Grant: Project Grant

High-quality data is the driver for accelerating innovation and improving value of health care in Canada. Administrative health data are routinely collected through every patients' encounter with the health system (e.g. hospitals, emergency departments, and primary cares). Hospital discharge abstract database (DAD) is one of the most important administrative health databases. To create the DAD, coders are trained to manually review charts and translate medical information into medical diagnosis and intervention codes. These codes enable comparisons of health outcomes within and between hospitals to determine reimbursements, funding, and resource planning. The current abstraction practices are costly and unsustainable due to the growing and aging population. Meanwhile, there is persistent issue of underreporting or misreporting of conditions in DAD. In turn, poor data quality poses a significant threat to the efforts in conducting meaningful health system evaluation and health services research. Innovation on DAD data collection is urgently needed. Although artificial intelligence (AI) has been quickly adopted into many industries, health care industry has lagged behind, partly due to the lack of digital data. However, with the continuous investment of electronic medical records (EMR) to digitally collect health information throughout Canada and advancement of computing power, we are ideally now positioned to leverage AI technologies to improve the healthcare. This project will develop a computer-assisted coding system to facilitate DAD collection. We will develop the system using state-of-the-art natural language processing tool in combination with machine learning and deep learning techniques. The proposed system will leverage existing database on medical knowledge and historical DAD data. The potential impact of this system will be to improve DAD data quality at a lower cost, which will support research and innovation to improve health care for all Canadians.