Machine learning based phenotyping of the response to mindfulness for chronic low back pain
Akshit Soota, Angela C. Incollingo Rodriguez, Benjamin C. Nephew et al. preprint
Machine learning applied to data from a clinical trial of mindfulness for chronic low back pain identified distinct phenotypes that predict who will respond to the intervention. Among 132 participants in the mindfulness arm, 49 were responders and 83 were non-responders. The top three responder phenotypes correctly identified 26 of the 49 responders with 92-100% precision, and the top three non-responder phenotypes identified 36 of the 83 non-responders with 100% precision. These results may help clinicians and patients decide whether mindfulness is likely to be beneficial, improving treatment efficiency and outcomes.