Anonymous healthcare cohort with the guarantee of data privacy and utility

Research Summary

Healthcare data ecosystem and four challenges

• 1 Consent of the Treatment of Personal Data

  Insufficient understanding of privacy notice

• 2 Business operator selection

  No personal data for testing,
  The quality of processing technology cannot be guaranteed

  The need for a fair benchmark

• 3 Security

  99.98% of personal attributes
  Americans are reidentifiable

  [Rocher 2019] “Estimating the success of reidentifications in incomplete datasets using generative models”, Nature Comm., 2019.

• 4 Utility

  Example of unnatural processing (age distribution)

  National Health and Nutrition Examination Survetion Survey CDC (Centers for Disease Control and Prevention) National Health and Nutrition Survey Program

Achievement Goals (Delivery) Healthcare Anonymous Cohort Infrastructure

Groups Collaboration

Research implementation system

Data name	Number of individuals	Details	Number of records	Number of attributes	Attribute exampe	Number of receipts
Health diagnosis data	198,740	Health checkup Recorded	964,636	49	Height, weight Health distribution	-
Health insurance claims data	288,568	The patient was diagnosed Injuries and illnesses have been recorded	39,363,878	15	Injury and illness classification Code	11,912,236
Pharmaceutical claims data	279,199	Patient was prescribed Drugs are recorded	31,465,504	21	Pharmaceuticals Classification code	9,000,249

Analysis 2. Prediction of morbidity by machine learning

Our goals