PeriGen CNO Featured in Femtech World

We need to learn to embrace new AI technologies if we want to improve perinatal outcomes.

There are several current threats to perinatal patient safety, a significant one being the nursing shortage and nurse knowledge gap. Healthcare facilities may struggle to maintain adequate nurse to patient ratios, and the majority of caregivers work twelve-hour shifts.

Human factors, such as environmental distractors, cognitive bias, sustained attention, and normalisation of deviance pose ongoing threats to situational awareness. The promotion of situational awareness through the use of an AI-driven early warning system cannot be overstated in the obstetric arena, where concerning trends over time, sometimes subtle, risk being missed.

Many of today’s healthcare technologies utilise AI. AI is a broad umbrella that encompasses the field of developing computers that are capable of simulating human intelligence.  If you had a bucket of AI concepts it would include things such as machine learning, natural language processing, data analytics, predictive analytics and computer vision, among others.

It is important to outline the difference between the concepts of decision-making algorithms and generative analytics. Decision-making algorithms are designed to make optimal decisions based on input data (typically large, if not massive, amounts of data) and a set of predefined rules or criteria. We see this often in healthcare; recommended treatment plans based on input data, for example, patient symptoms, medical history, lab results and imaging results.

Generative analytics, on the other hand, focuses on generating new insights from existing data and is not widely utilised in healthcare. Although the use of generative analytics in healthcare may hold promise, it faces challenges such as the need for large and diverse datasets, validation of the generated outputs, and ethical considerations around data privacy and patient consent.

If you were to take a healthcare AI deep dive, you would find that its implementation and use is growing exponentially and expected to continue to grow at that rapid pace.

As a long-term obstetrics nurse and chief nursing officer for an advanced healthcare technology company in the perinatal space, I am versed in the application of AI to improve childbirth outcomes.

It has been well established that greater than fifty per cent of birth related foetal injury events during labour and delivery are preventable. Of these events there is a sizeable portion that are related to interpretation of the electronic foetal monitor (EFM) tracing. For a system to get different results we must look at redesigning the system. There is currently only one available platform in the US utilising decision-making algorithms for EFM tracing, and the labour progression curve.

This platform uses a colour-coded system to draw attention to concerning trends developing in the foetal tracing, in contraction activity, and the labour progression curve. It is an early warning system for mums and babies.

Early warning systems are broadly used in many (non-obstetrical) service lines in healthcare facilities as a tool identifying patients at risk for deterioration. They bring a general increase in situational awareness to the frontline staff, which in turn promotes earlier recognition, notification and intervention.

The use of early warning systems has demonstrated a reduction in serious safety events. However, early warning systems have yet to be widely deployed in perinatal services lines; this is long overdue.

In the US, when a birthing person is in labour, the vast majority will have a foetal monitor applied to gain a continuous tracing of the foetal heart rate, uterine activity and maternal vital signs. The foetal heart rate tracing produced by the monitor is interpreted by clinicians utilising standardised language for the elements in the tracing.

These elements, when aggregated, lead to classification in three categories; category I defined as normal, category III defined as abnormal, and category II defined as indeterminate. Most birthing patients, eighty-four per cent, will be in category II at some point during their labour. This is problematic as it is an extremely broad category and contains tracings with varying acuity from low concern to high concern.

There has been the development of tools, category II algorithms, to assist in navigating the management of category II tracings. The Vigilance system utilises a model based on the work of an EFM scientist to classify category II tracings into three distinct colour categories. As automated software it is not dependent on the clinician to take the time to apply the elements in the current foetal tracing to a category II algorithm to determine the severity.

Further, when a birthing person is receiving an oxytocin (Pitocin) infusion, the foetal monitoring tracing must be evaluated and interpreted every fifteen minutes. Application of a category II algorithm would therefore need to be conducted every fifteen minutes while the EFM is demonstrating a category II tracing. This is the perfect example of AI automation and technology’s ability to bring efficiency to repeatable, data driven healthcare workflow cycles.

Interpretation of EFM tracings was historically, and is still currently in many facilities, a human visual interpretation. Therefore, open to human subjectivity. In other words, different highly skilled, experienced clinicians in foetal monitoring may look at the same EFM tracing and have different interpretations. This brings variance into care.

AI-driven technology brings an objective, automated, repeatable interpretation that mitigates the variances in human interpretation. This feature has been significant in supporting improved communication between caregivers.

At this time, we need to embrace new technologies that utilise AI to improve outcomes and focus on further optimisation and the use of data for reporting efficiency. We have to do better. There is no going back.

Originally published at femtechworld.co.uk