ClinicalAge Uses Medical Records to Track Aging and Flag Disease Risk
An AI aging estimate tracked disease and mortality risk across millions of UK adults, but its value for guiding treatment remains unproven.
Summary
Your medical history may contain clues about how quickly your health is changing. Researchers developed ClinicalAge, an artificial intelligence tool that estimates aging from routinely collected medical records. They developed it using 320,891 UK Biobank participants and evaluated it in primary care data covering 4,728,376 adults. ClinicalAge closely tracked an established aging measure and was associated with physical function, numerous diseases, and mortality. Aging patterns began separating in midlife among adults developing conditions including diabetes, high blood pressure, and depression. The findings suggest medical records could help identify adults who need earlier preventive care. However, this observational study does not show that using ClinicalAge improves outcomes or that lowering its score slows aging. This summary relies only on the supplied abstract, whose listed publication date is in the future.
Detailed Summary
Routine medical records may reveal how quickly health deteriorates before major diseases emerge. Researchers developed ClinicalAge, an artificial intelligence tool that turns years of health records into an aging estimate. The goal is to identify adults who might benefit from earlier prevention, rather than simply counting their birthdays.
The team developed the model using 320,891 UK Biobank participants and evaluated it in separate UK primary care data covering 4,728,376 adults. It learned to approximate PhenoAge, an established blood test based aging measure, using patterns in longitudinal medical records. Researchers then examined links with physical function, other aging biomarkers, disease diagnoses, and death.
ClinicalAge explained about 82 percent of the variation in PhenoAge. It was associated with all 17 functional measures and ten aging biomarkers assessed, plus all 20 noncancer diseases studied in both datasets. Associations also appeared for most cancers, although some ran in opposite directions. Higher ClinicalAge was associated with greater overall, cancer, and cardiovascular mortality after accounting for measured clinical and lifestyle factors.
Health trajectories diverged during midlife, with earlier acceleration among adults developing diabetes, hypertension, depression, or conditions described as lifestyle disorders. This suggests existing records could help flag patients for closer assessment of blood pressure, metabolic health, mental health, and preventive care. However, the study did not test whether acting on ClinicalAge improves health or extends life.
These are observational associations, not proof that the model measures a distinct biological aging process or that lowering its score prevents disease. Recordkeeping differences, unmeasured confounding, and limited geographic diversity could affect performance. The abstract does not specify the increment underlying mortality hazard ratios, so they cannot be interpreted as risk increases per year. This summary uses only the supplied abstract; its listed publication date is in the future and has not been independently verified against the original publication.
Key Findings
- ClinicalAge approximated PhenoAge with an R² of 0.819; this measures agreement with a reference biomarker, not accuracy in predicting lifespan.
- Development included 320,891 UK Biobank participants; evaluation used primary care data covering 4,728,376 adults.
- ClinicalAge was associated with all 17 functional measures, ten aging biomarkers, and all 20 noncancer diseases assessed.
- Midlife trajectories diverged earlier among adults developing diabetes, hypertension, depression, or conditions described as lifestyle disorders.
- The study did not establish whether ClinicalAge-guided prevention improves outcomes or whether reducing its score slows aging.
Methodology
This observational model development and validation study trained a transformer-based model on longitudinal UK Biobank electronic health records, using PhenoAge as its reference. Three CPRD Aurum cohorts supported external validation, trajectory analysis, and comparison with the electronic frailty index. Researchers assessed associations with functional measures, aging biomarkers, incident diseases, and mortality, adjusting mortality analyses for clinical and lifestyle factors.
Study Limitations
This summary is based on the abstract only; the supplied publication date of August 12, 2027 is future-dated relative to this assessment and has not been independently verified. Observational associations cannot establish causality, and UK-only cohorts, unmeasured confounding, and differences in healthcare recording may limit generalizability. The abstract omits the exposure increment for mortality hazard ratios and does not establish whether score-guided care improves outcomes.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe:
