Healthcare IT News | Bill Siwicki | April 27, 2022

“We have shoved so many things to the bedside nurse, between monitoring and documentation I am not sure how they manage the patient anymore.”

Since the 1970s, electronic fetal monitoring has been the go-to technology in perinatal obstetrics for the detection of fetal distress during labor. Although the technology has changed slightly in the ensuing years, the demands on perinatal nurses have escalated exponentially. Nursing burnout and the pandemic-driven shortage have exacerbated the situation.

Meanwhile, labor and delivery nurses with varying levels of experience must track and document as many as 10 different characteristics every 15 to 30 minutes, depending on the stage of labor – all with the added emotional overlay of treating multiple patients at once in a highly demanding, often stressful environment.

Matthew Sappern is CEO of PeriGen, a health IT company that applies artificial intelligence to improve safety in childbirth. Healthcare IT News sat down with Sappern to learn how AI and mobile technology are combining to help address these issues, as well as provide better access to safe labor and delivery for underfunded rural hospitals and newly acquired hospitals that are increasingly cutting NICU and labor and delivery departments.

Q. What’s a high-level history of EFM? How has it helped, and how have related demands on perinatal nurses changed?

A. The history of EFM is rife with controversy and debate. I believe this has more to do with how EFM data is being consumed, rather than issues with the technology itself.

Fetal monitoring itself dates back to the 19th century, and, while the equipment has advanced significantly, our collective challenge today is how to better interpret and use this data, overcoming human factors that contribute to avoidable adverse outcomes.

What began as a practice of using rudimentary stethoscopes to monitor the fetal heart advanced considerably through the 20th century, when the first bedside EFM was deployed in 1968. This monitor automatically generated fetal heart rate tracings drawn on a long strip of paper.

Frankly, the technology introduced in 1968 has not changed much, though now the tracings are displayed digitally, integrated into a surveillance program and interfaced with medical records.

From auscultation to EFM the clinical goal remains the same – identify fetuses with increased risk of injury during childbirth without facilitating an excessive number of unnecessary interventions. But the practice has been peppered with inconsistent outcomes and controversy since its introduction.

While EFM can indicate troubling fetal heart rate patterns, it is a nonspecific test. It does not definitively identify the time where injury occurs, and it does not isolate and identify the cause.

This has perpetuated questions as to its value. For decades, medical associations have endeavored to improve its usability and relevance through standardizing terminology and establishing management guidelines and protocols based on EFM data.

There have been notable improvements to outcomes, such as halving the rate of perinatal death or HIE from the 1980s to today, but that same time frame saw a significant increase in the rate of cesarean sections.

The disturbing reality is that interpreting fetal heart tracings is still too often considered more art than science. A single borderline fetal strip shown to a group of trained professionals will likely elicit multiple contradictory reactions.

Ultimately, despite well-intentioned guidelines, hours of training and fetal monitoring certification programs, the practice of interpreting tracings often falls victim to subjectivity, normalization of deviance and other human factors.

Over the last few decades, we have come to realize that these human factors and system failures play a substantial role in adverse outcomes. In childbirth specifically, loss of situational awareness and the miscommunication and delayed intervention that flow from it are present in nearly half of preventable birth-related brain injuries despite decades of focus on training programs.

Today, nurses on the labor floor navigate through a complex environment of increased documentation requirements, a cacophony of threshold alerts that have become meaningless, and a shortage of personnel and experience.

As well, maternal profiles are more problematic in general, with higher percentages of obesity, diabetes and other comorbidities that must be managed closely. Nurses must keep a vigilant watch over their patients, help their colleagues in emergencies, and defend their clinical insight to doctors and residents.

All of this with the backdrop of operating in the most litigious service line in a hospital, where their judgment might be called into question in a courtroom five years down the road – light-years from the controlled chaos of the day and lined up against witnesses who will try to discredit them.

As to the future of EFM, technology holds the most promise to mitigate many of the pitfalls that contribute to preventable injury. There are three tracks to pursue.

First, invest in advancements in sensors to better detect how a baby is tolerating labor. Many companies are pursuing wearable sensors to measure fetal heart rate and uterine activity. Although they provide for greater mobility and potential monitoring at home, they have not yet been shown to be more accurate than bedside monitors nor lessen birth-related injuries. This conclusion is not surprising, as they measure the same things that bedside monitors do.

Second, use today’s tracings in different and better ways. Use technology as a complement to human assessments to protect against inconsistency and subjectivity. Recognize that human assessment can be inconsistent, especially when carried over long periods of time in the presence of fatigue, distraction and normalization of deviance.

On this second note, employ data visualization techniques to improve situational awareness. Key components for good situational awareness are quick access to aggregated, relevant information and projection of what is likely to happen.

Projection is critical in obstetrics, to permit intervention before injury occurs and recovery is unlikely. Specialized displays can consolidate critical data in a single overview showing patients’ course over many hours, marking deviations from expected norms with color-codes. Such tools enable clinicians to quickly see the degree and duration of abnormalities and trends over time.

Also on this second note, reduce alarm fatigue by enhancing alarm meaningfulness. Alarming on simple, rudimentary threshold alerts has resulted in widespread fatigue and dismissal by clinicians. Amplifying this frustration, existing fetal tracing categorization methods are designed to be simple to remember and use, but this simplicity results in poor discrimination and very frequent alarms.

Success here means establishing evidence-based notifications based on the degree and duration of abnormalities. The capacity of computers along with automated analysis can make a more nuanced categorization of the tracing where notifications can be focused on more clinically meaningful findings.

And third, apply machine learning techniques to create better, more specific and timelier predictors of fetal injury or its immediate precursors. Health systems have exceptional data repositories and machine learning technologies are increasingly powerful.

It is very possible that combining clinical information such as EFM data with machine learning techniques can produce a better way to identify the fetus with impending injury. Such results would be easy to push to the bedside or wherever convenient for the clinicians.

Q. What must labor and delivery nurses track and document today, and what is the environment for them like?

A. I’ll start with the environment, which, given the shortage of experienced nurses, seems unsustainable without eventual technical intervention. “Controlled chaos” is a popular descriptor. Peaks and valleys of adrenaline, as any labor can go from normal to emergent quickly, which is why it is so critical to assess trends as opposed to thresholds.

A service line leader with a long history in bedside nursing, perinatal quality and nurse education recently commented to me, “We have shoved so many things to the bedside nurse, between monitoring and documentation I am not sure how they manage the patient anymore … sometimes they just shut down.”

Against that backdrop, the documentation requirements are basically the following: varying by levels of risk in the labor, nurses are assessing the fetal strip and uterine activity every 15 minutes and documenting the same at least every 30 minutes. Documentation requirements also include patient activity, tolerance of labor, pain level, meds, if administered, labs and interventions for labor management.

Q. How can artificial intelligence and mobile technologies help?

A. Machine learning, a type of AI, refers to mathematical techniques that use existing data to generate algorithms to make predictions on future data: they “learn” from training data. Neural networks – one machine learning approach in widespread use across many domains – is an approach that can approximate any mathematical function to characterize complex relationships between predictors and responses.

In contrast, rule-based systems prevalent today are little more than scoring systems developed by a clinical consensus. Most have not even been tested in a rigorous fashion to measure their performance.

By applying machine learning techniques to the big data sets available in labor and delivery today, one can develop predictors that are based on contemporary data and take into account many clinical factors like maternal age and coexisting medical conditions.

In addition, they can consider evolving factors like changing EFM patterns, the state of labor progression or a reduction in staffing levels. Digital signal processing can “see” and provide fetal tracing features that are not readily apparent to the human eye.

Machine learning systems are not limited by assumptions embedded in current rule-based systems. Furthermore, their performance can be calculated and compared to rule-based methods to actually measure – not speculate – about the increments of improvement.

Mobile technologies can do the same for healthcare as they have for communication in general. They can bring information to clinicians in various locations, and so they can interact with the data and with the bedside clinicians.

Pictures are worth a thousand words and miscommunication to offsite clini

cians is a common problem in birth injury. Projecting the fetal strip anywhere, with analytical markings to facilitate assessment and consensus, is a critical ingredient in an effective clinical infrastructure.

Twitter: @SiwickiHealthIT
Email the writer:
Healthcare IT News is a HIMSS Media publication.