Although it is sometimes medically appropriate to treat men and women differently—sex and gender* can affect epidemiology, symptom presentation, disease course, treatment effectiveness, and prognosis1–5—the evidence that men and women are treated differently even when there is no medical justification reveals a continuing bias in medicine and health care. Men’s illnesses are investigated and treated more extensively than are those of women, despite the same severity of symptoms, in a variety of conditions.6–11 Physicians are more likely to interpret symptoms as psychosocial,12–14 diagnose nonspecific symptoms, and prescribe psychoactive drugs with their female patients than with their male patients.14–17
The growing body of research on gender bias usually focuses on its identification and consequences for equity.18 Yet, although the effect is always a systematic, if unintended, neglect of either men or women, the underlying mechanisms of gender bias are varied and complex and must be viewed from different perspectives.18–20 Gender bias can be caused by overlooking biologically or culturally based differences or by assuming difference where there is sameness.20 It can also result from believing that equity exists when it does not or from presuming inequity where there is none.19 Beliefs about gender—concerning difference, similarity, or equity—play a key role in the processes underlying bias.
Research on people’s perceptions and social relations can explain the crucial role of gender categorization as an omnipresent, preconscious system of thought.21 To make sense of our complex social world, we automatically sort people into different categories, based on such things as gender, age, ethnicity, and class.22–27 Because gender is a cultural dichotomy that can be applied to anyone of any age, race, or class, the person’s status as a man or a woman cuts across all those categories.21,25–27 These immediate categorical judgments are based on presumptions and stereotypic ideas about gender.21 But, it will also influence the relationship between, for example, the physician and the patient in clinical practice and has consequences for how they behave, interact, and understand each other.28–31 Stereotypes that simplify, exaggerate, and generalize the differences between men and women in behavior, skills, emotions, and needs are widespread in society.21,32,33 For instance, men are associated with order, control, and individualism, whereas women are considered to be in touch with their feelings and as having a natural sense for the family.21,34–36
Neither gender norms nor stereotypes are permanent; they are created and recreated in practice and can vary by context and time.21,32,34 Because gender stereotypes tend to mirror a culturally constructed gender hierarchy, they may promote discrimination.21,34,35 Furthermore, because stereotypes are automatic, preconscious, and omnipresent, it is crucial to explore how they function in clinical practice.
In this mixed-method study,37 we explored medical students’ gender stereotypes. To accomplish this, we used narratives, stripped of information that revealed the author’s gender, in which patients described their health care experiences. We asked medical students to read these narratives, identify each patient as a man or woman, and explain their reasoning. We then analyzed their explanations, qualitatively and quantitatively, to highlight underlying stereotypes about male and female patients and to allow us to discuss the possible consequences for clinical practice. The mixed study design also allowed us to determine whether the students’ stereotypes helped them correctly identify patients as men or women.
We consecutively collected narratives at an oncology department in Sweden, asking all patients aged 18 to 70 who had received their cancer diagnosis two to eight months previously to “write a page or two describing how you received your diagnosis … including what the physician told you, how you reacted and how you felt afterwards.” We also asked them to “describe both what you perceived as beneficial and what was detrimental.” Of the 187 consecutively invited patients, 138 (74%) submitted a written narrative.38
The narratives were typed, and all names, places, and dates were systematically changed to prevent identification of patients and others concerned. We removed all information revealing the patient’s gender, changing the words “husband” and “wife” to “co-habiter,” and “mother” and “father” to “parent.” We reformulated descriptions that provided clues to a patient’s sex—for instance, references to gynecological or prostate symptoms. Abbreviations and spelling mistakes in the original were retained. We excluded all 53 narratives about breast cancer because this group of patients underwent a mammography screening program and received very specific treatment. We excluded 2 additional narratives that were too difficult to change without distorting the content, and another 2 because they were illegible. Eighty-one narratives remained, 42 written by men and 39 by women. The university’s ethics committee approved the study.
We invited all 160 first- and second-year medical students at Umeå University during the fall of 2005 to participate in the study. They were informed that they would be asked to read and share their impressions about patients’ narratives describing a health care experience and that the study would take between one and two hours. Eighty-seven students (54%) volunteered to take part in the study—65% of the women (62 of 96) and 40% of the men (25 of 64). The participating students were between 18 and 39 years of age (mean = 22.8, SD: 3.45). After a pilot study showed that it was too demanding for the students to read all 81 narratives, we divided the students into two test groups: Group A with 43 students (12 men, 31 women) read half of the narratives, and Group B with 44 students (13 men, 31 women) read the other half.
The students first provided information on their own gender, age, and social background. Then, each student in Group A read 41 narratives and, in Group B, 40 narratives. This gave a total of 3,523 cases in which a student could guess a patient’s gender, assign that gender to the narrative, and explain that guess. In 663 cases, data were missing because students had failed to indicate the patient’s gender, explain their guess, or both. The missing data were spread over narratives and students, and there is no reason to believe that they had any systematic effect on the results. Thus, 2,860 cases, complete with guess and explanation, were included in the analysis.
First, all three authors independently read and coded all explanations from two students (one from each group, A and B) with regard to content. We compared and discussed the codes, outlining a preliminary coding schedule. We then coded and discussed explanations from another two students, and revised the coding schedule. These steps of individual coding followed by joint discussion and code modification continued until we reached consensus that the elaborated codes covered the content of subsequent explanations. The final coding schedule consisted of 49 codes, and the first author (J.A.) coded all explanations using the schedule, consulting the other two authors in cases where coding was difficult.
To check the reliability of the coding, the last author (K.H.) coded the explanations from six randomly chosen students representing both groups, and compared the result with the first author’s coding of the same explanations. For each of the 243 explanations, there were 49 codes to consider. A total of 36 discrepancies were found between the researchers. This gave a coding error of less than 1%.
When we had completed the coding and checked the reliability, we decided to exclude from further analysis those explanations that concerned the writing styles and words used by the patients in their narratives. We did this because we wanted to explore medical students’ stereotypes about men and women’s behaviors, reactions, treatments, and emotions in the clinical context. In addition, in a preliminary analysis, we had found a substantial covariation between writing style and content. Therefore, we excluded another 646 cases in which the students used only writing style to explain their guess. Thus, we based the analysis in this article on the remaining 2,214 cases, where each explanation included one or more codes concerning the content of the letter.
To facilitate the presentation and interpretation of results, we grouped related codes in 21 categories. For example, we grouped codes describing behaviors and attitudes associated with gender in the category “Behavior of patient.” We then clustered the 21 categories into three themes: “Emotions” (2 categories), “Health care experience” (12 categories), and “Other aspects” (7 categories). The themes and categories are presented and further described in Table 1.
First, we explored the students’ background characteristics, the number and distribution of gender assignments, and the categories. Independent-samples t tests were used to compare means.
Second, we explored associations between categories and gender assignments through simple and multiple regressions—that is, if and how categories were connected to gender assignment. The categories were first explored individually through simple binary regression with the assigned gender as the dependent variable. Given the study design with repeated observations for individual students (each student read 40 or 41 letters for which he or she made a gender assignment that he or she then explained), we assumed an exchangeable correlation structure and estimated the parameters using generalized estimating equations (GEE). The method of GEE uses weighted combinations of observations to extract the appropriate amount of information from correlated data. Assigning “man” was used as the reference; that is, the odds ratio (OR) reflects the extent to which a category was used when assigning “woman” relative to the extent to which it was used when assigning “man.” Thereafter, we used a multiple regression model to assess the joint impact of all categories shown to be influential in the simple regression models at a significance level of 5%.
Third, we repeated the analyses described above with simple and multiple regressions with “success rate” as the dependent variable—that is, whether or not the assigned gender corresponded to the patient’s actual gender. “Incorrect assignment” was used as the reference, and the OR shows the odds of the category being used when the assignment is correct. Both simple and multiple logistic regressions were expressed as ORs/adjusted ORs (AORs) with 95% confidence intervals. We used PASW statistics version 18 (Chicago, Illinois) to analyze the data.
The results are divided into two parts: first, associations between categories and gender assignment, and, second, “success rate” of the categories—that is, whether or not a category guided the students to correctly guess the patient’s gender. We do not present the gender of the students themselves as a variable in the analysis because it was not related to outcome.
Table 2 shows the associations between the categories and gender assignments. Thirteen categories found in the students’ explanations were significantly associated with assigning either “man” or “woman” in the simple regression. Twelve categories were still significant when included in the multiple regression model. In the description below, we focus on the results from the multiple regression analysis in the last two columns of Table 2.
When the most frequent category, “Emotions described,” was found in the explanation, the student was more likely to identify the patient as female (AOR: 2.873). The other categories significantly associated with a student guessing that the patient was a woman were “Need for comfort” (AOR: 7.939), “Family and friends” (2.952), “Health care described” (2.253), and “Biomedical clue” (2.472).
The categories significantly associated with guesses that the patient was a man were: “Emotions not described” (AOR 0.041), “Health care not described” (0.058), “Independence” (0.134), “Passive approach” (0.366), “Direct information” (0.209), “Factual focus” (0.222), and “Work and career” (0.333).
Table 3 shows the success rate of the categories. The results from the multiple regression analysis, seen in the last two columns of Table 3, show that students were more likely to correctly guess patients’ genders when their explanations included one of the categories “Emotions not described” (AOR 2.018), “Family and friends” (1.581), and “Factual focus” (1.749), but less likely when they included the categories “Independence” (AOR 0.294) and “Work and career” (0.358). These two categories were instead significantly associated with incorrect guesses.
Two of the three categories associated with correct guesses—“Emotions not described” and “Factual focus”—were also strongly associated with identifying the patient as a man (Table 2 and 3). Thus, these categories seem to have some predictive value; that is, the narratives the students felt lacked emotions or focused on facts were more likely to have been written by male than female patients. Similarly, the success rate of the category “Family and friends” and its association with identifying the patient as female (Table 2) suggest that a narrative mentioning and emphasizing the family was, in fact, more likely to have been written by a woman.
The two categories significantly associated with incorrectly guessing gender—“Independence” and “Work and career”—were both associated with students identifying the patient as male (Table 2). The low success rate of these explanatory categories indicates that the narratives the students felt expressed independence or emphasized the importance of work and career were more likely to have been written by female patients, and that students’ generalized ideas that independence or importance of work and career are characteristics associated with a male patient misled them in their gender assignments.
In our study, medical students read narratives written by actual patients describing their health care experiences and then guessed the patients’ gender. The students’ explanations for their guesses revealed prevailing ideas about male and female patients. Overall, we found 21 different categories of explanations, 5 of which were significantly associated with the guess that a woman had written the narrative and 7 with the guess that a man had. The predictive value of the categories overall was 64%, but their success rates varied a great deal. Correctly identifying a patient’s gender was significantly associated with three categories: “Emotions not described,” “Factual focus,” and “Family and friends.” Incorrect assignment was significantly associated with two categories: “Independence” and “Work and career.”
Dichotomic categorization of the narratives
In line with theories of person perception,21–27 our results show that the majority of categories found in the students’ explanations could be paired (Table 1)—for instance, describing emotions versus leaving them out, or focusing on facts and practical information versus focusing on relationships with others. These opposing categories were then linked to either men or women (Table 2). More than half of the categories were significantly associated with students guessing one gender or the other, supporting the idea that gender stereotypes are widely shared. Whether we as individuals agree with them or not, these abstracted understandings of men and women are roughly consensual in that virtually everyone in society knows what they are.21,32,33
Stereotypes about male and female patients
The relevance of the results lies in the possible consequences of gender stereotypes for health care provision, as they tend to be connected to value and resources.21,34,35 According to the simplified picture associated with “maleness,” a man does not seek health care unless he has to (“Passive approach”); is independent, fact-focused, and in search of straight answers (“Independence,” “Factual Focus,” “Direct information”); and is more likely to discuss work and career than the experiential or emotional aspects of his health care (“Work and career,” “Health care not described,” “Emotions not described”). A stereotypic woman, in contrast, focuses on emotions, emphasizes her treatment and health care experiences, places importance on family and friends, and expresses a need for support from others (“Emotions described,” “Health care described,” “Family and friends,” “Need of comfort”).
The (un)predictive value of stereotypes
The students were able to predict the patient’s gender in only 64% of cases. Only 3 of the 21 categories led to significantly more correct than incorrect guesses, and 2 categories in fact led to more incorrect guesses. Because gender, in addition to something we are, is also something we constantly “do,” it is hardly surprising that the male and female patients “did” gender in their narratives. Nor is it surprising that beliefs corresponding to strong norms associated with femininity and masculinity have some predictive value. The lack of emotions and focus on facts in the male patients’ narratives accords with widespread expectations about strong, independent, and rational men. Likewise, the female patients’ descriptions of relationships to family and friends accord with the idea of women as communal and warm.21,34–36 But, despite these few examples, the narratives from the male and female patients largely overlapped. Even the most successfully predictive category led to a correct gender assignment in only 77% of the cases. Hence, we conclude that, even if gender stereotypes have some predictive value on a group level, they are not applicable to individuals, and that individual variation better characterizes the patients’ narratives than does any categorical “truth.” The students may have understood this implicitly: None of the categories in their explanations led them to guess exclusively one gender or the other.
Gender stereotypes about patients in clinical practice
Studies have shown that gender does influence the relationship between the physician and the patient12–14,28–31 and that beliefs, preconceptions, and assumptions about gender may be involved in the process of bias.18–20
The medical students in our study expressed gender stereotypes that could cause problems for both male and female patients. Many studies have shown how female patients who approach the health care system risk delayed or improper investigation or being viewed as demanding.6–11,31 Our students associated a need for comfort, descriptions of health care experience, and emotional reactions with female patients—characteristics that could explain the bias found in those studies. They could also explain why women’s symptoms are often interpreted as psychosocial and why female patients are prescribed more psychoactive drugs.14–17 In addition, our results show that female patients who strive for independence or express the importance of career and work—characteristics typically associated with men—risk being misunderstood.
Studies showing that men receive more extensive examinations and treatment may in part be explained by our findings, which reveal a widespread belief that men do not seek care unnecessarily and so, correspondingly, when they do, something must really be wrong. On the other hand, that stereotype may be a “mixed blessing.”1,19 Believing that male patients prefer factual, straight answers and that they avoid or ignore emotions and relationships may lead health care staff, when dealing with men, to neglect the importance of family circumstances and psychosocial networks in coping with disease.14,39
Methodological considerations, strengths, and limitations
Our study was not based on actual encounters but on a constructed situation. The use of written narratives allowed us to base our study on authentic patients and their stories while still creating a situation in which it was not obvious whether the author was a man or a woman. In this way, the study design put gender stereotypes in focus. To approximate the automatic process of categorization that occurs in clinical practice, we instructed the students to read through the narrative quickly, go with their first impression, and assign a gender even when uncertain. This was an attempt to keep students from expressing themselves in more guarded, politically correct, and distant terms. The use of qualitatively evolved categories grounded in the data when performing statistical calculations allowed us to avoid using fixed categories and thereby helped us highlight prevalent ideas with possible consequences for clinical practice.
In a previous study, we had analyzed these same narratives (in that case, unedited) for content and meaning.40 We found then that more women than men wrote personal, emotional narratives and included family members and other relatives in their stories. More men than women wrote impersonal narratives without references to emotions and feelings. However, the majority of letters, about 60%, were classified as neither personal/emotional nor impersonal/unemotional—instead, they were somewhere in between. These results are in good accordance with our current analysis of the success rate of the categories “Family and friends” and “Emotions not described” in the sense that they were helpful in predicting a patient’s gender. They contained an inherent element of “truth.” But, the majority of categories had low predictive value, suggesting that they were more biased ideas than faithful mirrors of male and female patients’ descriptions, examples, and statements.
We performed our study among first- and second-year medical students at a single site in Sweden in 2005. This may restrict the generalizability and actuality of the findings. Still, the essence of our findings are confirmed by other studies conducted at different occasions within Western, mostly Anglo-Saxon, contexts.21,31–36,39 Nevertheless, it is important that gender stereotypes are recurrently explored and questioned in different contexts.
As for its implications for medical education and academic medicine, our mixed-method study shows that medical students enter their training program with culturally shared beliefs about male and female patients, stereotypes that can cause bias during their careers as doctors. If not addressed, these gender stereotypes will probably affect their expectations, communication with patients, and interpretations in clinical practice. It is therefore necessary to create a curriculum that teaches students to distinguish those differences in treatment based on sex and gender that are medically necessary from those that arise from biased stereotypes. Such a gender-sensitive curriculum should also include an appreciation of sex, gender, and cultural variation, as well as reflection on implicit gender beliefs, their possible consequences, and aspects of power. Moreover, the impact of gender beliefs needs to be included in discussions and research on gender bias in health care.
Funding/Support: The research was supported by grants from the Swedish Research Council.
Other disclosures: None.
Ethical approval: The ethics committee at Umeå University, Sweden, approved the study.
Previous presentation: The study and preliminary results were presented at the annual conference arranged by the International Association for Medical Education, AMEE, in Vienna, Austria, August 2011.
* The term sex is used when discussing the biological differences between men and women. The term gender is used when discussing the socially constructed differences between them.
1. Doyal L. Sex, gender, and health: The need for a new approach. BMJ. 2001;323:1061–1063
2. Lagro-Janssen T. Sex, gender and health: Development in research. Eur J Womens Stud. 2007;14:9–20
3. Lorber J Gender and the Social Construction of Illness. 1997 Thousand Oaks, Calif Sage Publications, Inc.
4. Legato M Principles of Gender-Specific Medicine. 2004;Vol 1–2 London, UK Elsevier Academic Press
5. Hoffman E, Magrane D, Donoghue GD. Changing perspectives on sex and gender in medical education. Acad Med. 2000;75:1051–1055
6. McMurray RJ, Clarke OW, Barrasso JA, et al. Gender disparities in clinical decision making. JAMA. 1991;266:559–562
7. Arber S, McKinlay J, Adams A, Marceau L, Link C, O’Donnell A. Patient characteristics and inequalities in doctors’ diagnostic and management strategies relating to CHD: A video-simulation experiment. Soc Sci Med. 2006;62:103–115
8. Daly C, Clemens F, Lopez Sendon JL, et al.Euro Heart Survey Investigators. Gender differences in the management and clinical outcome of stable angina. Circulation. 2006;113:490–498
9. Hariz GM, Nakajima T, Limousin P, et al. Gender distribution of patients with Parkinson’s disease treated with subthalamic deep brain stimulation; A review of the 2000–2009 literature. Parkinsonism Relat Disord. 2011;17:146–149
10. Karim F, Islam MA, Chowdhury AM, Johansson E, Diwan VK. Gender differences in delays in diagnosis and treatment of tuberculosis. Health Policy Plan. 2007;22:329–334
11. Chang AM, Mumma B, Sease KL, Robey JL, Shofer FS, Hollander JE. Gender bias in cardiovascular testing persists after adjustment for presenting characteristics and cardiac risk. Acad Emerg Med. 2007;14:599–605
12. Bernstein B, Kane R. Physicians’ attitudes toward female patients. Med Care. 1981;19:600–608
13. Colameco S, Becker LA, Simpson M. Sex bias in the assessment of patient complaints. J Fam Pract. 1983;16:1117–1121
14. Hamberg K, Risberg G, Johansson EE, Westman G. Gender bias in physicians’ management of neck pain: A study of the answers in a Swedish national examination. J Womens Health Gend Based Med. 2002;11:653–666
15. Smith RC, Gardiner JC, Lyles JS, et al. Exploration of DSM-IV criteria in primary care patients with medically unexplained symptoms. Psychosom Med. 2005;67:123–129
16. Simoni-Wastila L. The use of abusable prescription drugs: The role of gender. J Womens Health Gend Based Med. 2000;9:289–297
17. Linden M, Lecrubier Y, Bellantuono C, Benkert O, Kisely S, Simon G. The prescribing of psychotropic drugs by primary care physicians: An international collaborative study. J Clin Psychopharmacol. 1999;19:132–140
18. Hamberg K. Gender bias in medicine. Womens Health (Lond Engl). 2008;4:237–243
19. Risberg G, Johansson EE, Hamberg K. A theoretical model for analysing gender bias in medicine. Int J Equity Health. 2009;8:28
20. Ruiz MT, Verbrugge LM. A two way view of gender bias in medicine. J Epidemiol Community Health. 1997;51:106–109
21. Ridgeway C, Correll S. Unpacking the gender system: A theoretical perspective on gender beliefs and social relations. Gend Soc. 2004;18:510–531
22. Macrae CN, Bodenhausen GV. Social cognition: Categorical person perception. Br J Psychol. 2001;92(pt 1):239–255
23. Bodenhausen GV. The role of stereotypes in decision-making processes. Med Decis Making. 2005;25:112–118
24. Fiske ST, Taylor SE Social Cognition. 19912nd ed New York, NY McGraw-Hill
25. Blair IV, Banaji MR. Automatic and controlled processes in stereotype priming. J Pers Soc Psychol. 1996;70:1142–1163
26. Brewer M, Layton L. The primacy of age and sex in the structure of person categories. Soc Cogn. 1989;7:262–274
27. Stangor C, Lynch L, Duan C, Glas B. Categorization of individuals on the basis of multiple social features. J Pers Soc Psychol. 1992;62:207–218
28. Zaharias G, Piterman L, Liddell M. Doctors and patients: Gender interaction in the consultation. Acad Med. 2004;79:148–155
29. Roter DL, Hall JA. Physician gender and patient-centered communication: A critical review of empirical research. Annu Rev Public Health. 2004;25:497–519
30. Hall JA, Roter DL. Patient gender and communication with physicians: Results of a community-based study. Womens Health. 1995;1:77–95
31. Foss C, Sundby J. The construction of the gendered patient: Hospital staff’s attitudes to female and male patients. Patient Educ Couns. 2003;49:45–52
32. Connell RW Gender. 2002 Cambridge, UK Polity Press
33. West V, Zimmerman D. Doing gender. Gend Soc. 1987;1:125–151
34. Talbot MHolmes J, Meyerhoff M. Gender stereotypes: Reproduction and challenge. In: The Handbook of Language and Gender. 2003 Oxford, UK Blackwell Publishing
35. Lupton D The Emotional Self. 1998 London, UK Sage
36. Tuana N The Less Noble Sex: Scientific, Religious, and Philosophical Conceptions of Woman’s Nature. 1993 Bloomington, Ind Indiana University Press
37. Tashakkori A, Teddlie C Handbook of Mixed Methods in Social & Behavioral Research. 2003 Thousand Oaks, Calif Sage Publications
38. Salander P. Bad news from the patient’s perspective: An analysis of the written narratives of newly diagnosed cancer patients. Soc Sci Med. 2002;55:721–732
39. Courtney W. Constructions of masculinity and their influence on men’s well-being: A theory of gender and health. Soc Sci Med. 2000;50:1385–1401
40. Salander P, Hamberg K. Gender differences in patients’ written narratives about being diagnosed with cancer. Psychooncology. 2005;14:684–695