How does social desirability bias influence survey-based estimates of the use of antenatal care in rural Nepal? A validation study

Abstract

Objectives Social desirability bias is often speculated to influence survey responses but seldom studied in healthcare. The objective was to explore whether social desirability scores (SDS) or the presence of interview observers is associated with inaccurate recall and overestimation of antenatal care (ANC) services. Design Longitudinal validation study comparing recalled receipt of ANC services and nutrition components of ANC against direct observations of care. An adapted short form Marlowe-Crowne questionnaire was used to generate an SDS, and the presence of interview observers was treated as a separate exposure. We assessed accuracy and overestimation of recalled receipt of ANC services against observed receipt using log-binomial regression, adjusting for age, education, first-pregnancy and socioeconomic status. Setting Rural Southern Nepal with recruitment from five government health posts. Participants 401 pregnant women. Results Social desirability scores did not significantly predict accuracy or overestimation of most types of ANC care except counselling on nausea. Higher SDS was associated with more accurate recall (adjusted RR, aRR 1.08 (95% CI 1.03, 1.12)) and less overestimation (aRR 0.85 (0.80, 0.91)). The presence of mothers-in-law or husbands during interviews was associated with greater overestimation of the number of ANC visits received by more than three visits (aRR 2.07 (1.11, 3.84)) and (aRR 4.19 (2.17, 8.10)), respectively. Those interviewed with friends present tended to overestimate the receipt of counselling on nausea, avoiding alcohol and not smoking. Conclusion The presence of observers can lead to overestimation of the receipt of ANC care and support the conduct of interviews in private settings despite challenges of doing so in village contexts. Findings that the SDS did not predict the accuracy of most types of ANC care might reflect a reality that such questions may not be sensitive from a social-norms perspective. Additional local adaptation of SDS is recommended.