Maternity care presents a unique and under-explored opportunity for technological innovation. Each year, millions of births take place in the UK and USA (3.6M livebirths in the USA and 0.6M in England and Wales in 2024), which follow care pathways that are often highly structured and repetitive. The main outcomes (positive or negative) are tracked across the system, and much of the data required to assess risk is collected routinely, both over the lifetime of the patient and throughout pregnancy.
Maternity care is also a field that attracts significant public scrutiny. Litigation costs run into the billions each year (£27.4 billion since 2019 in the UK alone), and poor outcomes in pregnancy and childbirth are often debated in the public sphere.
Despite this apparent structure, and the clinical, financial, and societal motivation to improve, care pathways are often still poorly implemented, patient datasets are still siloed, and technological support is lacking or outdated. Some of these challenges were detailed in two recently published seminal reports in the UK (the Amos and Ockendon reports).
At Ellescope, we focus on improving maternity care pathways by taking a holistic view of the expectant mother. Our approach is to build risk assessment and pathway recommendation models that proactively probe available data, identify relevant risk factors, and apply the best available clinical guidelines to support informed decision-making.
Data access and integration
The first challenge in the way of any data-driven AI innovation is obtaining sufficient data. In healthcare, and especially within SaMDs, the problem is not only securing sufficient training data but also ensuring we can access data of the same quality in production.
Our innovation was born within a healthcare organisation, and by consistently partnering with frontline organisations that can act as sponsors for data access, we have established trusted testing environments for every iteration of our technology.
Noise and variability in clinical data
Healthcare data is inherently noisy: different care providers have different standards in their use of clinical terminology, and staff members within those organisations adhere to those standards to varying degrees. Treating every individual clinical term as a separate category can lead to data sparsity and weak predictions. Over-grouping terms into broader categories can produce results that don’t discriminate between clinical outcomes and are no longer useful.
To tackle this challenge, we are building ML models that understand the relationships between clinical terms and can quantify the similarity between patients with similar medical history. Furthermore, alongside more specific prognostic models that predict the occurrence of specific adverse outcomes, we include broader risk severity models that are more robust to noise yet still provide actionable insights.
Mismatch between models and clinical needs
A further technical challenge is the frequent mismatch in the literature between what predictive models deliver and what healthcare professionals need. This often stems from insufficient workflow analysis, which fails to fully capture how patients move through the health system, and from a lack of decision cost-benefit analysis that accounts for the priorities, constraints, and pressures under which clinicians operate.
Our product was developed through multiple partnerships with healthcare organisations and tested in real-world settings from the very beginning. We have worked with GP networks, NHS trusts, charities, community services, and health commissioners: this has allowed us to identify clinical priorities early on, understand how they change in different care settings, and build a product that that adapt to those.
Conclusion
Despite these challenges, Ellescope is well positioned to lead in this space. Ellescope possesses the expertise, partnerships, datasets, and real-world evidence that many other organisations struggle to build.
Maternity care combines structured data, defined pathways, and a strong societal imperative to reduce adverse outcomes. With the right technical, clinical, and human-centred approaches, it represents one of the most promising areas for impactful innovation in healthcare.
