The Study of Potentially Stigmatizing Conditions: an Epidemiologic Perspective

Arthur L. Reingold, M.D.

University of California at Berkeley

Suneeta Krishnan. M.S., Ph.D.

http://www.stigmaconference.nih.gov/ReingoldPaper.htm 

Introduction

That various medical conditions or diseases and behaviors are stigmatizing within human society is attested to by numerous early written works, including the Bible.  In the Old Testament, a skin disease commonly believed to have been leprosy is clearly portrayed as a divine punishment for moral lapses and as a cause for removal from society and social isolation.  Some conditions, such as leprosy, appear to have been severely stigmatizing in virtually every known culture and time period for which evidence exists.  Other conditions and behaviors, however, such as mental retardation and abuse of women, have almost certainly been met by widely varying degrees of approbation or acceptance in different cultures and at different times in the history of or by different strata of the same culture.

Many different types of behaviors, conditions, and diseases may be stigmatizing.  In an effort to organize such behaviors and conditions into logical and distinct groups, we have divided them into behaviors; structural abnormalities; functional abnormalities; and contagious diseases.  (See table 1).  These groupings are not, however, mutually exclusive.  That is, a disease like leprosy, in addition to raising concern about contagion, produces structural and functional abnormalities through damage of the tissues of the face and extremities.  Similarly, individuals born with both cleft palate and cleft lip can have abnormalities of speech in addition to the readily apparent structural anomaly of the face.  Furthermore, some conditions that have been associated with stigma at various times and in various places, such as cancer, do not fit readily into any of these groups.

In this article, we have assembled the information available about the current (or, based on historical information, the likely current) global prevalence of only selected conditions and behaviors associated with varying levels of stigma, choosing conditions from each of the groupings.  We have selected the following conditions to review here:  alcohol and drug use, and spouse and child abuse (behaviors); cleft lip and palate (structural abnormalities); mental retardation, schizophrenia, and epilepsy (functional abnormalities); and HIV/AIDS (contagious diseases).  In selecting and summarizing these data from various published and unpublished sources, our focus has been on gathering information that can be used to help estimate the number of individuals with such conditions and, hence, the approximate magnitude of the problem of social stigma attached to selected medical conditions and behaviors worldwide.  However, we have not attempted to gather the information needed to describe the extent to which the various conditions under review are, in fact, stigmatizing in various cultures and geographic settings.  This information, which is equally necessary in order to estimate the global burden of such stigma, is more properly assembled and assessed by anthropologists and other social scientists who are expert in measuring stigma within a cultural framework.

The conditions we review here are all either multi-factorial in origin (e.g. mental retardation and cleft lip/cleft palate) or conditions for which there is a single causal agent (e.g. AIDS), but for which multiple (and often differing) behaviors, exposures, and other risk factors may play an important role in determining the prevalence of the condition and who is affected in a given part of the world.  However, a systematic review of the risk factors for and/or possible causes of the conditions being considered is beyond the scope of this review.  When appropriate, we do discuss briefly what is known about the factors that contribute to the risk of having or acquiring a given condition, primarily to focus attention on why the prevalence of the condition may vary geographically, culturally, or temporally.  Information concerning what is known about the causes and risk factors for a condition may also be useful in predicting the likely impact of prevention or intervention efforts on the prevalence of a condition in the future.

Choice of Measures for Estimating the Burden of Disease

In this article, we primarily review and discuss the prevalence (i.e. what proportion of a given population has a condition at a given point in time or during a given time period) rather than the incidence (i.e. the rate at which new cases of the condition occur in a population during a specified time interval) of a given condition.  In conducting analytic epidemiologic studies intended to elucidate the cause(s) of or contributing factors for a given condition, it is almost always preferable to calculate incidence rates rather than prevalence rates.  However, in this review we are more concerned with estimating the proportion of the population or the number of individuals who are or who may be at risk of being stigmatized as a result of having the condition, a purpose for which the prevalence is a more useful indicator.  However, for some conditions (e.g. HIV infection/AIDS), predictions about the likely future prevalence depend greatly on the incidence.  Therefore, when appropriate, we also review what is known about the incidence of a given condition.

Epidemiologic Approaches to Estimating the Burden of Disease

When attempting to estimate the prevalence or incidence of a given condition in a population, epidemiologists generally rely on one or more of the following five approaches:

            1)            passive surveillance

            2)            active surveillance/registries

            3)            surveys/cross-sectional studies

            4)            special studies involving closely monitored cohorts

            5)            use of administrative data bases assembled for other purposes

Each of these approaches can provide useful data about selected conditions and diseases under the proper conditions, but each has its limitations.  The advantages and limitations of each of these approaches are discussed in general terms here, to set the stage for a more detailed discussion of the accuracy of the data available for and to possible future approaches to studying the specific conditions under review.

Passive surveillance, which typically involves having health care providers report occurrences of disease or disability to a local government health agency and subsequent reporting by that agency to national and/or international agencies, is commonly used around the world for a variety of infectious diseases.  In some countries, such as the United States, there is also passive surveillance for selected non-infectious diseases (e.g. epilepsy and certain occupational or environmental illnesses considered "sentinel health events").  However, even when there are legal sanctions for failing to report and even in wealthy countries, passive surveillance systems generally capture only a subset of the occurrences of the disease being monitored.  While the sensitivity (the proportion of all cases that are reported) of such systems may be high in wealthy countries for rare and serious diseases of obvious public health significance (e.g. botulism or plague), the sensitivity is much lower for more common diseases (e.g. sexually transmitted infections); those cases that are reported are often not representative of those that aren't; and reporting may not occur at all in poor countries.  Furthermore, in addition to the fact that cases may go uncounted because they are not diagnosed or not reported, there may be political, economic, cultural, and social reasons why stigmatizing diseases such as AIDS or diseases whose occurrence can affect trade and tourism (e.g. cholera and plague) go unacknowledged.  Thus, data from passive surveillance systems are generally not available for many of the conditions under review and are unreliable for others.

Active surveillance systems or registries differ from passive surveillance systems primarily by having more resources available and having well-trained staff whose job it is to go out and find the cases of the condition under study and collect all needed data in a complete and consistent manner.  Such systems, while expensive, typically produce very high quality data.  The existence in numerous locations around the world of birth defects registries that are using consistent case definitions and case detection methods makes it possible to estimate quite accurately the prevalence of conditions such as cleft lip and palate.

Surveys and cross-sectional studies, in which the prevalence of a given condition at a given point in time (or during a given period of time) is estimated, can be very useful sources of data concerning conditions that persist over time and are detectable through interview, physical examination, or a laboratory test.  Historically, such surveys have been used to gather data about diverse conditions such as blindness and polio-induced lameness, and much of what we now know about the prevalence of mental retardation, epilepsy, HIV infection, alcohol abuse, and spouse and child abuse comes from such surveys.  However, such surveys tend to be expensive and time-consuming, particularly if truly representative samples of large populations are to be studied.  Also, if the prevalence of the condition under study is not stable over time, such surveys may need to be repeated with some frequency.

Highly accurate estimates of the incidence or prevalence of various conditions (e.g. HIV infection) can also come from epidemiologic studies of selected populations assembled for research purposes and followed over time (i.e. cohorts).  While such studies produce very accurate information about the incidence of the condition within the cohort, cohorts can be very expensive to assemble and follow over time, particularly if they need to be large.  Also, individuals agreeing to participate in such studies, like individuals agreeing to participate in surveys, may not be representative of the broader population (see below).

Finally, administrative data bases gathered for other purposes (e.g. provision of services or billing) or medical records from health care providers (e.g. health maintenance organizations, insurers, national health systems, etc.) can also be used to estimate the number of affected individuals or, when a denominator (i.e. the population at risk) is known, the prevalence of a condition such as mental retardation or schizophrenia.  However, such data bases are often inaccurate or incomplete, and the proportion of affected individuals captured in such data bases may vary by severity of the condition; changes in the treatment or management of the condition; referral patterns; and other factors, all of which may vary substantially over time or by geographic region.

Case Definitions

In order to estimate the prevalence or incidence of a given condition, it is necessary to establish a case definition or criteria for classifying individuals.  In order to make meaningful comparisons of the prevalence or incidence of a condition between different geographic regions, cultures, or time periods, it is important to assure that consistent case definitions are used (see below).  For some conditions, particularly dichotomous conditions (e.g. cleft lip and/or palate or HIV infection), it is relatively straightforward to establish and use a case definition or criteria to classify individuals as affected or not.  For other conditions, however, such as mental retardation, alcohol abuse, and spouse and child abuse, numerous case definitions are possible and can give rise to vastly different estimates of the number or proportion of individuals affected.  For such conditions, particularly those ascertained entirely through interview or questionnaire, it can be particularly challenging to establish, test, and assure the use of consistent classification criteria across multiple cultures, geographic regions, languages, and time periods.

Challenges Related to Accurate Measurement

In addition to the problems that can arise from using inconsistent sources of data or case definitions, the accuracy of the information concerning the frequency of a given condition in a population can be affected by selection bias, information bias, and problems associated with measuring prevalence rather than incidence.

Selection Bias

Selection bias occurs when the individuals selected for inclusion and agreeing to participate in a study differ systematically from the broader population about which one would like to draw conclusions.  High quality, population-based birth defects registries should produce measures of the prevalence of conditions like cleft lip and/or palate that are free or virtually free of selection bias.  On the other hand, the validity of the estimates of the prevalence of conditions like HIV infection; alcohol and other drug use; spouse and child abuse; and mental retardation and schizophrenia can be seriously jeopardized when surveys target and/or capture non-representative samples of the population under study.  Furthermore, the approaches to sampling a population that yield the most accurate estimates of the prevalence of a condition (i.e. sampling schemes that approximate simple random sampling) are invariably the most expensive and most difficult to implement.  Equally important, the willingness of individuals to participate in surveys or studies may vary substantially depending on whether they do or don't have the condition under study, particularly when the condition is stigmatizing or even illegal (e.g. illicit drug use). 

Information Bias

Information bias occurs when there are systematic errors in the accuracy of the data collected concerning the exposure(s) or outcome(s) under study.  As a result, individuals being studied may be mis-classified as to whether they do or don't have a particular condition.  While such errors can occur as a result of imperfect laboratory tests (e.g. for HIV infection), medical records, or clinical case definitions (e.g. for mental retardation), they are of particular concern when studies rely on individuals to provide accurate information about their behaviors, especially behaviors that are stigmatized or illegal.  Thus, individuals may not acknowledge illicit drug use, spouse/child abuse, or even alcohol/tobacco use because of fear, denial, or wanting to provide socially desirable answers.  While this problem can be reduced through a variety of means (e.g. collecting information from respondents in a private setting; asking questions about specific behaviors in a non-judgmental fashion; assuring the confidentiality of responses), it can rarely, if ever, be eliminated.

Examination of Prevalence vs. Incidence

As noted above, this review focuses largely on the prevalence of various conditions because we are interested in estimating the numbers of individuals at risk of being stigmatized by a given disease or condition.  However, for some conditions, such as child and spouse abuse, surveys or studies that estimate what proportion of respondents have abused a child or a spouse or been abused by a spouse or parent will systematically miss the most severe instances of such abuse - those instances in which a child or spouse died as a result of the injuries received.  When recognized as such, these episodes potentially can be studied through police or other legal records, but it is likely that many such deaths are never recognized or acknowledged to have been the result of abuse.

Comparisons and Compilations of Rates

Collection of data concerning the prevalence of potentially stigmatizing conditions in diverse geographic locations and cultures permits comparisons of the prevalence rates in different areas and compilation of the overall number of affected individuals.  Such information can help in planning the delivery of curative services (e.g. surgery to correct facial anomalies); programs to reduce the incidence or prevalence of a condition (e.g. drug treatment programs); and programs to reduce stigma.  Furthermore, such data can be used to establish baseline rates of these conditions, which can be invaluable when the impact of various interventions or prevention strategies is going to be assessed.

However, as alluded to above, estimating the prevalence of a medical condition or behavior in a given population, as well as compiling and comparing the estimated prevalences from various populations (or time periods) are beset by a number of difficulties, some of which may be peculiar to a given condition or behavior and others of which are more general in nature.  Thus, passive surveillance systems in which cases of a given illness (e.g. AIDS) are to be reported by health care providers almost invariably produce underestimations of the numbers and rates of cases in the population under study, due to misdiagnosis and failure to report.  Active surveillance systems or registries, for which specifically trained and dedicated staff gather data using standardized case definitions and have legally-mandated access to medical records (e.g. birth defects monitoring programs), typically produce very high quality information and virtually complete case ascertainment.  Population-based surveys, while expensive and cumbersome, and therefore often of limited size or scope, can also produce high quality data about the prevalence of a condition or behavior (e.g. mental retardation, alcohol or drug abuse, and spouse or child abuse), although fatalities may go uncounted, and conditions about which there is a feeling of shame or guilt are likely to be under-reported.  In this review, we rely as much as possible on active surveillance programs, disease registries, and population-based surveys for data concerning the prevalence (or incidence) of the conditions being discussed.  However, comparing or compiling results from different geographic areas, cultures, or time periods can produce meaningful results only if comparable case definitions, case finding methods, and approaches to data collection have been employed.  Therefore, when appropriate, we also discuss the comparability (or lack thereof) of these key aspects of the studies being reviewed.

Future Directions for Research

The conditions under review here can be divided into three groups with regard to the types of research needed in the future to define more accurately the global burden of disease and the numbers of individuals at risk of being stigmatized.

      1.   Conditions for which current approaches to estimating the number of affected individuals are adequate methodologically, but could be applied more broadly.

For some conditions, including cleft lip and palate and HIV infection, currently available approaches to estimating the prevalence are well-established and can give reasonably accurate estimates of the number of currently affected individuals when applied in a given country or region.  For these conditions, more resources would, if available, make it possible to apply these approaches to larger and more diverse populations, producing commensurate increases in the accuracy or certainty of the estimates of the number of individuals affected globally.  Additional research could, however, address whether there simpler, less expensive approaches to estimating prevalence that produce data of comparable accuracy.

2.   Conditions for which current approaches to estimating the number of affected individuals have been shown to work in some settings, but their applicability and feasibility in diverse cultural settings is uncertain.

      For some conditions, such as alcohol use, schizophrenia, epilepsy and mental retardation, approaches to estimating prevalence (largely based on surveys) have been developed and used in selected settings.  However, the applicability and feasibility of these approaches in diverse cultural settings has not been adequately tested.  For these conditions, additional research is needed to refine, standardize, and validate various aspects of these approaches (e.g. case detection methods and case definitions) to ensure that they can be applied in a consistent manner and can produce comparable results in diverse settings and cultures, particularly when the surveys have to be conducted by non-medically trained individuals.

3.   Conditions for which current approaches to estimating the number of affected individuals are or may be inadequate and better approaches may be needed.

      For some conditions, including use of illicit drugs and the more severe forms of spouse and child abuse, current approaches to estimating the number of affected individuals are or are likely to be inadequate, in large part because the conditions or behaviors under study are illegal and subject to punishment or because they are highly stigmatized and likely to the source of substantial feelings of shame and guilt.  For these conditions, current estimates of the numbers of affected individuals must be considered minimum estimates, and future research needs to focus on new or improved approaches to data collection that reduce or eliminate selection and information bias.

Table 1.  Classification Scheme for Possibly Stigmatizing Conditions

Email: