MARCH Outcome Evaluation
result we could not link the different components of the intervention with the recorded changes in the different sections of the populations.
Some limitations of this evaluation study deserve mentioning. First, both the outcome and baseline surveys were not based on experimental designs that often involve control group (comparison group). As a result, the study was heavily dependent upon a non-experimental static group comparison of the exposed and non-exposed in the same target populations. As a result we could not ascertain whether or not the comparison group (i.e. the non-exposed respondents) was completely free from being in someway exposed to the MARCH intervention. It may well be that those respondents who did not participate in the MARCH had exposed to some of the intervention messages by way of discussing with friends and family members that were participating in the MARCH (i.e. “contamination effect”). Thus, it is possible that the non-exposed respondents residing within the catchments’ areas of the MARCH intervention may have better knowledge, attitude, etc than the other population residing outside the catchments’ area. In such situation the observed effect of the MARCH intervention in this present evaluation could be an underestimate of the true effect. We couldn’t, however, tease out the “contamination effect” of the MARCH intervention, as this evaluation did not involve appropriate control group. Secondly, the poor quality of the baseline as well as part of the outcome survey data (such as data on condom use and HIV testing self-efficacy, and outcome expectancy, and those related to stigma and discrimination) precluded the evaluation of the full impact of the intervention in West Hararghe. Also in Addis Ababa data on psychosocial constructs concerning condom use and HIV testing (i.e. self-efficacy and outcome expectancy) and information on stigma and discrimination were measured differently at baseline, thus, not used in the present evaluation. Thirdly, the multivariate analyses adjusted for differences in respondents’ characteristics by controlling for age, sex, education, and marital status. Only this limited set of confounding variables was used in order to keep the statistical models simple, it is therefore possible that individual level differences between the exposed
Addis Ababa and West Hararghe