Case Definition Development (Version II)
Tyler Williamson, Neil Drummond, Stephanie Garies, Ashley Cornect-Benoit
Stephanie Garies (firstname.lastname@example.org)
Ashley Cornect-Benoit (email@example.com)
Ethics approval granted
An emerging concept in Canadian primary health care technology and research is data linkage, in which patient records in one data source are matched to their corresponding information in another discrete data source. In this proposed study, primary care EMR data will be linked with administrative data (i.e. hospital discharge abstracts, emergency department utilization, pharmacy information, census) to develop and validate case definitions for epilepsy, multiple sclerosis, Parkinson’s disease, affective disorders, psychotic disorders, ADHD/ADD, stroke, cardiovascular disease, congestive heart failure, community-acquired pneumonia, opioid dependency and diabetic retinopathy.
This study is innovative in that data linkage between primary care EMR and administrative data sources is only routinely conducted in a few provinces at present, due to the variability in healthcare and health legislation across Canada. This study will create and utilize a linked EMR-administrative database in Alberta, which is not a routine activity at this time.
We propose a new machine-learning method using linked EMR-administrative data to create and validate case definitions for neurological and mental health conditions. Specifically:
1. Use machine learning to develop case definitions for use in a linked primary care EMR-administrative database for the following conditions: epilepsy, multiple sclerosis, Parkinson’s disease, affective disorders, psychotic disorders, ADHD/ADD, stroke, cardiovascular disease, congestive heart failure, community-acquired pneumonia, opioid dependency and diabetic retinopathy.
2. Validate the case definitions and assess their performance in the EMR-only data and the EMR-administrative linked database.
3. Compare the validation metrics of the two data sources (EMR-only and EMR-admin linked).
We will undertake a cross-sectional case definition development and validation study using administrative data from Alberta Health Services (AHS) Analytics linked to EMR data from CPCSSN. A machine learning approach will be used to develop the algorithms for each case definition.