Development and validation of a prognostic tool for antenatal depressive symptoms: Statistical analysis plan for a two-stage model integrating self-reported and biomarker measures
Abstract
Background: Antenatal depression is common, under-detected, and harms mothers and infants. Pakistan has Asia’s highest perinatal depression prevalence and an estimated national economic burden of US$16.5 billion. About 9.7 million Pakistani women become pregnant each year, and roughly 3.6 million may experience antenatal depression. Routine detection relies on self-report scales (e.g., Edinburgh Postnatal Depression Scale [EPDS]); whether biomarkers add prognostic value is largely untested. Objective: Develop and validate a multivariable prognostic tool (MPT) to predict clinically significant depressive symptoms (EPDS ≥ 10) and assess the added value of biomarker data. Methods: We will use secondary data (N = 1,621) from the Pakistani Pregnancy Cohort at the Aga Khan Hospital for Women and Children, Sindh. Self-reported predictors across eight domains (demographic, socioeconomic, behavioural, mental health, clinical history, pregnancy-related, and social network) plus biomarkers will inform MPT development. In Stage 1 we will compare supervised classifiers (regularized regression, decision trees, support vector machines, and neural networks) to logistic regression. Models will be trained with stratified five-fold cross-validation and nested L1 selection within folds to retain the smallest predictor set within 0.01 area under the receiver operating characteristic curve (ROC-AUC) of the full mode. External cross-validation on an independent hold-out sample will report ROC-AUC, area under the precision-recall curve (relative to prevalence), calibration, sensitivity, specificity, and predictive values. Decision thresholds will be set on training data and applied unchanged. In Stage 2 we will build a biomarker-augmented model and compare it to the optimized self-report model using paired AUC tests and reclassification metrics; exploratory clustering will identify participant subgroups. Pre-specified criteria (ROC-AUC ≥ 0.70; calibration slope 0.8–1.2; sensitivity ≥ 0.70; positive net benefit) will determine progression to prospective validation studies in new samples. Expected Outcomes: A pragmatic self-report prognostic tool and evidence on whether biomarkers meaningfully improve prediction and justify added complexity.
Related
- Interpretable Machine Learning Prognostic Model Integrating Inflammatory-Nutritional Biomarkers for Post-Prostatectomy Biochemical Recurrence: Development, Internal Validation, and Translational Target Exploration.
- Development and Initial Validation of a Statistical Image Analysis Framework for Quantitative Bone SPECT/CT in Patients with Prostate Cancer
- Development and validation of a postoperative prognostic nomogram for early-stage pancreatic cancer
- Development and internal validation of a prognostic model for cancer-specific survival in elderly patients with metastatic non-small cell lung cancer: a SEER-based study
- Development and external validation of an XGBoost-based prognostic model for small cell lung cancer with brain metastases
BibTeX
@misc{ibrahim2026development,
title = {{Development and validation of a prognostic tool for antenatal depressive symptoms: Statistical analysis plan for a two-stage model integrating self-reported and biomarker measures}},
author = {Sheriff Ibrahim and Dr. Shahirose Sadrudin Premji and Shehzad Ali and Jaime Delgadillo},
howpublished = {PsyArXiv},
year = {2026},
month = nov,
doi = {10.31234/osf.io/hcgfx_v1},
url = {https://osf.io/preprints/psyarxiv/hcgfx_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Development and validation of a prognostic tool for antenatal depressive symptoms: Statistical analysis plan for a two-stage model integrating self-reported and biomarker measures
AU - Sheriff Ibrahim
AU - Dr. Shahirose Sadrudin Premji
AU - Shehzad Ali
AU - Jaime Delgadillo
PY - 2026
DA - 2026/11/02
PB - PsyArXiv
DO - 10.31234/osf.io/hcgfx_v1
UR - https://osf.io/preprints/psyarxiv/hcgfx_v1/
AB - Background: Antenatal depression is common, under-detected, and harms mothers and infants. Pakistan has Asia’s highest perinatal depression prevalence and an estimated national economic burden of US$16.5 billion. About 9.7 million Pakistani women become pregnant each year, and roughly 3.6 million may experience antenatal depression. Routine detection relies on self-report scales (e.g., Edinburgh Postnatal Depression Scale [EPDS]); whether biomarkers add prognostic value is largely untested. Objective: Develop and validate a multivariable prognostic tool (MPT) to predict clinically significant depressive symptoms (EPDS ≥ 10) and assess the added value of biomarker data. Methods: We will use secondary data (N = 1,621) from the Pakistani Pregnancy Cohort at the Aga Khan Hospital for Women and Children, Sindh. Self-reported predictors across eight domains (demographic, socioeconomic, behavioural, mental health, clinical history, pregnancy-related, and social network) plus biomarkers will inform MPT development. In Stage 1 we will compare supervised classifiers (regularized regression, decision trees, support vector machines, and neural networks) to logistic regression. Models will be trained with stratified five-fold cross-validation and nested L1 selection within folds to retain the smallest predictor set within 0.01 area under the receiver operating characteristic curve (ROC-AUC) of the full mode. External cross-validation on an independent hold-out sample will report ROC-AUC, area under the precision-recall curve (relative to prevalence), calibration, sensitivity, specificity, and predictive values. Decision thresholds will be set on training data and applied unchanged. In Stage 2 we will build a biomarker-augmented model and compare it to the optimized self-report model using paired AUC tests and reclassification metrics; exploratory clustering will identify participant subgroups. Pre-specified criteria (ROC-AUC ≥ 0.70; calibration slope 0.8–1.2; sensitivity ≥ 0.70; positive net benefit) will determine progression to prospective validation studies in new samples. Expected Outcomes: A pragmatic self-report prognostic tool and evidence on whether biomarkers meaningfully improve prediction and justify added complexity.
ER -