Abstract / Summary
Background/Objectives: A previous publication reported that routinely measured covariates explained approximately 10% of fetal mechanical PR interval (mPR) variance. This secondary analysis examined how explanatory information is distributed among correlated predictors, their incremental contributions beyond fetal heart rate (FHR), and performance across alternative model specifications. Methods: We analysed 903 complete cases with normal fetal echocardiographic findings. Predictors comprised FHR, gestational age (GA), biparietal diameter (BPD), fronto-occipital diameter (FOD), pulmonary valve diameter (PVD), and umbilical artery pulsatility index (UA PI). Hierarchical ordinary least squares regression was complemented by general dominance analysis across all 720 predictor entry orders, bootstrap uncertainty estimates, and sensitivity analyses using shared ten-fold outer cross-validation splits. Results: The full-model apparent R2 of 0.0986 was consistent with the previously reported low explanatory capacity. The additional analyses showed that FHR accounted for 51.23% of full-model explained variance in general dominance analysis. Adding GA after FHR increased R2 by 0.0237; subsequent additions of BPD, FOD and PVD, and then UA PI, increased R2 by 0.0061 and 0.0037, respectively (p = 0.112 and p = 0.057). Full-model pooled out-of-fold R2 was 0.0826. Exploratory gradient boosting increased apparent R2 to 0.1722 but yielded a lower out-of-fold R2 of 0.0554. Conclusions: This secondary analysis extends the previous report by quantifying incremental and shared explanatory information and comparing apparent fit with internal validation performance. FHR accounted for the largest average contribution across predictor entry orders, while the additional biometric and Doppler blocks provided limited incremental information after FHR and GA. The evaluated alternative specifications did not substantially improve internal validation performance. The low overall explained variance is consistent with the previous publication and is not a new finding. Model-unexplained variance cannot be attributed specifically to intrinsic atrioventricular conduction properties, because its biological and measurement-related components cannot be distinguished using the available data.