Time-surrogate variables enhance the association between cardiotocographic features and intrapartum hypoxic-ischemic...

Interruptions to the flow of oxygenated blood to the fetal brain during labor can lead to hypoxic-ischemic encephalopathy (HIE). Timely interventions for suspected hypoxemia are crucial to prevent neurological injury. Prior cardiotocography (CTG) based prediction systems have focused only on the end of labor, when preventive measures are unlikely to be effective. We previously demonstrated that accounting for proximity to birth improved the association between CTG features and outcome. Since proximity to birth cannot be known prospectively, other time-surrogate variables (TSVs) may be more clinically appropriate. We analyzed intrapartum data from 174,186 births (152,761 vaginal, 21,425 Caesarean): 171,246 healthy, 2,636 with perinatal acidosis, and 304 with confirmed HIE. Classical and novel CTG features were extracted across labor durations up to 72 hours. This dataset represents one of the largest cohorts of intrapartum CTG data to date by participant count and signal length. We evaluated four candidate TSVs alongside the reference proximity to birth to evaluate their ability to enhance the association of CTG features and the development of HIE. To do so, we applied information-theoretic methods to quantify their contribution to the association of fetal outcomes with CTG features. Incorporating TSVs significantly increased the normalized mutual information (NMI) between CTG features and outcomes, improving NMI by over 20% for the ten most informative baseline features. Cumulative contraction time (CCT) and the time from labor onset (TLO) were the most effective TSVs in both vaginal and Caesarean delivery cohorts, which demonstrates their robustness regardless of delivery mode. However, it is more practical to track TLO in clinical settings than to continuously monitor contractions. To conclude, TLO is the most suitable TSV for prospective intrapartum CTG evaluation. Incorporating it may substantially enhance the performance of automated systems for early detection of intrapartum HIE.