A new AI “digital twin” model suggests that loneliness, insomnia and poor mental health can dramatically raise a person’s risk of type 2 diabetes. The work could help doctors spot danger earlier and design more equitable prevention programs.
Loneliness, sleepless nights and poor mental health may be doing more than just making people feel bad in the moment. New research suggests they can dramatically increase the chances of developing type 2 diabetes — and an advanced “digital twin” artificial intelligence tool is helping scientists see how.
In a study led by Anglia Ruskin University (ARU) with collaborators at Cranfield University, the University of Portsmouth and Intelligent Omics Ltd, researchers used AI to simulate how changes in people’s daily lives could shape their long-term risk of type 2 diabetes.
Instead of relying on blood tests, step counts or data from wearable devices, the model drew only on behavioural, lifestyle and psychosocial information. It used health and lifestyle records from 19,774 adults in the UK Biobank, who were followed for up to 17 years.
The team’s digital twin system, developed by ARU, created a virtual version of each participant’s health profile, then ran “what if” scenarios to see how different patterns of stress, sleep and diet might play out over time.
The results, published in the journal Frontiers in Digital Health, were striking. Under the model’s assumptions, experiencing loneliness, insomnia or poor mental health was each linked to an estimated 35‑percentage‑point increase in the risk of developing type 2 diabetes in the future. When all three occurred together, the model predicted a 78‑percentage‑point jump in absolute risk.
The researchers noted that these patterns likely reflect the body’s response to chronic stress. Over time, elevated stress hormones, inflammation and disrupted sleep can interfere with how the body uses insulin and manages blood sugar, setting the stage for type 2 diabetes.
Those stress-related factors also appeared to shape what people ate. The AI uncovered strong links between loneliness, sleep disruption and mental health problems and higher consumption of salt, sugary cereals and processed meats — foods that are already associated with a higher risk of type 2 diabetes. Even small shifts in diet reinforced risk levels in the simulations.
The model also suggested that cheese might have some protective effect, but that this benefit shrank when mental health issues were present, underscoring how powerful psychosocial factors can be.
Type 2 diabetes affects more than 500 million people worldwide and is driven largely by preventable factors such as diet, physical activity, stress and social conditions. Unlike type 1 diabetes, which is an autoimmune disease, type 2 develops gradually and is often linked to lifestyle and environment.
Yet health professionals have long struggled to predict who will go on to develop type 2 diabetes early enough to intervene effectively. Many current tools focus on simple measures such as body mass index, age and blood pressure.
“Type 2 diabetes is a rising global health concern which we know is heavily influenced by lifestyle. However, current risk prediction models rely on BMI, age and blood pressure, which over-simplify this disease and overlook the more complex interconnected behavioural and emotional factors that precede and shape the onset of the condition,” co-author Barbara Pierscionek, the Deputy Dean for Research and Innovation in the Faculty of Health, Medicine and Social Care at ARU, said in a news release.
The new study takes a different approach by centering those behavioural and emotional factors. Instead of asking only how heavy or how old someone is, the AI looks at how they sleep, how connected they feel and what kinds of stress they live with — and then simulates how those patterns ripple through the years.
Lead author Mahreen Kiran, a postgraduate researcher at ARU, emphasized that these kinds of variables need to be built into health data from the start.
“This study shows the importance of including behavioural and psychosocial variables such as loneliness, sleep disruption and mental health history within health datasets used for risk prediction,” Kiran said in the news release.
By doing so, the researchers argue, AI tools can better capture the real-world complexity of people’s lives and provide more accurate and fair risk estimates.
The digital twin model also highlighted clear ethnic disparities. In the simulations, South Asian, African and Caribbean participants showed much higher estimated risks than white participants, echoing long-standing findings from the National Health Service and Public Health England. That suggests the tool may help health systems better understand and address inequities that have been documented for years but remain difficult to tackle.
One potential advantage of this approach is accessibility. Because the model does not depend on lab tests or constant streams of data from smartwatches and fitness trackers, it could be used in clinics and communities that lack expensive technology.
“Digital Twin model systems present a viable cost-effective way of diagnosis, testing and treatment for a number of conditions,” Pierscionek added.
In practice, that could mean using simple questionnaires about sleep, mood, social connection and diet, along with basic health information, to build a person’s digital twin. Clinicians could then explore different scenarios — for example, what might happen if someone received support for depression, joined a social group or made modest changes to their eating habits — and see how those changes might shift their long-term diabetes risk.
The study’s authors emphasized their work is an early step, and the model will need further testing and refinement before it can be used widely in clinics. But they argue that transparent, causal simulations like this can help build trust in AI in health care by showing not just what the risk is, but how different factors interact to create it.
As type 2 diabetes continues to rise globally, the findings point toward a future where mental health, social connection and sleep are treated as core parts of diabetes prevention, not afterthoughts — and where a digital twin might help people and their doctors see trouble coming years before it appears on a blood test.
Source: Anglia Ruskin University