Cognitive testing of questions about antenatal care and nutrition interventions in southern Nepal.

Large scale surveys such as the Demographic and Health Surveys (DHS) are used to measure the coverage and quality of antenatal care (ANC)-related services. Studies have increasingly validated questions from these surveys, though few have explored respondent comprehension or associated thought processes. This study aimed to use cognitive testing and validation approaches to understand how survey respondents understand questions related to ANC-related nutrition services. The study was nested within a larger validation study in southern Nepal. Pregnant women's receipt of ANC related services was directly observed at five health posts followed by a recall interview at 6 months postpartum. A week later, a survey module was re-administered to 30 women containing 15 questions about receipt of ANC care and specifically nutrition-related services. Detailed probing was used to identify cognitive challenges related to comprehension, retrieval, judgement, and response. Respondents accurately recalled the four specific ANC visits recommended by the government of Nepal but those with more visits struggled to estimate the total number of ANC visits they had made. A number of terms including "antenatal care, "nutrition" and "breastfeeding" were challenging for many respondents to understand. Visits to private providers including for ultrasounds were inconsistently included in ANC visit counts suggesting that question wording could better specify the type of care. Many respondents over-estimated the number of iron folic acid (IFA) supplements taken during pregnancy, and recall was challenging. Calculations were based on estimating the number of months between first ANC visit to delivery, and only sometimes factored in missed tablets. Opportunities exist to improve questions to facilitate better comprehension by respondents through a combination of using local terms and explanations, reordering some questions, and adapting questions to better match respondents' approaches to estimating numeric responses.