Grant: Project Grant
McGill University
$371,971.00
Making the most of Canada's "natural laboratory": advancing difference-in-differences methods for unpoolable data
FY 2020-21
Summary
Canadian provincial health systems have been called a "policy laboratory" where variations in health policies and health care system structure can be used to inform learning health systems and improve population health. Capitalizing on innovation and experimentation across Canadian health systems generates valuable opportunities for natural experiments and cross-jurisdictional comparative analyses to better understand what works - and what doesn't. One analytic approach that can be used to learn from policy variation in Canada is difference-in-differences (DiD) estimation, which is a simple and effective way to identify causal effects in the absence of randomization. This approach allows us to move beyond quantifying the basic difference in two units to quantifying the effect of the exposure itself, a crucially important parameter in policy analysis. Given its ease of use and interpretability, DiD is an increasingly common analytic strategy in Canadian health economics and health services and policy research. However, Canadian administrative health data are largely siloed by province and cannot be combined into a single data set, which is a significant barrier to DiD estimation. Despite the indispensability of DiD for health services and policy research, no
"best practices" currently exist for conducting DiD analyses in the absence of pooled data. The goal of this proposal is to develop strategies to overcome this barrier. Our team, which consists of experts in econometrics, biostatistics, and health services and policy research, will consider methods to generate DiD estimates using econometric/statistical theory, simulated data, and real Canadian data that is not siloed across provinces (e.g., the CCHS). The methods developed by this project will be relevant for both Canadian interprovincial comparisons and international comparisons.
Department
Program
Project Grant
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
CIHR Funding Decisions Database
As published
Id
433585
Name
Strumpf Erin C
Region
Québec
Country
Canada
Orgtype
Institution Paid; Research Institution
Instname2
Health Services and Policy Research
Theme2
Health systems / services
Prcname2
Health Policy & Systems Management Research
Projecttitle
Making the most of Canada's "natural laboratory": advancing difference-in-differences methods for unpoolable data
Programtype2
Operating Grants
Orgnameinp2
McGill University
Cihramount2
$371,971
Cihrequipment2
$0
Programname2
Project Grant
Keyworddelim
Administrative Health Data; Difference-In-Differences; Econometric Methods; Health Services And Policy Research
Abstract
Canadian provincial health systems have been called a "policy laboratory" where variations in health policies and health care system structure can be used to inform learning health systems and improve population health. Capitalizing on innovation and experimentation across Canadian health systems generates valuable opportunities for natural experiments and cross-jurisdictional comparative analyses to better understand what works - and what doesn't. One analytic approach that can be used to learn from policy variation in Canada is difference-in-differences (DiD) estimation, which is a simple and effective way to identify causal effects in the absence of randomization. This approach allows us to move beyond quantifying the basic difference in two units to quantifying the effect of the exposure itself, a crucially important parameter in policy analysis. Given its ease of use and interpretability, DiD is an increasingly common analytic strategy in Canadian health economics and health services and policy research. However, Canadian administrative health data are largely siloed by province and cannot be combined into a single data set, which is a significant barrier to DiD estimation. Despite the indispensability of DiD for health services and policy research, no
"best practices" currently exist for conducting DiD analyses in the absence of pooled data. The goal of this proposal is to develop strategies to overcome this barrier. Our team, which consists of experts in econometrics, biostatistics, and health services and policy research, will consider methods to generate DiD estimates using econometric/statistical theory, simulated data, and real Canadian data that is not siloed across provinces (e.g., the CCHS). The methods developed by this project will be relevant for both Canadian interprovincial comparisons and international comparisons.
Competitiondate
202010
Approvedterm2
3 yrs 0 mth
Pinamesdelim
Strumpf, Erin C
Approvedterm
3 ans 0 mois
Cihrcontribution2
~ $371,971~ $0~