The cumulative prominence by respondent (ranging from 0-5) was then used to calculate the mean prominence for each category. Thematically similar individual categories were grouped under specific headings (e.g. related to dehydration among somatic symptoms) for the analysis selleck of broader concepts of experience, meaning and behaviour. Calculation of the grouped prominence followed the same procedure as with the individual variables. To identify significant differences for cholera between the two sites and between sexes, a non-parametric statistic, the Wilcoxon rank-sum test, was used when comparing prominence variables; the Pearson Chi2 and Fisher’s exact test were applied when comparing proportions.
This particular approach to comparing prominence, which has been widely used in other cultural epidemiological studies, takes more information about a category into account than a simple comparison of frequencies of report without considering how they are reported. A similar series of questions were asked to elicit shigellosis-related illness experience, meaning and help-seeking behaviour. The same categories that were coded for cholera were also coded for shigellosis. Comparative analysis between the two conditions considered only spontaneously reported categories, because the interview coded only spontaneous responses for shigellosis. The proportion of positive responses by category was tabulated individually for each vignette, and for a report in both vignettes. To determine whether a category was associated more with one vignette than the other, McNemar’s Chi2 test for paired data was used.
To examine whether or not individual categories were differentiated between both conditions, Cohen’s kappa was calculated. The kappa statistic indicates the strength of agreement for a categorical assessment, corrected for agreement by chance. The analysis identified the two conditions as distinct for a category if the kappa coefficient was below 0.4, a level commonly accepted as a threshold for moderate agreement . Narrative information was written down during the interview in Kiswahili, then translated into English and typed in a word processor. The qualitative software MAXQDA, version 2007, was used for managing the textual data and to facilitate further analyses of findings from quantitative data. Quantitative data was entered twice and verified in Epi Info software, version 3.
4.3, and cleaned. Statistical analyses were done with Stata, version 10. Sample size The sample size calculation was based on comparison of mean prominence of categories of distress, perceived causes, self treatment and outside help seeking for peri-urban�Crural and female�Cmale differences. Entinostat The detection of a difference of 0.5 between prominence means with equal standard deviations of 1.