Grant: Master's Award: Canada Graduate Research Scholarships - Master's program
Université du Québec à Montréal
$27,000.00
Validation of disease classification systems using electronic health record data
FY 2025-26
Summary
Electronic health records (EHR) have provided a valuable resource for numerous types of medical research. However, the use of EHR data for the actual validation of disease classification systems has been limited. This is a lost opportunity as EHR data could provide a rich complementary source of information for validating existing clinical assumptions about diseases. In this context, this study will use a publicly accessible hospital EHR dataset and focus on clinical information.
Initially, any direct codified information about a patient's disease (i.e. their diagnosis) in the EHR dataset will be ignored. Instead, clustering and unsupervised modelling algorithms used in computer science and artificial intelligence will be applied to group observed clinical characteristics. These groupings will then be compared to the coded diagnoses for patients in the EHR data.
The analysis will focus on the consistency between the groupings from modelling observed clinical characteristics and the coded diagnoses in the EHR dataset. Complete consistency between modeling results and disease coding will support the hypothesis that our current disease classification system is fully reflective of observed groups of clinical characteristics. Inconsistency would suggest the need for further clinical investigation. Specific areas with inconsistency would be highlighted through this study and would provide rich information for researchers and clinicians alike.
Additionally, this study will test whether using EHR data alongside clustering and unsupervised modelling algorithms can be used to validate our current disease classification systems. If this is possible, these methods can be used on other EHR datasets to confirm clinical assumptions in the future.
Department
Program
Master's Award: Canada Graduate Research Scholarships - Master's program
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
CIHR Funding Decisions Database
As published
Id
573031
Name
Saksena Priyanka
Region
Québec
Country
Canada
Orgtype
Institution Paid; Research Institution
Prcname2
Special Cases - Awards Programs
Projecttitle
Validation of disease classification systems using electronic health record data
Programtype2
Studentship Programs
Orgnameinp2
Université du Québec à Montréal
Cihramount2
$27,000
Cihrequipment2
$0
Programname2
Master's Award: Canada Graduate Research Scholarships - Master's program
Keyworddelim
Algorithms; Artificial Intelligence; Clustering; Disease Classification; Electronic Health Data; Icd; International Classification Of Diseases; Machine Learning; Nosology; Topic Modelling
Abstract
Electronic health records (EHR) have provided a valuable resource for numerous types of medical research. However, the use of EHR data for the actual validation of disease classification systems has been limited. This is a lost opportunity as EHR data could provide a rich complementary source of information for validating existing clinical assumptions about diseases. In this context, this study will use a publicly accessible hospital EHR dataset and focus on clinical information.
Initially, any direct codified information about a patient's disease (i.e. their diagnosis) in the EHR dataset will be ignored. Instead, clustering and unsupervised modelling algorithms used in computer science and artificial intelligence will be applied to group observed clinical characteristics. These groupings will then be compared to the coded diagnoses for patients in the EHR data.
The analysis will focus on the consistency between the groupings from modelling observed clinical characteristics and the coded diagnoses in the EHR dataset. Complete consistency between modeling results and disease coding will support the hypothesis that our current disease classification system is fully reflective of observed groups of clinical characteristics. Inconsistency would suggest the need for further clinical investigation. Specific areas with inconsistency would be highlighted through this study and would provide rich information for researchers and clinicians alike.
Additionally, this study will test whether using EHR data alongside clustering and unsupervised modelling algorithms can be used to validate our current disease classification systems. If this is possible, these methods can be used on other EHR datasets to confirm clinical assumptions in the future.
Competitiondate
202512
Approvedterm2
1 yr 0 mth
Pinamesdelim
Saksena, Priyanka
Approvedterm
1 an 0 mois
Cihrcontribution2
~ $27,000~ $0~