This course is only offered in English | Le cours est offert en anglais seulement

Spaces limited. Register early to avoid disappointment.

This webinar has now passed.       

Date: March 19, 2014

Time: 1:30 – 4:00 PM EDT


Ruth Croxford, MSc, PStat Senior Research Coordinator Institute for Clinical Evaluative Sciences (ICES) ruth.croxford@ices.on.ca

By popular demand, Ruth Croxford will be leading this CSEB webinar on the use observational data. This webinar is considered a follow-up to the successful webinar “Introduction to the Use of Observational Data in Health Services Research” held in December 2013.


Using Observational Data in Research: Using the Propensity Score to Adjust for Bias

Course Description: Observational data are data that were not originally collected for research purposes.  This presentation will focus on the data contained in anonymized versions of the databases used to administer Ontario’s health care system. The timing of this webinar is fortuitous: this spring the Institute for Clinical Evaluative Sciences (ICES) will launch a program that will make the data in these databases available, at marginal cost, to Canadian researchers.   The ICES program is part of CIHR’s SPOR (Strategy for Patient-Oriented Research) initiative.

Course Outline

  1. What is observational data?
  2. A brief overview of the health care-related administrative databases
  3. What is a propensity score; how is it calculated
  4. How is the propensity score used to adjust for treatment bias
    1. Using the propensity score as a covariate in a regression model
    2. Stratification on the propensity score
    3. Matched analysis using the propensity score
    4. Inverse probability weighting

The course will use SAS programming examples to illustrate how the propensity score is calculated and how a propensity score-matched analysis is performed in order to compare two treatments. Although the programming examples will use SAS, familiarity with the programming syntax (not point and click, but the actual programming) of any statistical languages (e.g. Stata, SPSS, SAS, R) should be enough to follow the examples.

Recommended Readings