Completeness and utility of interview data from proxy respondents in prenatal care research in rural China.

In household surveys, the use of data provided by relatives can increase response rates and generalisability of research findings. This study assessed the quality of data from relatives and the impact of the data source on the association between the use of prenatal care and pregnancy outcomes. Data for 3,673 new mothers and 293 proxy respondents were available from a house-hold survey in 2008-2009 in rural China. Analyses were performed using chi-square test, ANOVA, Kruskal-Wallis test, and logistic regression models. Differences in the studied variables were small, but proxy respondents were slightly more likely to have missing data than the new mothers. Differences and missing data were more common for the use of prenatal care and outcome variables (mode of delivery, place of delivery, birth weight, use of postnatal care, and gestational age at birth) than for the background characteristics of the participants. Husbands' reports were closer to the index reports than that of the other proxies. The associations between the exposures and outcomes were mostly similar between the proxy and index respondents. Relatives can be interviewed instead of women to study prenatal care without a substantial negative impact on study results. Studies using proxy respondents should stratify the analysis by type of respondents.