A new paper published this month in the Journal of Medical Internet Research Research Protocols outlines MyVoice, an ambitious, new adolescent health research platform led by Tammy Chang, M.D., M.P.H., M.S., assistant professor in the department of family medicine. The publication’s lead author is family medicine postdoctoral research fellow Melissa DeJonckheere, Ph.D.
MyVoice was previously discussed in this post, after it was chosen as a technology “success story” by the tech firm Granicus.
MyVoice is a study that deploys weekly surveys to adolescents aged 14-24 via text message. The surveys are brief and open-ended--usually 3 to 5 questions--and focus on different topics each week. From mental health coverage under the Affordable Care Act, to sexual consent, to student loan debt and community policing, MyVoice seeks to give youth a voice on issues that affect them.
The protocol paper outlines the project’s unique approach to longitudinal mixed methods research design--including its utilization of 1.) mobile SMS technology to disseminate its open-ended surveys; 2.) targeted social media advertising--which has successfully recruited over 1,700 participants nationwide; and 3.) natural language processing algorithms to support the qualitative data analysis process.
Human-designed algorithms for rapid data analysis
About the use of natural language processing (NLP) in mixed methods research design, lead author and methodologist Melissa DeJonckheere, Ph.D., notes, “Natural language processing techniques allow us to rapidly understand thousands of responses from youth and augment more traditional qualitative data analysis.” Using NLP, the project can bridge mixed methods, harnessing both the richness of qualitative data and the metric quantity of responses from the over 1,700 consented participants, in rapid data analysis.
The algorithmic approach to data analysis is led by MyVoice primary investigator, Tammy Chang, and executed by a small team composed of MyVoice data scientists, who are based at the U-M School of Information and the Medical School’s Department of Learning Health Sciences, and by a core group of undergraduate research assistants, who are participants in the U-M Undergraduate Research Opportunity Program (UROP).
Strategic science and the sloth
“[With NLP], it means we can get really fast feedback on what is important to youth and how they are feeling about timely policy issues. Policymakers need information quickly, and we want them to have a real sense of what is impacting the health and well-being of youth,” adds DeJonckheere.
The broader goal of the MyVoice research project is to provide the type of timely and relevant information that can directly inform public policy. The recent paper calls the approach “strategic science,” and envisions “research that addresses gaps in knowledge important to policy decisions….derived from the reciprocal flow of information between researchers and policymakers.”
Dr. Chang and her co-investigators have described the “strategic science” of MyVoice to broad audiences of stakeholders, including researchers, policymakers, and youth. The team recently presented a MyVoice TED-style talk at the 2017 MCubed Symposium. On stage at Rackham Auditorium at the University of Michigan, Dr. Chang, along with co-investigators Michelle Moniz, M.D., assistant professor of obstetrics and gynecology, and Kendrin Sonneville, Sc.D., assistant professor of nutritional sciences at the U-M School of Public Health, make the case for why MyVoice’s unique methodology matters.
You can watch the full video here, where they present the problematic pace of traditional academic research.
Policymakers seek science-based evidence to inform their legislation. But traditional research output can take three years, when legislators may need it in three days.
In the video, Chang argues, “What’s wrong with this situation is that the pace and timeline of science doesn’t match what’s needed to inform policies in real time,” she said. “We do not get a good return on our investment for research and we really miss the mark.”
The protocol paper describes specific study design decisions the researchers have made, from the design of policy-specific surveys to the rapid analysis of data, to catch up with the pace of policy.
Accessible and Youth-Centered
Data quality is also a key concern for MyVoice’s impact mission. The project seeks to cultivate a sample that truly represents American youth. They plan to do so by employing new approaches to study recruitment and retention.
For example, the paper notes that previous efforts in adolescent health research tend to draw from high school, college, clinical, and community groups. As a result, the data collected often lacks the perspectives of underrepresented youth, including individuals not associated with social institutions like hospitals and schools.
In response, MyVoice has refined their online strategy for participant recruitment, including paid social media advertising to target underrepresented cohorts in their sample. The project also focuses on continued engagement of enrolled participants, providing “surprise and delight,” in the form of tailored graphics, nominal incentives, and youth-friendly reports on MyVoice data. The goal of these efforts is to make the project accessible to all youth and to center youth values and needs in multiple areas of the study’s design.
DeJonckheere notes, “Adolescents can be a difficult population to engage in research, particularly longitudinal research….MyVoice decided to meet youth where they are.”
Whether it is participant recruitment on Instagram, feedback via Spotify music playlists, or youth-friendly surveys delivered to their key technology devices (their phones), MyVoice makes youth central to the design and implementation of their population-level adolescent health and well-being study.