Japan Association for Medical Informatics

The 40th Joint Conference on Medical Informatics / APAMI2020

The 40th Joint Conference on Medical Informatics / APAMI2020

Nov 18, 2020 - Nov 23, 2020
Actcity Hamamatsu

[AP2-E2-4-05] An Electronic Phenotyping Algorithm to Identify Cases of Stevens-Johnson Syndrome/Toxic Epidermal Necrolysis in the MID-NET Database

*Rieko Izukura1, Takeshi Nakahara3, Takamichi Ito3, Chinatsu Nojiri2, Tadashi Kandabashi2, Takanori Yamashita2, Atsushi Takada2, Yoshiaki Uyama4, Naoki Nakashima2 (1. Social Medicine, Department of Basic Medicine, Faculty of Medical Sciences, Kyushu University, Japan, 2. Medical Information Center, Kyushu University Hospital, Japan, 3. Department of Dermatology, Graduate School of Medical Sciences, Kyushu University, Japan, 4. Pharmaceuticals and Medical Devices Agency, Japan)

Phenotyping Algorithm, Adverse Effects, Stevens-Johnson Syndrome/Toxic Epidermal Necrolysis

Stevens-Johnson syndrome/toxic epidermal necrolysis (SJS/TEN) is a well-known drug-induced adverse effect. We developed a phenotyping algorithm to identify SJS/TEN with high positive predictive value and sensitivity. The initial algorithm was modified by adding the principal variables based on the gradient boosting decision tree and clinical perspectives. Consequently, it was improved to the phenotyping algorithm to identify SJS/TEN with an increased positive predictive value from 8.5% to 76.5% and a sensitivity of 76.5%. It is necessary to further refine this algorithm by evaluating its robustness in other hospitals.