In the editorial Responding to Gender-based Violence in Disasters: Grappling With Research Methods to Clear the Way for Planning, Rosborough et al discuss challenges to measuring and reporting on gender-based violence (GBV) in disaster-affected populations.1 The authors also comment on our research contribution, which documented an increase in GBV among female internally displaced people following Hurricane Katrina well into a protracted phase of displacement.2 We greatly appreciate the editorialists' commentary and summary of the salient barriers to adequately assessing the prevalence of GBV in the context of disasters. Here, we seek to clarify our methods and choice of baseline GBV prevalence estimates.3
We assessed the change in the rate of GBV among internally displaced people in Mississippi by comparing rates between 2 phases of sampling, 1 conducted in 2006 and 1 in 2007.2, 4 The editorialists suggested that aggregate GBV statistics for stable populations—such as the 1996 National Violence Against Women Survey, the 2006 Behavioral Risk Factor Surveillance System (BRFSS) intimate partner violence (IPV) survey, or state crime statistics—were potential sources of “proxy statistics” to power our analyses.5, 6 The 1996 National Violence Against Women Survey was more than 10 years old at the time of our study, and the 2006 BRFSS in Mississippi did not include the optional IPV module in the years immediately preceding the hurricane.5, 7 Furthermore, national and state crime estimates are based on different modes of reporting, are typically relevant to 1 component of GBV for a specific catchment area, and are not based on definitions and measurement modalities such as those highlighted in the World Health Organization's Ethical and Safety Recommendations for Researching, Documenting, and Monitoring Sexual Violence in Emergencies.8, 9 Despite the important limitations of using baseline GBV estimates from general populations, we did indeed contextualize our findings by discussing how several of our GBV estimates compared with the state of Mississippi and the nation at large.10
Our decision to use baseline GBV rates from our own study originated from the same concern that the editorialists cite as a major problem: aggregate statistics from stable settings are often “underestimates of the population under study” and the “higher prevalence of vulnerability” among disaster-affected people often precludes direct comparisons to the general population. Given the methodology differences that were used to obtain baseline estimates (including sampling frames, respondent characteristics, item measurement, and representativeness of the results) in our study compared with those suggested above, the use of these statistics to establish our population baseline GBV prevalence estimate would have been particularly problematic. We thus used baseline GBV statistics derived from identical GBV-item wording in our successive surveys, both of which were conducted in similar populations, to provide the best baseline estimates with minimal variability and by using a lifetime recall period with conservative criteria. Although this measurement of GBV was not perfect, it was consistent and an improvement to using estimates drawn from a nonrepresentative population.
We ardently agree with the need for consistent definitions of GBV when conducting GBV research, particularly in cross-cultural conflict and disaster settings. Our survey items corresponded to internationally recognized standards for defining GBV.11, 12 We made clear distinctions in identifying specific variants of GBV (IPV and sexual violence) and the time period in which they occurred to better define our composite GBV measure.
Finally, it is important to note that we did not use major depressive disorder as a “proxy” measure for GBV, nor did we use it as a substitute for baseline GBV prevalence data. Rather, we used mental health measures to cross-validate our measures of GBV.13 Our overall measure of GBV and IPV showed consistent relations across time periods with major depressive disorder, severity of major depressive disorder, and suicidal ideation. Key to these associations was the insignificant interaction term between time of survey and mental health symptoms, suggesting that the relation between mental health and GBV was consistent across time periods. Although this relation remained stable between time periods, the overall rate of GBV increased. If a thermometer is off by 10°, it may not provide an accurate temperature but it will still tell you when the temperature changes.
Scientists should continue to grapple with research methods when measuring GBV in the context of disasters. In our case, complex actions were required to make a humble assertion about a change in the rates of GBV into the protracted phase of displacement following Hurricane Katrina. In the face of extreme measurement challenges, scientists should not shy away from attempting to research this public health issue, one which can have more devastating effects on its victims than the disaster that it follows.
Michael Anastario, PhD
Nadine Shehab, PharmD
Lynn Lawry, MD, MSPH, MSc
Uniformed Services University of the Health Sciences and Johns Hopkins University