This article was downloaded by:[University Library Utrecht] [University Library Utrecht] On: 2 July 2007 Access Details: [subscription number 768375558] Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Psychology, Health & Medicine Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713441652

Gender-sensitive epidemiological research: suggestions for a gender-sensitive approach towards problem definition, data collection and analysis in epidemiological research CJ Moerman a; J. van Mens-Verhulst b a Department of General Practice, Unit 'Gender and Health', Academic Medical Center/University of Amsterdam. Amsterdam b Research School Psychology and Health, Utrecht University. Utrecht. The Netherlands

Online Publication Date: 01 February 2004 To cite this Article: Moerman, CJ and van Mens-Verhulst, J. , (2004) 'Gender-sensitive epidemiological research: suggestions for a gender-sensitive approach towards problem definition, data collection and analysis in epidemiological research', Psychology, Health & Medicine, 9:1, 41 - 52 To link to this article: DOI: 10.1080/13548500310001637742 URL: http://dx.doi.org/10.1080/13548500310001637742

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. © Taylor and Francis 2007

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

PSYCHOLOGY, HEALTH & MEDICINE VOL. 9, NO. 1, FEBRUARY 2004

Gender-sensitive epidemiological research: suggestions for a gender-sensitive approach towards problem definition, data collection and analysis in epidemiological research C. J. MOERMAN1 & J.

VAN

MENS-VERHULST2

1

Academic Medical Center/University of Amsterdam, Department of General Practice, Unit ‘Gender and Health’, Amsterdam & 2Research School Psychology and Health, Utrecht University, Utrecht, The Netherlands

Abstract In this article we examine the importance of applying a gender-sensitive approach in epidemiological research and the way in which such an approach can be effected. After a brief investigation into the meaning of the variable ‘sex’ we will show that this meaning is not systematically addressed in epidemiological research. Even if a gender-conscious problem definition is achieved, many methodological hurdles still loom ahead. First, the data acquisition must take place gendersensitively—this involves the selection of population and measurement instruments, the operationalization of indicators and the screening of possible professional bias. Secondly, the analysis should systematically pay attention to possible sex and gender differences in risk profiles and should not regard sex as a determinant but as a category, comprizing biological, psychological and social-cultural factors that together determine the status of a person’s health. Given the interaction of these factors it is plausible that health problems, even if they are biologically the same, should be understood and treated differently in men and women.

Introduction Epidemiology offers information on the distribution of diseases and health in a population. In case of socio-demographical differences in health, scientific, moral and societal questions are called for. How can these differences be explained? Is action needed and, if so, is action possible? The answer to the first question is important for answering the following ones. After all, if an explanation is lacking or based on inadequate suppositions, the possibilities for intervention are most likely also based on a misinterpretation. One of the socio-demographic variables is ‘sex.’ Compared to variables such as ‘class’ and ‘age,’ ‘sex’ usually receives less attention unless the health problem under investigation is specific for women, such as breast cancer. But in case of non-sex-specific conditions the attention for sex as a variable is already lacking in the very early stages of research - when the Address for correspondence: C. J. Moerman, Department of General Practice, Unit ‘Gender and Health’, Room J3-320 Academic Medical Center/University of Amsterdam, Meibergdreef 15, 1105 AZ Amsterdam, The Netherlands. Tel: + 31 20 566 5105, Fax: + 31 20 566 9186, E-mail: [email protected] ISSN 1354–8506 print/ISSN 1465-3966 online/04/010041–12 # Taylor & Francis Ltd DOI: 10.1080/13548500310001637742

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

42

C. J. MOERMAN & J. VAN MENS-VERHULST

research questions are phrased and the aims are set (Klinge & Bosch, 2001). Even when sex is considered important, as in drug research where the lack of adequate information on women is acknowledged and measures have been taken to correct the errors (see for instance the policy of the National Institutes of Health (1994), an important sponsor of health research), a focus on sex has not yet led to an improvement in the conduct of studies or data analysis (Ramasubbu et al., 2001). In general, researchers are aware that it is important to recruit a sufficient number of women in their study, but they do not plan to collect and process the data with consideration for possible differences between the two sexes (Marrocco & Stewart, 2001). It is no surprise then that the published results on research populations are often not specified according to sex and that statistical analyses for sex are either missing or performed insufficiently as a result of inadequate data collection or a too straightforward data analysis schedule. The consequence of such gender-insensitive research is that no adequate information can be provided to health-care and prevention practices about a gender-specific approach in the prevention, diagnosis and treatment of health problems. Relevance of sex differences Some diseases, such as obstructive lung disease and schizophrenia, are more common in men, and others, such as urine-incontinence and depression, in women (e.g. Agenda for Research on Women’s Health, 1999; American Psychological Association, 1996; Pinn, 2003). These data are often derived from health surveys and vital statistics on morbidity and mortality. In contrast, knowledge on disease-specific aspects mostly originates from the clinical practice, which has only recently started to pay attention to sex differences. We now know, for example, that men and women with coronary heart disease or with schizophrenia present themselves at different ages and with a different clinical presentation and risk profile (Castle et al., 2000; Charney, 1999; Chiamvimonvat & Sternberg, 1998; Ha¨fner et al., 1998), which has an impact on the establishment of a diagnosis, the choice of the best possible treatment and the prognosis of the disease. Complexity of sex/gender As more facts are becoming available, the insight is growing that ‘sex’ is not the simple independent biological variable it seems to be at first glance. ‘Man’ or ‘woman’ must be understood in the context of psychological, interpersonal, cultural, social and biological factors that mutually influence each other during a person’s life (Bekker, 2003; Scott, 1988). ‘Sex’ implies a social arrangement (that may vary with the cultural context) on the basis of which male and female children are encouraged to pass through a specific process of socialization. In the course of this process they learn to discern what it means to be either ‘masculine’ or ‘feminine’, because they are rewarded for complying with the roles society has laid out for them. In due course these experiences are transformed into a more or less sex-specifically-coded way of thinking, judging, feeling and acting (Bussey & Bandura, 1999). Theoretically, this is expressed in the sex/gender distinction (see Figure 1). ‘Sex’ is thus ‘produced’ by parents, teachers, church functionaries, lawgivers, labour counsellors, employers, health-care providers and of course the individual him- or herself. The result of this process is that in general the experiences in daily life differ between women and men due to their different perception of financial and emotional responsibilities for family, work, and other relevant issues. These experiences are accompanied by different stress factors, different health risks and different health behaviours, which lead to different demands for help

HOW TO DO GENDER SENSITIVE EPIDEMIOLOGICAL RESEARCH

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

Social domain

43

Interpersonal domain

Cultural domain

Psychological domain

Gender Sex Bodily domain FIG 1. Sex/gender concept. The arrows express the mutual influence of the domains.

and a different use of the medical and mental health services (Doyal, 2001). Moreover, the conceptions of masculinity and femininity are subject to change over time. Shifts in these conceptions occur, although these shifts will not always keep pace with changes in life patterns. The resulting discrepancies between ideas, feelings and behavioural patterns may also be a source of stress for the man or woman involved (Twenge, 1999). These insights into the complexity of sex/gender—we prefer to make a distinction between ‘sex,’ referring to biological sex, ‘gender,’ referring to culturally shaped variations between men and women, and ‘sex/gender,’ referring to both—are not systematically included in research into specific health problems or the application of health measures in practice. They are often absent in the problem definition as well as in the ensuing research aims and questions. But even if the problem definition does take sex/gender into account, many mistakes can be made in the collection and analysis of the data. In the following part we will cast our light on the major pitfalls and offer suggestions for a more adequate approach with respect to the collection and analysis of quantitative data in a sex-mixed population. It is our aim to encourage gender-sensitive research in epidemiology, i.e. research into illness or treatment and prevention methods among men and women that is sensitive towards differences between men and women with regard to their health, related to biology and/or the cultural and social context, and sensitive towards gender bias in the researcher’s decisions about the design and conduct of the study. Hereafter we will describe where and how gender sensitivity might come into play. Defining the problem Each research project starts with a state-of-the-art assessment: what is known about the health problem under study? What are the new insights and what are the currently experienced problems? At this point, a systematic gender-sensitive exploration should start. The kind of questions that are most relevant to explore depends on the type of study and the health problem studied, but the following preliminary strategies will be helpful to define them:

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

44

C. J. MOERMAN & J. VAN MENS-VERHULST

(1) Obtain data on the health problem, broken down by sex. For example, regarding data on the occurrence of the disease: the age range in which the disease occurs or the prognosis once the disease has been diagnosed. (2) Collect up-to-date information on possible relevant sex/gender aspects by doing a literature search with a special focus on sex and gender differences, and repeat the search in several databases (Medline, Embase, Psychinfo, Sociofile and Womens’ Studies Abstracts). One can also consult gender-specific databanks. (3) Assemble ‘lived’ experiences from the field by asking patients or clients and health-care providers, men and women, about the health problem, in particular in relation to the new insights and the experienced problems. (4) Operationalize the sex and gender domains by considering—according to Fig. 1—what physical, psychological, interpersonal, cultural and/or social variables are relevant for a specific health problem. Examples can be found in Bekker (2003), Messing et al. (2003) and Twenge (1999). Information collected in this way will provide clues for a gender-specific definition of the problem’s nature and scope and will encourage the inclusion of these insights into the design and conduct of the research project, whether the study is descriptive or analytical in nature, observational or experimental in design, and etiologic, prognostic, diagnostic or therapeutic in aim. Designing the study From the perspective that sex/gender concerns all levels of existence, the implementation of gender as a neutral variable in the research procedures and measuring instruments suddenly seems naı¨ve. Gender bias seems almost inevitable, in particular because researchers are not sensitive towards the way in which their methodological decisions may either support or obstruct the findings on sex/gender differences. We will illustrate this by focusing on gendersensitive choices in the selection of the study population, the inclusion and exclusion criteria, the translation of health concepts into measurable variables, and the methods of data collection. Recruitment and level of health care If an epidemiological study reports that data were collected from patients encountered in a general practitioner’s waiting room or a medical specialist’s waiting room, such a reference to the research setting is usually not associated with the possibility of gender bias. Yet all processes preceding that visit to the general practitioner or the specialist do have a genderspecific course, which indicates the potential for a sex- or gender-specific selection bias. These processes range from a patient’s perception of his or her symptoms, the decision to seek help and the presentation of these symptoms in the consultation room, to a doctor’s response to the clinical presentation, including possible referral to other health-care providers or other health-care facilities (Ashton & Fuehrer, 1993; De Ridder, 2000; Gijsbers van Wijk et al., 1999; Hall et al., 1994; Hibbard & Pope, 1983; Kerssens et al., 1997; Kroenke & Spitzer, 1998; Lard et al., 2001; Orth-Gome´r, 2000). As the decision about the level of the health-care system at which the study subjects will be recruited for the collection of the data will affect the sex/gender differences researchers can discover, they had better take that into account in advance and should reflect on how their decision has affected the outcomes in the discussion.

HOW TO DO GENDER SENSITIVE EPIDEMIOLOGICAL RESEARCH

45

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

Inclusion and exclusion criteria Another type of gender bias often occurs at the time the inclusion and exclusion criteria for a study are set. Basically it is considered good practice to restrict the study participants to a population in which certain characteristics are constant or can be controlled so that the focus can be on the factors under study. Mainly for this reason there used to be a preference to recruit men for pharmacological studies. Their stable hormonal balance facilitated the possibility to establish a drug’s effectiveness. However, this practice resulted in insufficient knowledge on drug prescribing in children, the elderly and women, because the differences in their drug metabolism had remained unexplored. As a result the choice in type and recommended drug dosage was suboptimal, adaptations in a treatment already started needed to be made and serious side effects occurred (e.g. Robinson, 2002). One example is the rare but serious sex-specific side effect of life-threatening heart rhythm disturbances (‘Torsades de Pointes’), which could result in sudden death when using certain types of drugs. This effect is linked to the menstrual cycle (Rodriguez et al., 2001). Due to the under-representation of women in trial populations this side effect was not observed until after the drug had already become available on the market. The Torsades de Pointes effect was responsible for the recent withdrawal of four out of 10 prescription drugs from the US market (United States General Accounting Office, 2001). Thus, researchers should take special care in including women systematically and in sufficient numbers in all types of epidemiological studies. Exclusion criteria can also lead to gender bias in a more covert way. For example, if a researcher excludes somatic or psychiatric co-morbidity in order to determine the ‘genuine’ drug effect for a specific disease, it is highly possible that relatively more women than men are excluded, because women generally experience more co-morbidities (Brayne et al., 2001; Newman & Brach, 2001). In that case, the outcome is applicable to the average male patient more than to the average female patient. In addition, the type of the experienced co-morbid conditions may also vary with the sex of the patient, as for instance in depression (Fava et al., 1996). This brings along an extra problem with generalizing the observed effects to the overall group of male and female patients. In order to do gender-sensitive research, each research design should give insight into the gender-specificity of the underlying processes that are implied in the selection of the studied population and have a bearing on the applicability of the study results to men and women in other health-care settings and on the generalizability of the data to the general population. Measuring indicators of health The selection and measurement of indicators for health studies contains a range of pitfalls for gender-sensitive research. The habit of assessing social-economic status by measuring personal or household income and employment status or occupation, for instance, is not equally valid for men and women (Arber, 1997; Ballantyne, 1999). Men and women may have different access to their income. Many women are only partially attached or still unattached to the labour market, but if they are married the financial consequences may be moderated by the employment of their partner. Occupational group frequently hides gender-differentiated opportunity and wage structures that also vary over time (Ballantyne, 1999). For example, ‘vocation’ is not a good indicator for the actual workload. Men and women do not only work in different employment sectors, but their job content (in terms of demands and control) is also different, even if they are employed in the same profession or the same sector (Wamala et al., 2000). Educational qualifications seem to be a better indicator for social position. First,

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

46

C. J. MOERMAN & J. VAN MENS-VERHULST

because they are of major influence on occupational class and employment status (Arber, 1997) and secondly because they are a more sensitive measure for the condition of a person’s health. As Manor (1997) put it: ‘[E]ducation is linked to ‘‘cultural capital’’ which may identify differences in lifestyles, health related behaviours and illness behaviour’. All this means that the best applicable indicator of economic position and the most meaningful measuring level (individual or household) have to be determined for each study separately, taking into account the gender differentiation of the employment structure of a given society in a specific period. Methods of data collection Methods of data collection are less gender-neutral than they appear to be at first sight. Researchers who use self-reports on health symptoms may expect to elicit more symptoms from women than men. Women tend to notice physical changes at an earlier stage (Gijsbers van Wijk et al., 1999), are more inclined to interpret these as symptoms of illness, and are more willing to report them (Hibbard & Pope, 1983). Macintyre et al. (1999), however, have contested this last ‘fact’ in a recent study on the reporting of chronic morbidity by men and women. They propose that the supposed increase in symptom reporting since the 1980’s has either been superseded in time or has emerged from stereotyping in the past. Yet as they did not propose a ‘golden’ standard of actual existing morbidity in their study to set against the self-reports, their point still needs further elaboration. Questionnaires like the General Health Questionnaire, the Symptom Check-List, the Affect Balance Scale and the Questionnaire about Experienced Health are known to elicit higher scores from women than men. To correct this, the cut-off points (used to separate the abnormal from the normal) are often set at a higher level for women. However, such a solution does not address a potentially built-in gender bias. The inquiring style of questionnaires regarding stress is much more in line with the female than the male forms of expression: there are more questions about sleep disturbance, loss of concentration and tension than about alcohol abuse, expression of aggression or apathetic television watching. On the other hand, most questionnaires generally also do not cover all aspects of health and wellbeing that are important to women: experience with sexual and domestic violence is much less sought after than is to be expected on the basis of existing numerical data on their prevalence and consequences (Bachman & Saltzman, 1995; Grisso et al., 1991). Bias against men also occurs: the Narcissistic scales of the Millon Clinical Multiaxial Inventory-II, the Minnesota Multiphasic Personality Inventory and the Personality Diagnostic Questionnaire – Revised are found to bias against healthy men (Lindsay & Widiger, 1995). Their type of questions leads to discrimination based on sex and gender role rather than on personal dysfunction. In choosing the measuring instrument researchers should consider whether the domains that are relevant to women and men in relation to the research question are sufficiently well covered, so that the assessment of exposures or outcomes is accurate for each sex. They should also reflect on possible gender bias in the cut-off points, before deciding on how they can make the best assessment of what they want to measure. Interview results are influenced by the topic of the study, and by the relationship between interviewee and interviewer. In these, the gender variable will always play a role. In her study on pain, Bendelow (2000) pointed out that the responses of men in a questionnaire mainly covered aspects of physical pain; men recognised the emotional component of their pain in indepth interviews only. Bendelow’s assumption therefore is that men are more inclined to reveal information that is in conflict with their ideas of masculinity in a supportive setting, in particular in a non-public setting. Some evidence comes from a study by Robinson et al.

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

HOW TO DO GENDER SENSITIVE EPIDEMIOLOGICAL RESEARCH

47

(2001), using the Gender Role Expectations of Pain Questionnaire. They found that men may perceive more of a loss of image in association with admitting pain. The sex/gender of the interviewer is an important factor as well: it is frequently assumed that men, just like women, disclose their thoughts and feelings more readily to women than to men. Williams & Heikes (1993) have examined this assumption by comparing in-depth interviews with male nurses by a male and a female interviewer; they found that the framing of the nurses’ answers varied with the interviewers’ gender. Their results illustrate that the interviewers’ sex/gender inevitably will create a single- or mixed-gender interaction resulting in specific self-presentations and -responses. Professional assessment Professionals do not escape gender bias either. In establishing a diagnosis they have the difficult task to determine what disease best fits the symptoms presented. This so-called differential diagnosis depends, among other aspects, on a patient’s age and risk profile and differs between men and women. One is not always aware of the fact that the pattern of symptoms that is considered typical for a specific diagnosis has been derived from its manifestation in just one of the sexes—for instance in case of coronary heart disease or schizophrenia in men. As a result, these symptoms are less effectively understood and diagnosed in women (Beery, 1995; Bouma et al., 1999; Hochman et al. 1999). In research, professionals are not only involved in establishing a diagnosis; they may also be called in to monitor the course of a disease over time in order to see whether improvements or deteriorations can be noted. If, for instance, not all symptoms that are relevant to women are incorporated in the criteria to assess the presence, extent or severity of a disease, the monitoring will be less sensitive to changes occurring among women. But even gender-sensitive assessors will meet problems in categorising the symptoms of their patients. They have to work with the available instruments and cannot individually correct for the ‘gendered’ definitions of mental disorders in the Diagnostic Statistic Manual (Kupers et al., 1997). In addition, professional assessors also impose their own personal sex stereotypes and judgements on their observations. The labelling of symptoms appears to be dependent on the way they are presented and thus on the patient’s sex. Birdwell’s (1993) study, for example, revealed that a woman (actress) presenting with chest pain evoked the (mostly male) doctors to consider coronary heart disease when she presented her symptoms in a businesslike (‘masculine’) manner. The doctors were less ready to consider heart disease when she presented her symptoms in a more affective and interactive (‘feminine’) manner, often referred to as ‘histrionic’. All in all, methods of data collection cannot be expected to be gender-neutral. Possible gender bias related to how information is collected can be demonstrated by applying a triangulation of methods, thus creating the possibility that men and women will show different responses to the distinct methods. This suggestion is meant as an invitation for further discussion.

Analysing and reporting of data Steps in a gender-sensitive analysis The analysis of the data eventually shows whether a health problem is different for men and women. When quantitative data are explored it is important to always keep in mind that sex as

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

48

C. J. MOERMAN & J. VAN MENS-VERHULST

a variable comprises all sorts of biological, psychological, and societal determinants and cannot be equated with one single determinant. In epidemiological research, the researcher deals with determinants and outcome variables related to the health problem under study, whether the study is observational or experimental in design and etiologic, prognostic, diagnostic or therapeutic in aim. The first step in a gender-sensitive analysis is to accurately describe the research population according to sex. A table displaying the background characteristics and the basic measurements for women and men separately will give the reader insight into the socio-demographical characteristics and the risk profiles of both sexes. It will immediately show whether the research population adequately reflects the patients and clients normally found at that level of the health-care system. It will also show whether the numbers of men and women included in the study are sufficient to draw reliable conclusions on both groups. The next step in the analysis depends on the type of design. In an observational design, the analysis usually focuses on exploring the—a priori hypothesized—relationships of each of the determinants with the health problem under study, followed by the building of a causal model starting from the relevant determinants, using multivariate regression techniques. Through such modelling the determinants most strongly related to the health outcome are identified and the extent to which these factors explain this outcome are assessed. In both stages of analysis sex is generally seen as a potentially distorting influence which has to be eliminated by applying a correction (sex as a confounder). The possibility that the direction (which indicates the relation between determinant and outcome variable and whether this relation is either risk-increasing or protective) or magnitude of a relationship among women can differ from that among men (sex as an effect modifier) is often not considered. In this way important relationships can be overlooked. A gender-sensitive analysis assesses the single effect of the determinants on the health outcome for women and men separately in order to evaluate whether heterogeneity is present according to sex. If so, the overall analysis, including both women and men, needs to be adapted accordingly. In order to check the overall model, it is also important to assess the determinant profiles of women and men separately and to compare the outcomes for the differences and the similarities. Likewise, when the health outcomes in an experimental design are evaluated for a preventive, diagnostic or therapeutic intervention, the analyses must be further extended with a subgroup analysis for sex differences. This can be done by evaluating whether there is heterogeneity in the effect of the intervention according to sex or through a separate outcome assessment for women and men and a mutual comparison of the results (see for example two recently published articles on sex-specific subgroup analyses of trial results; Simons-Morton, 2001; Rathore et al., 2002). By looking at the data in more detail, the researcher will find new, practice-relevant venues that may lead to a more in-depth study of health problems. A third step in the analysis is the interpretation of the results. If sex is introduced in a multivariate model that combines data on women and men, the variable may initially display a significant relationship with the health outcome that weakens or even disappears after adding other variables. What actually happens, is that the interplay of the factors comprized in the variable sex is decomposed into separate components of biological, psychological or social origin, which play a role in the health problem under study. Concluding that sex/gender is not an issue in this case would be an error of reasoning. On the contrary, it is recommendable to highlight the differences between men and women in the origin and scope of a health problem by emphasizing the distinct ‘determinants’ and testing their importance. In the end, in order to draw practical conclusions and recommendations from a study, it is of prime importance to translate the results back to the women and men of flesh and blood. Back to the women and men at risk, to the socio-demographical and other relevant

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

HOW TO DO GENDER SENSITIVE EPIDEMIOLOGICAL RESEARCH

49

background characteristics, and to the most important determinants, so that the potential for treatment and prevention can be clarified, preferential points of action for intervention can be prioritized, and, at the same time, the group of healthy or diseased individuals that these conclusions apply to can be established. The data analysis and report of a cohort study on the probability of getting infected with HIV among IV drug users (Strathdee et al., 2001) can be taken as a successful example of the above-mentioned steps. The study assessed the HIV status of the participants every 6 months and collected information on drug-using practices and sexual behaviour in the preceding 6 months of each assessment over a period of 10 years. By reporting the number of person years of follow-up for aspects of drug use and sexual practice for men and women separately, differences between the sexes in living conditions and risk behaviour became clear. A sexspecific report of the number of seroconversions per 100 person years revealed that the probability of becoming infected, given a specific behaviour, differed between men and women. From the separate analyses for men and women a difference in risk profile emerged: the profile of men was strongly determined by needle sharing and homosexual activity, whereas the profile of women was strongly determined by heterosexual activity. Based on the outcomes a gender-specific prevention programme was recommended, attuned to the genderspecific risk profiles. Concluding remarks The time of non-critical acceptation of empirical evidence on sex and health lies behind us (Hunt & Annandale, 1999; Inhorn & Whittle, 2001). It is no longer sufficient to include the variable ‘sex’ in a regression analysis or to test for sex differences and to restrict the reporting of results to the significant variables. A more fruitful insight into health problems and their solutions can be obtained by also considering biological differences between men and women and by studying health problems, not only those typical of women, but also those afflicting both men and women in a socio-economical, cultural and/or psychological context (Bekker, 2003; Doyal, 2001; Messing et al., 2003). Doing real gender-sensitive research means systematically checking on the lack of attention for gender aspects or on hidden imbalances in the attention paid to aspects relevant to men and women (gender bias), and doing so in all steps of designing and conducting a study. It means putting a focus on sex and gender issues in defining the problem and choosing the variables accordingly. It means paying attention to gender bias in both the choice of population and the methods of data collection. It also means analysing in a way that tries to retrieve the complex interdependence of biological, psychological, social and cultural factors—and not dealing with sex as if it were a confounding variable. Doing gender-sensitive research can be greatly enhanced if the principles of the approach are included in the criteria for funding studies and publishing results. In this way researchers would receive the strong reminder most of them need to perform a systematic check on existing omissions or gender bias. Apart from sponsors of health research, editorial boards of scientific journals have the opportunity to play a critical role in the promotion of gendersensitive epidemiological research. These boards set criteria for the publication of manuscripts and use them to decide whether a manuscript is suitable for publication or not. A good example are the so-called uniform requirements of the editors of biomedical journals which are made available to authors for the preparation of their manuscripts. The requirements pay attention to gender aspects by asking for a specification of the study population according to sex and a justification when sex is part of the inclusion or exclusion criteria in the method section (International Committee of Medical Journal Editors, 2001). First, it is important that

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

50

C. J. MOERMAN & J. VAN MENS-VERHULST

the boards have the existing sex criteria systematically considered in the review of submitted manuscripts. In a next step the publication criteria can be further refined towards the gendersensitive research approach. It is high time for editorial boards of psychological journals to start developing a similarly transparent publication strategy that incorporates attention for gender issues and is used to review manuscripts and decide what to publish. Presumably a gender-sensitive research approach will tend to decrease the number of general truths about men and women as categories. The eventual gain will be a more real-life picture of the differences between men and women as well as a set of similarities between men and women that is no longer based on assumptions but on sound evidence. This will benefit the practice of prevention and care as well as the individual male and female patients. A gender-sensitive approach of epidemiological studies, as outlined above, invites to further reflect on how to deal with other categories in epidemiological research showing biopsycho-socio-cultural differences such as age, ethnicity and socio-economic status. In heterogeneous populations, with respect to these categories, epidemiologists should reflect on how their population is selected, which choices they have made in the data collection methods, how they are going to analyse the data, and which generalizations will be allowed on the basis of these data.

References AGENDA FOR RESEARCH ON WOMEN’S HEALTH. (1999). Reports of the task force on the NIH women’s health research agenda for the 21st century, 6 volumes. Bethesda, MD: US National Institutes of Health. Also available on the Internet: http:// www4.od.nih.gov/orwh/pubs.html. Latest date of access on website: 11-06-2003. AMERICAN PSYCHOLOGICAL ASSOCIATION. (1996). Research agenda for the psychosocial and behavioral factors in women’s health. Washington: APA. ARBER, S. (1997). Comparing inequalities in women’s and men’s health: Britain in the 1990s. Social Science & Medicine, 44, 773 – 787. ASHTON, W. A. & FUEHRER, A. (1993). Effects of gender role and gender role identification of participant and type of social support resource on support seeking. Sex Roles, 28, 461 – 476. BACHMAN, R. & SALTZMAN, L. E. (1995). Violence against women: Estimates from the resigned survey. USA Bureau of Justice Statistics, Aug. Special Report, NCJ-1543. BALLANTYNE, P. J. (1999). The social determinants of health: a contribution to the analysis of gender differences in health and illness. Scandinavian Journal of Public Health, 27, 290 – 295. BEERY, T. A. (1995). Diagnosis and treatment of cardiac disease. Gender bias in the diagnosis and treatment of coronary artery disease. Heart & Lung, 24, 427 – 435. BEKKER, M. H. J. (2003). Investigating gender within health research is more than sex disaggregation of data: a MultiFacet Gender and Health Model. Psychology, Health and Medicine, 8, 231 – 244. BENDELOW, G. (2000). Pain and gender. Harlow: Prentice Hall. BIRDWELL, B. G. (1993). Evaluating chest pain. The patient’s presentation style alters the physician’s diagnostic approach. Archives of Internal Medicine, 153, 1991 – 1995. BOUMA, J., BROER, J., BLEEKER, J., VAN SONDEREN, E., MEYBOOM-DE JONG, B. & DEJONGSTE, M. (1999). Longer prehospital delay in acute myocardial infarction in women because of longer doctor decision time. Journal of Epidemiology and Community Health, 53, 459 – 464. BRAYNE, C., MATTHEWS, F. E., MCGEE, M. A. & JAGGER, C. (2001). Health and ill-health in the older population in England and Wales. Age & Ageing, 30, 53 – 62. BUSSEY, K. & BANDURA, A. (1999). Social cognitive theory of gender development and differentiation. Psychological Review, 106, 676 – 713. CASTLE, D. J., MCGRATH, J. & KULKARNI, J. (eds.) (2000). Women and schizophrenia. Cambridge: Cambridge University Press. CHARNEY, P. (ed.) (1999). Coronary artery disease in women. What all physicians need to know. Philadelphia: American College of Physicians. CHIAMVIMONVAT, V. & STERNBERG, L. (1998). Coronary artery disease in women. Canadian Family Physician, 44, 2709 – 2717.

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

HOW TO DO GENDER SENSITIVE EPIDEMIOLOGICAL RESEARCH

51

DE RIDDER, D. T. (2000). Gender, Stress and Coping: Do Women Handle Stressful Situations Differently from Men? In: L. SHERR & J. S. ST LAWRENCE (Eds), Women, Health and the Mind (pp. 115 – 135). London: Wiley & Sons. DOYAL, L. (2001). Sex, gender, and health: the need for a new approach. British Medical Journal, 323, 1061 – 1063. FAVA, M., ABRAHAM, M., ALPERT, J., NIERENBERG, A. A., PAVA, J. A. & ROSENBAUM, J. F. (1996). Gender differences in Axis I comorbidity among depressed outpatients. Journal of Affective Disorders, 38, 129 – 133. GIJSBERS VAN WIJK, C. M. T., HUISMAN, H. & KOLK, A. M. (1999). Gender differences in physical symptoms and illness behavior. A health diary study. Social Science & Medicine, 49, 1061 – 1074. GRISSO, J. A., WISHNER, A. R., SCHWARTZ, D. F., WEENE, B. A., HOLMES, J. H. & SUTTON, R. L. (1991). A populationbased study of injuries in inner-city women. American Journal of Epidemiology, 134, 59 – 68. HA¨FNER, H., HAMBRECHT, M., LO¨FFLER, W., MUNK-JORGENS, P. & RIECHER-ROSSLER, A. (1998). Is schizophrenia a disorder of all ages? A comparison of first episodes and early course across the life cycle. Psychological Medicine, 28, 351 – 365. HALL, J. A., IRISH, J. T., ROTER, D. L., EHRLICH, C. M. & MILLER, L. H. (1994). Gender in medical encounters: An analysis of physician and patient communication in a primary care setting. Health Psychology, 13, 384. HIBBARD, J. H. & POPE, C. R. (1983). Gender roles, illness orientation and use of medical services. Social Science & Medicine, 17, 129 – 137. HOCHMAN, J. S., TAMIS, J. E., THOMPSON, B. S., WEAVER, W. D., WHITE, H. D., VAN DE WERF, F., AYLWARD, P., TOPOL, E. & CALIFF, R. M. (1999). Sex, clinical presentation, and outcome in patients with acute coronary syndromes. New England Journal of Medicine, 341, 226 – 232. HUNT, K. & ANNANDALE, E. (1999). Relocating gender and morbidity: examining men’s and women’s health in contemporary Western societies. Introduction to Special Issue on Gender and Health. Social Science & Medicine, 48, 1 – 5. INHORN, M. C. & WHITTLE, K. L. (2001). Feminism meets the ‘new’ epidemiologies: toward an appraisal of antifeminist biases in epidemiological research on women’s health. Social Science & Medicine, 53, 553 – 567. INTERNATIONAL COMMITTEE OF MEDICAL JOURNAL EDITORS. (2001). Uniform requirements for manuscripts submitted to biomedical journals. Latest update October 2001. http://www.icmje.org/index.html. Latest date of access on website: 17-06-2003. KERSSENS, J. J., BENSING, J. M. & ANDELA, M. G. (1997). Patient preferences for genders of health professionals. Social Science & Medicine, 44, 1531 – 1540. KLINGE, I. & BOSCH, M. (2001). Gender in research. Gender Impact Assessment of the specific programmes of the Fifth Framework Programme—Quality of Life and Management of Living Resources. Brussels: European Commission (EUR 20017). KROENKE, K. & SPITZER, R. L. (1998). Gender differences in the reporting of Physical and Somatoform Symptoms. Psychosomatic Medicine, 60, 150 – 155. KUPERS, T. A., ROSS, R., GRANCES, A. & WIDIGER, T. A. (1997). Is there gender bias in the 1994 ‘Diagnostic and Statistic Manual’ (DSM-IV)? In: M. R. WALSH (ed.). Women, men, & gender: Ongoing debates (pp. 337 – 357). New Haven: Yale University Press. LARD, L. R., HUIZINGA, T. W. J., HAZES, J. M. W. & VLIET VLIELAND, T. P. M. (2001). Delayed referral of female patients with rheumatoid arthritis. Journal of Rheumatology, 28, 2190 – 2192. LINDSAY, K. A. & WIDIGER, T. A. (1995). Sex and gender bias in self-report personality disorder inventories: item analyses of the MCMI-II, MMPI and PDQ-R. Journal of Personality Assessment, 65, 1 – 20. MACINTYRE, S., FORD, G. & HUNT, K. (1999). Do women ‘over-report’ morbidity? Men’s and women’s responses to structured prompting on a standard question on long-standing illness. Social Science & Medicine, 48, 89 – 98. MANOR, O., MATTHEWS, S. & POWER, C. (1997). Comparing measures of health inequality. Social Science & Medicine, 45, 761 – 771. MARROCCO, A. & STEWART, D. (2001). ‘We’ve come a long way, maybe: recruitment of women and analysis of results by sex in clinical research’. Journal of Women’s Health & Gender-Based Medicine, 10, 175 – 179. MESSING, K., PUNNETT, L., BOND, M., ALEXANDERSON, K., PYLE, J., ZAHM, S., WEGMAN, D., STOCK, S. R. & DE GROSBOIS, S. (2003). Be the fairest of them all: challenges and recommendations for the treatment of gender in occupational health research. American Journal of Industrial Medicine, 43, 618 – 629. NATIONAL INSTITUTES OF HEALTH. (1994). NIH guidelines on the inclusion of women and minorities as subjects in clinical research NIH guide, Volume 23, Number 11, March 18. http://grants1.nih.gov/grants/guide/notice-files/not94100.html. Latest date of access on website: 04-12-2002. NEWMAN, A. B. & BRACH, J. S. (2001). Gender gap in longevity and disability in older persons. Epidemiologic Reviews, 23, 343 – 350. ORTH-GOME´R, K. (2000). New light on the Yentl syndrome. European Heart Journal, 21, 874 – 875. PINN, V. W. (2003). Sex and gender factors in medical studies. Implications for health and clinical practice. Journal of the American Medical Association, 289, 397 – 400.

Downloaded By: [University Library Utrecht] At: 13:29 2 July 2007

52

C. J. MOERMAN & J. VAN MENS-VERHULST

RAMASUBBU, K., GURM, H. & LITAKER, D. (2001). ‘Gender bias in clinical trials: do double standards still apply?’ Journal of Women’s Health & Gender-Based Medicine, 10, 757 – 764. RATHORE, S. S., WANG, Y. & KRUMHOLZ, H. M. (2002). Sex-based differences in the effect of digoxin for the treatment of heart failure. New England Journal of Medicine, 347, 1403 – 1411. ROBINSON, M. E., RILEY, J. L., MYERS, C. D., PAPAS, R., WISE, E., WAXENBERG, L. B. & FILLINGIM, R. (2001). Gender role expectations of pain: relationship to sex differences in pain. Journal of Pain, 2, 252 – 257. ROBINSON, G. E. (2002). Women and psychopharmacology. Medscape Women’s Health eJournal, 7. http://www. medscape.com/viewarticle/423938?srcmp = wh-030102. Latest date of access on website: 17-06-2003. RODRIGUEZ, I., KILBORN, M. J., LIU, X. K., PEZZULLO, J. C. & WOOSLEY, R. L. (2001). Drug-induced QT prolongation in women during the menstrual cycle. Journal of the American Medical Association, 285, 1322 – 1326. SCOTT, J. W. (1988). Gender: a useful category for historical analysis. In J. W. SCOTT. Gender and the politics of history. New York: Columbia University Press. SIMONS-MORTON, D. G. (2001). Effects of physical activity counseling in primary care: The activity counseling trial: a randomized controlled trial. Journal of the American Medical Association, 286, 677 – 687. STRATHDEE, S. A., GALAI, N., SAFAIEAN, M., CELENTANO, D. D., VLAHOV, D., JOHNSON, L. & NELSON, K. E. (2001). Sex differences in risk factors for hiv seroconversion among injection drug users: a 10-year perspective. Archives of Internal Medicine, 161, 1281 – 1288. TWENGE, J. M. (1999). Mapping gender. The Multifactorial Approach and the Organization of Gender-Related Attributes. Psychology of Women Quarterly, 23, 485 – 502. UNITED STATES GENERAL ACCOUNTING OFFICE. (2001). Drug safety: most drugs withdrawn in recent years had greater health risks for women. Washington DC: The Office, Jan. Report No: GAO-01-286R. WAMALA, S. P., MITTLEMAN, M. A., HORSTEN, M., SCHENK-GUSTAFSSON, K. & ORTH-GOME´R, K. (2000). Job stress and the occupational gradient in coronary heart disease risk in women. The Stockholm Female Coronary Risk Study. Social Science & Medicine, 51, 481 – 489. WILLIAMS, C. L. & HEILES, E. J. (1993). The importance of researcher’s gender in the in-depth interview: Evidence from two case studies of male nurses. Gender and Society, 7, 280 – 291.