What is epistemic trust?
Imagine for a moment that computer interpretation of the CTG has become so good that it now significantly outperforms skilled maternity professionals ability to detect impending fetal hypoxia. As a maternity professional, you can’t see what the computer is seeing, because your brain simply lacks the capacity to interpret the trace the same way the computer does. What criteria would need to be met for you to reach a point where you would stop looking at the CTG, and simply wait for the computer to tell you when intervention was advised? This concept is known as epistemic trust – having faith that the knowledge being communicated is trustworthy and fit for purpose.
New research
A team of researchers in the United Kingdom recently published research looking at what would be needed for midwives and obstetricians to trust a computer interpretation of a CTG recording (Dlugatch et al., 2024). They interviewed five midwives and eight medical practitioners providing obstetric care, none of whom had previous exposure to a system using computer interpretation of the CTG during labour. Having reflected on how they used information from the CTG in their clinical decision making, the principles of how a particular artificial intelligence system providing CTG interpretation (OxySys 3.0)were explained. They were then asked what they would need in order to trust the information the system was providing them.
What would it take to trust the computer?
Participants described four main criteria they would need to achieve epistemic trust:
- Accuracy and efficiency. Participants wanted a system that would accurately predict when an intervention would lead to better outcomes and for this to happen in a time-effective manner. Even if this could be achieved, participants still believed the final decision, and accountability for outcomes, rested with them rather than a computer. One obstetrician noted: “let’s say if you had to go to court you can’t say, ‘Oh the computer told me it’s okay’“.
- Individualisation. The ability to input clinical information pertaining to an individual woman and her fetus and to have that included in the decision-making algorithm was considered important. Participants rejected a standardised approach to CTG interpretation in the absence of personalising information. Even with this, participants considered that developing a sense of connectedness to the woman in labour and the instinctive sense of something being not quite right would remain an important aspect of care.
- Outcomes rather than reputations. Participants were asked whether it was important who was developing the system – for example whether it as a university based team or a for-profit organisation. This was considered less important than the question of whether the system delivered accurate information.
- Transparency, ethics, and rigour. While the issue of who developed the system was not of great significance to participants, how the system was developed was important. They wanted reassurance that high research standards had been employed when testing the system and that research was done ethically; and they wanted to be able to verify this for themselves.
How far away from having a trustworthy system are we?
The authors point out that there is no widely accepted process to validate artificial intelligence systems for use in healthcare settings. The most recent systematic review of randomised controlled trials of computer interpretation of the CTG continues to demonstrate that, when compared to standard CTG monitoring, such systems offer no improvement in perinatal outcomes (Tsipoura, et al., 2023). We are nowhere near being able to demonstrate the required accuracy and efficiency of such systems.
Things I worry about…
The idea of computer interpretation of the CTG as a form of artificial intelligence or decision support concerns me deeply for several reasons. I remain unconvinced that such systems will ever work. It seems overly optimistic to expect human trained and developed computer systems will be able to over come the inherent flaws in the theoretical assumptions behind fetal heart rate monitoring. If 60 years of human expertise has not managed to get CTGs to improve perinatal outcomes – why do we think computers will do better?
I’m concerned about the large sums of money currently being invested by technology developers to come up with a better CTG machine, when approaches to fetal monitoring that aren’t based around the fetal heart rate remain under investigated.
I wonder what women and those close to them will make of such systems? Will it be enough that because they trust their care provider, and the care provider trusts the CTG interpreting computer, they are then comfortable giving the computer permission to make recommendations about their care? This current research is a good start, but we should not stop at asking maternity professionals what they need from such systems. A co-design process that includes users of maternity care and centres their values, priorities, and concerns is required, and I’m yet to see evidence that this is happening.
It is important to remember that once it has been introduced, we lose sight of technology very quickly, and the work it is doing becomes largely invisible. Ask yourself – when was the last time you pulled out a watch and counted the fetal heart rate while looking at the numbers on the CTG screen to be sure the computer in your CTG machine is accurately recording the actual heart rate? And did you then look to see whether the number on the screen is being plotted correctly on the graph on the display? Have you ever gone one step further and checked that the information displayed in the birth room is identical to the display at the central fetal monitoring station (if one is in use at your place of work)?
I’ll bet good money you have never done any of these things. You just assume it works.
How quickly will the fancy behind the scenes bits in a computer interpretation system disappear from view and everyone jump to action when the computer says intervention is needed – without any critical thought about whether it is true?
What impact might this have on obstetric authority?
CTG use has always reinforced obstetric authority. When there’s any doubt about how to interpret a CTG – the voice that “wins” will (almost always) be an obstetric one. What if I’m wrong and CTG interpretation by computer algorithms is proven to be stunningly more reliable than human interpretation? Relying on the computer’s recommendation would rapidly become established as the professional and legal expectation for duty of care. Rather than being criticised for having done as the computer said to, the obstetrician quoted earlier in the story would be facing legal scrutiny for not having done what the computer said.
In this new future – obstetricians trained in CTG interpretation would no longer be needed. Their presence in birth suite could be replaced by employing a technician to input clinical data, adjust the sensors, and notify the obstetric staff. Sure – obstetric staff would still have other roles to play, but given that CTG abnormalities constitute the main driver of obstetric intervention once labour is underway, the impact would be significant.
If obstetricians role in relation to the CTG were reduced to their technical skill in performing instrumental births and caesarean sections when the system said to – what impact would that have on the reputation and authority of the profession?
Sign Up for the BirthSmallTalk Newsletter and Stay Informed!
Want to stay up-to-date with the latest research and course offers? Our monthly newsletter is here to keep you in the loop.
By subscribing to the newsletter, you’ll gain exclusive access to:
- Exciting Announcements: Be the first to know about upcoming courses. Stay ahead of the curve and grab your spot before anyone else!
- Exclusive Offers and Discounts: As a valued subscriber, you’ll receive special discounts and offers on courses. Don’t miss the chance to save money while investing in your professional growth.
Join the growing community of birth folks by signing up for the newsletter today!
References
Dlugatch, R., Georgieva, A., & Kerasidou, A. (2024, Jan 6). AI-driven decision support systems and epistemic reliance: a qualitative study on obstetricians’ and midwives’ perspectives on integrating AI-driven CTG into clinical decision making. BMC Medical Ethics, 25(1), 6. https://doi.org/10.1186/s12910-023-00990-1
Tsipoura, A., Giaxi, P., Sarantaki, A., & Gourounti, K. (2023). Conventional cardiotocography versus computerized CTG analysis and perinatal outcomes: A systematic review. Maedica: A Journal of Clinical Medicine, 18(3), 483-489. https://doi.org/10.26574/maedica.2023.18.3.483
- Five questions to ask yourself when deciding about fetal monitoring in labour
- Why I’m not on a guideline writing group
Categories: CTG, EFM, New research
Tags: Artificial intelligence, computerised interpretation, decision making, Decision support systems, epistemology, trust
Birth Small Talk 
Dear Kirsten, I love your approach and remembering questioning Fetal monitoring 40 years ago .. as it was creeping in. I was a member of ARM and challenged my midwifery tutors as to why there was no research comparing pinards and Ultra sound. As a student, it could have affected my grades if I asked – does any listening improve the outcome. Latterly, I listen in if the woman requests it or if I think what I might hear will confirm my thoughts that we should be transferring in!
Love Chris Chris Warren Midwife York Storks Midwifery Support 01423360460
LikeLike