Machine Learning to Identify Patients with Lower-Risk MDS

The use of machine learning methods is a viable option for identifying patients with myelodysplastic syndrome (MDS), according to a study presented at the 2022 American Society of Clinical Oncology (ASCO) Annual Meeting.

In this study, a research team led by Colden Johanson developed a machine-learning model to discern erythropoietin-stimulating agent (ESA)-treated, lower risk (LR)-MDS patients. They used electronic health records (EHRs) to assess a sample of 1,549 patients from the Syapse Learning Health Network (SLHN), of which 25% were confirmed as ESA-treated LR-MDS patients. The population of interest were divided into a training set (80% of patients), and validation set (20%), both stratified by outcomes. Predictive variables built into the model included age, sex, diagnosis codes, clinical lab tests, and evidence of bone marrow biopsy. The model was subsequently applied to the unscreened SLHN population.

According to the results, the machine learning model was able to identify an additional cohort of 157 patients based on prediction likelihood, of which 44% were CTR-confirmed ESA-treated LR-MDS patients. “The application of machine learning methods increased the rate of ESA-treated MDS patient identification even after the expertly-determined population was depleted,” the researchers concluded. They added that the findings suggest “the application of machine learning models using EHR data may improve the efficiency of MDS patient identification and screening efforts for research, quality improvement, and clinical care.”AI