The study, published in Frontiers in Drug Safety and Regulation, was led by Omkar Basnet of Golden Community in Lalitpur, Nepal, together with collaborators at Uppsala University and the University of Gothenburg. It took place in seven public referral hospitals, each representing one of Nepal’s provinces, over an eight-month period from October 2020 to June 2021. The team enrolled 1,091 women during a control period in which nurses checked fetal heart rates intermittently with the Moyo device, and 8,687 women during an intervention period in which the same device was used continuously until delivery. In total, 32,269 deliveries occurred in the participating hospitals during the study window, from which the analyzed cohort was drawn after exclusions and consent procedures.

The technology at the center of the trial is deliberately simple. Moyo is a portable fetal heart rate monitor built around a nine-crystal Doppler ultrasound sensor that straps onto the mother’s abdomen. Unlike a handheld Doppler, which a midwife must pick up and position every 30 minutes during the first stage of labor and every 15 minutes during the second, Moyo tracks the fetal heart rate continuously and displays it through sound waves, numerical readouts, and color coding. Crucially, it also sounds an alarm if an abnormal heart rate, defined as above 160 or below 110 beats per minute, persists for three minutes. That alarm allows a single midwife, who in some Nepali hospitals may be responsible for roughly 900 deliveries a year, to keep watch over several laboring women at once.

The results on timing were striking. The median time from the first detection of an abnormal fetal heart rate to delivery fell from 132 minutes under intermittent monitoring to 73 minutes under continuous monitoring. After adjusting for maternal and fetal complications and for intrauterine resuscitation, the researchers calculated a 30 percent reduction in that interval, corresponding to an adjusted reduction of 66.5 minutes. The time from the last abnormal heart rate check to delivery shortened by about 43 minutes, and the total time from first monitor placement to delivery fell by 67 minutes. Only the interval from first monitor placement to first abnormal detection, which shrank by 35.5 minutes, missed the threshold of statistical significance.

Because the exposure was not randomized, the team took unusual statistical care. They used generalized linear models for the time intervals, then applied causal mediational modeling to separate the direct effect of monitoring type on newborn outcomes from its indirect effect through intrapartum resuscitation, treating intrapartum complications as confounders. Risk estimates were standardized with the Mantel–Haenszel homogeneity method. The researchers also ran sensitivity analyses using propensity scores, including inverse probability weighting, and found that the direction and magnitude of the associations held, though some effect estimates were attenuated. Meta-regression across the seven hospitals showed low heterogeneity for the key delivery-timing outcomes, suggesting the findings were not driven by one unusually well-run facility.

The downstream clinical outcomes are where the findings become most consequential. Non-crying at birth, a clinical sign that a newborn needs resuscitation within the first fifteen seconds of life, occurred in 11.5 percent of babies in the control group versus 10.0 percent in the intervention group. After mediational adjustment, the risk was 31 percent lower with continuous monitoring, a statistically significant result with a risk ratio of 0.69. The need for neonatal ventilation, at 5.1 percent versus 4.3 percent, likewise showed a 31 percent lower adjusted risk. Intrapartum stillbirth fell from 4.5 to 3.0 per 1,000 births, a 35 percent crude reduction, but with such a rare outcome the confidence interval stretched across the null, and the authors caution that the analysis may simply have been underpowered to confirm a mortality effect.

The mechanism the authors propose is physiologically straightforward. When a fetus experiences hypoxic stress during labor, prolonged delay before intervention allows injury to accumulate. Continuous monitoring compresses two critical delays: the time it takes to notice that something is wrong, and the time it takes the clinical team to act once the alarm sounds. In the Nepali hospitals studied, where a single provider may face eight laboring women and two active deliveries at any moment, intermittent checks every half hour are difficult to sustain. The device’s histogram review of the previous 30 minutes of heart rate tracing, combined with its automatic alarm, effectively gives each midwife a continuous surveillance capability that would otherwise require far more staff.

The findings align with evidence from Tanzania, where randomized trials comparing continuous Moyo monitoring with intermittent Doppler assessment found earlier detection of abnormal fetal heart rates and fewer newborns requiring resuscitation in the continuous group. Earlier studies that compared cardiotocography with intermittent auscultation in high-resource settings often failed to show improvements in perinatal death, which makes the consistency of the low-resource findings notable. The Nepali authors suggest that the difference may lie in the alarm function and the portability of the device, features designed specifically for settings where continuous electronic monitoring with wired equipment is impractical.

The study is not without limitations, and the authors are candid about them. Blinding was impossible, monitoring frequency itself differed by design between the two periods, and some time intervals were recorded through routine clinical documentation, which can vary across settings. There was also selection bias: women who were eligible but not enrolled differed in obstetric characteristics from those who were, though the mediational analysis adjusted for measured differences. The small control group of 1,091 women, relative to 8,687 in the intervention arm, further limits the precision of comparisons. The authors also note that the high uptake of continuous monitoring may partly reflect the presence of research staff supporting the quality improvement process, and that neonatal mortality itself was not among the measured outcomes.

Even with those caveats, the policy implications are hard to ignore. Poor-quality intrapartum care has been linked to nearly six million deaths over the past 25 years, and although institutional births in Nepal rose from 9 percent in 2001 to 78 percent in 2024, only about a third of women receive high-quality care during labor. The authors argue that continuous fetal heart rate monitoring with low-cost portable devices is a practical, sustainable investment: it reduces the workload per provider, shortens the window of fetal vulnerability, and lowers the number of newborns who arrive needing resuscitation. In settings where the ratio of births to birth attendants is extreme, a device that watches every labor continuously and shouts when something goes wrong may be the closest thing to adding staff without hiring them.

Subject of Research: Implementation of continuous versus intermittent fetal heart rate monitoring during labor and its effects on intrapartum care and neonatal outcomes in Nepal

Article Title: Evaluating the implementation of intermittent versus continuous fetal heart rate monitoring on intrapartum care and early neonatal outcomes using causal mediational modeling

Article References: Basnet, O., Gurung, R., Bhattarai, P., Hofert, M. M., Pokharel, S. M., Paudel, P., & K. C., A. (2026). Evaluating the implementation of intermittent versus continuous fetal heart rate monitoring on intrapartum care and early neonatal outcomes using causal mediational modeling. Frontiers in Drug Safety and Regulation, 6, Article 1755487. https://doi.org/10.3389/fdsfr.2026.1755487

Image Credits: AI Generated

DOI: 10.3389/fdsfr.2026.1755487

Keywords: fetal heart rate monitoring, Moyo device, intrapartum care, neonatal resuscitation, Nepal, stillbirth, causal mediational modeling, low-resource settings, obstetrics, newborn health, Doppler monitoring, maternal health

Researchers at Weill Cornell Medicine searched PubMed and Embase for studies published between January 2015 and December 2025, combining clinical terms for obstetric anesthesia and delivery with methodological terms for artificial intelligence and machine learning. From 3,135 records identified, two reviewers with consensus adjudication by board-certified anesthesiologists narrowed the field to 172 studies that met inclusion criteria: 154 original research studies, 17 systematic or narrative reviews, and a single case report. Twelve additional records, including nine registered clinical trials and three workforce-focused studies, were analyzed separately. The timing tells its own story. Roughly 69 percent of the included publications appeared after 2022, and more than half were published in 2024 or 2025 alone, a pace that mirrors the public release of large language models and the broader AI boom in medicine.

The geographic spread was equally revealing. The United States contributed 40 studies and China 35, followed by Israel, Italy, India, Singapore, and Spain, with eighteen multinational collaborations rounding out the picture. When the researchers classified the studies by machine learning approach, classification tasks dominated, followed by regression, neural networks, and large language models. Unsupervised learning, reinforcement learning, and dedicated radiomics pipelines appeared only rarely. The work clustered into distinct domains: maternal risk and complications prediction led with 61 studies, delivery mode prediction followed with 43, patient education and conversational AI claimed 15, fetal heart rate tracing analysis 12, imaging and radiomics 8, and provider education and simulation just 2.

The single largest domain, maternal risk and complications prediction, targets some of the most feared events in obstetrics. Postpartum hemorrhage accounts for 11 percent of U.S. maternal mortality and remains a leading preventable cause of maternal death, making accurate risk assessment a matter of life and death. Machine learning models predicting hemorrhage and transfusion have reported widely varying performance, with areas under the curve ranging from 0.67 to 0.99. Large language models have even been used to extract risk factors from prenatal notes and electronic health records, identifying features like multiple gestation and prior cesarean delivery. But the review uncovered a sobering technical problem: differing hemorrhage definitions across health systems can dramatically degrade model performance. One model’s C-statistic fell from 0.93 to 0.57 when applied to a contemporary cohort rather than an early-2000s cohort, a collapse that underscores how fragile these tools can be outside their training environment.

Placenta accreta spectrum, a life-threatening condition in which the placenta invades the uterine wall, illustrates both the power and the limits of current approaches. Although its prevalence is low at roughly 0.17 percent, it drives severe peripartum morbidity, with pooled incidences of 52.2 percent for hysterectomy and 46.9 percent for transfusion among affected women. Deep learning and MRI-based radiomics have shown genuine promise for earlier recognition, with one 2025 study using 15 radiomic features reaching an area under the curve of 0.95 for predicting the condition. Yet the same study found that random forest and k-nearest neighbor models using identical variables achieved AUCs of only 0.704 to 0.863, a reminder that algorithm choice matters enormously and that headline numbers can mask substantial intra-study variation.

Hypertensive disorders of pregnancy represent another major target. Pre-eclampsia affects 555 per 10,000 deliveries, and hypertensive disorders account for 6.8 percent of pregnancy-related deaths. A 2024 pilot study used routine biomarkers and clinical characteristics to predict severe pre-eclampsia requiring intensive care, achieving an AUC of 0.91 with an XGBoost model in an 81-patient cohort. Aspartate aminotransferase, uric acid, and body mass index emerged as the strongest indicators of severe disease. Meanwhile, peripartum cardiomyopathy, which complicates roughly 1 in 2,000 pregnancies worldwide and carries 76-fold higher odds of cardiogenic shock, has become the focus of one of the field’s most compelling real-world tests: a recent trial found that an AI-guided digital stethoscope doubled the diagnosis of pregnancy-related cardiomyopathy compared with usual care, with a number needed to screen of 47 to identify one new case of perinatal left ventricular systolic dysfunction.

Beyond risk prediction, AI is being applied to the mechanics of birth itself. Approximately 30 percent of the reviewed articles addressed prediction of delivery mode and timing, spanning emergency cesarean, successful trial of labor after cesarean, and vaginal versus cesarean delivery. Because about one in three U.S. births occurs by cesarean and most of those are unplanned, carrying higher complication rates than planned surgery, accurate forecasting could transform labor ward planning. Machine learning models for intrapartum delivery mode prediction showed variable performance, with AUCs from 0.69 to 0.932, and one study found a k-nearest neighbor model predicting vaginal delivery four hours after admission with an AUC of 0.812, outperforming traditional logistic regression. For trial of labor after cesarean, three machine learning models significantly outperformed the established Maternal Fetal Medicine Unit calculator, though the margins were narrow and external validation remained limited. The Artificial Intelligence Dystocia Algorithm, designed to assess transverse fetal head position and cesarean risk, reached 95.5 percent accuracy on just 135 cases, a tantalizing but preliminary result.

Fetal heart rate tracing analysis, the backbone of intrapartum monitoring and the primary indication behind 27 percent of cesarean deliveries in the United States, has attracted its own wave of machine learning. Deep learning and neural network models for cardiotocography analysis have reported strong performance, with AUCs between 0.87 and 0.987, and one model categorized delivery mode from fetal heart rate signals with an AUC of 0.967. Imaging and radiomics studies clustered around two themes: MRI and ultrasound models for placenta accreta, and procedural ultrasound for anesthesia itself, including software for lumbar level identification and a neural network for real-time cricothyrotomy landmark identification using wireless handheld ultrasound, relevant to failed-airway preparation on labor and delivery. Notably, fusion models combining radiomic features with clinical and laboratory data consistently outperformed either data type alone, suggesting the most useful clinical applications will not rely on imaging in isolation.

The most rapidly emerging domain is patient education and conversational AI, with 15 of 16 studies published in 2024 or 2025. Multiple evaluations of ChatGPT and other chatbots found that responses to common questions about labor epidural analgesia were generally accurate and appropriately framed but often exceeded the recommended sixth-to-eighth grade reading level, and occasionally included factual errors or omitted clinically important caveats. A 2025 evaluation of ChatGPT-5 in obstetric and gynecologic emergencies highlighted reliability gaps that persist even in the newest models. Innovative tools like the Mothers’ Assistant for Labor analgesia, an interactive avatar-based education platform, point toward more sophisticated applications, but no included study prospectively evaluated whether large language model-supported education actually changed patient knowledge, anxiety, or satisfaction compared with standard counseling. On the provider side, a 2025 study comparing four large language models on complex anesthesia decision-making across 30 high-risk scenarios found that all four generated structured plans, with DeepSeek-R1 performing best in the obstetric subgroup, though the study was limited by small sample size and no outcome-based validation.

Perhaps the most important finding of the entire review is what is missing. Of the 154 original research studies, only 22, or 14 percent, reported external validation in an independent dataset. Prospective evaluations were rare, and almost none described integration into clinical workflow, meaning that even the most accurate models have rarely been tested in the hands of actual anesthesiologists caring for actual patients. The nine registered clinical trials identified cover narrow topics such as labor analgesia sequelae, lung ultrasound interpretation, and automated spinal landmark identification, a pipeline that looks thin relative to the sheer volume of model development activity. The authors also flag serious ethical concerns: algorithmic biases arising from unbalanced training datasets, historical health inequities, race-based clinical adjustments, and opaque black-box methodology could exacerbate the very racial and ethnic disparities in maternal outcomes that persist in the United States, where maternal mortality reached 32.6 deaths per 100,000 live births in 2022. The path forward, the researchers argue, requires a deliberate shift from building models to validating them externally, testing them prospectively, and embedding them responsibly into clinical workflows, with obstetric anesthesiologists serving as designers, evaluators, and stewards of the tools that will increasingly shape their practice. Until then, the extraordinary numbers on paper remain exactly that: numbers on paper.

Subject of Research: Applications of artificial intelligence and machine learning in obstetric anesthesia, including maternal risk prediction, delivery mode forecasting, fetal monitoring, and patient education

Article Title: Artificial intelligence and obstetric anesthesia: a scoping review

Article References: Givens, M. F., Rodriguez Caceres, E., Jotwani, R., Rubin, J. E., Goldstein, P. A., & White, R. S. (2026). Artificial intelligence and obstetric anesthesia: a scoping review. Frontiers in Anesthesiology, 5, Article 1896595. https://doi.org/10.3389/fanes.2026.1896595

Image Credits: AI Generated

DOI: 10.3389/fanes.2026.1896595

Keywords: artificial intelligence, machine learning, obstetric anesthesia, maternal mortality, postpartum hemorrhage, placenta accreta, pre-eclampsia, fetal heart rate monitoring, large language models, external validation, clinical decision support, scoping review

The research introduces what its authors call a cardiotocography age, an AI-derived measure modeled loosely on the idea of biological age. Rather than asking whether a fetal heart tracing looks reassuring or non-reassuring in the moment, the algorithm was trained to estimate where a given pregnancy sits along a developmental trajectory inferred from thousands of prior recordings. The gap between the cardiotocography age the model assigns and the actual gestational age of the fetus serves as the digital biomarker. When that gap widens in one direction, the researchers report, it flags pregnancies at elevated risk of adverse outcomes such as hypertensive complications, fetal growth restriction, and emergency operative delivery.

The underlying insight is that a fetal heart rate pattern is not merely a snapshot of moment-to-moment oxygenation but a composite signature of autonomic nervous system maturation, placental function, and intrauterine environment. Cardiologists learned something analogous when they discovered that a machine can read an electrocardiogram and estimate a patient’s physiological age; hearts that look older than their chronological age tend to belong to people at higher cardiovascular risk. The npj Women’s Health study transfers that logic to the fetus, arguing that the heart rate variability, baseline trends, accelerations, and decelerations inscribed on a cardiotocography strip encode a quantifiable signature of developmental maturity.

Technically, the approach rests on deep learning applied to raw cardiotocographic signals rather than to the summary indices clinicians typically compute. Conventional analysis reduces a tracing to a handful of features, such as baseline heart rate, variability bands, and the count of accelerations and decelerations, each defined by consensus criteria drafted before modern computing. These parameters are individually weak predictors, and large trials of computerized cardiotocography interpretation have historically failed to show that automated alerts improve outcomes, partly because the features were chosen by human intuition decades ago. The new model instead learns its own representations directly from the continuous signal, using architectures capable of detecting temporal patterns too subtle or too diffuse for the standard feature set to capture.

Training such a model requires careful handling of a notorious problem in fetal monitoring: the labels are noisy and the populations are imbalanced. Adverse outcomes are, fortunately, rare relative to the number of recordings collected, which means an algorithm can achieve deceptively high accuracy by simply predicting that everything is fine. The researchers addressed this through appropriate validation design, holding out complete cohorts for testing and evaluating the model against outcomes that had not yet occurred at the time of the recording. This temporal separation matters enormously. Many earlier machine learning studies in obstetrics tested their algorithms on tracings taken during labor and then predicted outcomes of that same labor, a design that risks simply detecting the crisis already underway. The cardiotocography age concept is explicitly forward-looking, derived from antenatal recordings and validated against complications arising later in the pregnancy.

The reported performance suggests the biomarker has genuine discriminative power. Pregnancies whose AI-assigned cardiotocography age lagged behind their gestational age clustered disproportionately among those that later developed hypertensive disorders of pregnancy, delivered growth-restricted infants, or required urgent cesarean sections for fetal compromise. The signal held up after adjustment for conventional risk factors, meaning the model was not merely rediscovering maternal age, body mass index, or pre-existing hypertension, information clinicians already collect. Instead, the fetal heart tracing itself appeared to contain additive information, a conclusion that, if replicated across diverse populations, would justify incorporating the measure into routine antenatal assessment alongside ultrasound and blood pressure monitoring.

Why should a heart rate pattern recorded weeks before symptoms predict a disorder like preeclampsia, which is conventionally understood as a placental disease manifesting through maternal hypertension? The most plausible explanation lies in the shared biology of placental insufficiency. When the placenta fails to establish adequate perfusion, the fetus adapts, redistributing blood flow and modulating autonomic tone in ways that subtly reshape heart rate dynamics long before growth falters or maternal blood pressure rises. These adaptations leave traces in the variability and periodicity of the fetal electrocardiographic signal captured at the maternal abdomen. In effect, the algorithm learns to recognize the physiological fingerprint of a struggling placenta, converting it into a number that clinicians can act on.

The digital biomarker framing carries implications beyond any single disease. Precision-medicine researchers have increasingly argued that pregnancy is an ideal proving ground for continuous, physiological, computationally interpretable data, because the obstetric population is monitored intensively and the outcomes are discrete and consequential. A validated cardiotocography age could function as a longitudinal vital sign for the fetus, tracked across visits like blood pressure or fetal growth percentiles. Deviation from the expected trajectory would trigger targeted surveillance: serial growth scans, Doppler studies of umbilical artery flow, or low-dose aspirin prophylaxis in cases where risk models currently rely on coarse demographic and biophysical screening. The measure could also serve as an endpoint in clinical trials of interventions designed to improve placental function, replacing months of waiting with a computable surrogate readout.

Significant hurdles remain before the metric enters practice. The study’s training data reflect the equipment, acquisition protocols, and population characteristics of the institutions that contributed recordings, and cardiotocographic signals are sensitive to gestational age in ways that demand robust calibration across the full range of pregnancy. Regulatory pathways for adaptive algorithms in medical devices are still maturing, and clinicians will reasonably ask what they should do with a risk score in the absence of trials demonstrating that acting on it improves outcomes. History counsels humility: electronic fetal monitoring itself was adopted worldwide on physiological plausibility alone, and only later did randomized trials reveal that its widespread use had reduced neonatal seizures while dramatically increasing cesarean rates without lowering cerebral palsy incidence. A new biomarker built on the same signal must clear a higher evidentiary bar than its predecessor did.

Yet the conceptual achievement is difficult to overstate. The study demonstrates that an ordinary, ubiquitous clinical recording, taken with equipment already installed in virtually every maternity unit on earth, contains information about future pregnancy health that decades of expert interpretation never systematically extracted. If subsequent multi-center validation confirms the findings across populations and devices, the cardiotocography age could join the growing family of AI-derived organ ages, alongside brain age, heart age, and retinal age, that are reshaping how medicine quantifies risk. For a field in which the central tools have remained essentially unchanged since the 1960s, the prospect of turning the labor ward’s oldest instrument into a predictive window on the pregnancy’s future marks a genuinely new chapter in fetal medicine.

Subject of Research: An AI-derived cardiotocography age biomarker that predicts future adverse pregnancy outcomes from routine fetal heart rate recordings

Article Title: AI-derived cardiotocography age as a digital biomarker for predicting future adverse pregnancy outcomes

Article References: Gu, J., Lin, Z., Ma, J., Wang, J., Zhang, L., Bai, R., Tu, Z., Jiang, Y., Xie, D., Zhou, Y., Liu, G., & Hong, S. (2026). AI-derived cardiotocography age as a digital biomarker for predicting future adverse pregnancy outcomes. npj Women's Health. https://doi.org/10.1038/s44294-026-00165-4

Image Credits: AI Generated

DOI: 10.1038/s44294-026-00165-4

Keywords: artificial intelligence, cardiotocography, digital biomarker, pregnancy outcomes, fetal heart rate monitoring, preeclampsia, fetal growth restriction, deep learning, maternal-fetal medicine, predictive analytics, placental insufficiency, npj Women's Health