A Relational Method for Determining Eventual Causality in Electronic Health Records Bilal El-Hajj-Diab School of Computer Science and Software Engineering University of Wollongong Wollongong, Australia b.diab@alumni.uottawa.ca Bassam Hussein Department of Industrial Engineering Lebanese International University Beirut, Lebanon bassam.hussein@liu.edu.lb Ali Hage-Diab Department of Biomedical Engineering Lebanese International University Beirut, Lebanon ali.hagediab@liu.edu.lb Mohamad Raad Department of Computer and Communication Engineering Lebanese International University Beirut, Lebanon mohamad.raad@liu.edu.lb Hassan M. Khachfe Center for Quality Assurance, Institutional Assessment & Scientific Research (QAIASR) Lebanese International University Beirut, Lebanon hassan.khachfe@liu.edu.lb Abstract— the use of electronic health records (EHR) has been increased substantially to improve quality of healthcare outcomes. That’s why recent EHR systems have been eagerly evolving from patient documentation systems toward intellectual tools for physicians to accomplish their tasks. Such evolution would mean continuous changes to the current EHR systems which would be a real challenge since usually the designed software for these systems is customized based on prior requirements and thus having the correct workflow design is an essential key to allow system intellectuality and system automation. Such upgrade would give more realistic view to the patient medical profile by associating all related medical entities in a hierarchical order without losing the usability and performance of these systems.) Keywords-component; workflow; electronic health record; causality; logical modeling. I. INTRODUCTION Consider a simple chat regarding a missing boy; the mother who is worried about her child not coming home sends a message to her friend asking “my boy is missing” (message 1), then the mom sends another message stating that her boy is home and no need to worry (message 2); her friend replies saying “thanks god” (message 3). So for these three messages it’s obvious that “message 3” was caused by “ message 2” and the correct workflow of causality of these messages concluded that the “thanks god” message makes sense since now the boy is safe and she is happy about it. Now assume that for some reason “message 2” is missing then the whole concept would change and therefore “message 3” which is “thanks god” would look as if it was caused by the “message 1” and in this case the message would be completely misinterpreted because “thanks god” would be an answer to “my boy is missing” message. In medical profiling, similar scenarios may apply but in much more serious conditions [1]. Consider a patient who had a minor head injury two days before going to a clinic for a severe swelling sinus and headache. After several examinations by the physician, he was diagnosed with flu. The head injury was not considered since it had nothing to do with his symptoms so it was ignored. After few years, the same patient was taken to the clinic again with serious symptoms of amnesia. After careful examination, the doctor concluded that this patient might be developing Alzheimer disease. However, there could be a possibility that the minor head injury which occurred a couple of years ago had something to do with this patient memory loss, but since this fact was not eliminated before, neither the doctor nor the patient would even consider such possibility. The flow of the symptoms and diagnosis can be thought of as messages or processes of causality. For example, head injury is “message 1”, then flu symptoms is “message 2”, followed by amnesia, which is “message 3”. Therefore, if we had a knowledgebase to remove message 2 from the flow, then it would have been clear to the physician that amnesia (message 3) might have been caused by the head injury (message 1). In this paper, we introduce a causal workflow protocol which will act as shim (a library that transparently intercepts API calls) on top of the EHR system and produce symptoms as well as diagnoses based on causality to assist physicians to have as much as possible accurate presentation of the patient’s electronic profile.