Cohort Study Ties Glucose Monitors to Lower Death, Hospitalization Risk in Insulin-Treated Dementia Patients

A Medicare-based cohort study of more than 2,000 older adults with Alzheimer disease and related dementias found continuous glucose monitor users had a 43 percent lower death rate and fewer hospital stays than those using fingerstick checks, though several safety outcomes showed no clear difference.

A nationwide cohort study of insulin-treated older adults with diabetes and Alzheimer disease and related dementias (ADRD) found that those who used a continuous glucose monitor (CGM) had a 43 percent lower risk of all-cause death (hazard ratio 0.57, 95 percent confidence interval 0.48 to 0.67) and a 14 percent lower risk of all-cause hospitalization (hazard ratio 0.86, 95 percent CI 0.76 to 0.96) than those who checked blood sugar by fingerstick. The study, led by Pareeta Kotecha and Jingchuan Guo of the University of Florida, was published online in JAMA Network Open on December 2, 2025.

The researchers used a 15 percent random sample of Medicare fee-for-service claims, from January 2016 through December 2020, to identify 1,011 CGM users age 66 and older with diabetes and ADRD, then matched each to a fingerstick user of similar age, sex, comorbidities and insulin regimen using propensity scores. It is a retrospective, claims-based comparison, not a randomized trial, and it follows an earlier feasibility report this paper covered on fitting blinded glucose sensors to hospitalized dementia patients; that work asked whether the devices could be worn and read in this population, while this new analysis asks whether their ordinary use tracks with fewer hospital stays and deaths.

Not every measure moved in CGM's favor with statistical confidence. Hospitalizations specifically coded for hypoglycemia were less frequent among CGM users (hazard ratio 0.66, 95 percent CI 0.40 to 1.08) and falls were slightly less frequent (hazard ratio 0.86, 95 percent CI 0.68 to 1.08), but both confidence intervals crossed 1.0, meaning the difference could be due to chance. Hyperglycemia crises ran higher among CGM users (hazard ratio 1.38, 95 percent CI 0.99 to 1.94), also not statistically significant. The authors note that CGM users in the unmatched population were younger and more often prescribed rapid-acting insulin, the kind of underlying difference that propensity matching tries, but cannot fully guarantee, to erase.

The study's authors write that their findings support running pragmatic clinical trials to confirm whether CGM use itself, rather than the characteristics of people who get prescribed it, accounts for the lower hospitalization and mortality figures, and to test how feasible the devices are for patients whose cognitive impairment may limit their ability to read or respond to glucose alerts without caregiver help.