Format

Send to

Choose Destination
J Clin Epidemiol. 2017 Sep;89:53-66. doi: 10.1016/j.jclinepi.2017.02.017. Epub 2017 Mar 29.

Quasi-experimental study designs series-paper 7: assessing the assumptions.

Author information

1
Institute of Public Health, Faculty of Medicine, Heidelberg University, Heidelberg, Germany; Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Boston, USA; Africa Health Research Institute, Somkhele, South Africa. Electronic address: tbaernig@hsph.harvard.edu.
2
Francis I. Proctor Foundation, University of California San Francisco, San Francisco, USA.
3
Department of Medicine, University of Ottawa, Ottawa, Canada.
4
Department of Economics, University of Göttingen, Göttingen, Germany.
5
Instituto de Saúde Coletiva, Federal University of Bahia, Salvador, Brazil; National Institute in Science, Technology and Innovation in Health (CITECS), Salvador, Brazil.
6
International Initiative for Impact Evaluation, London, UK; International Initiative for Impact Evaluation, Washington DC, USA.
7
Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Boston, USA.
8
Center for Global Health and Development, Boston University, Boston, USA.
9
Institute for Fiscal Studies, London, UK.
10
School of Public Health, Boston University, Boston, USA.
11
College of Education and Human Development, University of Louisville, Louisville, USA.
12
Impact Evaluation, World Bank, Washington DC, USA.
13
Hendrix College, Conway, AR, USA.
14
Centre for Medicine and Society, Freiburg University, Freiburg, Germany.

Abstract

Quasi-experimental designs are gaining popularity in epidemiology and health systems research-in particular for the evaluation of health care practice, programs, and policy-because they allow strong causal inferences without randomized controlled experiments. We describe the concepts underlying five important quasi-experimental designs: Instrumental Variables, Regression Discontinuity, Interrupted Time Series, Fixed Effects, and Difference-in-Differences designs. We illustrate each of the designs with an example from health research. We then describe the assumptions required for each of the designs to ensure valid causal inference and discuss the tests available to examine the assumptions.

PMID:
28365306
DOI:
10.1016/j.jclinepi.2017.02.017
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Elsevier Science
Loading ...
Support Center