Developing digital tools for health surveys in low- and middle-income countries: Comparing findings of two mobile phone surveys with a nationally representative in-person survey in Bangladesh.

Abstract

Non-communicable disease (NCD) risk factor data from low- and middle-income countries (LMICs) are inadequate, mostly due to the cost and burden of collecting in-person population-level estimates. High-income countries regularly use phone-based surveys, and with increasing mobile phone subscription in developing countries, mobile phone surveys (MPS) could complement in-person surveys in LMICs. We compared the representativeness and prevalence estimates of two MPS (i.e., interactive voice response (IVR) and computer-assisted telephone interview (CATI)) with a nationally representative household survey in Bangladesh-the STEPwise approach to NCD risk factor surveillance (STEPs) 2018. This cross-sectional study included 18-69-year-old respondents. CATI and IVR recruitments were done by random digit dialing, while STEPs used multistage cluster sampling design. The prevalence of NCD risk factors related to tobacco, alcohol, diet, and hypertension was reported and compared by prevalence differences (PD) and prevalence ratios (PR). We included 2355 (57% males), 1942 (62% males), and 8185 (47% males) respondents in the CATI, IVR, and STEPs, respectively. CATI (28%) and IVR (52%) had a higher proportion of secondary/above-educated people than STEPs (13%). Most prevalence estimates differed by survey mode; however, CATI estimates were closer to STEPs than IVR. For instance, in CATI, IVR, and STEPs, respectively, the prevalence was 21.4%, 17.9%, and 23.5% for current smoking; and 1.6%, 2.2%, and 1.5% for alcohol drinking in past month. Compared to STEPs, the PD ranged from '-56.6% to 0.4%' in CATI and '-41.0% to 8.4%' in IVR; the PR ranged from '0.3 to 1.1' in CATI and '0.3 to 1.6' in IVR. There were some differences and some similarities in NCD indicators produced by MPS and STEPs with differences likely due to differences in socioeconomic characteristics between survey participants.