Categories

Idea: Collecting and analyzing data requires categories: Have we omitted relevant categories or mixed different phenomena under one label? What basis do we have for subdividing a continuum into categories? How do we ensure correct diagnosis and assignment to categories? What meaning do we intend to give to data collected in our categories?

PT's initial notes on the Cases

Comparative methods for studying socioeconomic position and health in different ethnic communities, Davey-Smith et al. 2000 -- Does SES mean the same thing for different communities?

Marriage and divorce by class in the USA, Hymowitz 2007. I heard her on the radio make the case that divorce rates hid that there were different phenomena and trends in different social classes. (I looked for a more scholarly piece by her, but her book did not include any references even though many people were cited.)

George Brown (UK) and Bruce Dohrenwend (USA) have done research for decades on the relationship between mental illness and life events or difficulties. Brown (as described by Birley and Goldberg 2000) developed methods that tried to expose the meaning of an event for the person and was critical of the US emphasis on "objective" surveys (where the same event, e.g, death of a spouse, might have very different meanings and significance for the subject). Dohrenwend describes his group's eventual realization of this issue, but they still wanted to measure events without having the context fused into the rating of the event.


Class assignment:
The first part of class will be spent discussing the articles in smaller groups. To guide the discussions, we are asking you to come to class with one question that relates to the week's topic on at least 2 of the readings. For example, "in what ways might Brown and Dohrenwend differ in their answer to the question 'Have we omitted relevant categories or mixed different phenomena under one label?'" This could include a question that you are puzzled over and/or a question that interests you and you think would lead to good discussion. Please read the draft substantive statement as well as the readings before preparing their questions. And please write your question on an index card or a piece of paper.

questions (from 9/26/07):
on Brown and Dohrenwend
How does Brown come up with the idea that social events/situations can cause a psychiatric illness and can also be used as treatment? What kind of studies did he do?
What values/judgement com into Brown & Dohrenwend's methods to assess associations between stressful life events and depression/schizophrenia? Implications of these values/judgements?
When Dohrenwend prefers the "normative" ratings over the "subjective" assessment of the pain sufferers, is he implying that the subjective has no causla role in explaining their health outcomes (p. 181)?
Why does Brown not consider genetic factors on depression?
Are dichotomous variables adequate for considering social stress?


on Davey-smith et al. on race/ethnicity

on Hymowitz and marriage/parenthood categories
- In her argument for two-parent families, Hymowitz states that married couples have two of several things, income being one. That assumption may be incorrect; often only high income 2-worker families earn enough to pay for day care. Maybe it's that higher income is the defining factor-not marriage?
- Would not some intra-categories of marriage be feasible to assess children's progress? Such as marriage income, level of education, age at conception.


Substantive statement

In the process of epidemiological research, as in most research, there are many

points at which a researcher’s decision how to interpret a phenomenon may be subject to questioning not the least of which is developing or selecting measurement categories. This stage cannot be taken lightly; it requires background research on the phenomena you are studying, a review of measures used in other studies and validation procedures to ensure your measures and their categories are appropriate. What are ‘appropriate’ categories? From this week’s readings and other investigation and experience, ‘appropriate’ can mean several things, some of which include that: the categories are valid, i.e. they accurately represent what you are intending to measure and represent all options available (e.g. the answers given to survey questions are truthful; to measure diet, all frequencies of meat consumption need to be included); they are tested, either through the use of statistical tests or through or qualitative inquiry (e.g. focus groups); they have relevant explanations for being included (e.g. is including religion really relevant to what you are studying?); they are culturally sensitive; and they have high reliability, that is, they could be applied to similar populations and come up with similar results. Epidemiologically speaking, how good are your categories in identifying who in fact is 'sick'? Gordis (2004) refers to the challenges of determining of a cut-off point of blood sugar count to test/measure positive or negative diagnosis for diabetes. Set it too high and many diabetics will test negative; set it too low, many non-diabetics will test positive. More broadly speaking, from my own experience in helping to develop a screening protocol for depression, substance abuse and domestic violence, it took a planning group of experts and 4 focus groups with a sample of health center patients to come up with what we considered valid and culturally-sensitive measures. The better the measure, the least amount of error in your results. But as we could from this week’s readings, you don’t always see the error unless you ask more questions about your data.

This week’s articles ask us to do just that by digging deeper in examining disease phenomena at the category development stage, suggesting the concept of eco-epidemiology cited in an article from last week (Schwartz et al, 1999). This concept asks us to look more broadly at the influences of social and physical environments and historical contexts on the incidence of disease. In digging deeper we can identify whether the meanings and uses of categories describing the nature of disease distribution are in fact accurate or misleading. Davey Smith et al (2000), for example, argue and exemplify that socioeconomic indicators/categories may not have the same or consistent meanings for different ethnic groups and one must “acknowledge the problems with SES indicators when studying the contribution of SES position to health differences between ethnic groups,” otherwise posing the risk of producing results that are based on assumed genetic or cultural differences. For example, one of the commonly used SES indicators is education, which Davey-Smith found in their research to be confounded by migration status. That is, some immigrants may have received high levels of education in their native country but are not able to transfer their qualifications to the UK, therefore resulting in them taking employment with lower qualifications. As a result, their higher education levels do not index well with, say, income (which may be lower than expected) as a measure of SES. Relatedly, in a study on cancer-related health disparities in women, Karen Glanz et al (2003) refers to subgroup variation within different race/ethnic groups which affects their health outcomes differently: ethnic groups can vary by geographic location, country of origin, SES, and cultural beliefs. She strongly advocates for supplementing published data with primary sources of surveillance and needs assessment information in order to identify the unique characteristics of different ethnic groups…..a more ecological approach. The term ecological suggest that contributing factors go beyond the biological to include the social and environmental contexts of a particular phenomenon. This approach may help overcome what Nazroo cited in Davey Smith et al call “untheorized ethnicity” which can lead to the pathologizing of minority ethnic status.

Examining and addressing disparities in health outcomes has become a stated national priority in this country (and very much in the city of Boston). Building on our discussions last week, hypotheses about what causes the phenomena of health disparities does not just come from looking at the data (although that does happen too) but from previous experience, notions and study. These motivations incline researchers to look at the data showing disparities and ask so what? What does that really mean that, e.g., more Black babies in Boston die in their first year of life compared with White babies, even controlling for level of education of parents? What categories need to be added to the study to elucidate the causes? Access to prenatal care, cause of death, SES, ethnicity? Racist medical practice? How do we measure racism?

Disturbingly, the topic of race/ethnic health disparities is so important that, according to the Scientific American article Louisa sent us (thank you!), the FDA approved a congestive heart failure treatment drug (BiDil) as an ethnic-specific drug that was not only based on bad science and but also resulted in pooling ALL African-American people into one solitary risk group.

George Brown brought the concepts of finding accurate meaning and diagnosis to the relationship between social factors and depression and schizophrenia. Birley and Goldberg (2000) and Dohrenwend et al. describe social psychiatrist George Brown’s premise that similar events (i.e. life course /stress events) on a checklist can have different meanings for different individuals. As a result, he developed a methodology that built upon the checklist format using pre-selected lists of possible stressful events, but instead of placing ‘arbitrary values’ to the events, he employed detailed interviews to identify each patient’s life circumstances, and ‘objective’ raters, who were part of the research team but not the patient of the clinical staff, to assess the significance of the event. The latter method was designed to avoid intracategory variablility, that is, the different interpretations that can manifest regarding the meaning of just one category. Brown’s approach of emphasizing social factors makes sense because it is thorough and comprehensive, however, the labor intensity of rating may not be practical and surely be expensive on the part of the researcher. Further, Birley and Goldberg list some of Brown’s shortcomings which I see as pretty critical to valid research, particularly his preference for dichotomizing social variables. Dichotomizing variables calls for collapsing categories and blurring distinctions of meaning.

Finally, in a lecture to the Heritage Foundation, Kay Hymowitz cites marriage and its accompanying family structure as the primary protective factor to the health and welfare of children. She interprets marriage as a “human universal” institution that defines the rights and responsibilities of parenthood. In short, her meaning of marriage suggests more of a child-rearing arrangement. It appears that even her critics agree that marriage has changed (see Kane and Lichter 2006 and Berdick, 2007) over the years from traditional roles of husband, wife and parent to an arrangement more focused on finding a soul-mate for life. Of relevance to this class is how is marriage defined today? And, given today’s circumstances, how should marriage be defined? Is this like Brown’s life-events research, where we need to explore what marriage means to individual couples before examining it?

Hymowitz attributes the “sickness” of marriage to a change in the 1960s when the bond between marriage and childbearing was erased. The resultant increase in divorce rate and out-of-wedlock births has given rise to social ills caused by single parenthood. She cites statistics regarding the welfare of children to single (never married and divorced) mothers: 36% of single female-headed families vs. 6% of married couple families are in poverty; 92% of children whose families make over $75k are living with their two parents, when only 20% of children under $15k are living with both parents; children from single parent families (from both divorce and unwed child-bearing) are more prone to social problems: school failure, delinquency, crime, early pregnancy, emotional difficulties, etc. In the lecture, Hymowitz does not expound on what she advocates for interventions to fix a ‘sick’ marriage, but one can assume that she would advocate the promotion of marriage, similar to the prevailing Bush agenda. And while critics agree that unwed parenthood may lead to issues with children, some researchers, including those affiliated with the Campaign to prevent Teen Pregnancy, would say that the problem is not lack of marriage , but out-of-wedlock childbearing: marriage is not necessarily the solution for teenagers, and for older women, unwed mothers tend to lack good marriage prospects because single mothers tend to marry men with lower than average employment and incomes that will not help them financially (Kane and Lichter, 2006). Funding that currently goes toward marriage and responsible fatherhood programs may be better applied to intervene earlier by educating teens about healthy relationships, marriage and pregnancy prevention. This case addresses one of Prof Taylor’s idea questions: What meaning do we intend to give to data collected in our categories? Perhaps Hymowitz’s views on the sanctity of marriage has prevented her from examining the potential disadvantages of marriage to select populations.

References
Andrea Kane and Daniel T. Lichter, Reducing Unwed Childrearing: The Missing Link in Efforts to Promote Marriage, Brooking Institute Report, April 2006.

Berdick, Chris, The Greedy Marriage: Two scholars argue that good spouses can make bad neighbors. The Boston Globe, September 16, 2007.

Birley, J. and D. Goldberg (2000). George Brown's contribution to psychiatry: The effort after meaning. Where Inner and Outer Worlds Meet. T. Harris. London, Routledge: 55-60.

Davey-Smith, G. et al. (2000). Ethnicity, health and the meaning of socio-economic position Pp. 25-37 In Graham, H., Ed. Understanding health inequalities. Buckingham [England], Open University Press.

Dohrenwend, B. P., K. G. Raphael, et al. (1993). The structured event probe and narrative rating method for measuring stressful life events. Handbook of Stress: Theoretical and Clinical Aspects. L. Goldberg and S. Breznitz. New York, Free Press: 174-199.

Glanz, K. Croyle, Rt, Chollette, VY, Pinn, VW. Cancer-related Health Disparities in Women. American Journal of Public Health. February 2003, Vol. 93, No. 2.

Hymowitz, K. S. (2007). "Marriage and Caste in America: Separate and Unequal Families in a PostMarital Age." Heritage Lecture #1005.


Response

As Elizabeth and Peter have pointed out in their comments and substantive statement, categories pose a significant challenge to the study of epidemiology and public health practice. As the controversies over the changing “race” categories in the last decennial census and the use of ethnically-targeted medications indicate, classification by race/ethnicity is a primary mode by which researchers and practitioners draw conclusions about health outcomes. Referring to the article by Davey Smith et al., Elizabeth points out some of the more salient problems that may accompany epidemiological studies employing ethno-racial differences as the principal means of categorizing findings. Socioeconomic status (SES), or class, is another primary means of classifying health outcomes. However, there is debate as to whether class and race confound one another in researching health disparities, and similar to migration status interfering with a clean interpretation of educational outcomes as in the Davey Smith et al. article, factors such as marriage, to which Hymowitz assigns the responsibility for societal ills or advantages, may confound indicators of SES.

Furthermore, attempting to draw conclusions about health outcomes through the use of such broad categories as race and SES requires a thoughtful and detailed understanding of what comprises each of those areas. For example, when discussing health disparities between Latinos and non-Latino whites in the United States, are we talking about all Latinos or one particular nativity group? Are we talking about adults, adolescents, children, or all of the above? Are we talking about immigrants? If so, are we lumping all immigrants into one category, or are we trying to determine how duration of stay in the U.S. affects health in addition to whatever we may assume to be the additional implications of immigrant status and Latino ethnicity? These questions just glance off the tip of the iceberg, but they further elucidate the arguments some of the articles for this week invoke, i.e. it is essential to not only have a clear understanding of what you intend to measure and how you are defining the objects of measurement, but additionally to have a solid understanding of the issues surrounding your hypotheses (as they relate to classification) and a willingness to acknowledge that you may always have more work to do in order to capture information comprehensively and with integrity.

As some of the articles for this week discuss, however, race and class are just two broad divisions we draw when attempting to determine risk factors and ultimately causes of disease. Within these overarching categories, there are many additional divisions, and beyond these categories, there are many other questions about how to classify factors related to health. The articles for this week provide a good basis on which to begin thinking about the many aspects of health research that require us to draw lines, with Davey Smith pointing out the problems that may arise as a result of drawing lines in the wrong places, Hymowitz landing on a point of the spectrum distant from Davey Smith where she feels quite comfortable drawing a very specific line, and the other authors struggling with exactly how to conceptualize and connect the many lines they wish to include in their research.

Developing appropriate frameworks and instruments as Dohrenwend and Brown and colleagues attempt to do, is the health researcher’s means of assessing and validating the choices he or she makes about what categories are important. Such instruments also provide excellent bases on which to witness the expansive nature of categories and the many ways these categories may be employed. Elizabeth mentions her efforts to develop a screening protocol and the Dohrenwend piece describes the SEPRATE scale to measure life events. These two examples of the myriad scales and methodologies that exist to measure certain aspects of health provide a convenient bridge to probing the issue of stress a little more deeply as it relates to the challenges and questions inherent to a process of categorization.

The Birley and Goldberg article and the Dohrenwend chapter present two different means of assessing stressful life events in order to determine the impact of those events on health manifestations and outcomes. Each piece refers to the Holmes and Rahe stressful life events checklist previously developed and the methods of Adolf Meyer earlier in the century. These references paint a picture of evolving, and we might say, increasingly complex visions of categorization as employed to study the physiological effects of stress. Since the publication of these readings, many additional tools have been developed to study the relationship between life events and stress; some target certain ethno-racial groups, some tackle the challenges that Dohrenwend and colleagues raise with respect to their own instrument, some focus more on the interaction between specific biological indicators and self-reports of stress, etc.; and this refers only to those instruments that use stressful life events as their organizing principle. (1-6)

Beyond these, there are countless other manners in which people are currently studying stress and its effects. For example, the ideas of allostasis and allostatic load have been gaining traction in the epidemiology literature, where allostatic load is a physiological state constructed to tell us something important about the effects of stress on a body’s ability to maintain healthy functioning.(7-9) Another example in which researchers have used the Dohrenwend and Brown approaches to practically legitimate the use of again more general measures of stress is the Kessler-6 (K6) scale, which the Centers for Disease Control and Prevention (CDC) employ in the National Health Interview Survey, a major source of health information in the U.S. The K6 uses a simple cutoff score to determine levels of distress in respondents.(4) The proliferation of such tools underlines the slippery nature of categories – they change depending on individual and group perception and interpretation, which in turn may be shaped by the theoretical foundations specific to certain fields of research, the cultural orientation or geographic location of the researchers/practitioners, researcher limitations (e.g. lack of cultural competence, monetary restrictions), etc.

As we see from the example of stress, the process of categorization is manifest at every stage of the research endeavor. It is not simply a matter of choosing your population and health outcome and selecting categories from that point (which of course is not actually simple at all). Rather every action you take across your stages of inquiry requires some level of categorization or classification. As such, my thoughts above are only an entrée to a discussion which in itself may be difficult to guide in one direction versus another depending on the different epidemiologically-oriented topics the articles and questions for this week inspire us to examine further; a very discussion on health and its dissection requires us to undertake a process of categorization in deciding where to focus our comments in group discussion and what pieces of each reading we took to be most significant.

So I leave you with a question: how do we reconcile the necessity of creating categories or drawing classifications in order to get our arms around the study of such an immensely complex field as epidemiology without compromising the integrity of our findings and the health of our subjects? (lh)

References
  1. Belli RF, Shay WL, Stafford FP. Event History Calendars and Question List Surveys. Public Opinion Quarterly 2001;65:45-74.
  2. Caspi A, Moffitt TE, Thornton A, et al. The Life History Calendar: A Research and Clinical Assessment Method for Collecting Retrospective Event-History Data. International Journal of Methods in Psychiatric Research 1996;6.
  3. Ice GH, James GD eds. Measuring Stress in Humans: A Practical Guide for the Field Cambridge: Cambridge University Press, 2007.
  4. Kessler RC, Andrews G, Colpe LJ, et al. Short screening scales to monitor population prevalences and trends non-specific psychological distress. Psychol Med 2002;32:950-76.
  5. Kessler RC, Wethington E. The reliability of life event reports in a community survey. Psychol Med 1991;21:723-38.
  6. Lyketsos CG, Nestadt G, Cwi J, et al. The Life Chart Interview: A Standardized Method to Describe the Course of Psychopathology. International Journal of Methods in Psychiatric Research 1994;4:143-55.
  7. Seeman TW, Singer BH, Rowe JW, et al. Price of Adaptation -- Allostatic Load and Its Health Consequences. Archives of Internal Medicine 1997;157.
  8. Massey DS. Segregation and Stratification: A Biosocial Perspective. Du Bois Review 2004;1:7-25.
  9. McEwen BS, Lasley EN. The End of Stress as We Know It. Washington D.C.: Joseph Henry Press, 2002.



Annotated additions by students

Hunt LM, Schneider S, Comer B. Should ‘‘acculturation’’ be a variable in health research? A critical review of research on US Hispanics. Social Science and Medicine 2004;59:973-986.

The article for this week’s class by Davey Smith et al. questions the wisdom of using broad assignations of race and ethnicity to definitively explain health disparities across large, multi-ethnic populations. In this article, Hunt and colleagues add to this debate by further examining the use of the variable and not always well-defined construct of acculturation in health research among Latinos in the U.S. The article traces the recent increase in published studies on how acculturation may help to explain differences in health behavior and outcomes between Latinos and other ethno-racial population groups in the U.S. The authors argue that this increase has occurred despite lack of a common definition of acculturation in the field of public health, or an understanding of the mechanisms through which acculturation, in its alternating definitions, impacts health. Furthermore, grouping all peoples of Hispanic origin or lineage into one population is neither warranted nor advised in health research as intra-group variability may be quite significant. Thus, acculturation becomes a construct built on an already problematic population grouping. Additionally, the article points out the need for more mixed method approaches to such studies, echoing the tradition of George Brown and Dohrenwend and colleagues to entertain contextual factors in analyzing the natural history of disease. (lh)

Pickles, A. and A. Angold (2003). "Natural categories or fundamental dimensions: On carving nature at the joints and the rearticulation of psychopathology." Development and Psychopathology 15: 529-551.

Should we think of psychopathologies in terms of discrete categories (e.g., you are either clinically depressed or you are not) or a continuum (e.g., everyone is somewhere along a scale of being clinically depressed). The authors conclude that there are circumstances where it behaves each way and our methods need to allow for both. The discussion covers many, often technical issues. (PT)

Poland, J. Bias and schizophrenia. pp. 149-161 in P. J. Caplan and L. Cosgrove, eds. Bias in Psychiatric Diagnosis. NY: Rowman & Littlefield.

Poland argues that the category schizophrenia does not "have anything useful to add to clinical practice concerned with severe mental illness." It lacks construct or predictive validity, is heterogeneous with respect to criteria and biological processes, and "obscures the complexity that must be managed if clinical goals are to be pursued effectively."

Roberts, S. 25th Anniversary Mark Elusive for Many Couples. New York Times, September 20, 2007

More than half the Americans who might have celebrated their 25th wedding anniversaries since 2000 were divorced, separated or widowed before reaching that milestone, according to the latest census survey. The trends are downwards for every anniversary (10th, 15th, 20th..)


Trostle, J. A. (2005). Epidemiology and Culture. Cambridge, Cambridge University Press.

"Interpreting epidemiology as a cultural practice helps to reveal the ways in which measurement, causal thinking, and intervention design are all influenced by belief, habit, and theories of power." (from the bookends)