Clinical study applying machine learning to detect a rare disease: results and lessons learned. Issue 2 (30th June 2022)

• Record Type: Journal Article
• Authors: Hersh, William R; Cohen, Aaron M; Nguyen, Michelle M; Bensching, Katherine L; Deloughery, Thomas G
• Is Part Of: JAMIA open. Volume 5:Issue 2(2022)
• Publication Date: 2022-06-30
• Subjects: machine learning -- porphyria -- acute intermittent -- electronic health records -- clinical study; Medical informatics -- Periodicals; 610.285
• DOI: 10.1093/jamiaopen/ooac053
• Languages: English
• ISSNs: 2574-2531

Abstract

Abstract: Machine learning has the potential to improve identification of patients for appropriate diagnostic testing and treatment, including those who have rare diseases for which effective treatments are available, such as acute hepatic porphyria (AHP). We trained a machine learning model on 205 571 complete electronic health records from a single medical center based on 30 known cases to identify 22 patients with classic symptoms of AHP that had neither been diagnosed nor tested for AHP. We offered urine porphobilinogen testing to these patients via their clinicians. Of the 7 who agreed to testing, none were positive for AHP. We explore the reasons for this and provide lessons learned for further work evaluating machine learning to detect AHP and other rare diseases. Lay Summary: This study aimed to determine if patients identified by a machine learning algorithm applied to the electronic health record data had the rare disease, acute hepatic porphyria (AHP). The algorithm had identified 22 patients who had a clinical presentation consistent with AHP but had never been tested for the disease. We attempted to contact and invite all 22 patients through their primary care or other providers to have a simple urine porphobilinogen test for AHP. A total of 7 patients agreed to testing, and all tested negative. A number of lessons were learned for the challenges of assessing machine learning algorithms in clinical settings.