Evaluating the impact of health policies: using a difference-in-differences approach
- 82 Downloads
Constrained healthcare resources worldwide have made evaluating the impact of population health interventions increasingly important to maximize health and equity, while minimizing costs. However, the effects of population-level exposures such as health policies can seldom be evaluated through randomized controlled trials (RCTs). The following article will examine how the difference-in-differences method can be used to estimate the causal effect of such interventions. While this method was formalized and is extensively used in the field of economics (Meyer 1995), its first application is believed to have originated in the field of public health in 1855 (Snow 1855). The difference-in-differences method emulates a randomized design by measuring changes in outcomes over time between exposed and control groups. But unlike an RCT where the researcher randomly assigns exposure status; in a difference-in-differences design, researchers use “natural experiments” to assign exposure status,...
This study was funded through support by Doctoral Awards funded to SS by the Canadian Institutes of Health Research and the Canadian Hepatitis C Network. ECS and EEMM are supported by a Chercheur boursier Junior 2 from the Fonds de Recherche Santé (FRQ-S). The Canadian HIV-HCV Coinfection Cohort Study is supported by the Fonds de recherche du Québec-Santé (FRQ-S); Réseau SIDA/maladies infectieuses, the Canadian Institutes of Health Research (CIHR FDN 143270) and the CIHR Canadian HIV Trials Network (CTN222).
Compliance with ethical standards
Conflict of interest
Authors SS, EEMM and ECS declare that they have no conflicts of interest. None of the authors feel in conflict of interest with regard to this study, and there was no pharmaceutical industry support to conduct this study although MBK has received research grants for investigator-initiated trials from Merck and ViiV Healthcare and consulting fees from ViiV Healthcare, Bristol-Meyers Squibb, Merck, Gilead and AbbVie.
- Bertrand M, Duflo E, Mullainathan S (2002) How much should we trust differences-in-differences estimates? The National Bureau of Economic Research, 8841Google Scholar
- Kmenta J (2010) Mostly harmless econometrics: an empiricist’s companion. Springer, BerlinGoogle Scholar
- Liang KY, Zeger SL (1993) Regression analysis for correlated data. Annu Rev Public Health 14:43–68. https://doi.org/10.1146/annurev.pu.14.050193.000355 CrossRefPubMedGoogle Scholar
- Little RJ, Rubin DB (2000) Causal effects in clinical and epidemiological studies via potential outcomes: concepts and analytical approaches. Annu Rev Public Health 21:121–145. https://doi.org/10.1146/annurev.publhealth.21.1.121 CrossRefPubMedGoogle Scholar
- McCormick D, Hanchate AD, Lasser KE, Manze MG, Lin M, Chu C, Kressin NR (2015) Effect of Massachusetts healthcare reform on racial and ethnic disparities in admissions to hospital for ambulatory care sensitive conditions: retrospective analysis of hospital episode statistics. BMJ (Clin Res Ed) 350:h1480. https://doi.org/10.1136/bmj.h1480 CrossRefGoogle Scholar
- Snow J (1855) On the mode of communication of cholera, 2nd edn. John Churchill, LondonGoogle Scholar
- Strumpf EC, Harper S, Kaufman JS (2017) Fixed effects and difference in differences. In: Oakes JM and Kaufman JS (eds) Methods in social epidemiology. Jossey-Bass, San FranciscoGoogle Scholar