From spotting damage to preventing it: AI at the start of labour
Our invited commentary argues for catching at-risk babies early enough to prevent injury, by combining several kinds of data.
The question
When a baby is in trouble during labour, the problem is often recognised only after damage has been done, when it is too late to change the outcome. Can we recognise the babies at risk much earlier?
What this is
This is an invited Comment, not a research study. It accompanies a study called Fit4Labour by Bozhilova and colleagues, which uses the first 20 minutes of computerised fetal heart rate monitoring, together with routine risk factors, to flag serious fetal compromise at the start of labour.
What we argue
We argue the field should shift from diagnosing damage to finding at-risk babies early enough to prevent injury. We see Fit4Labour as one layer of a wider picture that could add three more kinds of data developed mostly in isolation: linked health records over time, the mother’s oral health and oral microbiome, and continuous wearable monitoring. We also point to the AI challenges of combining them, such as knowing when a model is reliably extrapolating, and whether a signal is truly biological or only a stand-in for factors like smoking and deprivation.
Why it matters
Moving from reacting to damage to preventing it could change outcomes, but only if the AI behind it is dependable and fair.