Finally, mixed methods—combining quantitative and qualitative research—is all the rage.
That’s impossible.” “That’s not reliable or scientific.” “That’s soft science.”
Yet its time has come.
Mixed methods research, which combines the data-driven power of quantitative science with the perspective-changing insights of qualitative research, can add a valuable dimension to implementation science, randomized trials and public health research.
Yet this powerful approach has often been avoided. Quantitative purists see mixed methods as something you do before the real research happens. They’re wrong. Mixed methods can be applied throughout the research process and provide insights unavailable to quantitative methods alone.
Let me give you an example:
The Precursors Study, which has been called “the granddaddy of longitudinal health surveys,” has followed more than 1,000 Johns Hopkins School of Medicine students from the 1948 through 1964 graduating classes. The study receives annual updates from participants and has helped researchers link high cholesterol in young adults with heart disease later in life, depression with increased heart attack risk, and vascular risk factors with late-onset depression, to mention a few studies.
“When people tell their stories in their own words, they share what’s most salient to them—not what we researchers think is important.”
Beginning around 1999, and every three years thereafter, my colleagues and I added questions to the survey about the physicians’ end-of-life preferences. Most did not want aggressive measures. (This was featured in a Wall Street Journal editorial and became the basis for a Radiolab piece.) What the survey questions could not address was why physicians answered the way they did or what concerns they have about end-of-life care. To explore their concerns, we began interviewing participants and family members on the phone.
We learned they are worried that advance directives will not be followed! If physicians cannot direct their own care, who can? These interviews provided new insights into health, treatment, the goals of care, and wishes for how decisions might be made. None of these insights would be possible from standardized questionnaires.
When people tell their stories in their own words, they share what’s most salient to them—not what we researchers think is important. Listening to people should force us to question the assumptions we are making from the ivory tower. It reminds me of the joke about the old fish who comes across a couple of young fish and asks, “How’s the water today, boys?” One young fish turns to the other and says, “What’s water?” In other words, the assumptions we make about our methods are part and parcel of what we do, and we don’t always question what we’ve been trained to believe.
More researchers today are embracing mixed methods because the intractable public health problems we face demand it. A couple decades ago, it was commonly assumed that you couldn’t blend quantitative and qualitative research because the underlying assumptions were so different. Yet the history of anthropology and epidemiology demonstrates that investigators have combined both approaches. They have studied the health effects of social or cultural change, how culture influences exposure to risks for disease, as well as what motivates certain behaviors. Colleagues in International Health have long been using mixed methods because they work in diverse cultures and cannot make assumptions about context, but listening to people can be important in any setting—even in your own city.
Today, I hear more and more from students that they want to learn about mixed methods and incorporate them in their research. It’s become a desired qualification, a marketable skill. And, in my work on NIH study sections, I see more and more proposals that include mixed methods. I and others from the School joined a working group convened by the Office of Behavioral and Social Sciences Research (OBSSR) at NIH to create “Best Practices for Mixed Methods Research in the Health Sciences,” a frequently visited website. In addition to a Summer Institute mixed methods course at the School, I direct an NIH-supported national program for mixed methods research training under OBSSR auspices.
So how do mixed methods work?
Think of it as a “before, during and after” approach. Most investigators realize the value of interviewing potential participants before launching a study to work out recruitment procedures, improve survey questions or zero in on key issues. It’s essential to figure out things from the perspective of persons in the settings you’re working in. However, such “formative research” isn’t the only way to deploy mixed methods.
They are increasingly being used during an intervention to understand how participants are experiencing the intervention or why some drop out. Mixed methods may provide clues to how an intervention works by identifying potential mediators from the participant’s point of view. In implementation research, mixed methods may be central to finding out how medical practices or providers adapt an intervention to specific settings or circumstances.
After a mainly quantitative study is concluded, mixed methods designs with informative sampling (for example, selecting persons who did and did not respond or adhere) could be useful to explain variations in outcomes. Why some participants don’t respond to an intervention may become clearer if they are asked for their own insights. The value of even “null” trials would be increased.
A more complete picture of how an intervention works in the real world may emerge. I hope that investigators will consider the added value of mixed methods. Researchers should consider how this strategy might enhance their work, whether it be in implementation science, community-based participatory research, behavioral interventions or other areas.
The challenge for us all as we approach a research project is to think through our conceptual models and assumptions and deploy the methods—quantitative, qualitative or mixed—that best answer the research questions at hand.