Epidemiology 1, 2, 3: study and sample design.

M H HusseinDepartment of Community Medicine, Faculty of Medicine, Cairo University, Egypt.

Search for other papers by M H Hussein in
Current site
Google Scholar
PubMed
Close
,
M K El-SayedDepartment of Community Medicine, Faculty of Medicine, Cairo University, Egypt.

Search for other papers by M K El-Sayed in
Current site
Google Scholar
PubMed
Close
,
M TalaatDepartment of Community Medicine, Faculty of Medicine, Cairo University, Egypt.

Search for other papers by M Talaat in
Current site
Google Scholar
PubMed
Close
,
A El-BadawyDepartment of Community Medicine, Faculty of Medicine, Cairo University, Egypt.

Search for other papers by A El-Badawy in
Current site
Google Scholar
PubMed
Close
, and
F D MillerDepartment of Community Medicine, Faculty of Medicine, Cairo University, Egypt.

Search for other papers by F D Miller in
Current site
Google Scholar
PubMed
Close

Bad sample designs and selection bias have plagued studies on schistosomiasis, and as a result some believe that schistosomiasis is too focal, making it difficult to draw reliable samples. The Epidemiology 1, 2, 3 (EPI 1, 2, 3) sample design, although complex, demonstrates that sampling theory is readily applicable to epidemiologic studies of schistosomiasis. The EPI 1, 2, 3 sampling scheme was designed to achieve the smallest feasible standard errors given EPI 1, 2, 3 objectives and certain logistical constraints. The sample design is a multi-stage selection of villages (ezbas, which were stratified by size) and households within each of 9 purposely selected Egyptian governorates. Villages and households were systemically selected from census frames. The sampling of ezbas was especially difficult because of the lack of complete sampling frames and their wide variation in population size. Ultimately, ezbas were stratified by size and then randomly selected from each stratum. Sample sizes for villages and ezbas and individuals within ezbas were calculated based on EPI 1 and 2 objectives, respectively. No re-selection was made for non-respondents. A 20% subsample of the full sample was drawn for clinical and ultrasonographic examinations. The sample selected from individual governorates closely parallel the age structure of the 1986 census of the respective rural populations. Details of the study design and related methods are given below.

Save