JMIR Publications

ABSTRACT

Artificial patients’ technology has the potential to transform healthcare by improving and potentially accelerating diagnosis and treatment, and by mapping clinical pathways or patient care processes. Deep learning methods for generating artificial data in healthcare include data augmentation by variational autoencoders (VAE) technology. The aim of our study was to test the feasibility of generating artificial patients with reliable clinical characteristics by using a geometry-based VAE applied, for the first time, on high dimension low sample size tabular data. Real patients’ tabular data were extracted from the “MAX” digital conversational agent created for preparing patients for anesthesia (BOTdesign®, Toulouse, France). A 3-stage methodological approach was implemented to generate up to 10,000 artificial patients: training the model and generating artificial data, assessing the consistency and confidentiality of artificial data, and validating the plausibility of the newly created artificial patients. We demonstrated for the first time the feasibility to transpose the VAE technique from imaging to tabular data for the generation of a large number of artificial patients. Our digital patients’ cohort was highly consistent. Moreover, artificial patients could not be matched with real patients, thus guaranteeing the essential ethical concern of confidentiality. Further studies integrating dynamic changes (and their variability) are needed to map trends and identify patient trajectories.