Abstract
The increased subscription and ownership of mobile phones have created opportunities to improve health, education, or economic outcomes, including mobile phone surveys (MPS) to collect health data. Most MPS used cross-sectional survey designs. We explored the potential of MPS to collect panel data using anonymous surveys with agreement in age and gender, and participants' retention across survey waves in three low- and middle-income countries (LMICs): Bangladesh, Colombia, and Uganda. Using random digit dialing, participants were recruited from 6 age-gender strata (i.e., 18-29-, 30-44-, and 45+-year-old males and females). Three interactive voice response survey waves were sent at two-week intervals. In Wave 1, the number of complete interviews in Bangladesh, Colombia, and Uganda was 2693, 5912, and 4813, respectively. In all waves, the proportion of 18-29-year-olds responding to the surveys was higher than that of 30-44- or 50+-year-olds. Bangladesh (83.7% in Wave 1) and Uganda (70.1% in Wave 1) had a higher proportion of males than females, while it was different in Colombia (45.6% in Wave 1). Regarding the reporting of age and gender in survey waves, we observed a high agreement in all three countries; the Kappa statistic was 0.89 (agreement: 93.7%) from Wave 1 to Wave 2 and 0.90 (agreement: 94.5%) from Wave 1 to Wave 3. In Wave 1, the response and refusal rates were, respectively, 0.26% and 0.19% in Bangladesh; 0.65% and 0.89% in Colombia; and 2.63% and 0.71% in Uganda. From Wave 1 to Wave 2, the attrition rate was 37.2% in Bangladesh, 43.7% in Colombia, and 39.2% in Uganda. From Wave 1 to Wave 3, the attrition rate was 64.2%, 62.8%, and 58.4% in Bangladesh, Colombia, and Uganda, respectively. Despite high attrition across survey waves, the agreement about responses was substantial in all countries and MPS has the potential to be implemented in LMICs. More research is required to improve the retention and increase enrollment in some sociodemographic groups (e.g., older people or women). Future studies could also be benefitted from adding validation questions to ensure the participation by the same respondent.