Can AI pinpoint hidden primary aldosteronism before it’s diagnosed?

Clinical Takeaway: In patients with hypertension, AI-driven risk stratification using routine EHR data may help identify who should be prioritized for primary aldosteronism screening, especially when universal screening is difficult to operationalize.

Primary aldosteronism may affect up to 20% of patients with hypertension and carries higher cardiovascular risk than primary hypertension, but it remains widely underdiagnosed despite effective treatments.

An artificial intelligence model trained on nearly 40 years of Mayo Clinic electronic health record data identified patients at risk for primary aldosteronism up to 1 year before diagnosis, according to findings being presented June 13 at the Endocrine Society annual meeting.

Researchers developed an XGBoost-based model using de-identified EHR data from 22,264 adults, including 1,833 patients with primary aldosteronism and 20,431 controls with negative renin/aldosterone screening tests. Inputs included age, sex, hypertension- and hypokalemia-related ICD diagnoses, systolic blood pressure, potassium levels, and prescriptions for antihypertensive medications or potassium supplements.

Compared with controls, patients with primary aldosteronism had higher rates of hypokalemia diagnoses (15.5% vs 8.9%), higher median systolic BP (132 vs 126 mm Hg), and lower median potassium levels (3.9 vs 4.2 mmol/L). Hypertension-related ICD diagnoses were common in both groups (74.6% vs 71.2%), and both cohorts had a median of 2 prescribed antihypertensive medications.

The model achieved an AUROC of 0.71 for 30-day prediction and 0.67 for 1-year prediction, with moderate calibration across intermediate risk probabilities. Threshold selection substantially changed the screening yield: in a holdout cohort of 225,887 adults with hypertension and no prior primary aldosteronism diagnosis or screening, a 0.3 threshold flagged 153,561 patients (68.0%), while a 0.6 threshold identified 13,431 patients (5.9%). At the lower threshold, the model correctly flagged more than 90% of primary aldosteronism cases while missing fewer than 10%, according to a press release.

“Clinicians have been challenged to screen primary aldosteronism effectively,” said lead researcher Frank Lee, MD, of Mayo Clinic. “The tool developed by our team could offer a solution based on routine information available in a patient’s medical records.”

Source: Lee F, et al. Endocrine Society Annual Meeting, Abstract SAT-830. June 13, 2026. AI-Based Approach to Screening Primary Aldosteronism