What should the sample size be?

Anemia and Use the HemoCue  from multiple project areas should include at least 450 pregnant and 450 nonpregnant women. Table III-2. Potential problems, consequences, and alternatives for using pregnant women as survey subjects. Problem Consequence Alternative It may be difficult to find enough pregnant women in each and every community surveyed. The sample size may be too small to estimate the prevalence of anemia among pregnant women. Go to the next nearest community to find more pregnant women to complete the sample. In some cultures some women may be reluctant to admit that they are pregnant. Some women may not know that they are pregnant early in pregnancy. The sample of women who report that they are pregnant may not be as anemic as the sample of women who are reluctant to admit that they are pregnant e.g., unmarried mothers. Provide training on how to ask questions about pregnancy in a sensitive manner, test the survey questions extensively, and consider whether female survey workers might elicit more truthful responses. The percentage of women who are pregnant at any one time may be low. For example, only about 12 43350 of the sample of women of reproductive age were identified as pregnant in a Project HOPE survey in Peru. It may cost too much money for travel and per diems to visit enough households to find an adequate sample size of pregnant women. For example, survey workers would have had to visit almost 3,000 households with women of reproductive age to find 350 pregnant women in the Project HOPE survey. Consider using community health workers to identify all pregnant women in the community i.e., create a pregnancy register and select the sample using a random sampling method. If community health workers are expected to identify pregnant women in the community, they may only identify those who use their services. Pregnant women who have access to health services might not be as anemic as women who do not use health services. Provide training and incentives for community health workers to identify all pregnant women in their service catchment areas. Preschool children. As discussed in Chapter I, preschool children often have levels of anemia similar to those of pregnant women. One approach, therefore, is to use data from women of reproductive age and assume that preschool children have a prevalence of anemia that is somewhat higher. But since iron interventions are not as commonly provided to preschool children, data on this group are generally needed to convince donors, government counterparts, project personnel, health care workers, or other decision-makers that interventions for this target group are a priority. Do not include infants who are younger than six months, as they are unlikely to be anemic unless they were premature. In addition, the standard cutoff values for hemoglobin concentration are not well established for infants younger than six months. Mothers may also be reluctant to have a blood sample taken from a very young infant. The group of survey subjects should be limited to infants and young children from six to fifty-nine months of age. 31 School-age children. School-age children may also be selected as survey subjects if school-based iron supplementation is being considered as a means of improving school performance. In countries where school attendance is good, this may be the ideal group to improve the iron stores and status of adolescent girls before pregnancy. At least twenty-five percent of adolescent girls in developing countries will have their first child by the time they are nineteen years old. 6 Men. Men have lower needs for iron than children and women of reproductive age and are likely to consume more iron because the distribution the intrahousehold distribution of iron- rich foods often favors men. Thus, they do not generally have a high prevalence of anemia unless they have conditions other than low dietary intake of iron. If the prevalence of anemia among men is low around five percent, this indicates that low dietary intake of iron is likely to be the main cause of iron deficiency among women and children. If the prevalence of anemia among men is higher and the distribution curve for men shows the same skewed pattern as that for women and children, then factors other than low dietary iron intake are likely to contribute to the prevalence of anemia in the target population. In developing areas of the world, the most commonly occurring nondietary factors contributing to anemia are parasites see Diagram III-2 below. Diagram III-2. Common factors that contribute to anemia among men in developing areas. Iron deficiency Loss of red blood cells Hookworm, schistosomiasis, whipworm Blood loss Anemia Destruction of red blood cells other mechanisms Malaria The addition of men to a survey can be considered a qualitative assessment of whether parasitic infections associated with anemia are likely in the project area. If there is no information about parasites in the project area and the project does not have the human and financial resources to collect and analyze stool samples or there are cultural constraints to collecting stool samples, consider collecting data on anemia among men see Decision Tree III-1 on the following page. If anemia is prevalent among men, then malaria, hookworm, whipworm, andor schistosomiasis are likely to be prevalent in the project area. Step 2. Define the survey area. For most projects, the project area defines the survey area. However, ethnicity, economic development, urbanization, and climate can influence the types of foods that are eaten and exposure to parasites and hence, the prevalence of anemia. If these factors differ greatly in certain sections of the project area and there are adequate resources to conduct more than one survey, divide the project area into subareas and conduct a survey in each subarea. Logistics, funding, or other practical considerations may limit the size and number of the survey areas. 32