting reasonably good ability to tell higher- from lower-risk people, but methodological weaknesses mean these models need better-designed studies and more external validation before they can be recommended for clinical use.
Why it matters
- Postpartum post-traumatic stress disorder (PP-PTSD) can cause lasting maternal morbidity; identifying people at high risk could enable targeted monitoring or early intervention.
- This review summarizes available prediction models (including machine learning models) and their external validation performance, which informs whether any models are ready for clinical use or further validation.
- The pooled AUC of externally validated logistic models (0.86) suggests potentially useful discrimination, but methodological concerns may limit trust and transportability to other settings or populations.
Primary outcomes
- Predictive performance (AUC) of externally validated models for postpartum post-traumatic stress disorder
Effect summary
Among seven studies that performed external validation, pooled AUC for externally validated logistic regression models was 0.86 (95% CI: 0.79-0.90) with moderate heterogeneity (I2 = 70.8%). However, all included studies were judged to have high risk of bias, mainly due to data source quality and inadequate reporting.
Benefit-cost lens
| Quick take | Promising discrimination reported for some externally validated logistic models (pooled AUC 0.86), but pervasive high risk of bias and limited external validation mean benefits are uncertain; do not deploy clinically without local validation and cost-effectiveness assessment. |
|---|---|
| BCR anchor | 2 |
| Time horizon | 3 |
| Discount rate | 0.03 |
| Assumptions | Assessment based only on PubMed metadata and abstract; assumes pooled AUC refers to logistic regression models with external validation as stated. Full text needed to extract calibration, risk thresholds, population characteristics, and harms. |
Benefit-cost fields are assumptions-based unless explicitly source-derived. Treat them as prompts for deeper economic review.
Risk of bias
| Tool | PROBAST (as reported in paper) and rapid-abstract-screen |
|---|---|
| Verdict | High risk of bias |
| Notes | The abstract reports that all included studies were judged high risk of bias (issues: data source quality, inadequate outcome and analysis reporting). Assessment here is limited to the abstract and metadata; full-text review needed to confirm details and to appraise calibration and applicability factors. |
Harms, equity, conflicts & implementation
| Implementation | Full-text review; extraction of model predictors, coefficients, population characteristics; local external validation; assessment of calibration and decision thresholds; workflow, training, and cost analysis before any clinical implementation. |
|---|---|
| Equity impact | Unclear from abstract; equity effects depend on whether models were developed and validated in diverse populations and on access to follow-up interventions. Full-text subgroup analyses required. |
| Harms | Not reported in abstract. Potential harms include false reassurance or unnecessary interventions for false positives; requires extraction from full text and consideration of downstream interventions. |
| Funding | XJK2303B General science and technology program of Quanzhou Medical College; 2021N137S Quanzhou guiding science and technology program |
| Registration | CRD42025630468 |
| Replication | Unknown from automated PubMed triage; replication not reported in abstract. |
Source links — verify original
Use PubMed and full-text links to confirm methods, population, outcomes, harms, conflicts, and applicability.