Discovery Health is using data science, machine learning and artificial intelligence tools to create a new system to provide individualised health care to its adult medical aid members.
Premised on the principle of “hyper-personalisation”, Personal Health Pathways uses the idea of a digital twin, or a “clinic twin” in the company’s jargon, to give real-time recommendations to members with the idea of creating a proactive, curated approach to health care.
According to Discovery Health CEO Ron Whelan, who was speaking to TechCentral in a recent interview, the use of advanced machine learning and AI models is allowing a targeted approach, which improves outcomes for patients while, over time, reducing costs for Discovery Health.
“Healthcare is inordinately complex and has always run on a set of algorithms, rules, protocols and guidelines. But with the advent of artificial intelligence, we are able to see new patterns and trends. More importantly, we are able to use those patterns on an individualised basis to deliver health care that is more precise,” Whelan said.
“We saw an opportunity to use Discovery Health’s big data set, together with an advanced AI model, to create these precise and personalised pathways; this was not possible before the advent of artificial intelligence.”
According to Whelan, a major benefit of AI-infused health care is the ability to administer the right treatment at the right time. This avoids “wasting time and minimises the risk of adverse events from any particular drugs or interventions”. This is because the system reduces much of the trial and error that doctors go through when diagnosing patients and prescribing treatments.
Clinical twin
A member’s clinical twin is not a static entity. Each time a recommended action is performed (or not performed), it is fed back into the system, which incorporates the new information into the model to suggest the next best action.
Whelan said the model creates many possible actions and ranks them according to their efficacy, but this alone is not what determines whether a recommendation will be made. Another factor is the likelihood that a medical aid member will perform the action, or stick to a given routine. The highest-value action with the greatest chance of being completed is then recommended to the user. Users are rewarded when their actions are completed, and just like the recommendations, the rewards are also individualised.
“The same action can have a different reward value to different members according to their circumstances,” said Whelan.
Read: Discovery Vitality ditches Pick n Pay for Checkers
Given the sensitivity of health data, Whelan said Discovery Health had to take strict precautions in how it used the data mined from its client base, which consists of nearly three million South African beneficiaries.
To comply with the Protection of Personal Information Act, the data used to create the causal inference model and the others supporting was anonymised. Children are also not included in the programme, in part because the causal inference models based on child health-care data are still under development.
Users interact with Personal Health Pathways via the Discovery Health app or the Discovery Corporate app. As with any machine learning algorithm, the model gets better the more data it is fed.
“The pathway is updated in a live and intelligent way because it adds to the data set, enriches it and then suggests the next best appropriate action. It’s a completely closed-loop and iterative cycle,” he said.
The importance of this closed loop to a member’s livelihood and longevity makes the accuracy of the inferences it makes of critical importance – a matter of life and death, in fact. This makes ensuring that the data fed into the model is accurate absolutely critical.
Data is fed into the system from a variety of sources, including wearable devices such as smart fitness bands, member claims submissions and electronic health records. Owning and using a wearable device is one of the prerequisites for registering for Personal Health Pathways, and the system is compatible with various brands including Apple, Samsung, Garmin and Suunto.
According to Whelan, exercise band data adds a dynamic element to Discovery Health’s models because, for the first time, patients can be monitored outside of the clinical environment through their devices.
360-degree view
“The challenge with hospital data is you only have data from the time of the admission to the end of the admission. It is certainly difficult to know from hospital data what happened before and after that. The power of our data set is we have a 360-degree view of any member through the wearables,” said Whelan. – © 2025 NewsCentral Media
Get breaking news from TechCentral on WhatsApp. Sign up here.