Quality carbs and dementia; health concerns drive most involuntary retirements; biased AI – Morning Medical Update

Low–glycemic index diets linked to lower dementia risk

The quality of carbohydrates in the diet may play an important role in long-term brain health, according to a study published in the International Journal of Epidemiology. Investigators followed more than 200,000 dementia-free adults for an average of 13 years and found that diets emphasizing low–glycemic index foods were associated with a reduced risk of Alzheimer’s disease and other dementias. Compared with higher–glycemic index diets, low to moderate GI patterns were linked to a 16% lower risk of Alzheimer’s, while higher GI intake was associated with a 14% increase in risk.

Poor health drives most involuntary retirements

Poor health is the leading reason middle-aged Americans leave the workforce earlier than planned, according to a new study published in Rehabilitation Counseling Bulletin. Reviewing data from more than 12,700 adults who reported retiring, Penn State researchers found that over half of forced retirements were driven primarily by health problems, often tied to the onset of chronic illness or disability in midlife. The study also identified disparities, with racial and ethnic minority workers more likely to report involuntary retirement and college-educated workers significantly less likely to be forced out early.

Demographic bias in pathology AI

Artificial intelligence (AI) models increasingly used to analyze pathology slides for cancer diagnosis may perform unevenly across patient populations, according to a new study from Harvard Medical School published in Cell Reports Medicine. Researchers found that several widely used pathology AI models showed lower diagnostic accuracy based on patients’ self-reported race, gender and age, with disparities affecting nearly one-third of diagnostic tasks. The team identified multiple drivers of bias, including unequal training data, differences in disease incidence and subtle molecular patterns correlated with demographics rather than disease. They then tested a new framework, FAIR-Path, which reduced diagnostic disparities by nearly 90% without requiring perfectly representative datasets.