As the world accelerates the development of powerful artificial intelligence, it faces urgent public health challenges that demand AI-driven solutions. These include preparing for future pandemics, combating drug-resistant infections, reversing the obesity crisis, addressing rising mental health issues, and improving care for aging populations living with multiple chronic conditions.
An international network of innovative health systems could accelerate progress toward AI-enhanced public health.
However, training AI with health data poses unique difficulties. Unlike clear visual data from cameras or sensors used for autonomous vehicles, medical records offer only fragmented snapshots of patients already ill, missing the full trajectory of their health. These records contain gaps and measurement errors, making AI training akin to peering through obscured, incomplete views of patient health journeys.
Critical early information on how illnesses develop or why people seek care is largely absent. This data is essential for prevention and promoting well-being—especially as healthcare costs rise with aging populations. Fortunately, AI is beginning to harness daily life patterns through new data streams and user interactions, offering fresh opportunities to support health and prevent disease.
Currently, patient apps and devices tend to focus on isolated diseases or treatments, creating a confusing array of tools for those with complex conditions. A paradigm shift is needed—from a fragmented “medical app store” model to an integrated “healthcare skill store” centered on a patient’s digital twin. This shift is crucial for enabling more predictive, preventive, and personalized care.
For instance, patients on medication for serious mental illness face risks like weight gain, hypertension, diabetes, and cardiovascular events. Effective AI allies would support not only medication adherence but also healthy lifestyle choices—diet, exercise, social engagement—to mitigate side effects and enhance quality of life and daily functioning, including work performance.
These AI allies must learn from patients’ interactive digital twins or health avatars, which reflect a far broader range of health, social factors, habits, and treatment use than medical records capture.
Achieving AI-powered personalized and preventive care requires close collaboration among patients, providers, payers, researchers, and AI engineers. Data powering patient-facing health avatars can also improve services for healthcare and social care providers.
Moreover, data on individuals’ paths to care helps systems allocate resources effectively, targeting the most vulnerable—a concept known as the “learning health system.”
The University of Liverpool’s Civic Health Innovation Labs and partners are pioneering this approach with a triple digital twin model for the patient, the provider, and the population. The patient-level twin tracks medication experiences and early warning signs, empowering self-management. The provider-level twin gives doctors a concise summary of key information alongside prescribing guidance. The population-level twin identifies patients who would benefit from medication reviews and facilitates fast-track care.
This work addresses critical issues: medication-related harm is common, drug costs strain health budgets, and adherence to prescriptions is poorly understood, with estimates suggesting one-third to half of medicines go unused as directed.
Like AI systems that improve with vast, diverse data, effective AI allies for patients and health systems must learn from a global network of learning health systems. Tackling antimicrobial resistance, promoting prevention, and managing complex chronic conditions amid limited resources are global challenges that cross borders. To meet these pressing health needs, worldwide cooperation is essential to optimize AI for human health.
Related Topics
