This blog is the second in a three-part blog series written by Babylon's Head of Corporate Development, Brian Skiba. The series explores how Babylon’s global technology may be effective in reducing health costs of populations we serve while increasing the quality of care.
In the first blog in this series, we identified the HealthGraph, a key technology building block for Babylon’s approach to population risk assessment and stratification. Importantly, it emphasized the ability for Babylon to bring in data from numerous disparate data sources previously siloed to data lakes and analytical platforms. This helps us develop a 360-degree view of each Babylon member available for our products and AI platform to leverage in real time. Once the data has been streamed and ingested, the next step is to carry out automated assessment through AI and begin predicting outcomes to members most at-risk and enriching that information back into our members’ Health Graphs.
Babylon has made significant investments to accelerate the development of its end-to-end AI for healthcare. By investing in the development of its (bespoke) data and AI platforms, Babylon has applied to healthcare best-in-class frameworks that power some of the leading technology companies familiar to us all today. This approach ensures that its products, services - and ultimately its members - benefit from the scalable development, monitoring of deployed solutions, ensuring its AI continues to push the state of the art to deliver deep predictive insights to improve clinical and operational outcomes.
Purpose of Machine Learning in analysis and discovery
Machine Learning (ML) is a subset of Artificial Intelligence (AI). Machine learning methods learn models from available data to solve for specific clinical (and operational) outcomes such as predicting disease risk, identifying diagnostic codes from medical records or identifying the presence/absence of diagnostic markers from images of the skin. Leveraging large, complex, real-world data - from labs, monitoring devices and health records - ingested/connected via our real-time health graph, to deliver continual, precision-based insights.
Babylon’s HealthIQ delivers predictive insights relating to risk for clinical (disease risk and progression) and operational outcomes e.g. risk stratification, which are used to identify and recommend optimal evidence based interventions with the Babylon Advisor.
Moving from rule-based population risk assessment to an ML-based approach and why it matters
The majority of existing methods for assessing population risk still rely on a retrospective assessment of risk, which categorize members into different risk categories (e.g. High, Medium or Low) based on thresholds applied to different clinical criteria. Whilst effective in providing a coarse-grained understanding of risk at the population level, these methods are prone to mean reversion… Due to the availability of large, complex, real-world health records (as discussed with the Babylon individual Health Graph), ML-based approaches deliver an optimal approach, and ensure that available up-to-date data can be utilized - including data relating to social determinants of health (SDOH) where available - to deliver insights - including recommendations for reducing risk/improving health - that are tailored to the specific population under management.
Relying on legacy retrospective assessments can miss rapidly emerging issues brought on by changes in contemporary health practices (i.e. an increased use of prescription opioids for pain medication) or the likes of a global pandemic.
As such, everything from risk prediction to treatments are likely to evolve from a machine learning (ML) based approach. Rather than assume that historical rules are (a) comprehensive, and (b) the best foundation for prediction, learning from data is essential to address specific clinical outcomes, as defined by our care advisors and population health management teams.
Heath Graph and HealthIQ – Data, AL and ML as part of an approach to improving population health
Babylon’s Health Graph builds upon an event-based architecture that captures complex real world data captured in real-time, in addition to historic health data.
We then leverage our best in class AI / ML models which are developed / trained, deployed, and monitored at scale using our AI platform, which applies and extends the best features of platforms from leading technology companies to HealthIQ.
Data, AI and ML are central to our approach to building a platform for healthcare that delivers upon our mission of accessible, affordable healthcare for all.
Try Babylon today
Babylon offers high-quality, 24/7 comprehensive health care. Let us help:
Learn more about our services
