Exploring barriers to IVR surveys and the effectiveness of human follow-up calls: insights from a mixed methods study in Uganda.

Background: This study explored reasons why respondents neither initiate nor complete an interactive voice response (IVR) survey and whether call-backs by a human can increase subsequent IVR survey participation.

Methods: We conducted a mixed methods study. Using random digit dialing (RDD), participants were sent an IVR survey (IVR-RDD) to their mobile phone. Participants from the IVR-RDD who either did not pick the phone or terminated the survey within two questions were contacted for a computer-assisted telephone interview (CATI) survey to assess reasons for non-participation. Following CATI completion, a similar IVR survey was sent (post-CATI IVR). Descriptive statistics and adjusted logistic regression models were conducted to assess differences in survey outcomes between the IVR-RDD and the post-CATI IVR groups.

Results: A total of 23 288 IVR-RDD, 9740 CATI and 1000 post CATI IVR calls were made to yield 1.9%, 11.8% and 44.9% response rates, respectively. The most common reasons for non-response or drop-off to the IVR-RDD were being busy, misunderstanding IVR instructions and mistrust of the IVR caller. Compared to the IVR-RDD, the post-CATI IVR increased both contact rate, from [(2.9%; 669/23062) to (7.74%; 1758/22704); adjusted odds ratio (AOR) 2.81, 95% confidence interval (95%CI) 2.56, 3.08, P < 0.001] and response rate, from [(2.25%; 518/23062) to 4.54% (1031/22704); AOR 2.07, 95%CI 1.86, 2.30, P < 0.001], but no impact on the cooperation rate.

Conclusions: Understanding reasons for survey non-response can allow for interventions to improve survey response. Introducing a human interviewer to those who did not complete the IVR survey improves subsequent IVR survey participation rates.