RESUMEN
BACKGROUND: Text messaging systems are used to collect data on symptom prevalence. Using a text messaging system, we evaluated the effects of question load, question frequency, and financial incentive on response rates and reported infant diarrhoea rates in an infant diarrhoea survey. METHODS: We performed a factorial cross-over randomised controlled trial of an SMS surveying system for infant diarrhoea surveillance with treatments: financial incentive (yes/no), question load (1-question/3-question), and questioning frequency (daily/fortnightly). Participants progressed through all treatment combinations over eight two-week rounds. Data were analysed using multivariable logistic regressions to determine the impacts of the treatments on the response rates and reported diarrhoea rates. Attitudes were explored through qualitative interviews. RESULTS: For the 141 participants, the mean response rate was 47%. In terms of percentage point differences (ppd), daily questioning was associated with a lower response rate than fortnightly (- 1·2[95%CI:-4·9,2·5]); high (3-question) question loads were associated with a lower response rate than low (1-question) question loads (- 7·0[95%CI:- 10·8,-3·1]); and financial incentivisation was associated with a higher response rate than no financial incentivisation (6·4[95%CI:2·6,10·2]). The mean two-week diarrhoea rate was 36·4%. Daily questioning was associated with a higher reported diarrhoea rate than fortnightly (29·9[95%CI:22·8,36·9]); with little evidence for impact by incentivisation or question load. CONCLUSIONS: Close to half of all participants responded to the SMS survey. Daily questioning evoked a statistically higher rate of reported diarrhoea, while financial incentivisation and low (1-question) question loads evoked higher response rates than no incentive and high (3-question) question loads respectively. TRIAL REGISTRATION: The protocol was prospectively registered on ISRCTN on the 20th of March 2019 under number ISRCTN11410773 .