Modified two-stage cluster sampling: Methodology for post-Katrina health outcomes research

Hurricane Katrina was one of the worst natural disasters in U.S. history and the public health impacts of such a natural disaster are immense. As in many post-disaster communities, infrastructure and population are disrupted creating a need to modify research methods used to obtain primary data. The UCLA Center for Public Health and Disasters is conducting research using an IRB approved protocol to assess the public health impacts of Hurricane Katrina in the communities of New Orleans and Slidell, LA; and Biloxi and Gulfport, MS. Using a modified two-step cluster sample, modeled on the WHO Expanded Programme on Immunization method, we created a sampling frame that retains probability within both steps. Thirty random block groups were selected for each community, based upon 2000 Census Tract data. Households within the selected block groups were then assessed for occupancy. A complex algorithm was developed to help assess occupancy in New Orleans and other affected communities. Occupied household addresses were utilized to create a new sampling frame from which seven random addresses were selected for each block group. These addresses represent a cluster from which data regarding structural damage, exposure to potentially hazardous conditions (environmental contaminants), length of time exposed, and length of time with limited access to food/water and medical care is being collected. Interviewers are conducting on-going data collection via face-to-face household interviews in these affected communities. The modified two-stage cluster sampling model that is currently being used to conduct research in the post-Katrina Gulf Coast study will be presented.