Grant: Team Grant: Embracing Diversity to Achieve Precision & Health Equity-Aging - LOI
University of Toronto
$20,000.00
Improving equity and precision of AI models in hospital based care - the critical role of incorporating social determinants of health
FY 2023-24
Summary
As healthcare systems integrate artificial intelligence (AI) tools to enhance patient care there is an urgent need to better understand when, how, and on whom these models perform well or fail. Many patient populations are highly diverse in terms of age, sex and gender, socioeconomic factors, as well as medical factors like coexisting chronic diseases. This makes it a challenge to develop models that work well across all patients and to identify subgroups wherein predictive models might fail.
Our proposed research takes an interdisciplinary approach to generate evidence and create actionable mechanisms to ensure the equity of predictive systems in healthcare. In collaboration with our patient partners, we will develop a patient and caregiver advisory panel that will reflect the diversity of patients cared for on general medical wards and that will inform all aims of our research. Our project will focus on AI models that predict death or critical illness in hospital for patients with serious bacterial infections. We will focus particularly on older adults and on people living with diabetes, who are at the greatest risk for severe infections. AI prediction models can help clinical teams intervene earlier and prevent patient harm. Our study will have 3 aims. First, we will describe the patient groups (based on age, sex, and socioeconomic factors) where predictive models perform well, and where they do not. Second, we will construct predictive models that incorporate the key factors from Aim 1 to perform more accurately in all groups. Finally, our third aim will develop an automated dashboard that identifies when AI models are starting to fail in real-time and helps explain why failures are occurring. This tool will help users of AI models monitor their performance and ensure equity over time. Taken together, this research will have a major impact on improving our ability to use AI models that embrace diversity and advance precision healthcare for all.
Department
Program
Team Grant: Embracing Diversity to Achieve Precision & Health Equity-Aging - LOI
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
CIHR Funding Decisions Database
As published
Id
500684
Name
Gopalkrishnan Rahul
Region
Ontario
Country
Canada
Orgtype
Institution Paid; Research Institution
Instname2
Population and Public Health
Theme2
Clinical
Prcname2
Team Grant: Embracing Diversity to Achieve Precision & Health Equity
Projecttitle
Improving equity and precision of AI models in hospital based care - the critical role of incorporating social determinants of health
Programtype2
Operating Grants
Orgnameinp2
University of Toronto
Cihramount2
$20,000
Cihrequipment2
$0
Programname2
Team Grant: Embracing Diversity to Achieve Precision & Health Equity-Aging - LOI
Keyworddelim
Equity In Predictive Models; Machine Learning/Artificial Intelligence In Healthcare; Social Determinants Of Health
Abstract
As healthcare systems integrate artificial intelligence (AI) tools to enhance patient care there is an urgent need to better understand when, how, and on whom these models perform well or fail. Many patient populations are highly diverse in terms of age, sex and gender, socioeconomic factors, as well as medical factors like coexisting chronic diseases. This makes it a challenge to develop models that work well across all patients and to identify subgroups wherein predictive models might fail.
Our proposed research takes an interdisciplinary approach to generate evidence and create actionable mechanisms to ensure the equity of predictive systems in healthcare. In collaboration with our patient partners, we will develop a patient and caregiver advisory panel that will reflect the diversity of patients cared for on general medical wards and that will inform all aims of our research. Our project will focus on AI models that predict death or critical illness in hospital for patients with serious bacterial infections. We will focus particularly on older adults and on people living with diabetes, who are at the greatest risk for severe infections. AI prediction models can help clinical teams intervene earlier and prevent patient harm. Our study will have 3 aims. First, we will describe the patient groups (based on age, sex, and socioeconomic factors) where predictive models perform well, and where they do not. Second, we will construct predictive models that incorporate the key factors from Aim 1 to perform more accurately in all groups. Finally, our third aim will develop an automated dashboard that identifies when AI models are starting to fail in real-time and helps explain why failures are occurring. This tool will help users of AI models monitor their performance and ensure equity over time. Taken together, this research will have a major impact on improving our ability to use AI models that embrace diversity and advance precision healthcare for all.
Competitiondate
202402
Approvedterm2
1 yr 0 mth
Pinamesdelim
Gopalkrishnan, Rahul; Razak, Fahad; Siddiqi, Arjumand; Verma, Amol
Approvedterm
1 an 0 mois
Cihrcontribution2
III~ $20,000~ $0~