Analysis of missing responses to the sexual experience question in evaluation of an adolescent HIV risk reduction intervention

In evaluating the effectiveness of risk reduction interventions, a methodological issue is how to treat missing data on key sexual outcomes. Population estimates and the evaluation results are affected by whether we drop missing cases, or treat them as ‘yeses' or ‘nos'. We ask how low income African American young adolescent girls in an intervention study panel who legitimately decline to answer whether they have had sex differ from those who answer affirmatively or negatively on: background characteristics, risk related characteristics, intervention group assignment, and subsequent panel attrition. The girls with missing responses were younger than those who affirmed having had sex, and had stricter curfews than those who denied having had sex. Compared with those who answered either affirmatively or negatively to having had sex, girls declining to answer had significantly lower HIV transmission knowledge, lower self-efficacy to protect themselves (through abstinence or condoms), and lower intentions to do so, factors that were positively related to their likelihood of having sex. After the intervention, they were over-represented in the control group that did not receive the HIV risk reduction intervention. Response group interacted with intervention assignment in affecting HIV transmission knowledge and self-efficacy and intention to use condoms. Also, panel attrition for the missing-response group by 6-months post-intervention was 46%, compared with 27% overall. Methodologically, the girls who declined to answer should be included in the analysis, with a control variable for missing responses. In addition, their high risk profile dictates a need for HIV risk reduction intervention.