Interactive voice response surveys as a method for increasing the representativeness of rural respondents in a mortality mobile phone survey: Findings from Malawi.

Objectives: Our study aims to (i) evaluate the yield and costs of a fully automated interactive voice response survey as a screening tool for identifying rural respondents for participation in a mortality mobile phone survey, and (ii) compare mortality survey call outcomes among interactive voice response pre-screened and unscreened numbers.

Methods: In order to identify respondents living in rural areas, a short interactive voice response survey was conducted among 24,924 unique mobile phone numbers to determine place of residence (Rural vs. Other). We calculated the proportion of rural numbers derived from the interactive voice response survey among all numbers dialled. Mobile phone numbers screened with interactive voice response were then combined with those generated via random digit dialling and used in a national mortality mobile phone survey in Malawi. Final dispositions for each mobile number dialled were compared for both groups by testing the difference in proportions.

Results: Approximately half of all phone numbers dialled in the interactive voice response survey were answered, and among them, 33.9% indicated that they lived in a rural area. The cost per completed interactive voice response was US$8.75 and just under half of the numbers screened by interactive voice response later resulted in a completed mortality mobile phone survey, at a cost of US$17.4 per completed mortality survey. In comparison, less than a quarter of the numbers that were not screened through interactive voice response resulted in a completed mortality survey (45.3% vs. 22.3%, p <0.001). On average, 12 call attempts were required to complete a mortality survey interview in the unscreened group, compared to 6.3 call attempts among the interactive voice response pre-screened numbers.

Conclusions: Interactive voice response surveys can be used to increase the representation of rural respondents in mobile phone surveys at an acceptable cost. Modifications to the interactive voice response survey process (e.g., survey timing and number of call attempts) should be explored further to increase engagement.