Georgieva, A; Redman, CWG; Papageorghiou, AT
(2017)
Computerized data-driven interpretation of the intrapartum cardiotocogram: a cohort study.
Acta Obstet Gynecol Scand, 96 (7).
pp. 883-891.
ISSN 1600-0412
https://doi.org/10.1111/aogs.13136
SGUL Authors: Papageorghiou, Aris
|
PDF
Accepted Version
Available under License ["licenses_description_publisher" not defined]. Download (303kB) | Preview |
Abstract
INTRODUCTION: Continuous intrapartum fetal monitoring remains a significant clinical challenge. We propose using cohorts of routinely collected data. We aim to combine non-classical (data-driven) and classical cardiotocography features with clinical features into a system (OxSys), which generates automated alarms for the fetus at risk of intrapartum hypoxia. We hypothesize that OxSys can outperform clinical diagnosis of "fetal distress", when optimized and tested over large retrospective data sets. MATERIAL AND METHODS: We studied a cohort of 22 790 women in labor (≥36 weeks of gestation). Paired umbilical blood analyses were available. Perinatal outcomes were defined by objective criteria (normal; severe, moderate or mild compromise). We used the data retrospectively to develop a prototype of OxSys, by relating its alarms to perinatal outcome, and comparing its performance against standards achieved by bedside diagnosis. RESULTS: OxSys1.5 triggers an alarm if the initial trace is nonreactive or the decelerative capacity (a nonclassical cardiotocography feature), exceeds a threshold, adjusted for preeclampsia and thick meconium. There were 187 newborns with severe, 613 with moderate and 3197 with mild compromise; and 18 793 with normal outcome. OxSys1.5 increased the sensitivity for compromise detection: 43.3% vs. 38.0% for severe (p = 0.3) and 36.1% vs. 31.0% for moderate (p = 0.06); and reduced the false-positive rate (14.4% vs. 16.3%, p < 0.001). CONCLUSIONS: Large historic cohorts can be used to develop and optimize computerized cardiotocography monitoring, combining clinical and cardiotocography risk factors. Our simple prototype has demonstrated the principle of using such data to trigger alarms, and compares well with clinical judgment.
Statistics
Actions (login required)
![]() |
Edit Item |