Enhanced data extraction and modelling from electronic medical records and phenotyping for clinical care, and research: Case studies in management of medication stewardship

Start Date

30 June 2020

End Date

29 May 2025

Project Participants

Metro North Hospital and Health Service, University of Queensland

Status: Completed

Tags: acute care, artificial intelligence (AI), clinical decision support, dashboards, data analytics, data extraction, data visualisation, deterioration, electronic health records, hospital care, machine learning, medication management, medication stewardship, pharmacy, phenotyping, public health, quality and safety, workforce capability

Opportunity 

Hospitals generate large volumes of clinical data in real time, but much of this information remains difficult to extract, search and use for service improvement. Structured information may be fragmented across systems, while valuable information in free-text clinical notes can remain unindexed and inaccessible for analysis.  

There was an opportunity to demonstrate how electronic medical record (EMR) data could support safer and more proactive hospital care. The project looked to develop new, efficient capability for data extraction from the EMR, along with demonstration of EMR data use within clinical analytic methods that address clinical workflow activities and evaluate impact. 

Project Objectives  

This project aimed to develop and demonstrate efficient methods for data extraction from electronic medical records, starting with Queensland’s integrated Electronic Medical Record (iEMR). 

It then looked to deliver real-time analytics into selected clinical workflows, establishing governance and evaluation approaches that support the safe use of these analytics in clinical practice. 

The project also investigated further applications of machine learning and language models in phenotyping, patient-deterioration prediction and clinical question answering. 

The project delivered validated clinical data extraction methods, along with a clinical data dictionary and ontology-mapping capability.  

This enabled two operational analytical dashboards, for opioid-prescribing and insulin-safety, which were implemented at the Surgical Treatment and Rehabilitation Service (STARS). The insulin-safety dashboard was implemented subsequently at five additional Queensland hospitals. The project provided training for effective use of the dashboards to nursing, medical and pharmacy staff across all hospitals with dashboard implementation, with each site then adopting a train-the-trainer model. 

The project also generated clinical AI research assets, methods, technical outputs, and codebases, advancing understanding of the data and infrastructure foundations required for safe and effective clinical AI in healthcare.  

This project also supported the successful completion of two PhD programs and a Masters.

This work demonstrated how real-time analytics can be implemented in routine hospital environments and extended across multiple sites, as a foundation for moving from retrospective review towards earlier identification of clinical and medication-safety risks. 

For clinicians and health services, the project strengthened medication stewardship, clinical decision support and workforce capability. For researchers and technology teams, it produced reusable methods, datasets, code and lessons concerning the application of artificial intelligence to electronic medical record data. 

The challenges encountered also highlighted a critical issue for Australia’s digital-health ecosystem: advanced analytics depend on reliable clinical data infrastructure, extraction, linkage and access. The lessons learned are helpful for planning future clinical AI research and implementation. 

Integrity, Excellence,
Teamwork and Authenticity

Hand holding a smartphone against a colourful, defocused background