Molecular data, wearable devices, and predictive systems are moving parts of health care closer to the point before symptoms emerge, but the model still faces questions around evidence, workflow, and payment.
Mayo Clinic and Thermo Fisher Scientific launched Precure, LLC, in September 2026 to study biological changes that occur as people move from health toward disease. The initiative will combine genomic, proteomic and longitudinal clinical data from approximately one million biospecimens. The goal is to identify early biological signals that may eventually support new approaches to prevention, diagnosis and treatment.
Disease detection is moving upstream
The Precure initiative will combine molecular data with longitudinal health information to study diseases affecting the brain, heart, kidneys, liver and lungs. Its research program also brings together biospecimens, wearable-device data, environmental information and clinical records. It is still a research effort, not a routine diagnostic service. Even so, the approach points toward a health care model in which risk may be identified before symptoms send a patient to a hospital.
That could change what happens after a risk signal appears. A patient considered high risk might need additional testing, closer monitoring or preventive care rather than emergency treatment. But earlier detection does not automatically mean better or less expensive care. The value depends on whether clinicians can act on the information and whether those actions improve outcomes.
Wearables can extend monitoring beyond the hospital
Research on inflammatory bowel disease offers one example of where wearable data may fit. A study of 309 participants across 36 states found changes in heart rate, resting heart rate, heart-rate variability, daily steps and oxygenation during inflammatory and symptomatic flares. Several measures also changed significantly as much as seven weeks before a flare.
The findings suggest that long-term wearable data could help identify changes before patients notice that their condition is worsening. There is an important caveat, though: the study did not establish a validated alert that could reliably predict an individual patient's flare. For now, the technology appears more useful as another source of information than as a stand-alone warning system.
Continuous monitoring also changes the workload for clinicians. Instead of reviewing health data only during scheduled appointments, they could potentially look for meaningful changes from a patient's usual baseline. That creates another challenge: systems must distinguish meaningful changes from normal variation, limit false alerts and determine who is responsible for follow-up.
Hospitals are already using predictive tools
The idea is already being tested inside hospitals. A JAMA Internal Medicine study of 9,938 patients at a single academic medical center examined an artificial intelligence model designed to identify patients at higher risk of clinical deterioration. Exposure to the intervention was associated with a 10.4-percentage-point absolute reduction in a composite outcome that included rapid response activation, intensive care transfer or cardiopulmonary arrest.
The result is notable, but the evidence has limits. The analysis covered four general internal medicine units at one academic hospital and used data from 2021 and 2022. The researchers said further testing is needed across other care settings.
That distinction matters as predictive systems move from research into clinical environments. A model can perform well in one setting without producing the same results across hospitals, patient populations or workflows.
Screening still has to prove that earlier means better
Cancer screening highlights another problem with the push toward earlier detection. The NHS-Galleri randomized trial enrolled 142,250 people ages 50 to 77 in England and tested whether adding a multicancer early detection blood test to usual care could reduce late-stage cancer diagnoses. After three screening rounds, the trial did not meet its primary endpoint for reducing the combined incidence of stage III and IV cancers across 12 prespecified cancer types.
A prespecified secondary analysis did find a lower incidence of stage IV cancers in the intervention group. That finding is relevant, but it does not by itself establish that the screening approach improves overall health outcomes.
There is also an operational consequence for hospitals. Earlier screening can increase the number of people entering the diagnostic pathway. Positive or uncertain results may lead to additional imaging, biopsies and specialist appointments, increasing demand before treatment even begins.
Payment will shape how remote monitoring scales
Medicare already pays for some remote physiologic and remote therapeutic monitoring services. In its proposed 2027 physician fee schedule, the Centers for Medicare & Medicaid Services proposed changes involving initiating visits, staffing requirements, and the valuation of some remote monitoring services.
The policy question is closely tied to the economics of continuous care. Devices need support. Data must be reviewed. Abnormal findings may require clinical interpretation and follow-up. Reimbursement will help determine whether health systems incorporate remote monitoring into routine care or continue to run these services mainly as limited programs and pilots.
Data quality will set practical limits
More data can also create more room for error. Wearable devices may produce inconsistent measurements, while predictive models can generate false alerts or miss important changes. In the IBD study, researchers combined wearable measurements with symptom surveys and clinical tests rather than treating device data as an independent diagnostic tool.
Regulation is expanding alongside the technology. The U.S. Food and Drug Administration said more than 1,600 AI-enabled medical devices had been authorized for marketing in the United States as of September 2026.
That figure demonstrates the breadth of AI-enabled medical technology, but it does not establish that those devices can predict disease before symptoms appear. Authorization also covers a wide range of clinical applications, making the aggregate number a poor proxy for the maturity of predictive diagnostics.
Hospitals may shift toward earlier intervention
Molecular research, wearable monitoring and predictive systems are beginning to address different points along the path from health to acute illness. The common thread is a move toward identifying changes earlier, when there may still be an opportunity to intervene.
Whether that shift becomes a practical model at scale is still unclear. Health systems will need to show that early signals can lead to validated interventions without creating excessive workload, follow-up demands or cost.
The central question is no longer simply whether hospitals can detect risk earlier. It is whether they can determine, with enough confidence, which signals require action—and prove that acting on them improves patient outcomes.
