A Documentary and Numerical Audit of Reported Incremental Heart Rate Variability Prediction in the 3PAD Study

This Zenodo record contains Part I of an ongoing documentary and numerical research audit entitled “A Documentary and Numerical Audit of Reported Incremental Heart Rate Variability Prediction in the 3PAD Study.” The audit examines selected published numerical summaries and methodological descriptions from Eriksson et al. (2026), “Exploring early and late pregnancy heart rate variability as incremental predictors of postpartum depression and anxiety symptoms,” published in Psychoneuroendocrinology, volume 188, article 107813. The central question is whether the published numerical summaries of the incremental predictive contribution of heart rate variability (HRV) are internally reconcilable, and whether the publicly documented methods are sufficiently detailed to permit independent reproduction. The report compares the performance differences implied by Supplementary Table 9a with the pooled differences reported in Supplementary Table 9b. It evaluates whether observed differences can be explained by the displayed rounding precision, and assesses the transparency of the reported HRV imputation, Random Forest modelling, feature-selection, and internal-validation procedures. The numerical reconciliation finds that seven of twelve comparisons exceed the approximate rounding tolerance of 0.0105: depression MSE; state-anxiety MSE, RMSE, and MAE; and trait-anxiety MSE, RMSE, and MAE. These findings identify a numerical-reconciliation question. They do not, by themselves, prove a statistical error, methodological defect, or misconduct. The published pooled results indicate statistically supported MAE improvements for depression and state anxiety, but not trait anxiety. All R-squared confidence intervals include zero. The reported incremental effects are modest, and statistical significance does not automatically establish external validity or clinical utility. This report was prepared and authored by Omar Magdy Mohamed Abdelkader, Egypt. The author is solely responsible for the scope, interpretation, conclusions, and limitations of this documentary audit. An AI assistant was used as a research-support tool for literature discovery, document review, numerical checking of published summary values, organization of the research findings, and manuscript preparation. The AI system is not listed as an author. The sole named human author remains responsible for verifying the sources, numerical transcriptions, references, and final manuscript. This is Part I of an ongoing research investigation. It is not the original 3PAD study, an independent replication, a completed clinical trial, or an original participant study. No participant-level data were used or generated.