Realistic Synthetic Health Condition Timelines: Generating the Patient History using Contextually Appropriate Disease Burden and Health Statistics

EasyChair Preprint 5173

Abstract

Synthetic patient populations and their electronic healthcare records (EHR) have been recognised to be valuable in many secondary uses including pandemic modelling while avoiding access to real health records, which breaches patient privacy. The problem of generating realistic synthetic EHR has remained an elusive challenge partly due to its knowledge- intensive and computationally expensive nature. Central to this challenge is the problem of generating the realistic health condition timelines (RS-HCT) for synthetic patients spanning from cradle to current age or to grave. This position paper is part of ongoing work, addresses this problem and presents an innovative approach to, and an algorithm for, generating the RS-HCT over the lifetimes of synthetic individuals within a given population without using real patient data. Statistics on disease burdens as well as clinical vocabulary, clinical expertise and population demographics across age groups are taken into consideration. This work is significant in that achieving the RS- HCT results in a skeletal realistic synthetic electronic healthcare record (RS-EHR) that would then be developed into a full RS- EHR using inexpensive methods that do not require access to the actual EHR for real patients.

Keyphrases: Electronic Health Records, Synthetic Data Generation, synthetic health record

@booklet{EasyChair:5173,
  author    = {Scott McLachlan and Kudakwashe Dube and Clement Puybareau and Maxime Pitard and Derek Buchanan and Patience Chiketero and Thomas Gallagher and Bridget Daley and Graham Hitman and Norman Fenton},
  title     = {Realistic Synthetic Health Condition Timelines: Generating the Patient History using Contextually Appropriate Disease Burden and Health Statistics},
  howpublished = {EasyChair Preprint 5173},
  year      = {EasyChair, 2021}}