Читать книгу The Social Causes of Health and Disease - William C. Cockerham - Страница 15

The recognition of social factors inherent in causing sickness or mortality has been late in coming. One reason may be methodological difficulties in specifying the exact underlying explanatory social mechanisms that affect health, since these mechanisms are multiple, complex, and difficult to isolate in order to determine their precise effects. Such difficulties are increased when trying to determine the direct effects of social structures on individuals because of the possible role of other variables that may intervene in the relationship. Nevertheless, the scientific method requires proof that independent variables have specific and measurable effects on dependent variables whether they are structural or not.

Qualitative methods such as participant observation that concentrate on individuals face shortcomings in determining the effects of structure on people. Roger Sibeon (2004) suggests there are limits to what can be achieved by micro-level methods in addressing structural questions, since such methods are not equipped theoretically or methodologically to measure macro-phenomena. However, the Lutfey and Freese (2005) study of diabetic clinics was qualitative, relying on ethnographic data consisting of observations at the clinics; consultations with patients and medical practitioners; video recordings; semi-structured interviews with physicians, nurses, dietitians, social workers, and diabetes educators; and telephone interviews with patients. The study was narrowly focused on capturing the pervasiveness of the causal relationship between socioeconomic status and outcomes for diabetes. Consequently, qualitative studies that are alert to structural influences on the individual behaviors they observe and monitor do have potential for increasing our understanding.

Determining the effects of structure in quantitative studies requires the construction of independent variables having collective properties indicative of such structures. The most powerful structural predictor of poor health is social class or SES typically consisting of measures of income, education, and occupational prestige. Even though each of these variables is distinct and reflects differing dimensions of social stratification, they are nevertheless interrelated and structurally connected (Adler et al. 1994; Wolfe 2015). They can be viewed separately or in combination with each other as structural variables within which class-based behaviors and norms are created and imposed on individuals through socialization and experience. Age, gender, and race can also be measured as structural variables influencing health. Another strategy is to apply class categories to the family/household rather than the respondent/individual. The status of the person (or perhaps persons) in the family/household with the highest level of labor market participation can be conceptualized as providing a master social status to the household representing its collective position vis-à-vis the marketplace. This outcome is evident when the parent’s social standing is passed to their children and the household as a whole is accorded a particular social standing in the community.

As for the effects of neighborhoods on health, an index of living conditions can be constructed from the value of homes in particular neighborhoods or census tracts and the extent of basic utilities, modern plumbing, heating, air-conditioning, hot water, and the like, as well as the presence of parks, recreational facilities, restaurants, pharmacies, and grocery stores. Other health-related variables are the ready availability of physicians, clinics, and hospitals, along with crime rates and various measures of public safety. Variables such as these are not the properties of similar individuals, but those of structures that constrain or enable individuals to lead healthy lives.

Recent developments in statistics for estimating hierarchical linear models now provide efficient estimations for a wider range of applications than previously possible. Hierarchical linear modeling (HLM) makes it feasible to test hypotheses about relationships occurring at different levels and also assess the amount of variation at each level (Raudenbush and Bryk 2002). Briefly stated, HLM tests the strength of the interaction between variables that describe individuals at one level (level 1), structural entities (like households) at the next level (level 2), and sequentially higher levels (e.g., neighborhoods, communities, social classes, nations), if necessary, depending on the variable’s conceptual position in a structural hierarchy. By comparing changes in the regression equations, the relative effects of each level of variables on health outcomes can be simultaneously determined. As Stephen Raudenbush and Anthony Bryk (2002: 5) point out, “the barriers to the use of an explicit hierarchal modeling framework have now been removed.” Therefore, the capability to examine complex social dynamics and the links extending from society to the individual is now possible.

A caveat concerning multilevel measures of social settings involving both collectives and individuals is that such measures in the past have often been subject to problems of ecological inference. What this means is that the association of two variables at aggregated levels may not reflect the association between the same two variables at the individual level. However, rather than treat structural variables as an aggregation or sum of individual-level variables, problems of ecological inference can be overcome by using structural variables that are a direct measure of the structure itself, such as measures of neighborhood characteristics that reflect the neighborhood not the individuals who reside in it (Thisted 2003). This way the direct effects of the neighborhood on individuals can be determined since they are not confounded by individual characteristics. Properly employed, multilevel measures can ascertain the effects of higher levels of social organization on individuals.

There are other multilevel statistical techniques like variance component analysis by maximum likelihood (VARCL) and procedures like MLn and MLWIN that can be used. The point is that adequate statistical methods now exist that allow sociologists to test hierarchical models that better reflect the reality of everyday situations in which individuals experience the layers of social structures that exist in their lives and affect their health.