Development and evaluation of an algorithm for computer analysis of maternal heart rate during labor

Abstract

Background: Maternal heart rate (MHR) recordings are morphologically similar and sometimes coincident with fetal heart rate (FHR) recordings and may be useful for maternal-fetal monitoring if appropriately interpreted. However, similarly to FHR, visual interpretation of MHR features may be poorly reproducible.

Methods: A computer algorithm for on-line MHR analysis was developed based on a previously existing version for FHR analysis. Inter-observer and computer-observer agreement and reliability were assessed in 40 one-hour recordings obtained from 20 women during the last 2h of labor. Agreement and reliability were evaluated for the detection of basal MHR, long-term variability (LTV), accelerations and decelerations, using proportions of agreement (PA) and Kappa statistic (K), with 95% confidence intervals (95% CI). Changes in MHR characteristics between the first and the second hour of the tracings were also evaluated.

Results: There was a statistically significant inter-observer and computer-observer agreement and reliability in estimation of basal MHR, accelerations, decelerations and LTV, with PA values ranging from 0.72 (95% CI: 0.62-0.79) to 1.00 (95% CI: 0.99-1.00), and K values ranging from 0.44 (95% CI: 0.28-0.60) to 0.89 (95% CI: 0.82-0.96). Moreover, basal MHR, number of accelerations and LTV were significantly higher in the last hour of labor, when compared to the initial hour.

Discussion: The developed algorithm for on-line computer analysis of MHR recordings provided good to excellent computer-observer agreement and reliability. Moreover, it allowed an objective detection of MHR changes associated with labor progression, providing more information about the interpretation of maternal-fetal monitoring during labor.

Keywords: Agreement; Cardiotocography; Computer analysis; Fetal heart rate; Maternal heart rate; Reliability.