Related Experiment Video
Updated: Apr 16, 2026
Frailty Assessment in an Aging Mouse Model
Published on: September 23, 2025
Predicting the risk of mild cognitive impairment in the Mayo Clinic Study of Aging
V Shane Pankratz1, Rosebud O Roberts1, Michelle M Mielke1
1From the Department of Internal Medicine (V.S.P.), University of New Mexico Health Sciences Center, Albuquerque; Division of Epidemiology, Department of Health Sciences Research (R.O.R., M.M.M., W.A.R., R.C.P.), and Departments of Neurology (R.O.R., M.M.M., D.S.K., R.C.P.) and Radiology (C.R.J.), Mayo Clinic, Rochester, MN; and the Departments of Psychiatry, Psychology, and Neurology (Y.E.G.), Mayo Clinic, Scottsdale, AZ.
Objective:
We sought to develop risk scores for the progression from cognitively normal (CN) to mild cognitive impairment (MCI).
Methods:
We recruited into a longitudinal cohort study a randomly selected, population-based sample of Olmsted County, MN, residents, aged 70 to 89 years on October 1, 2004. At baseline and subsequent visits, participants were evaluated for demographic, clinical, and neuropsychological measures, and were classified as CN, MCI, or dementia. Using baseline demographic and clinical variables in proportional hazards models, we derived scores that predicted the risk of progressing from CN to MCI. We evaluated the ability of these risk scores to classify participants for MCI risk.
Results:
Of 1,449 CN participants, 401 (27.7%) developed MCI. A basic model had a C statistic of 0.60 (0.58 for women, 0.62 for men); an augmented model resulted in a C statistic of 0.70 (0.69 for women, 0.71 for men). Both men and women in the highest vs lowest sex-specific quartiles of the augmented model's risk scores had an approximately 7-fold higher risk of developing MCI. Adding APOE ε4 carrier status improved the model (p = 0.002).
Conclusions:
We have developed MCI risk scores using variables easily assessable in the clinical setting and that may be useful in routine patient care. Because of variability among populations, validation in independent samples is required. These models may be useful in identifying patients who might benefit from more expensive or invasive diagnostic testing, and can inform clinical trial design. Inclusion of biomarkers or other risk factors may further enhance the models.
