Computerised analysis of intrapartum fetal heart rate patterns and adverse outcomes in the INFANT trial

Abstract

Objective To assess if a computerised decision support system reliably identified abnormal fetal heart rate (FHR) patterns in fetuses with adverse neonatal outcomes in the INFANT trial, and whether its use reduced substandard care. Design Prospective cohort study within a randomised controlled trial. Setting Twenty‐four maternity units in the UK and Ireland. Population or sample A total of 46 614 labours between January 6 2010 and August 31 2013 in the INFANT trial. Methods Panel review of intrapartum and neonatal care in infants with adverse outcome, and an assessment of the effectiveness of computerised interpretation of fetal heart rate in reducing substandard care. Descriptive analysis of other factors associated with adverse outcome. Main outcome measures Incidence and detection rate of abnormal fetal heart rate patterns, other characteristics associated with perinatal adverse outcome, and frequency of substandard care. Results Computer interpretation of FHR patterns was deemed to be completely valid in only 24 of 71 (33.8%) cases of adverse outcome. On a scale of 0–10 (completely invalid to completely valid), 28 cases (39.4%) had a score of 6 or less, mainly due to lack of recognition of decelerations (15 cases), or reduced variability (seven cases), or failure to recognise tachysystole (five cases). There were multiple associated factors that modified the clinical assessment of FHR patterns. There was substandard care in 45/71 cases (63%). Conclusion A significant proportion of abnormal fetal heart rate patterns were not detected accurately by computer analysis, and its use did not reduce the incidence of substandard care. Funding UK National Institute for Health Research Health Technology Assessment Programme (project number 06.38.01). Tweetable abstract Improved recognition of abnormal fetal heart rate patterns is insufficient to reduce the incidence of substandard care.

Authors' reply re: Computerised analysis of intrapartum fetal heart rate patterns and adverse outcomes in the INFANT trial

“…However, the purpose of the INFANT trial was to address the specific hypothesis that some intrapartum care givers are unable to interpret fetal heart rate (FHR) patterns and this hypothesis was not supported by the results. Moreover, abnormal CTG patterns were common in labours with no adverse outcome (73.9% had at least one yellow alert and 10.5% at least one red alert) . Instead, our review suggests that in most cases of avoidable substandard care, insufficient weight was given to the interpretation of FHR abnormality in the context of often multiple coexisting risk factors, for example fetal growth restriction, preterm labour, oxytocin administration, meconium staining of the amniotic fluid, prolonged labour and pyrexia.…”

Re: Computerised analysis of intrapartum fetal heart rate patterns and adverse outcomes in the INFANT trial

“…A recent paper and the one‐sided BJOG debate unreservedly concluded that ‘computerising intrapartum cardiotocography (CTG) does not improve predictive value’ because the benefits are unproven by statistical significance tests. Remarkably, more than 800 scientists and statisticians have now called for the retiring of the ‘dichotomous’ nature of statistical significance tests, questioning ‘How statistics so often leads the scientists to deny differences that are clear to see or experience?’ This seems particularly relevant to trials on CTG and ancillary technologies .…”

“…The INFANT trial showed that the computer‐aided analysis was equivalent to visual CTG interpretation, but (crucially) performed by rigorously trained clinicians under ‘trial conditions’ with not only the benefit of the Hawthorne effect but also ‘crossover‐learning’ from the computerised CTG arm. The paper and the debate make no mention that the ‘crossover‐learning’ has been a major criticism/invalidation of the trial results. Hence, it is possible that the computer‐aided CTG could perform uniformly better than median clinical judgment in the absence of Hawthorne effect and in places around the world with less rigorous training or different staffing resources/structures.…”

“…It is disappointing but not surprising that the retrospective review of CTG in 71 select cases with the hindsight of known poor perinatal outcome detected three‐times as many abnormal cases/CTGs as compared with study and control groups looking at 46 614 cases under practice conditions . The accompanying editorial rightly calls CTG interpretation an ‘art’, but it is not an uncommon gift.…”

“…It is longstanding knowledge that a subset of higher‐risk cases (quite a substantial group, not a small minority) will have more abnormal CTGs and adverse outcomes. Conversely, a cohort of cases with adverse outcomes and suboptimal CTG interpretation would naturally contain a higher proportion of higher‐risk cases . But this does not necessarily mean that clinicians are more specifically failing to recognise abnormal CTGs in higher‐risk cases.…”