Keywords: CATI, Africa, Asia, nonresponse, measurement, bias

Researchers increasingly use computer-assisted telephone interviewing (CATI) in low- and middle-income countries (LMIC). A nascent methodological literature explores representation and measurement error and bias in CATI surveys. However, this literature is disparate, siloed across disciplines, countries, and research designs. Given that we are in the nascent era of CATI in LMIC, synthesizing methodological literature is critical to understand constraints and benefits of CATI.

Using the Total Survey Error framework, this Research Synthesis summarizes findings from peer-reviewed methodological research on CATI in LMIC. We used a scoping review methodology to identify and review 38 peer-reviewed journal articles. These 38 articles were largely based in Africa (n = 22) and focus on public health (n = 27). Out of the 38 articles, 29 were published since 2020.

Our key findings regarding representation are as follows. Existing literature shows that CATI surveys using a FTF follow-up design are more representative than RDD surveys. However, FTF follow-up designs still face some challenges with representation, even when weights are applied. RDD surveys were representative in two middle-income countries (Colombia and Myanmar), but RDD surveys in Africa and South Asia substantially underrepresented lower socioeconomic status groups, women, rural people, people from less populous regions, and (to a lesser extent) older people.

In general, we find that weights were not successful in addressing coverage and nonresponse errors in these countries. Five studies found the weights created a sample that was representative of other demographic factors not used in the construction of the weights, but four studies were not successful.

From a measurement perspective, evidence showed that CATI surveys produced reliable measurements for questions using binary responses about clearly defined constructs. However, measurement quality was lower for harder-to-define constructs (e.g., physical activity), very detailed questions, numeric questions, knowledge questions, and proxy reports. Evidence about the impact of questionnaire length on data quality was mixed, with studies reporting conflicting results.

In conclusion, our Research Synthesis highlights when CATI surveys can be "fit for purpose" to meet the needs of data users. Our paper shows that CATI can be fit for purpose when face-to-face surveys are not feasible (in remote or inaccessible areas, conflict zones, high frequency data, longitudinal data collection, conducting surveys during health emergencies, or studying multiple countries.) Further, CATI can be fit for purpose for research that does not require nationally representative samples, in the case of studying skilled professionals, respondents recruited from a specific location (school or healthcare facility), or describing changes over time. Finally, our Research Synthesis shows the value of CATI as a complement (not a replacement for face-to-face surveys.)