A group of researchers from the OHSU Department of Medical Informatics & Clinical Epidemiology (DMICE), collaborating with clinical experts, has demonstrated an approach to detect possible cases of a rare disease by applying machine learning to patient data in the electronic health record (EHR). The approach developed has potential applicability to additional rare diseases that are often not diagnosed in a timely manner. The initial application of the methodology to one rare disease, acute hepatic porphyria (AHP), has been described in a paper published (this week) in the journal, PLoS ONE.
Occurring in about 1 per 200,000 people, AHP is characterized by a triad of intermittent and severe abdominal pain, neurological dysfunction, and psychiatric disturbances. Because the disease is uncommon, and the symptoms non-specific, diagnosis is often delayed and sometimes never made. Recently, a highly effective new treatment has become available, giving more impetus to identifying all patients with the disease.
Led by Aaron Cohen, MD, MS, professor of medical informatics and clinical epidemiology, the researchers applied machine learning to 200,000 OHSU EHR records to determine whether this approach could be effective in identifying patients not previously tested for AHP, and who could be good candidates to receive a diagnostic workup for AHP. The algorithm “learned” from the 30 known patients in the OHSU system and identified 100 patients whose records indicated AHP might be present yet had never been considered as a diagnosis. Manual review of the 100 patients’ records identified four patients where AHP diagnostic testing was likely indicated and 18 patients where AHP diagnostic testing was possibly indicated. Based solely on the reported prevalence of AHP, the analysis of manually reviewed patients would have expected to find only 0.001 cases, demonstrating the ability of the methodology to identify possible cases of this rare disease.
Senior author William Hersh, MD, professor and chair of medical informatics and clinical epidemiology, will extend the work in collaboration with Dr. Cohen and several OHSU clinical leaders to validate the output of the machine learning algorithm with the patients it has identified. The research team will also extend the work by manually reviewing more patients identified by the algorithm, assessing additional machine learning approaches, and applying this methodology to other rare diseases. The research was funded by Alnylam Pharmaceuticals.