June 13, 2016
In a recently published editorial, researchers from the UNC Gillings School of Global Public Health departments of epidemiology and biostatistics argue that a “causal impact” framework – one that includes internal and external validation and examines the effects of population-level interventions – could improve public health decision-making.
Gillings School co-authors are Daniel Westreich, PhD, associate professor, Jessie K. Edwards, PhD, research assistant professor, and Stephen R. Cole, PhD, professor, all in epidemiology, and Michael G. Hudgens, PhD, professor of biostatistics.
The article, “Causal Impact: Epidemiological Approaches for a Public Health of Consequence,” appears in the June issue of the American Journal of Public Health.
Westreich and colleagues offer the example of a randomized trial of antiretroviral agents used to prevent HIV acquisition, in which one group of volunteers without HIV are given the drugs, and one group is given a placebo. In the ideal study, comparing the rates of HIV acquisition in the control group (those who receive antiretroviral therapy) and placebo group (who receive no therapy) should inform us about the percentage of people likely to be protected by antiretroviral therapy.
The researchers note that this difference in outcomes between the two groups is sometimes referred to as the treatment’s “effectiveness” – or the public health effect – of the treatment. More work is needed, the authors write, to translate “effectiveness” in ways that reflect the impact of an intervention in a real population, specifically by considering external validity and population intervention effects.
External validity requires a consideration of the impact of possible differences between the study sample and the larger population in which the intervention is to be implemented.
In the HIV example, for instance, it may be that a drug regimen is more effective in preventing HIV in women than in men. If women participate in the study in disproportionately greater numbers, it could appear that the drug regimen is more effective than it actually would be in that larger population.
Westreich and colleagues suggest that researchers must “make efforts to understand the effect of the intervention under real-world conditions.” Although randomized trials and nonexperimental studies continue to be valuable, they only examine questions of internal validity.
“Considering all three [internal and external validity and population intervention impact] – as in the causal impact framework – may help us produce research more relevant to policy making, and thus help produce,” they conclude, “a ‘public health of consequence.’”
Other co-authors are Elizabeth T. Rogawski, PhD, Gillings School alumna and WHIL Innovation Postdoctoral Fellow in infectious diseases and global health at the University of Virginia in Charlottesville, and Elizabeth A. Stuart, PhD, professor of biostatistics, health policy and management and mental health at the Johns Hopkins Bloomberg School of Public Health.
The authors are commended in an invited review of the June issue of American Journal of Public Health, which appears in that issue.
Written by Sandro Galeo MD, DrPH, dean and professor in the Boston University School of Public Health, and Roger Vaughan, DrPH, an AJPH editor and vice dean and professor of biostatistics in Columbia University’s Mailman School of Public Health, the review notes that the framework proposed by Westreich and colleagues “elevates the importance of external validity” by asserting that research will be of most consequence 1) when used to make inferences about the population whose health it aims to improve and 2) when population health science findings are more accessible.