- Email: [email protected]
Effects of contact with treatment users on mental illness stigma: Evidence from university roommate assignments
Social Science & Medicine 75 (2012) 1122e1127
Contents lists available at SciVerse ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Effects of contact with treatment users on mental illness stigma: Evidence from university roommate assignments Daniel Eisenberg a, *, Marilyn F. Downs b, Ezra Golberstein c a
Department of Health Management & Policy, School of Public Health, University of Michigan, M3517 SPH II, MC 2029, 1415 Washington Heights, Ann Arbor, MI 48109-2029, USA Counseling and Mental Health Service, Tufts University, USA c Division of Health Policy & Management, School of Public Health, University of Minnesota, USA b
a r t i c l e i n f o
a b s t r a c t
Article history: Available online 1 June 2012
Mental illness stigma refers to negative stereotypes and prejudices about people with mental illness, and is a widespread phenomenon with damaging social, psychological, and economic consequences. Despite considerable policy attention, mental illness stigma does not appear to have declined significantly in recent years. Interpersonal contact with persons with mental illness has been identified as a promising approach to reducing mental illness stigma. This study investigates the effect of contact with mental health treatment users on stigma using an observational research design that is free of self-selection bias. The research design is based on the quasi-experiment in which university students are assigned to live together as roommates. Survey data were collected from first-year undergraduates at two large universities in the United States (N ¼ 1605). Multivariable regressions were used to estimate the effect of assignment to a roommate with a history of mental health treatment on a brief measure of stigmatizing attitudes. Contact with a treatment user caused a modest increase in stigma (standardized effect size ¼ 0.15, p ¼ 0.03). This effect was present among students without a prior treatment history of their own, but not among those with a prior history. The findings indicate that naturalistic contact alone does not necessarily yield a reduction in mental illness stigma. This may help explain why stigma has not declined in societies such as the United States even as treatment use has risen substantially. The findings also highlight the importance of isolating the specific components, beyond contact per se, that are necessary to reduce stigma in contact-based interventions. Ó 2012 Elsevier Ltd. All rights reserved.
Keywords: USA Stigma Social contact Adolescents Mental health
Introduction Mental illness stigma refers to negative stereotypes and prejudices about people with mental illness, and is a widespread phenomenon with damaging social, psychological, and economic consequences (Corrigan, 2004; Phelan, Link, Stueve, & Pescosolido, 2000). Stigma is associated with negative views about help-seeking and a lower use of mental health services (Penn et al., 2005; Van Voorhees et al., 2005), and may contribute to underfunding of mental health services and programs (Saraceno et al., 2007). For all of these reasons there have been significant policy and research efforts to improve attitudes in many countries, including the United States (Sartorius, 2002; U.S. Department of Health and Human Services, 1999). Despite this attention, however, public stigma about mental illness did not decline significantly between 1996 and 2006 in national surveys in the United States (Pescosolido et al., 2010). * Corresponding author. Tel.: þ1 734 615 7764; fax: þ1 734 764 4338. E-mail address: [email protected] (D. Eisenberg). 0277-9536/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2012.05.007
Interpersonal contact with persons with mental illness has been identified as one of the most promising approaches to reducing stigma (Corrigan & Penn, 1999). This proposition is supported by a large number of studies that have found significant associations between various indicators of stigma and interpersonal contact with mentally ill persons and treatment users (Corrigan, Morris, Michaels, Rafacz, & Rusch, under review; Couture & Penn, 2003; Holzinger, Dietrich, Heitmann, & Angermeyer, 2008). Previous studies also suggest that reductions in stigma are more likely when the interpersonal contact is a deliberate choice (Couture & Penn, 2003; Kolodziej & Johnson, 1996), the quality of contact is high (Couture & Penn, 2006; Pettigrew & Tropp, 2006), and the contact experience only moderately challenges existing stereotypes (Reinke, Corrigan, Leonhard, Lundin, & Kubiak, 2004). The previous studies of interpersonal contact are based on two types of study designs: 1) naturalistic, non-experimental studies measuring the correlation between stigma and contact, controlling for potential confounding variables; and, 2) experimental studies that manipulate the amount and type of contact and then compare
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
stigma outcomes across groups. In the naturalistic studies, persons who self-select into contact with persons with mental illness may have lower stigma, or larger decreases in stigma, due to factors that are difficult to control for statistically, such as a greater openness to different experiences. Experimental studies, on the other hand, can avoid this type of confounding by randomly assigning participants to contact conditions; however, these studies typically involve a limited duration and intensity of contact, such as presentations or testimonials by a person with a mental illness. The present study adds a new type of evidence to this literature. The effect of interpersonal contact on stigma is estimated among first-year college roommates who are assigned to each other. In contrast to previous naturalistic, non-experimental studies, this research design avoids potential confounding due to self-selection into contact. Also, in contrast to most experimental studies, the design involves sustained, close contact; college roommates share a small living space over a period of approximately seven months, and are therefore likely to develop friendships and have frequent one-on-one interactions. A priori it is unclear what effect this type of contact with mental health services users should be expected to have on stigma. Although contact-based interventions have shown promise, as noted above, these approaches typically involve a purposeful and thoughtful disclosure of mental illness, including some form of education or guidance. The present study, in contrast, evaluates naturalistic contact, without any structured intervention regarding how mental illness or use of services is presented. Thus, the present study complements previous evidence by addressing a somewhat different angle both in terms of research design and substantive question. Method Sample selection and recruitment The sample consisted of first-year students who began college in fall 2009 at two large universities in the United States. At both universities first-year students were required to live in campus housing, except in unusual circumstances. Students who did not request specific roommates were assigned to their roommates. All first-year students with assigned roommates were recruited for the study, with the exception of students who were still under 18 years old as of the follow-up survey (0.9% of the initial sample). The students were invited to complete two brief online surveys: a baseline survey in August 2009 (1e3 weeks before students arrived at college) and a follow-up survey in MarcheApril 2010 (2e4 weeks before the end of the academic year). The survey data were linked to administrative data on housing preferences, room assignments, and academic and demographic characteristics. For both the baseline and follow-up surveys, students were recruited with an introductory letter including a $10 bill, followed by up to four email invitations to those who had yet to respond, spaced by 3e5 days each. All communications included a web link to the survey and a unique, randomly assigned log-in code for each student. The recruitment messages also informed students that they were entered into a sweepstakes for cash prizes regardless of participation. Informed consent was obtained at the beginning of the online survey. All aspects of data collection were approved by Institutional Review Boards at both universities. The primary analytic sample consisted of students with complete data necessary for the statistical analyses. These were students who completed both baseline and follow-up surveys and whose roommate(s) also completed the baseline survey. The potential for bias due to differential survey non-response was investigated, as described in the technical Appendix.
1123
Measures The primary outcome was a brief measure of general stigmatizing attitudes toward mental health treatment. This was measured in both baseline and follow-up surveys with an item asking, “Please indicate whether you agree or disagree with the following statement: I think less of someone who has received mental health treatment.” The answer choices were on a Likert scale (strongly agree, agree, somewhat agree, somewhat disagree, disagree, and strongly disagree), which was coded as a 0e5 linear score. This measure was adapted from an item on the widely used Discrimination-Devaluation Scale (Link, 1987; Link, Cullen, Struening, Shrout, & Dohrenwend, 1989). The adapted item has been used in a national analysis of stigma and help-seeking among university students (Eisenberg, Downs, Golberstein, & Zivin, 2009). In that analysis, the item had a significant negative correlation (r ¼ 0.58, p < 0.001) with the sum of other items adapted from the Discrimination-Devaluation Scale, and was a significant negative predictor of treatment use (odds-ratio ¼ 0.65, p < 0.001). The key predictor variable was the treatment history of assigned roommates. This variable was constructed based on roommates’ responses to three survey questions about prior treatment and diagnoses, which were taken from the Healthy Minds Study, a national survey study of mental health and service utilization among university students (Eisenberg, Hunt, Speer, & Zivin, 2011). Specifically, these questions asked about lifetime diagnoses of mental disorders by a health professional, use of psychiatric medication in the past six months, and use of counseling or therapy in the past six months. The primary measure of roommate’s treatment history was a composite variable equal to one if the roommate had any previous diagnosis or treatment use, and zero if not. The three components of this variable were also examined separately in additional analyses. For students with multiple roommates, the mean value of roommates was used. The key covariates in the analysis are preferences expressed in students’ housing applications, which were used by the housing administrators to match roommates. The primary variables used for matching roommates included preferences about room type (double, triple, or quad), same-sex versus mixed-sex hallway, and smoker status. These variables were completely controlled for, so any remaining variation in roommate characteristics should be random and thus uncorrelated with stigma or any factor that could affect stigma, as detailed in the Appendix. Statistical analysis In the primary analysis, a linear regression model was estimated as in this equation:
Stigmaðtþ1Þ ¼ b0 þ b1 Prefst þ b2 RMtxHistoryt þ b3 Stigmat þ b4 Xt þ 3 tþ1 (1) The subscript t denotes a measurement in the baseline survey, and t þ 1 denotes a measurement in the follow-up survey. Prefs refers to a vector of housing preferences and other variables used to make roommate assignments (as detailed in the Appendix), and X is a vector of individual covariates including gender, age (exact to the day), race/ethnicity (white, Black, Asian, Hispanic/Latino, mixed race, or other), and parents’ highest education level (less than a college degree, college degree, or graduate degree). These covariates were included because prior studies have found them to correlate significantly with mental health and help-seeking behavior in college populations (Blanco et al., 2008; Eisenberg, Golberstein, & Gollust, 2007). The key coefficient is b2, which
1124
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
represents the effect of roommate treatment history on the stigma score at follow-up. Heteroskedasticity-robust standard errors were estimated, to correct for correlated outcomes among roommates. The statistical significance of comparisons of means was assessed using two-tailed t-statistics and the significance of the regression coefficients was assessed using two-tailed z-statistics. In order to interpret b2 as an unbiased estimate of the causal effect of roommate’s treatment history on stigma, the key assumption is that the roommate’s treatment history is uncorrelated with the error term, 3 tþ1 , in equation (1). In effect this assumption means that there were no unmeasured roommate characteristics that would confound the analysis; this assumption is highly plausible because all variables used for matching roommates were included in the analysis. The Appendix describes how this assumption was examined empirically. Sensitivity analyses examined whether the relationship between stigma and roommate’s treatment history is robust to the inclusion of several additional baseline roommate covariates related to mental health and other factors. These covariates include: the PHQ-2 depression score (Löwe, Kroenke, & Gräfe, 2005); an index of anxiety symptoms measured by the sum of two items in the K-6 psychological distress scale (Kessler et al., 2003); parents’ education (highest level obtained by either parent); how religious one is (very, somewhat, a little, not at all); frequency of binge drinking in the past 30 days; frequency of exercise in the past 30 days; average hours per day spent studying in the last year of high school; standardized admissions test score (total ACT and/or SAT, converted to a z-score based on the withinuniversity distribution); and high school GPA (also converted to a z-score). Another sensitivity analysis defined the key predictor variable as equal to one if any roommate had a treatment history, rather than the average across roommates, in cases of multiple roommates. In a final set of sensitivity analyses, ordered probit regressions were estimated instead of linear regressions. Subgroups In addition to the primary analysis of the overall sample, a number of exploratory subgroup analyses were conducted to examine the heterogeneity of contact effects. First, it was hypothesized that students with a prior treatment history would experience less change in stigma due to contact with a roommate with a treatment history, because they would be more likely to have previous knowledge and well-formed attitudes about mental health treatment. Similarly, it was hypothesized that students with higher self-reported knowledge about mental illnesses (as measured by the question, “Relative to the average person, how knowledgeable are you about mental illnesses (such as depression and anxiety disorders) and their treatments? (Please just give your best guess)”) would be less affected by contact than those who reported average or lower knowledge. A number of other subgroup analyses were conducted for which the hypotheses were ambiguous, due to potentially offsetting factors. Three of these analyses defined subgroups by the following baseline characteristics, respectively: own stigma level (high or low); roommates’ tendency to disclose distress (high, medium, low on an item in the Distress Disclosure Index asking about agreement with the statement, “When I feel depressed or sad, I tend to keep those feelings to myself.”) (Kahn & Hessling, 2001); and roommates’ symptoms of psychological distress (high or low K-6 score). Students with higher stigma at baseline may be more resistant to changing their attitudes, but they also have more room for a potential decrease in stigma. Roommates with more symptoms may provide more salient cues, which could either trigger negative perceptions or evoke more sympathy and support. Similarly, roommates who
tend to disclose distress may engender negative perceptions, but may also generate more openness and interpersonal understanding. Finally, a subgroup analysis by university was conducted to assess the robustness of results across settings. Subgroup heterogeneity was tested by adding a covariate for the interaction between roommate treatment history and subgroup membership (separately for each of the above subgroups) to the main regression model. Results The basic characteristics of the full analytic sample (N ¼ 1605) are shown in Table 1. The average level of stigma was almost identical between the baseline and follow-up surveys. Mental health, on the other hand, worsened between the time points, with significant increases in depression and psychological distress scores. The response rate to the baseline survey was 70%, and among baseline participants 74% had at least one roommate who also participated. Among baseline participants with at least one roommate who also participated at baseline, 63% completed the followup survey. Across these levels of sample attrition the sample remained nearly identical (and statistically indistinguishable) in terms of stigma and all other measures except gender (women had higher response rates than men) (eTable 1). Also, roommate treatment history at baseline did not predict response at follow-up, indicating that there was not differential attrition between the “treatment” and “control” groups in this natural experiment. The baseline characteristics of the “treatment” and “control” groups are displayed in Table 2. There were no significant Table 1 Characteristics of primary analytic sample (N ¼ 1605). Baseline
Follow-up
University A (large public) University B (large private) Double room Triple room Quad room Age Female White Asian Black Hispanic Other Multi
0.69 0.31 0.79 0.17 0.04 18.4 (0.41) 0.54 0.70 0.17 0.03 0.05 0.02 0.04
Parents’ education Less than college degree College degree Graduate degree
0.16 0.27 0.56
Depression (PHQ-2 screen) Score (0e6) Positive screen (score >¼ 2)
0.84 (1.05) 0.24
1.07 (1.22) 0.33
p < 0.01 p < 0.01
4.13 (3.28)
5.13 (3.91)
p < 0.01
1.20 (1.20) 0.10
1.18 (1.22) 0.13
p ¼ 0.64 p ¼ 0.03
0.07
0.09
p ¼ 0.01
0.08
0.09
p ¼ 0.19
0.16
0.19
p ¼ 0.01
Psychological distress (K-6 score) (0e24) Personal stigma (0e5) Ever diagnosed with mental disorder Psychotropic medication in past 6 months Counseling/therapy in past 6 months Any tx history (any of the above three categories)
19.0 (0.41)
All values are proportions, except for variables with standard deviations in parentheses. The p-values are for two-tailed tests of equal means (t-test) or proportions (chi-square) between baseline and follow-up.
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
1125
Table 2 Baseline demographic, mental health, and stigma characteristics of students assigned to no roommates with prior treatment experience, versus assigned to at least one roommate with prior treatment experience. Characteristic
Age (years) PHQ-2 Score (0e6) K-6 Score (0e24) Personal stigma (0e5)
No roommates with prior treatment experience (N ¼ 1329)
At least one roommate with prior treatment experience (N ¼ 276)
Mean
Mean
SD
18.4 0.85 4.06 1.22
0.41 1.05 3.27 1.21
p
SD
18.4 0.81 4.35 1.12
0.37 1.03 3.3 1.19
N
%
N
%
Race/ethnicity White Asian Black Hispanic Other Multi
934 226 41 59 24 45
70.3 17.0 3.1 4.4 1.8 3.4
200 36 7 14 5 14
72.5 13.0 2.5 5.1 1.8 5.1
Parents’ education Less than college degree College degree Graduate degree
215 355 750
16.3 26.9 56.8
45 81 148
16.4 26.9 56.8
Ever diagnosed with mental disorder Psychotropic medication in past 6 months Counseling/therapy in past 6 months Any tx history (any of the above three categories)
144 92 101 213
10.9 6.9 25.0 16.1
21 17 25 41
7.6 6.2 9.1 14.9
0.33 0.63 0.19 0.23 p 0.47
0.64
0.11 0.65 0.41 0.63
All p-values correspond to 2-tailed tests of means (t-tests) or proportions (chi-squared tests).
differences between the two groups in the baseline demographic, mental health, or stigma characteristics. The validity of the research design was further supported by the lack of correlation in other baseline characteristics among roommates. These correlations were small and not statistically significant for all measures that we examined (eTable 2), supporting the premise that the assignment of roommates with or without treatment history was random, conditional on the factors controlled for in the analysis. The main results (Table 3) indicate that being assigned to a roommate with a history of mental health treatment or diagnosis caused a statistically significant increase in personal stigma (p ¼ 0.03). The size of the increase was equivalent to 15% of the mean (and of the standard deviation, which happened to be identical to the mean) of the baseline stigma score. The estimated increase in stigma was larger from contact with roommates with previous diagnosis or therapy/counseling, as compared to medication use, although the differences in these effects were not statistically significant. In sensitivity analyses the effect of roommate treatment history on stigma remained positive and significant in an ordered probit specification (p ¼ 0.02). Also, the results remained nearly identical
(B ¼ 0.16, p ¼ 0.04) when controlling for additional roommate characteristics. The main result also remained nearly unchanged (B ¼ 0.14, p ¼ 0.04) when the key predictor variable was redefined as equal to one, rather than the roommate average, when any of multiple roommates had a treatment history. Table 4 shows results from the exploratory subgroup analyses. For ease of interpretation the table shows results from separate regressions by subgroup, and the p-values in the final column refer to statistical significance of the subgroup interaction terms added to the main model to test for heterogeneous effects of contact. The only statistically significant difference in effects across subgroups was for own treatment history (p ¼ 0.02). Students without a previous treatment history experienced a significant increase in stigma when assigned to a roommate with a treatment history (p ¼ 0.01), whereas the estimated effect on students with a previous treatment history was negative and not significant (p ¼ 0.42). The estimated impact on stigma was also larger among students with lower previous knowledge about mental illness, as hypothesized, although this difference across subgroups was not statistically significant.
Table 3 Effect of roommate (RM) treatment history on stigma. Coeff
SE
RM: any tx history (0/1) RM: ever diagnosed (0/1) RM: medication past 6 mos (0/1) RM: therapy in past 6 mos (0/1)
0.162
0.072
Student’s stigma level at baseline
0.605
0.024
p
Coeff
SE
0.200
0.089
p
Coeff
SE
0.121
0.108
p
Coeff
SE
p
Coeff
SE
p
0.161 0.020
0.110 0.133
0.14 0.88
0.133
0.104
0.20
0.03
<0.01
0.604
0.024
0.03
<0.01
0.606
0.024
0.26
<0.01
0.199
0.095
0.04
0.604
0.024
<0.01
Joint F(3,1038) test: p ¼ 0.09 0.604 0.024 <0.01
N ¼ 1605. Each set of three columns corresponds to a separate linear regression, and only the estimates for the key coefficients are shown. The regression coefficients are unstandardized and represent the estimated effect on the personal stigma score at follow-up, which has a mean of 1.18 and standard deviation of 1.22. All p-values correspond to 2-tailed t-tests with df ¼ (1, 1038). All regressions also include controls for variables used for housing assignments, baseline level of the dependent variable, gender, age, race/ethnicity, parents’ education.
1126
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
Table 4 Subgroup analysis of effect of roommate (RM) treatment history on stigma.
Men Women Own treatment history ¼ yes Own treatment history ¼ no Self-reported knowledge ¼ above average Self-reported knowledge ¼ average or below Own stigma ¼ high Own stigma ¼ low RM tendency to disclose distress ¼ high RM tendency to disclose distress ¼ medium RM tends to disclose distress ¼ low RM symptoms of psych. distress ¼ high RM symptoms of psych. distress ¼ low University A University B
N
%
DV mean
Coeff.
SE
p-Value
Interaction p-value
779 873 264 1381 731 918 554 1095 248 245 795 226 1423 1101 504
0.47 0.53 0.16 0.84 0.44 0.56 0.34 0.66 0.19 0.19 0.62 0.14 0.86 0.69 0.31
1.42 0.98 0.73 1.27 1.02 1.32 2.14 0.70 1.08 1.25 1.2 1.32 1.17 1.15 1.27
0.213 0.138 0.166 0.212 0.111 0.237 0.231 0.136 0.217 0.163 0.186 0.045 0.170 0.145 0.264
0.116 0.093 0.205 0.083 0.117 0.104 0.179 0.077 0.212 0.221 0.101 0.202 0.082 0.083 0.148
0.07 0.14 0.42 0.01 0.34 0.02 0.20 0.08 0.31 0.46 0.07 0.82 0.04 0.08 0.08
0.52 0.02 0.26 0.22 0.75
0.69 0.52
Each row corresponds to a separate linear regression, and only the estimate for the key coefficient (on RM treatment history) is shown. The regression coefficients are unstandardized and represent the estimated effect on the personal stigma score at follow-up. All regressions also include controls for variables used for housing assignments, baseline level of the dependent variable, gender, age, race/ethnicity, parents’ education. Also, analyses by RM tendency to disclose distress are limited to students in double rooms (with just one roommate). “Interaction p-value” refers to the significance of the coefficient on the interaction term between RM treatment history and the independent variable of the respective row.
Discussion This study provides novel evidence on the effect of interpersonal contact with treatment users on mental illness stigma, based on data from assigned university roommates. The main strength of this study is in examining the effect of naturalistic contact with treatment users, while also closely mimicking a randomized study design. The main result was that interpersonal contact caused a modest increase in stigma. This result runs counter to previous research in a number of respects. First, previous studies have generally found that interpersonal contact was associated with a reduction in stigma, across a wide range of settings and research designs (Corrigan et al., under review; Couture & Penn, 2003; Holzinger et al., 2008). Second, the context of university roommates seems to meet the conditions hypothesized by social contact theory to facilitate stigma reductions (Allport, 1954): equal status, sustained and close contact, and socially sanctioned relationships. Third, previous studies of assigned roommates have found beneficial effects of interpersonal contact in the context of attitudes and empathy toward different racial/ethnic groups (Boisjoly, Duncan, Kremer, Levy, & Eccles, 2006; Van Laar, Levin, Sinclair, & Sidanius, 2005). One possible interpretation of the present study’s findings is that previous non-experimental research does not measure a true causal effect of contact on stigma, due to the presence of uncontrolled confounding factors in naturalistic settings. This seems plausible, given that confounding factors due to self-selection into contact are likely to overstate the effect of contact in reducing stigma: people in voluntary contact with mentally ill persons may be more inclined, on average, to have lower stigma (or greater decreases in stigma over time), for reasons not fully captured by measured variables. Previous randomized experimental studies, however, also find that contact significantly reduces stigma (Corrigan et al., under review). As noted earlier, contact-based interventions typically present the contact in a purposeful manner, with information and guidance regarding mental illness and treatment. The present study evaluates a somewhat different question, and suggests that naturalistic contact alone, without a carefully planned process of disclosure and education, does not necessarily reduce stigma and may actually increase stigma in some contexts. This is consistent with the persistence of stigma over time in countries such as the United States, despite substantial increases in treatment and the accompanying increase in contact with treatment users.
The findings in the present study may also reflect other differences in context as compared to previous studies. Most previous studies focused on contact with persons with severe mental illnesses such as schizophrenia or bipolar disorder, whereas the present study examined contact with persons who had a previous diagnosis or use of services but did not necessarily have severe, active, or even self-acknowledged mental illness. The fact that roommate assignments represent a form of involuntary contact may also be important. Although previous studies have found stigma reductions associated with both involuntary (not chosen) and voluntary (deliberately chosen) contact (Link & Cullen, 1986), the evidence is generally stronger for voluntary contact (Couture & Penn, 2003). In addition, the age group in the present study (most students were 18 or 19) may experience different effects than other age groups. Prior research on the association between contact with mentally ill persons and stigmatizing attitudes among adolescents and young adults has mixed conclusions about the direction of the relationship (Corrigan et al., 2005; Jorm & Wright, 2008), and a recent meta-analysis found that the contact-based interventions may be somewhat less effective for adolescents as compared to adults (Corrigan et al., under review). Limitations This study’s findings should be interpreted with several limitations in mind. First, the effects observed in the context of university students may not generalize to other types of interpersonal contact. University students are nevertheless an important population in their own right, as they represent a large proportion of the next generation of adults. Second, information was not collected in the follow-up survey about roommates’ awareness or perceptions of each other’s mental health history, due to concerns that such questions could provoke discomfort among roommates. Thus, the study was not able to characterize fully the nature of contact among roommates. Third, the stigma measure was based on a single survey item. Whereas most previous studies focused on more marked stigmatizing attitudes about people with mental illness (e.g., dangerousness and desire for social distance), the present study examined a more generic measure of feelings about people who have used mental health treatment. In designing the study, this measure was considered more appropriate for assessing stigma in a population where depression and anxiety are far more common conditions than bipolar disorder and schizophrenia.
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
Although this item was chosen based on its high correlation with other items adapted from a well-established stigma scale, this measure did not allow for an examination of the multiple, complex dimensions of stigma. Collectively, these limitations point to important directions for future research. Conclusions The results of this study indicate that naturalistic contact with users of mental health services does not inevitably lead to improved attitudes about mental illness. The effects of contact are likely to depend on a variety of factors at the individual, social network, and macro-societal levels, as emphasized in a recently proposed theoretical framework for mental illness stigma (Pescosolido, Martin, Lang, & Olafsdottir, 2008). Given that contact alone may not be effective in some contexts, it is essential to identify which combination of intervention strategies can yield the largest desired effects. Proposed intervention strategies typically include not only interpersonal contact but also some form of education or guidance to help people interpret their experiences in this contact (Pinfold, Thornicroft, Huxley, & Farmer, 2005). A priority for future research should be to improve understanding of both the separate and interactive effects of the key components within contact-based programs to reduce stigma. It will also be valuable to evaluate intervention strategies in the contexts of both severe mental illness as well as more common mental health problems such as depression and anxiety. While stigma associated with the most severe and persistent mental illnesses has understandably been the focus of most previous research on contact-based strategies, stigma associated with a broader set of mental health problems is relevant to a larger proportion of the population. The present study’s findings raise the possibility that the effects of contact may differ in some ways for less severe but more common mental disorders. Acknowledgments This study was funded by a research grant from the William T. Grant Foundation. The survey data collection was administered by Survey Sciences Group, LLC. The authors are also grateful to campus staff members who assisted with the collection of administrative data and to the students who participated in the surveys. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.socscimed.2012.05.007. References Allport, G. W. (1954). The nature of prejudice. Cambridge, Massachusetts: AddisonWesley Publishing Co., Inc. Blanco, C., Okuda, M., Wright, C., Hasin, D. S., Grant, B. F., Liu, S., et al. (2008). Mental health of college students and their non-college-attending peers: results from the national epidemiologic study on alcohol and related conditions. Archives of General Psychiatry, 65(12), 1429e1437. Boisjoly, J., Duncan, G. J., Kremer, M., Levy, D. M., & Eccles, J. (2006). Empathy or antipathy? The impact of diversity. American Economic Review, 96(5), 1890e1905. Corrigan, P. W. (2004). How stigma interferes with mental health care. The American Psychologist, 59(7), 614e625. Corrigan, P. W., Lurie, B. D., Goldman, H. H., Slopen, N., Medasani, K., & Phelan, S. (2005). How adolescents perceive the stigma of mental illness and alcohol abuse. Psychiatric Services, 56(5), 544e550.
1127
Corrigan, P. W., Morris, S. B., Michaels, P. J., Rafacz, J. E., & Rusch, N. (2011). Challenging the public stigma of mental illness: A meta-analysis of outcome studies. Psychiatric Services, in press. Corrigan, P. W., & Penn, D. L. (1999). Lessons from social psychology on discrediting psychiatric stigma. American Psychologist, 54(9), 765e776. Couture, S. M., & Penn, D. (2003). Interpersonal contact and the stigma of mental illness: a review of the literature. Journal of Mental Health, 12(3), 291e305. Couture, S. M., & Penn, D. L. (2006). The effects of prospective naturalistic contact on the stigma of mental illness. Journal of Community Psychology, 34(5), 635e645. Eisenberg, D., Downs, M. F., Golberstein, E., & Zivin, K. (2009). Stigma and help-seeking for mental health among college students. Medical Care Research and Review, 66(5), 522e541. Eisenberg, D., Golberstein, E., & Gollust, S. E. (2007). Help-seeking and access to mental health care in a university student population. Medical Care, 45(7), 594e601. Eisenberg, D., Hunt, J., Speer, N., & Zivin, K. (2011). Mental health service utilization among college students in the united states. Journal of Nervous and Mental Disease, 199(5), 301e308. Holzinger, A., Dietrich, S., Heitmann, S., & Angermeyer, M. (2008). Evaluation of target-group oriented interventions aimed at reducing the stigma surrounding mental illness: a systematic review. Psychiatrische Praxis, 35(8), 376e386. Jorm, A. F., & Wright, A. M. (2008). Influences on young people’s stigmatising attitudes towards peers with mental disorders: national survey of young Australians and their parents. The British Journal of Psychiatry, 192, 144e149. Kahn, J. H., & Hessling, R. M. (2001). Measuring the tendency to conceal versus disclose psychological distress. Journal of Social and Clinical Psychology, 20(1), 41e65. Kessler, R. C., Barker, P. R., Colpe, L. J., Epstein, J. F., Gfroerer, J. C., Hiripi, E., et al. (2003). Screening for serious mental illness in the general population. Archives of General Psychiatry, 60(2), 184e189. Kolodziej, M. E., & Johnson, B. T. (1996). Interpersonal contact and acceptance of persons with psychiatric disorders: a research synthesis. Journal of Consulting and Clinical Psychology, 64(6), 1387e1396. Link, B. G. (1987). Understanding labeling effects in the area of mental disorders: an assessment of the effects of expectations of rejection. American Sociological Review, 52(1), 96e112. Link, B. G., & Cullen, F. T. (1986). Contact with the mentally ill and perceptions of how dangerous they are. Journal of Health and Social Behavior, 27, 289e303. Link, B. G., Cullen, F., Struening, E., Shrout, P., & Dohrenwend, B. (1989). A modified labeling theory approach to mental disorders: an empirical assessment. American Sociological Review, 54(3), 400e423. Löwe, B., Kroenke, K., & Gräfe, K. (2005). Detecting and monitoring depression with a two-item questionnaire (PHQ-2). Journal of Psychosomatic Research, 58(2), 163e171. Penn, D. L., Judge, A., Jamieson, P., Garczynski, J., Hennessy, M., & Romer, D. (2005). Stigma. Treating and preventing adolescent mental health disorders: What we know and what we don’t know. New York: Oxford University Press with the Annenberg Foundation Trust at Sunnylands and the Annenberg Public Policy Center at the University of Pennsylvania. Pescosolido, B. A., Martin, J. K., Lang, A., & Olafsdottir, S. (2008). Rethinking theoretical approaches to stigma: a framework integrating normative influences on stigma (FINIS). Social Science & Medicine, 67(3), 431e440. Pescosolido, B. A., Martin, J. K., Long, J. S., Medina, T. R., Phelan, J. C., & Link, B. G. (2010). “A disease like any other?” A decade of change in public reactions to schizophrenia, depression, and alcohol dependence. American Journal of Psychiatry, 167, 1321e1330. Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Personality and Social Psychology, 90(5), 751e783. Phelan, J. C., Link, B. G., Stueve, A., & Pescosolido, B. A. (2000). Public conceptions of mental illness in 1950 and 1996: what is mental illness and is it to be feared? Journal of Health and Social Behavior, 41(2), 188e207. Pinfold, V., Thornicroft, G., Huxley, P., & Farmer, P. (2005). Active ingredients in anti-stigma programmes in mental health. International Review of Psychiatry, 17(2), 123e131. Reinke, R. R., Corrigan, P. W., Leonhard, C., Lundin, R. K., & Kubiak, M. A. (2004). Examining two aspects of contact on the stigma of mental illness. Journal of Social and Clinical Psychology, 23(3), 377e389. Saraceno, B., van Ommeren, M., Batniji, R., Cohen, A., Gureje, O., Mahoney, J., et al. (2007). Barriers to improvement of mental health services in low-income and middle-income countries. Lancet, 370(9593), 1164e1174. Sartorius, N. (2002). Iatrogenic stigma of mental illness. BMJ, 324, 1470. U.S. Department of Health and Human Services. (1999). Mental health: A report of the surgeon general. Rockville, MD: National Institute of Mental Health. Van Laar, C., Levin, S., Sinclair, S., & Sidanius, J. (2005). The effect of university roommate contact on ethnic attitudes and behaviors. Journal of Experimental Social Psychology, 41(4), 329e345. Van Voorhees, B. W., Fogel, J., Houston, T. K., Cooper, L. A., Wang, N., & Ford, D. E. (2005). Beliefs and attitudes associated with the intention to not accept the diagnosis of depression among young adults. Annals of Family Medicine, 3, 38e46.
Contents lists available at SciVerse ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Effects of contact with treatment users on mental illness stigma: Evidence from university roommate assignments Daniel Eisenberg a, *, Marilyn F. Downs b, Ezra Golberstein c a
Department of Health Management & Policy, School of Public Health, University of Michigan, M3517 SPH II, MC 2029, 1415 Washington Heights, Ann Arbor, MI 48109-2029, USA Counseling and Mental Health Service, Tufts University, USA c Division of Health Policy & Management, School of Public Health, University of Minnesota, USA b
a r t i c l e i n f o
a b s t r a c t
Article history: Available online 1 June 2012
Mental illness stigma refers to negative stereotypes and prejudices about people with mental illness, and is a widespread phenomenon with damaging social, psychological, and economic consequences. Despite considerable policy attention, mental illness stigma does not appear to have declined significantly in recent years. Interpersonal contact with persons with mental illness has been identified as a promising approach to reducing mental illness stigma. This study investigates the effect of contact with mental health treatment users on stigma using an observational research design that is free of self-selection bias. The research design is based on the quasi-experiment in which university students are assigned to live together as roommates. Survey data were collected from first-year undergraduates at two large universities in the United States (N ¼ 1605). Multivariable regressions were used to estimate the effect of assignment to a roommate with a history of mental health treatment on a brief measure of stigmatizing attitudes. Contact with a treatment user caused a modest increase in stigma (standardized effect size ¼ 0.15, p ¼ 0.03). This effect was present among students without a prior treatment history of their own, but not among those with a prior history. The findings indicate that naturalistic contact alone does not necessarily yield a reduction in mental illness stigma. This may help explain why stigma has not declined in societies such as the United States even as treatment use has risen substantially. The findings also highlight the importance of isolating the specific components, beyond contact per se, that are necessary to reduce stigma in contact-based interventions. Ó 2012 Elsevier Ltd. All rights reserved.
Keywords: USA Stigma Social contact Adolescents Mental health
Introduction Mental illness stigma refers to negative stereotypes and prejudices about people with mental illness, and is a widespread phenomenon with damaging social, psychological, and economic consequences (Corrigan, 2004; Phelan, Link, Stueve, & Pescosolido, 2000). Stigma is associated with negative views about help-seeking and a lower use of mental health services (Penn et al., 2005; Van Voorhees et al., 2005), and may contribute to underfunding of mental health services and programs (Saraceno et al., 2007). For all of these reasons there have been significant policy and research efforts to improve attitudes in many countries, including the United States (Sartorius, 2002; U.S. Department of Health and Human Services, 1999). Despite this attention, however, public stigma about mental illness did not decline significantly between 1996 and 2006 in national surveys in the United States (Pescosolido et al., 2010). * Corresponding author. Tel.: þ1 734 615 7764; fax: þ1 734 764 4338. E-mail address: [email protected] (D. Eisenberg). 0277-9536/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2012.05.007
Interpersonal contact with persons with mental illness has been identified as one of the most promising approaches to reducing stigma (Corrigan & Penn, 1999). This proposition is supported by a large number of studies that have found significant associations between various indicators of stigma and interpersonal contact with mentally ill persons and treatment users (Corrigan, Morris, Michaels, Rafacz, & Rusch, under review; Couture & Penn, 2003; Holzinger, Dietrich, Heitmann, & Angermeyer, 2008). Previous studies also suggest that reductions in stigma are more likely when the interpersonal contact is a deliberate choice (Couture & Penn, 2003; Kolodziej & Johnson, 1996), the quality of contact is high (Couture & Penn, 2006; Pettigrew & Tropp, 2006), and the contact experience only moderately challenges existing stereotypes (Reinke, Corrigan, Leonhard, Lundin, & Kubiak, 2004). The previous studies of interpersonal contact are based on two types of study designs: 1) naturalistic, non-experimental studies measuring the correlation between stigma and contact, controlling for potential confounding variables; and, 2) experimental studies that manipulate the amount and type of contact and then compare
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
stigma outcomes across groups. In the naturalistic studies, persons who self-select into contact with persons with mental illness may have lower stigma, or larger decreases in stigma, due to factors that are difficult to control for statistically, such as a greater openness to different experiences. Experimental studies, on the other hand, can avoid this type of confounding by randomly assigning participants to contact conditions; however, these studies typically involve a limited duration and intensity of contact, such as presentations or testimonials by a person with a mental illness. The present study adds a new type of evidence to this literature. The effect of interpersonal contact on stigma is estimated among first-year college roommates who are assigned to each other. In contrast to previous naturalistic, non-experimental studies, this research design avoids potential confounding due to self-selection into contact. Also, in contrast to most experimental studies, the design involves sustained, close contact; college roommates share a small living space over a period of approximately seven months, and are therefore likely to develop friendships and have frequent one-on-one interactions. A priori it is unclear what effect this type of contact with mental health services users should be expected to have on stigma. Although contact-based interventions have shown promise, as noted above, these approaches typically involve a purposeful and thoughtful disclosure of mental illness, including some form of education or guidance. The present study, in contrast, evaluates naturalistic contact, without any structured intervention regarding how mental illness or use of services is presented. Thus, the present study complements previous evidence by addressing a somewhat different angle both in terms of research design and substantive question. Method Sample selection and recruitment The sample consisted of first-year students who began college in fall 2009 at two large universities in the United States. At both universities first-year students were required to live in campus housing, except in unusual circumstances. Students who did not request specific roommates were assigned to their roommates. All first-year students with assigned roommates were recruited for the study, with the exception of students who were still under 18 years old as of the follow-up survey (0.9% of the initial sample). The students were invited to complete two brief online surveys: a baseline survey in August 2009 (1e3 weeks before students arrived at college) and a follow-up survey in MarcheApril 2010 (2e4 weeks before the end of the academic year). The survey data were linked to administrative data on housing preferences, room assignments, and academic and demographic characteristics. For both the baseline and follow-up surveys, students were recruited with an introductory letter including a $10 bill, followed by up to four email invitations to those who had yet to respond, spaced by 3e5 days each. All communications included a web link to the survey and a unique, randomly assigned log-in code for each student. The recruitment messages also informed students that they were entered into a sweepstakes for cash prizes regardless of participation. Informed consent was obtained at the beginning of the online survey. All aspects of data collection were approved by Institutional Review Boards at both universities. The primary analytic sample consisted of students with complete data necessary for the statistical analyses. These were students who completed both baseline and follow-up surveys and whose roommate(s) also completed the baseline survey. The potential for bias due to differential survey non-response was investigated, as described in the technical Appendix.
1123
Measures The primary outcome was a brief measure of general stigmatizing attitudes toward mental health treatment. This was measured in both baseline and follow-up surveys with an item asking, “Please indicate whether you agree or disagree with the following statement: I think less of someone who has received mental health treatment.” The answer choices were on a Likert scale (strongly agree, agree, somewhat agree, somewhat disagree, disagree, and strongly disagree), which was coded as a 0e5 linear score. This measure was adapted from an item on the widely used Discrimination-Devaluation Scale (Link, 1987; Link, Cullen, Struening, Shrout, & Dohrenwend, 1989). The adapted item has been used in a national analysis of stigma and help-seeking among university students (Eisenberg, Downs, Golberstein, & Zivin, 2009). In that analysis, the item had a significant negative correlation (r ¼ 0.58, p < 0.001) with the sum of other items adapted from the Discrimination-Devaluation Scale, and was a significant negative predictor of treatment use (odds-ratio ¼ 0.65, p < 0.001). The key predictor variable was the treatment history of assigned roommates. This variable was constructed based on roommates’ responses to three survey questions about prior treatment and diagnoses, which were taken from the Healthy Minds Study, a national survey study of mental health and service utilization among university students (Eisenberg, Hunt, Speer, & Zivin, 2011). Specifically, these questions asked about lifetime diagnoses of mental disorders by a health professional, use of psychiatric medication in the past six months, and use of counseling or therapy in the past six months. The primary measure of roommate’s treatment history was a composite variable equal to one if the roommate had any previous diagnosis or treatment use, and zero if not. The three components of this variable were also examined separately in additional analyses. For students with multiple roommates, the mean value of roommates was used. The key covariates in the analysis are preferences expressed in students’ housing applications, which were used by the housing administrators to match roommates. The primary variables used for matching roommates included preferences about room type (double, triple, or quad), same-sex versus mixed-sex hallway, and smoker status. These variables were completely controlled for, so any remaining variation in roommate characteristics should be random and thus uncorrelated with stigma or any factor that could affect stigma, as detailed in the Appendix. Statistical analysis In the primary analysis, a linear regression model was estimated as in this equation:
Stigmaðtþ1Þ ¼ b0 þ b1 Prefst þ b2 RMtxHistoryt þ b3 Stigmat þ b4 Xt þ 3 tþ1 (1) The subscript t denotes a measurement in the baseline survey, and t þ 1 denotes a measurement in the follow-up survey. Prefs refers to a vector of housing preferences and other variables used to make roommate assignments (as detailed in the Appendix), and X is a vector of individual covariates including gender, age (exact to the day), race/ethnicity (white, Black, Asian, Hispanic/Latino, mixed race, or other), and parents’ highest education level (less than a college degree, college degree, or graduate degree). These covariates were included because prior studies have found them to correlate significantly with mental health and help-seeking behavior in college populations (Blanco et al., 2008; Eisenberg, Golberstein, & Gollust, 2007). The key coefficient is b2, which
1124
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
represents the effect of roommate treatment history on the stigma score at follow-up. Heteroskedasticity-robust standard errors were estimated, to correct for correlated outcomes among roommates. The statistical significance of comparisons of means was assessed using two-tailed t-statistics and the significance of the regression coefficients was assessed using two-tailed z-statistics. In order to interpret b2 as an unbiased estimate of the causal effect of roommate’s treatment history on stigma, the key assumption is that the roommate’s treatment history is uncorrelated with the error term, 3 tþ1 , in equation (1). In effect this assumption means that there were no unmeasured roommate characteristics that would confound the analysis; this assumption is highly plausible because all variables used for matching roommates were included in the analysis. The Appendix describes how this assumption was examined empirically. Sensitivity analyses examined whether the relationship between stigma and roommate’s treatment history is robust to the inclusion of several additional baseline roommate covariates related to mental health and other factors. These covariates include: the PHQ-2 depression score (Löwe, Kroenke, & Gräfe, 2005); an index of anxiety symptoms measured by the sum of two items in the K-6 psychological distress scale (Kessler et al., 2003); parents’ education (highest level obtained by either parent); how religious one is (very, somewhat, a little, not at all); frequency of binge drinking in the past 30 days; frequency of exercise in the past 30 days; average hours per day spent studying in the last year of high school; standardized admissions test score (total ACT and/or SAT, converted to a z-score based on the withinuniversity distribution); and high school GPA (also converted to a z-score). Another sensitivity analysis defined the key predictor variable as equal to one if any roommate had a treatment history, rather than the average across roommates, in cases of multiple roommates. In a final set of sensitivity analyses, ordered probit regressions were estimated instead of linear regressions. Subgroups In addition to the primary analysis of the overall sample, a number of exploratory subgroup analyses were conducted to examine the heterogeneity of contact effects. First, it was hypothesized that students with a prior treatment history would experience less change in stigma due to contact with a roommate with a treatment history, because they would be more likely to have previous knowledge and well-formed attitudes about mental health treatment. Similarly, it was hypothesized that students with higher self-reported knowledge about mental illnesses (as measured by the question, “Relative to the average person, how knowledgeable are you about mental illnesses (such as depression and anxiety disorders) and their treatments? (Please just give your best guess)”) would be less affected by contact than those who reported average or lower knowledge. A number of other subgroup analyses were conducted for which the hypotheses were ambiguous, due to potentially offsetting factors. Three of these analyses defined subgroups by the following baseline characteristics, respectively: own stigma level (high or low); roommates’ tendency to disclose distress (high, medium, low on an item in the Distress Disclosure Index asking about agreement with the statement, “When I feel depressed or sad, I tend to keep those feelings to myself.”) (Kahn & Hessling, 2001); and roommates’ symptoms of psychological distress (high or low K-6 score). Students with higher stigma at baseline may be more resistant to changing their attitudes, but they also have more room for a potential decrease in stigma. Roommates with more symptoms may provide more salient cues, which could either trigger negative perceptions or evoke more sympathy and support. Similarly, roommates who
tend to disclose distress may engender negative perceptions, but may also generate more openness and interpersonal understanding. Finally, a subgroup analysis by university was conducted to assess the robustness of results across settings. Subgroup heterogeneity was tested by adding a covariate for the interaction between roommate treatment history and subgroup membership (separately for each of the above subgroups) to the main regression model. Results The basic characteristics of the full analytic sample (N ¼ 1605) are shown in Table 1. The average level of stigma was almost identical between the baseline and follow-up surveys. Mental health, on the other hand, worsened between the time points, with significant increases in depression and psychological distress scores. The response rate to the baseline survey was 70%, and among baseline participants 74% had at least one roommate who also participated. Among baseline participants with at least one roommate who also participated at baseline, 63% completed the followup survey. Across these levels of sample attrition the sample remained nearly identical (and statistically indistinguishable) in terms of stigma and all other measures except gender (women had higher response rates than men) (eTable 1). Also, roommate treatment history at baseline did not predict response at follow-up, indicating that there was not differential attrition between the “treatment” and “control” groups in this natural experiment. The baseline characteristics of the “treatment” and “control” groups are displayed in Table 2. There were no significant Table 1 Characteristics of primary analytic sample (N ¼ 1605). Baseline
Follow-up
University A (large public) University B (large private) Double room Triple room Quad room Age Female White Asian Black Hispanic Other Multi
0.69 0.31 0.79 0.17 0.04 18.4 (0.41) 0.54 0.70 0.17 0.03 0.05 0.02 0.04
Parents’ education Less than college degree College degree Graduate degree
0.16 0.27 0.56
Depression (PHQ-2 screen) Score (0e6) Positive screen (score >¼ 2)
0.84 (1.05) 0.24
1.07 (1.22) 0.33
p < 0.01 p < 0.01
4.13 (3.28)
5.13 (3.91)
p < 0.01
1.20 (1.20) 0.10
1.18 (1.22) 0.13
p ¼ 0.64 p ¼ 0.03
0.07
0.09
p ¼ 0.01
0.08
0.09
p ¼ 0.19
0.16
0.19
p ¼ 0.01
Psychological distress (K-6 score) (0e24) Personal stigma (0e5) Ever diagnosed with mental disorder Psychotropic medication in past 6 months Counseling/therapy in past 6 months Any tx history (any of the above three categories)
19.0 (0.41)
All values are proportions, except for variables with standard deviations in parentheses. The p-values are for two-tailed tests of equal means (t-test) or proportions (chi-square) between baseline and follow-up.
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
1125
Table 2 Baseline demographic, mental health, and stigma characteristics of students assigned to no roommates with prior treatment experience, versus assigned to at least one roommate with prior treatment experience. Characteristic
Age (years) PHQ-2 Score (0e6) K-6 Score (0e24) Personal stigma (0e5)
No roommates with prior treatment experience (N ¼ 1329)
At least one roommate with prior treatment experience (N ¼ 276)
Mean
Mean
SD
18.4 0.85 4.06 1.22
0.41 1.05 3.27 1.21
p
SD
18.4 0.81 4.35 1.12
0.37 1.03 3.3 1.19
N
%
N
%
Race/ethnicity White Asian Black Hispanic Other Multi
934 226 41 59 24 45
70.3 17.0 3.1 4.4 1.8 3.4
200 36 7 14 5 14
72.5 13.0 2.5 5.1 1.8 5.1
Parents’ education Less than college degree College degree Graduate degree
215 355 750
16.3 26.9 56.8
45 81 148
16.4 26.9 56.8
Ever diagnosed with mental disorder Psychotropic medication in past 6 months Counseling/therapy in past 6 months Any tx history (any of the above three categories)
144 92 101 213
10.9 6.9 25.0 16.1
21 17 25 41
7.6 6.2 9.1 14.9
0.33 0.63 0.19 0.23 p 0.47
0.64
0.11 0.65 0.41 0.63
All p-values correspond to 2-tailed tests of means (t-tests) or proportions (chi-squared tests).
differences between the two groups in the baseline demographic, mental health, or stigma characteristics. The validity of the research design was further supported by the lack of correlation in other baseline characteristics among roommates. These correlations were small and not statistically significant for all measures that we examined (eTable 2), supporting the premise that the assignment of roommates with or without treatment history was random, conditional on the factors controlled for in the analysis. The main results (Table 3) indicate that being assigned to a roommate with a history of mental health treatment or diagnosis caused a statistically significant increase in personal stigma (p ¼ 0.03). The size of the increase was equivalent to 15% of the mean (and of the standard deviation, which happened to be identical to the mean) of the baseline stigma score. The estimated increase in stigma was larger from contact with roommates with previous diagnosis or therapy/counseling, as compared to medication use, although the differences in these effects were not statistically significant. In sensitivity analyses the effect of roommate treatment history on stigma remained positive and significant in an ordered probit specification (p ¼ 0.02). Also, the results remained nearly identical
(B ¼ 0.16, p ¼ 0.04) when controlling for additional roommate characteristics. The main result also remained nearly unchanged (B ¼ 0.14, p ¼ 0.04) when the key predictor variable was redefined as equal to one, rather than the roommate average, when any of multiple roommates had a treatment history. Table 4 shows results from the exploratory subgroup analyses. For ease of interpretation the table shows results from separate regressions by subgroup, and the p-values in the final column refer to statistical significance of the subgroup interaction terms added to the main model to test for heterogeneous effects of contact. The only statistically significant difference in effects across subgroups was for own treatment history (p ¼ 0.02). Students without a previous treatment history experienced a significant increase in stigma when assigned to a roommate with a treatment history (p ¼ 0.01), whereas the estimated effect on students with a previous treatment history was negative and not significant (p ¼ 0.42). The estimated impact on stigma was also larger among students with lower previous knowledge about mental illness, as hypothesized, although this difference across subgroups was not statistically significant.
Table 3 Effect of roommate (RM) treatment history on stigma. Coeff
SE
RM: any tx history (0/1) RM: ever diagnosed (0/1) RM: medication past 6 mos (0/1) RM: therapy in past 6 mos (0/1)
0.162
0.072
Student’s stigma level at baseline
0.605
0.024
p
Coeff
SE
0.200
0.089
p
Coeff
SE
0.121
0.108
p
Coeff
SE
p
Coeff
SE
p
0.161 0.020
0.110 0.133
0.14 0.88
0.133
0.104
0.20
0.03
<0.01
0.604
0.024
0.03
<0.01
0.606
0.024
0.26
<0.01
0.199
0.095
0.04
0.604
0.024
<0.01
Joint F(3,1038) test: p ¼ 0.09 0.604 0.024 <0.01
N ¼ 1605. Each set of three columns corresponds to a separate linear regression, and only the estimates for the key coefficients are shown. The regression coefficients are unstandardized and represent the estimated effect on the personal stigma score at follow-up, which has a mean of 1.18 and standard deviation of 1.22. All p-values correspond to 2-tailed t-tests with df ¼ (1, 1038). All regressions also include controls for variables used for housing assignments, baseline level of the dependent variable, gender, age, race/ethnicity, parents’ education.
1126
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
Table 4 Subgroup analysis of effect of roommate (RM) treatment history on stigma.
Men Women Own treatment history ¼ yes Own treatment history ¼ no Self-reported knowledge ¼ above average Self-reported knowledge ¼ average or below Own stigma ¼ high Own stigma ¼ low RM tendency to disclose distress ¼ high RM tendency to disclose distress ¼ medium RM tends to disclose distress ¼ low RM symptoms of psych. distress ¼ high RM symptoms of psych. distress ¼ low University A University B
N
%
DV mean
Coeff.
SE
p-Value
Interaction p-value
779 873 264 1381 731 918 554 1095 248 245 795 226 1423 1101 504
0.47 0.53 0.16 0.84 0.44 0.56 0.34 0.66 0.19 0.19 0.62 0.14 0.86 0.69 0.31
1.42 0.98 0.73 1.27 1.02 1.32 2.14 0.70 1.08 1.25 1.2 1.32 1.17 1.15 1.27
0.213 0.138 0.166 0.212 0.111 0.237 0.231 0.136 0.217 0.163 0.186 0.045 0.170 0.145 0.264
0.116 0.093 0.205 0.083 0.117 0.104 0.179 0.077 0.212 0.221 0.101 0.202 0.082 0.083 0.148
0.07 0.14 0.42 0.01 0.34 0.02 0.20 0.08 0.31 0.46 0.07 0.82 0.04 0.08 0.08
0.52 0.02 0.26 0.22 0.75
0.69 0.52
Each row corresponds to a separate linear regression, and only the estimate for the key coefficient (on RM treatment history) is shown. The regression coefficients are unstandardized and represent the estimated effect on the personal stigma score at follow-up. All regressions also include controls for variables used for housing assignments, baseline level of the dependent variable, gender, age, race/ethnicity, parents’ education. Also, analyses by RM tendency to disclose distress are limited to students in double rooms (with just one roommate). “Interaction p-value” refers to the significance of the coefficient on the interaction term between RM treatment history and the independent variable of the respective row.
Discussion This study provides novel evidence on the effect of interpersonal contact with treatment users on mental illness stigma, based on data from assigned university roommates. The main strength of this study is in examining the effect of naturalistic contact with treatment users, while also closely mimicking a randomized study design. The main result was that interpersonal contact caused a modest increase in stigma. This result runs counter to previous research in a number of respects. First, previous studies have generally found that interpersonal contact was associated with a reduction in stigma, across a wide range of settings and research designs (Corrigan et al., under review; Couture & Penn, 2003; Holzinger et al., 2008). Second, the context of university roommates seems to meet the conditions hypothesized by social contact theory to facilitate stigma reductions (Allport, 1954): equal status, sustained and close contact, and socially sanctioned relationships. Third, previous studies of assigned roommates have found beneficial effects of interpersonal contact in the context of attitudes and empathy toward different racial/ethnic groups (Boisjoly, Duncan, Kremer, Levy, & Eccles, 2006; Van Laar, Levin, Sinclair, & Sidanius, 2005). One possible interpretation of the present study’s findings is that previous non-experimental research does not measure a true causal effect of contact on stigma, due to the presence of uncontrolled confounding factors in naturalistic settings. This seems plausible, given that confounding factors due to self-selection into contact are likely to overstate the effect of contact in reducing stigma: people in voluntary contact with mentally ill persons may be more inclined, on average, to have lower stigma (or greater decreases in stigma over time), for reasons not fully captured by measured variables. Previous randomized experimental studies, however, also find that contact significantly reduces stigma (Corrigan et al., under review). As noted earlier, contact-based interventions typically present the contact in a purposeful manner, with information and guidance regarding mental illness and treatment. The present study evaluates a somewhat different question, and suggests that naturalistic contact alone, without a carefully planned process of disclosure and education, does not necessarily reduce stigma and may actually increase stigma in some contexts. This is consistent with the persistence of stigma over time in countries such as the United States, despite substantial increases in treatment and the accompanying increase in contact with treatment users.
The findings in the present study may also reflect other differences in context as compared to previous studies. Most previous studies focused on contact with persons with severe mental illnesses such as schizophrenia or bipolar disorder, whereas the present study examined contact with persons who had a previous diagnosis or use of services but did not necessarily have severe, active, or even self-acknowledged mental illness. The fact that roommate assignments represent a form of involuntary contact may also be important. Although previous studies have found stigma reductions associated with both involuntary (not chosen) and voluntary (deliberately chosen) contact (Link & Cullen, 1986), the evidence is generally stronger for voluntary contact (Couture & Penn, 2003). In addition, the age group in the present study (most students were 18 or 19) may experience different effects than other age groups. Prior research on the association between contact with mentally ill persons and stigmatizing attitudes among adolescents and young adults has mixed conclusions about the direction of the relationship (Corrigan et al., 2005; Jorm & Wright, 2008), and a recent meta-analysis found that the contact-based interventions may be somewhat less effective for adolescents as compared to adults (Corrigan et al., under review). Limitations This study’s findings should be interpreted with several limitations in mind. First, the effects observed in the context of university students may not generalize to other types of interpersonal contact. University students are nevertheless an important population in their own right, as they represent a large proportion of the next generation of adults. Second, information was not collected in the follow-up survey about roommates’ awareness or perceptions of each other’s mental health history, due to concerns that such questions could provoke discomfort among roommates. Thus, the study was not able to characterize fully the nature of contact among roommates. Third, the stigma measure was based on a single survey item. Whereas most previous studies focused on more marked stigmatizing attitudes about people with mental illness (e.g., dangerousness and desire for social distance), the present study examined a more generic measure of feelings about people who have used mental health treatment. In designing the study, this measure was considered more appropriate for assessing stigma in a population where depression and anxiety are far more common conditions than bipolar disorder and schizophrenia.
D. Eisenberg et al. / Social Science & Medicine 75 (2012) 1122e1127
Although this item was chosen based on its high correlation with other items adapted from a well-established stigma scale, this measure did not allow for an examination of the multiple, complex dimensions of stigma. Collectively, these limitations point to important directions for future research. Conclusions The results of this study indicate that naturalistic contact with users of mental health services does not inevitably lead to improved attitudes about mental illness. The effects of contact are likely to depend on a variety of factors at the individual, social network, and macro-societal levels, as emphasized in a recently proposed theoretical framework for mental illness stigma (Pescosolido, Martin, Lang, & Olafsdottir, 2008). Given that contact alone may not be effective in some contexts, it is essential to identify which combination of intervention strategies can yield the largest desired effects. Proposed intervention strategies typically include not only interpersonal contact but also some form of education or guidance to help people interpret their experiences in this contact (Pinfold, Thornicroft, Huxley, & Farmer, 2005). A priority for future research should be to improve understanding of both the separate and interactive effects of the key components within contact-based programs to reduce stigma. It will also be valuable to evaluate intervention strategies in the contexts of both severe mental illness as well as more common mental health problems such as depression and anxiety. While stigma associated with the most severe and persistent mental illnesses has understandably been the focus of most previous research on contact-based strategies, stigma associated with a broader set of mental health problems is relevant to a larger proportion of the population. The present study’s findings raise the possibility that the effects of contact may differ in some ways for less severe but more common mental disorders. Acknowledgments This study was funded by a research grant from the William T. Grant Foundation. The survey data collection was administered by Survey Sciences Group, LLC. The authors are also grateful to campus staff members who assisted with the collection of administrative data and to the students who participated in the surveys. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.socscimed.2012.05.007. References Allport, G. W. (1954). The nature of prejudice. Cambridge, Massachusetts: AddisonWesley Publishing Co., Inc. Blanco, C., Okuda, M., Wright, C., Hasin, D. S., Grant, B. F., Liu, S., et al. (2008). Mental health of college students and their non-college-attending peers: results from the national epidemiologic study on alcohol and related conditions. Archives of General Psychiatry, 65(12), 1429e1437. Boisjoly, J., Duncan, G. J., Kremer, M., Levy, D. M., & Eccles, J. (2006). Empathy or antipathy? The impact of diversity. American Economic Review, 96(5), 1890e1905. Corrigan, P. W. (2004). How stigma interferes with mental health care. The American Psychologist, 59(7), 614e625. Corrigan, P. W., Lurie, B. D., Goldman, H. H., Slopen, N., Medasani, K., & Phelan, S. (2005). How adolescents perceive the stigma of mental illness and alcohol abuse. Psychiatric Services, 56(5), 544e550.
1127
Corrigan, P. W., Morris, S. B., Michaels, P. J., Rafacz, J. E., & Rusch, N. (2011). Challenging the public stigma of mental illness: A meta-analysis of outcome studies. Psychiatric Services, in press. Corrigan, P. W., & Penn, D. L. (1999). Lessons from social psychology on discrediting psychiatric stigma. American Psychologist, 54(9), 765e776. Couture, S. M., & Penn, D. (2003). Interpersonal contact and the stigma of mental illness: a review of the literature. Journal of Mental Health, 12(3), 291e305. Couture, S. M., & Penn, D. L. (2006). The effects of prospective naturalistic contact on the stigma of mental illness. Journal of Community Psychology, 34(5), 635e645. Eisenberg, D., Downs, M. F., Golberstein, E., & Zivin, K. (2009). Stigma and help-seeking for mental health among college students. Medical Care Research and Review, 66(5), 522e541. Eisenberg, D., Golberstein, E., & Gollust, S. E. (2007). Help-seeking and access to mental health care in a university student population. Medical Care, 45(7), 594e601. Eisenberg, D., Hunt, J., Speer, N., & Zivin, K. (2011). Mental health service utilization among college students in the united states. Journal of Nervous and Mental Disease, 199(5), 301e308. Holzinger, A., Dietrich, S., Heitmann, S., & Angermeyer, M. (2008). Evaluation of target-group oriented interventions aimed at reducing the stigma surrounding mental illness: a systematic review. Psychiatrische Praxis, 35(8), 376e386. Jorm, A. F., & Wright, A. M. (2008). Influences on young people’s stigmatising attitudes towards peers with mental disorders: national survey of young Australians and their parents. The British Journal of Psychiatry, 192, 144e149. Kahn, J. H., & Hessling, R. M. (2001). Measuring the tendency to conceal versus disclose psychological distress. Journal of Social and Clinical Psychology, 20(1), 41e65. Kessler, R. C., Barker, P. R., Colpe, L. J., Epstein, J. F., Gfroerer, J. C., Hiripi, E., et al. (2003). Screening for serious mental illness in the general population. Archives of General Psychiatry, 60(2), 184e189. Kolodziej, M. E., & Johnson, B. T. (1996). Interpersonal contact and acceptance of persons with psychiatric disorders: a research synthesis. Journal of Consulting and Clinical Psychology, 64(6), 1387e1396. Link, B. G. (1987). Understanding labeling effects in the area of mental disorders: an assessment of the effects of expectations of rejection. American Sociological Review, 52(1), 96e112. Link, B. G., & Cullen, F. T. (1986). Contact with the mentally ill and perceptions of how dangerous they are. Journal of Health and Social Behavior, 27, 289e303. Link, B. G., Cullen, F., Struening, E., Shrout, P., & Dohrenwend, B. (1989). A modified labeling theory approach to mental disorders: an empirical assessment. American Sociological Review, 54(3), 400e423. Löwe, B., Kroenke, K., & Gräfe, K. (2005). Detecting and monitoring depression with a two-item questionnaire (PHQ-2). Journal of Psychosomatic Research, 58(2), 163e171. Penn, D. L., Judge, A., Jamieson, P., Garczynski, J., Hennessy, M., & Romer, D. (2005). Stigma. Treating and preventing adolescent mental health disorders: What we know and what we don’t know. New York: Oxford University Press with the Annenberg Foundation Trust at Sunnylands and the Annenberg Public Policy Center at the University of Pennsylvania. Pescosolido, B. A., Martin, J. K., Lang, A., & Olafsdottir, S. (2008). Rethinking theoretical approaches to stigma: a framework integrating normative influences on stigma (FINIS). Social Science & Medicine, 67(3), 431e440. Pescosolido, B. A., Martin, J. K., Long, J. S., Medina, T. R., Phelan, J. C., & Link, B. G. (2010). “A disease like any other?” A decade of change in public reactions to schizophrenia, depression, and alcohol dependence. American Journal of Psychiatry, 167, 1321e1330. Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Personality and Social Psychology, 90(5), 751e783. Phelan, J. C., Link, B. G., Stueve, A., & Pescosolido, B. A. (2000). Public conceptions of mental illness in 1950 and 1996: what is mental illness and is it to be feared? Journal of Health and Social Behavior, 41(2), 188e207. Pinfold, V., Thornicroft, G., Huxley, P., & Farmer, P. (2005). Active ingredients in anti-stigma programmes in mental health. International Review of Psychiatry, 17(2), 123e131. Reinke, R. R., Corrigan, P. W., Leonhard, C., Lundin, R. K., & Kubiak, M. A. (2004). Examining two aspects of contact on the stigma of mental illness. Journal of Social and Clinical Psychology, 23(3), 377e389. Saraceno, B., van Ommeren, M., Batniji, R., Cohen, A., Gureje, O., Mahoney, J., et al. (2007). Barriers to improvement of mental health services in low-income and middle-income countries. Lancet, 370(9593), 1164e1174. Sartorius, N. (2002). Iatrogenic stigma of mental illness. BMJ, 324, 1470. U.S. Department of Health and Human Services. (1999). Mental health: A report of the surgeon general. Rockville, MD: National Institute of Mental Health. Van Laar, C., Levin, S., Sinclair, S., & Sidanius, J. (2005). The effect of university roommate contact on ethnic attitudes and behaviors. Journal of Experimental Social Psychology, 41(4), 329e345. Van Voorhees, B. W., Fogel, J., Houston, T. K., Cooper, L. A., Wang, N., & Ford, D. E. (2005). Beliefs and attitudes associated with the intention to not accept the diagnosis of depression among young adults. Annals of Family Medicine, 3, 38e46.