A multi-country comparison between mobile phone surveys and face-to-face household surveys to estimate the prevalence of non-communicable diseases behavioural risk factors in low- and middle-income settings.

BACKGROUND

Although mobile phone surveys (MPS) are routinely used to collect health information in high-income countries, concerns remain about the impact of bias on population-level estimates in low-income settings and validation studies are lacking. This study aims to compare non-communicable diseases (NCDs) risk factor estimates obtained from MPS and nationally representative face-to-face household surveys in six low- and middle-income settings.

METHODS

The MPS contained core questions from the standard STEPwise approach to NCD risk factor surveillance questionnaire. MPS sampling frames were generated by random digit dialling, while data collection was done by interactive voice response and SMS. At the same time, a nationally representative household survey (WHO STEPS) was conducted using multi-stage sampling. Participants aged 18 and older were included. Absolute differences and prevalence ratios, with 95% CIs, were analysed. The distribution of the differences between estimates by sex, age and education was also explored.

RESULTS

MPS and STEPS surveys were conducted in Ecuador, Malawi, Morocco, Zambia, Mumbai (India) and Sri Lanka between 2017 and 2022. Overall, MPS estimates of NCDs were most similar to STEPS estimates in Ecuador and Sri Lanka, and most dissimilar in Mumbai and Malawi. Broadly, smoking tobacco, fruit and vegetable consumption, and current drinking questions performed similarly across settings, whereas questions on smokeless tobacco, salt intake and hypertension yielded dissimilar results.

CONCLUSIONS

MPS estimates were most similar to household estimates in settings with high levels of mobile phone ownership. MPS have the potential to serve as a valuable tool to monitor and address NCD risk factors, in addition to traditional face-to-face household surveys. However, producing nationally representative MPS estimates requires careful adjustments to sampling strategies, addressing coverage biases and overcoming technological limitations. Currently, face-to-face household surveys reach a more representative sample of the population, including those in remote and lower educational demographics.