Study population, sample size determination and sampling procedure

Adult pregnant women (> 18 years) who self-reported being healthy in their second or third trimester were recruited to participate in the study. Women who were not healthy were excluded because some chronic diseases require special diets, which could affect dietary diversity. Pregnant women in their first trimester were excluded because most pregnant women are physiologically unstable and experience dietary disturbances, which could affect their dietary diversity.

A sample size of 612 pregnant women was chosen as the appropriate sample size to estimate the proportion of pregnant women with adequate dietary diversity to within 5%, using the formula [27]: n=Z2P(1−P)d2

We assumed that the proportion of women with adequate dietary diversity would be approximately 50%, which would make our sample size calculation slightly conservative. A sample size of 385 was considered sufficient to estimate the proportion to within ± 5% (as measured by the 95% CI for the proportion). This sample size assumed that a simple random sample of women was recruited; thus, allowances needed to be made for any clustering effects of women within a clinic. Assuming a moderate design effect of 1.5 increased the sample size to 578, and allowing for a 6% of refusal rate led to the final sample size of 612.

A multistage sampling scheme was used to sample 612 pregnant women. At the first stage four out of the eight districts in the Southern Province were selected using simple random sampling. At the second stage in each of the four sampled districts one urban health centre and one rural health centre were selected using stratified random sampling (with the strata being “rural” and “urban”). This gave a selected sample of eight health centres. Since each health centre was expected to receive about 200 pregnant women monthly, this would give an expected total of approximately 1600 pregnant women. Since we required a sample of 612 women, the sampling interval was approximately 3. In each of the sampled health centres a systematic sample of pregnant women was selected with a random start after which every third woman was sampled.