AI model could improve screening for primary aldosteronism

A study of a new AI model examining 30 years of routine electronic health records (EHR) data could improve screening for primary aldosteronism, a leading cause of high blood pressure that is often unrecognized but increases patients' risk of cardiovascular complications, according to a study being presented Saturday at ENDO 2026, the Endocrine Society's annual meeting in Chicago, Ill.

Primary aldosteronism occurs when the adrenal glands-the small glands located on the top of each kidney-produce too much of the hormone aldosterone. This causes aldosterone, which helps balance levels of sodium and potassium, to build up in the body. People with primary aldosteronism face a higher risk of cardiovascular disease than those with primary hypertension.

The true prevalence of primary aldosteronism is unknown, but it is estimated that up to 20 percent of patients with hypertension have primary aldosteronism, according to the study's lead researcher, Frank Lee, M.D., of Mayo Clinic in Rochester, Minn. Because effective treatments exist for primary aldosteronism, early diagnosis can prevent future complications and reduce healthcare costs, Lee explained.

The Endocrine Society's "Primary Aldosteronism: An Endocrine Society Clinical Practice Guideline" released in 2025 calls for more widespread screening for primary aldosteronism. This cause of high blood pressure increases patients' risk of cardiovascular complications, including stroke, coronary artery disease, atrial fibrillation, heart failure, and renal disease.

Using de-identified data from more than 22,000 patients collected between 1986 and 2025 in the Mayo Clinic Platform-a federated, privacy-preserving infrastructure with multimodal clinical data - researchers developed an AI screening model that analyzed variables including age, gender, hypertension- and hypokalemia-related ICD diagnoses, systolic blood pressure measurements, potassium blood levels, and prescribed antihypertensive or potassium supplement medications. They then tested it on data of 225,887 adults with hypertension. A XGBoost architecture, a type of machine learning library, predicted patients at risk for primary aldosteronism 12 months prior to diagnosis.

The model showed an AI-based approach to screen for primary aldosteronism may be feasible, Lee said. When researchers set the threshold to identify people at low risk, the model correctly flagged more than 90% of primary aldosteronism cases while missing fewer than 10%. At this setting, about two-thirds of the study participants were identified as candidates for screening.

During testing on patients with high blood pressure who had never been screened previously for primary aldosteronism, our model identified approximately two out of every three patients for further work-up. Clinicians have been challenged to screen primary aldosteronism effectively. The tool developed by our team could offer a solution based on routine information available in a patient's medical records."

Frank Lee, M.D., of Mayo Clinic in Rochester, Minn.