[3Yin2] Interactive session 1

[3Yin2-28] Expanding Features of Outcome Prediction using Estimated Hospitalization Progress.

〇Shotaro Misawa¹, Taiki Furukawa², Shintaro Oyama², Kikue Sato², Ryuji Kano¹, Hirokazu Yarimizu¹, Tomoki Taniguchi¹, Kohei Onoda¹, Tomoko Ohkuma¹, Yoshimune Shiratori² (1.FUJIFILM Corporation, 2.Medical IT Center, Nagoya University Hospital.)

Keywords:Integrative data analysis, Outcome prediction

Outcome prediction using clinical data such as mortality prediction, length-of-stay prediction is applicable to acute change prediction, early treatment, and prediction of treatment effects. However, it is difficult to predict the long-term future status of patients. To improve the performance of the prediction model, we first estimate the short-term future and leverage the estimated value to predict the long-term future status of patients. Such hospitalization progress in the short-term future can be estimated by constructing another estimation model. In this study, we propose the feature expansion using estimated hospitalization progress for the outcome prediction model. We conduct experiments on clinical data of pneumonia cases aggregated in "CITA Clinical Finder", the integrated medical support platform. The result shows our model can predict more accurately than the model without feature expansion.