A number of studies have suggested that text message-based interventions can help people improve their health and change their behavior, but we still have much to learn about how users engage with these interventions.
Eric Green, assistant professor of global health and co-founder of tech startup Nivi, is among the digital health intervention developers seeking these kinds of data. He recently led a user engagement analysis of askNivi, a free sexual and reproductive health information service in Kenya and India, to glean insights that will be instrumental in guiding future content development, tailoring and automation of the product.
Green and his colleagues—including Alexandra Whitcomb MS’17—employed an approach called text mining, which uses automated tools to examine large amounts of free-form text, summarize the contents and uncover interesting patterns. They analyzed more than 179,000 messages between users and live agents from September 2017 through January 2019. The study is the first reproducible example in the evaluation literature of how to apply text mining techniques to text message-based interventions.
One notable finding was that while the majority of messages from askNivi users related to the health topics targeted by the product, interest in other health issues also emerged among users. The authors noted that “as artificial intelligence is increasingly incorporated into text message-based interventions like askNivi, opportunities for intervention personalization, tailoring and interaction will grow.” They conclude that text mining can provide valuable data for maximizing the potential of artificial intelligence in digital interventions.
Read the article in Gates Open Research.