Abstract
BACKGROUND: Face-to-face interviews by trained field workers are commonly used in household surveys. However, this data collection method is labor-intensive, time-consuming, expensive, prone to interviewer and recall bias and not easily scalable to increase sample representativeness. OBJECTIVE: To explore the feasibility of using text messaging to collect information on infant and young child feeding practice in rural China. METHODS: Our study was part of a clustered randomized controlled trial that recruited 591 mothers of children aged 12 to 29 months in rural China. We used the test-retest method: first we collected data through face-to-face interviews and then through text messages. We asked the same five questions on standard infant and young child feeding indicators for both methods and asked caregivers how they fed their children yesterday. We assessed the response rate of the text messaging method and compared data agreement of the two methods. FINDING: In the text messaging survey, the response rate for the first question and the completion rate were 56.5% and 48.7%, respectively. Data agreement between the two methods was excellent for whether the baby was breastfed yesterday (question 1) (kappa, κ = 0.81), moderate for the times of drinking infant formula, fresh milk or yoghurt yesterday (question 2) (intraclass correlation coefficient, ICC = 0.46) and whether iron fortified food or iron supplement was consumed (question 3) (κ = 0.44), and poor for 24-hour dietary recall (question 4) (ICC = 0.13) and times of eating solid and semi-solid food yesterday (question 5) (ICC = 0.06). There was no significant difference in data agreement between the two surveys at different time intervals. For infant and young child feeding indicators from both surveys, continued breastfeeding at 1 year (P = 1.000), continued breastfeeding at 2 years (P = 0.688) and minimum meal frequency (P = 0.056) were not significantly different, whereas minimum dietary diversity, minimum accepted diet and consumption of iron-rich or iron fortified foods were significantly different (P < 0.001). CONCLUSIONS: The response rate for our text messaging survey was moderate compared to response rate of other studies using text messaging method and the data agreement between the two methods varied for different survey questions and infant and young child feeding indicators. Future research is needed to increase the response rate and improve data validity of text messaging data collection.