Research Areas

Health systems generate huge amounts of data every day. Every hospital admission, diagnosis code, prescription, and lab test leaves a digital record. This research area focuses on making that data useful and trustworthy for answering big questions about population health.

Foundational Contributions

These landmark papers helped build the global toolkit for using administrative health data, including Charlson and Elixhauser comorbidity algorithms, chronic disease case definitions, and ICD validation frameworks.

Comorbidity & Risk Adjustment Indices

Developing and refining scoring tools, such as Charlson and Elixhauser indices, that use administrative data to measure how sick patients are when they arrive at hospital. These tools make fair comparisons possible between hospitals, countries, and time periods.

Validating ICD Codes & Case Definitions

Testing whether diagnosis codes in administrative databases actually capture the diseases they are intended to represent. This work supports research across conditions ranging from COVID-19 to depression to breast cancer recurrence.

Data Quality & Coding Accuracy

Making sure health data are accurate, complete, and trustworthy. This theme looks under the hood, assessing how well hospital and EMR data reflect clinical reality, developing quality indicators, and creating methods to detect errors at scale.

Patient Safety Indicators & Adverse Events

Using routinely collected hospital data to detect harm that happens during care, such as infections, complications, falls, and safety events. This work helps hospitals and health systems identify where patients may be experiencing harm, support quality improvement, and compare safety across organizations and countries.

Applied Population Health Studies

Using administrative data to answer real-world questions about how diseases affect populations, from heart attacks to pregnancy outcomes to kidney disease to rare genetic conditions.

Electronic medical records contain information that administrative data cannot capture, such as free-text clinical notes, lab values, vital signs, medications, and the nuance of how care unfolds. This research area focuses on unlocking that information using machine learning, natural language processing, and rigorous validation.

EMR Case Finding & Phenotyping

Using EMR data, including free-text clinical notes, to identify patients with specific conditions more accurately than diagnosis codes alone. This includes large language model work for disease detection, explainable natural language processing algorithms for hypertension, phenotyping chronic conditions, and methods to disambiguate clinical abbreviations in notes.

EMR Data Quality & Validation

Before EMR data can power research or AI, we need to know it is accurate, complete, and comparable across systems. This theme includes studies comparing EMR data to the gold-standard hospital abstract database, assessing missing discharge summaries, validating automated data extraction, and measuring the quality of primary care EMR data across Canadian and international sites.

Heart Failure Studies with AI/ML Applications

Using EMR data and machine learning to predict who will be readmitted to hospital after heart failure and to find ways to prevent it. This integrated program combines cardiac MRI phenotyping, EMR-based case finding, environmental scans of Alberta initiatives, and Delphi expert panels.

Anchored by the Alberta Innovates-funded PRRAM-HF grant led by Dr. Cathy Eastwood.

Modern modeling approaches that work across borders, time, and data sources. This theme covers methods that share intelligence without sharing data, capture how clinical patterns unfold over time, and connect Alberta to national algorithm-sharing initiatives.

Machine Learning for Hospital Adverse Event Detection

Training modern AI models, including BERT, large language models, and unsupervised clustering, to find patient safety events buried in clinical notes. Administrative codes miss many adverse events; this work uses the full text of the EMR to detect falls, surgical site infections, pulmonary embolisms, and sepsis more accurately.

Supported by a CIHR grant focused on EMR-based adverse event detection.

Federated Learning and Privacy-Preserving AI

Federated methods allow institutions to collaboratively develop AI models while keeping patient-level data within local environments. This supports multi-site learning where governance, privacy, and trust are central design requirements.

Health data work is global work. The lab collaborates with the World Health Organization, international consortiums, and partners across many countries to shape how disease classification and health data systems evolve worldwide.

ICD-11 Development & WHO Partnerships

Shaping the WHO's next-generation disease classification system so it better captures patient safety events, healthcare-related harms, and quality of care. This work directly influenced what ICD-11 can do for every country that adopts it.

ICD-11 Field Trials & Transition to Adoption

Testing the new ICD-11 system in real-world Canadian settings and planning how health systems can transition from ICD-10. This includes Canada's first ICD-11 field trial, mapping tools, physician acceptance studies, and cost/outcome evaluation frameworks.

International Consortiums & Cross-Country Research

QUAN-CORE works with partners across countries, including through the International Methodology Consortium for Coded Health Information, to compare how data are collected, coded, and used internationally.