If quantitative methods are the “head” of a study, qualitative methods are the “heart.” Each performs a critical function, and helps researchers glean a richer, more contextual understanding of an issue so they can better serve the health of their communities.
At the New Connections 2017 Symposium in July, panel discussants and workshop facilitators shared with participants how they successfully incorporated mixed methodologies into their research.
Below are a few highlights:
Food Purchasing Behaviors
Symposium and Regional Meeting co-chair, Dharma Cortes, PhD, led a workshop where she shared details of two studies in which she used quantitative and qualitative methods. Both studies focused on improving the food purchasing behavior among low-income Spanish-speaking Latinos. Cortes, a senior research associate at Harvard Medical School, recruited families in this population, conducted baseline interviews with them to better understand their food purchasing decisions, and conducted a quantitative analysis of their grocery visits. To engage the whole family in the study, she gave the kids cameras to document what they ate at home. At the end of the study, she found families decreased the total number of calories per dollar before and after their education on nutrition. Cortes is currently working on a mobile app known as “Lista” that helps families develop healthy shopping lists, keep track of their expenses, and monitor their weight.
Mixed method studies can shed light on why certain treatments are not effective. Robert Dunigan, PhD, a senior research associate at Schneider Institutes for Health Policy at Brandeis University, conducted a study on the impact and effectiveness of treatment provider incentives for African American and Latino consumers. While conducting interviews, Dunigan learned that smaller programs serving Latino males with a history of opioid abuse and criminal justice involvement provided a residential setting in which they could live. Many people who had graduated from the program returned to mentor the current residents; Latino males thrived in this environment. When it was time to transfer them to the next level of care, which did not offer that cultural support, it was much more difficult to get the same individuals to stick with the program.
Grief and Illness
“You see the most intimate images in the most public spaces,” said Rebeca Pardo, PhD,referring to her project on documenting illness narratives that are shared online. When Pardo, a visiting scholar at Harvard University and professor of photography at the Universitat de Barcelona, studied Alzheimer’s visual narratives in Barcelona, she used mixed methods such as direct observation, interviews, ethnography, and quantifying online reactions such as likes and followers. These were images of people and caregivers in the Alzheimer’s community shared on public networks. She was interested in understanding how documenting these narratives online helped change the social perception of them, and built an online support community. A relatively new field of study, Pardo acknowledged the challenges of quantifying images and the ethics associated with privacy and permission.
“Think about the ethical ramifications of machine learning.” This was the constant refrain from Benjamin Cook, PhD, as he shared the value of using machine learning for everything from identifying suicide risk to identifying claim fraud patterns. Machine learning allows software applications to become more accurate in predicting outcomes without specifically being programmed to do so. Cook, an assistant professor of psychiatry at Harvard University, explained how natural language processing (NLP) at its core converts unstructured data, such as clinician notes, into structured data (e.g., variables), allowing researchers to run prediction models. These models can be used to track sentiment about a patient and find words that are predictive of suicide. Much like Pardo, Cook reminded attendees to think about the ethical issues surrounding NLP, such as safeguarding the privacy of their patients, making leaps with their predictions, and building unfair biases into their models.