Design of an Automated Voice Application for Home Monitoring of Diabetes

Given the significant increased incidence of various chronic diseases and conditions such as diabetes, there are enormous challenges to provide high-quality healthcare within the constraints of finite budgets and available resources. This paper describes the issues involved in the design of a voicebased system to support the home monitoring by patients with Type 2 diabetes of their condition. Patients use the telephone to input key readings in either voice or DTMF mode and they receive immediate verbal feedback on their condition. Two versions of the system were developed and compared – a conversational interface, which allowed morenatural input and a more intuitive dialogue flow, and a system-directed interface with carefully designed prompts and confirmations and in which the user input was more constrained. Evaluations showed that the system-directed version performed better in terms of recognition accuracy, transaction success, and call completion time. Some conclusions are drawn from these findings regarding the importance of user-centered design along with implicationsfor solutions involving speech technology,