AI model flags hidden hormone disorder behind some high blood pressure cases

An AI screening tool aims at a common diagnosis doctors still miss

Researchers presenting at ENDO 2026 say a new artificial intelligence model could help clinicians identify patients with primary aldosteronism, a frequently underrecognized cause of high blood pressure that carries added cardiovascular risk. The study used 30 years of electronic health record data to build a screening approach that can flag high-risk patients before a formal diagnosis is made.

Primary aldosteronism occurs when the adrenal glands produce too much aldosterone, a hormone involved in balancing sodium and potassium. Excess aldosterone can push blood pressure higher and is associated with greater risk of stroke, coronary artery disease, atrial fibrillation, heart failure, and renal disease. The study notes that effective treatments exist, making earlier detection clinically important.

According to the researchers, the condition may affect up to 20% of patients with hypertension, yet it remains widely underdiagnosed. That gap is part of why the Endocrine Society's 2025 clinical practice guideline called for broader screening. In practice, however, expanding that screening is difficult because large health systems must sort through many patients with overlapping symptoms, medication histories, and lab patterns.

How the model was built

The team, led by Mayo Clinic's Frank Lee, used de-identified data from more than 22,000 patients gathered between 1986 and 2025 through the Mayo Clinic Platform. The model analyzed variables including age, sex, diagnoses related to hypertension and hypokalemia, systolic blood pressure measurements, potassium levels, and prescriptions for antihypertensive drugs or potassium supplements.

Researchers then tested the model on 225,887 adults with hypertension. The best-performing approach used XGBoost, a machine learning framework often applied to structured clinical data. In the reported result, the model predicted patients at risk for primary aldosteronism 12 months before diagnosis.

That lead time matters. A year of earlier recognition could give clinicians time to order confirmatory tests, adjust treatment, and reduce exposure to avoidable cardiovascular complications. It also suggests that AI may be particularly useful not for replacing diagnosis, but for narrowing a large population into a smaller group that should be evaluated more closely.

Why this could matter beyond one disorder

The study highlights a practical use case for health AI: surfacing patients hidden in routine care data rather than generating entirely new forms of clinical evidence. High blood pressure is common, but its causes are not uniform. If systems can distinguish patients whose hypertension is driven by an endocrine disorder from those with more typical primary hypertension, care can become more targeted and less reactive.

The findings also reflect a larger trend in medicine toward using long-span record sets to identify treatable disease earlier. Because the model relied on variables already captured in ordinary care, the barrier to adoption may be lower than for tools that require new imaging, wearable data, or specialized testing. Even so, a screening model is only the first step. It still has to be integrated into workflow, validated across health systems, and used in ways that improve real-world outcomes.

For now, the work adds weight to the case for broader, smarter screening. Primary aldosteronism is both consequential and treatable. A model that can move patients onto a diagnostic pathway before the condition is formally recognized could help close one of hypertension care's most persistent blind spots.

This article is based on reporting by Medical Xpress. Read the original article.