and twenty-first centuries transformed education into a social determinant of illness and disease in the US (Link and Phelan 1995; Masters et al. 2015; Phelan and Link 2015; Wolfe et al. 2018a; 2018b; Wolfe 2019). To extend this line of inquiry, we might elaborate on our initial question and ask whether the association between education and health changed for people who lived before and after the rapid expansion of higher education in the US. Although this is just a slight alteration to our original question, it has major implications for our research methods. Our point is that researchers should always strive to inform a broader theoretical model of health and society, and their choice of research methods should reflect and explicitly acknowledge that model.
The third requirement is reproducibility and transparency (Freese 2007; Long 2009). Reproducibility means that the investigators maintain all of the data and analysis files necessary to reproduce the results they publicly report, and transparency includes reporting pertinent information and decisions alongside the results. These are necessary for ensuring scientific accumulation and enhancing the credibility of findings. We should note that a study can be reproducible without being replicable. A study is replicable when multiple studies using different data (more on data coming up) come to the same overall conclusions. Reproducibility simply means that the researchers have preserved the instructions for recreating their study’s results so that other researchers can extend or correct findings. Maintaining a study’s reproducibility can be a time consuming and thankless job, but it’s a necessary part of conducting good research and contributing to the scientific community.
Our examples thus far are quantitative, but these three requirements apply equally to qualitative work and even descriptive research exploring new areas without explicit guidance from sociological theory. Related, although we organize key topics in quantitative and qualitative research separately, we believe in a fluidity of research designs, taking a dynamic and flexible approach to our empirical investigating of the social world. Science is a cumulative effort, and different research methods shed light on sociological problems from different angles (Collins 1989; Gross 2009; Lieberson and Horwich 2008).
With our basic requirements out of the way, we can now provide a short outline of the sections to come. We begin our discussion by reviewing data and related issues such as the importance of conceptualization and choosing the appropriate level of analysis. We then discuss methods for analyzing data, beginning with quantitative issues like measurement and causality. Next we consider several qualitative methods, including special topics that offer interesting new avenues for research in medical sociology.
DATA REIGN
Assuming you have a health-related question inspired by sociological theory, how do you investigate it empirically? You need data. Without data – and good data at that – you’ll never be able to empirically answer your questions. Unfortunately, data collection is one of the most challenging aspects of conducting original research in medical sociology. Researchers are often forced to make sacrifices as they move from the limitless world of ideas to the restrictive nature of rigorous, scientific observation. All research is pressed on either side by efficiency and bias. Normally these are statistical terms, but we think efficiency and bias are conceptually useful here, too. On the one hand, researchers must collect enough data to accurately answer their question but not so much that they waste precious resources, time and money, collecting extraneous information (we’ll call this inefficiency). On the other hand, if researchers fail to collect enough information to sufficiently observe a sociological phenomenon, they will likely get incorrect (i.e. biased) answers to their questions. We review several of the challenges of collecting data and discuss why addressing them carefully is important for correctly answering research questions.
First and foremost, sociological phenomena are difficult to observe and even harder to measure consistently. This makes conceptualization a critical part of any sociological endeavor, and most researchers spend a large part of their time reviewing prior research in an effort to clearly define the unique pieces of their research questions. Human capital, for example, is a widely referenced concept in medical sociology, especially among scholars studying health behaviors (Mirowsky and Ross 1998). The basic idea is straightforward – human capital improves the constellation of behaviors that maintain health (e.g. avoiding sugar, staying physically active, etc.). Although years of schooling offers a sensible measure of human capital in the US, educational attainment is only an indirect measure of the wide array of cognitive and non-cognitive abilities (e.g. intelligence and self-discipline) that could potentially provide health-related advantages. By clearly conceptualizing these two distinct components of human capital, Herd (2010) found that cognitive human capital played a greater role than non-cognitive capital in mediating education and health among older adults in the Wisconsin Longitudinal Sample.
This brings us to the next challenge, choosing a level of analysis. Social action can be observed at micro, meso, and macro levels (Collins 1981; Fine 1991). Although some sociological theories lend themselves more clearly to specific levels of analysis, most of the major theoretical frameworks in medical sociology can be examined at multiple levels. For our investigation of education and health, should we collect rich descriptions of human experiences with health problems through a few dozen in-depth interviews from people with different educational backgrounds? Or, should we collect less detailed information from a sample of people who represent the entire US population? The answer usually depends on your research question. For example, the question “is the rate of college graduates in US cities associated with cardiovascular mortality rates?” would require a macro-level assessment of city demographics, whereas the question “how does physician-perceived human capital among patients affect patient-physician interactions in free health clinics?” would require micro-level data based on observations of social interactions. Both of these questions are inspired by sociological work on education and health, but each requires a much different approach to data collection. Studies of micro-level sociological phenomena often collect data on social interactions, attitudes, cognition, emotions, and decision-making (e.g. Spencer 2018). Studies of meso-level phenomena generally collect data on individuals, social groups, and social networks (e.g. Hallgren et al. 2020). Studies of macro-level phenomenon often consider country-level data on population trends, institutional and cultural change, laws, and health inequalities (e.g. Bakhtiari et al. 2018).
Data Collection
Research in medical sociology uses a variety of data collection techniques. What follows is a brief outline of the techniques contemporary medical sociologists use to obtain data. The most common tool for data collection is the questionnaire. Questionnaires are used in all types of research, ranging from in-person interviews to simple online surveys, and there’s an entire science about developing them in ways that will improve the chances of actually getting respondents to answer questions (Couper 2017; Schaeffer and Presser 2003).
Non-experimental surveys that collect information with questionnaires are the most common source of data in sociology. Surveys like the General Social Survey (GSS, https://gss.norc.org) provide a wide array of information about the US population (currently around 330 million), including health information, by asking a carefully selected sample of only several thousand people. The GSS is also an example of secondary data, which are data collected by prior researchers. There are several prominent secondary data sources that appear frequently in medical sociology research, e.g. National Longitudinal Study of Adolescent to Adult Health (Add Health, https://www.cpc.unc.edu/projects/addhealth), National Longitudinal Surveys (NLS, https://www.bls.gov/nls), Behavioral Risk Factor Surveillance System (BRFSS, https://www.cdc.gov/brfss). These examples, like many other popular secondary datasets, are publicly available. Although researchers are unable to operationalize their theories exactly as they would please (increasing the potential for bias), secondary data removes the burden of data collection. Most social scientists don’t have resources required
