| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the University of Washington School of Medicine, Department of Psychiatry and Behavioral Sciences, Seattle, WA.
Address correspondence and reprint requests to Eric D. Strachan, PhD, University of Washington School of Medicine, Department of Psychiatry and Behavioral Sciences, Harborview Medical Center, Box 359911, Seattle, WA 98104. E-mail: erstrach{at}u.washington.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Methods: A sample of psychiatric outpatients (N = 373) from a large, urban HIV clinic was assessed for level of sexual orientation and HIV status disclosure as well as absolute CD4 cell counts over time. Mixed-effects random regression analysis was used to build a predictor model that included biobehavioral covariates.
Results: Consistent disclosure of both sexual orientation and HIV status independently predicted increased CD4 cell counts over time controlling for important biobehavioral covariates. The only other significant effects in the model were baseline CD4 cell count and number of days between assessment of disclosure and assessment of CD4 cell count.
Conclusions: Relieving potential psychological distress by disclosing sexual orientation and HIV status has a positive impact on CD4 cell counts over time even among outpatients stressed by psychiatric illness and economic disadvantage. Additional research is needed to understand whether and under what conditions disclosure should be part of HIV disease management.
Key Words: psychological inhibition • CD4 cell counts • mixed-effects random regression • HIV • sexual orientation
Abbreviations: AIDS = acquired immune deficiency syndrome; BASIS-32 = Behavior and Symptom Identification Scale; HAART = highly active antiretroviral therapy; HIV = human immunodeficiency virus; MCS = Mental Component Summary; PCS = Physical Component Summary; RSO = Relationship to Self and Others; SO = sexual orientation.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
The construct of psychological inhibition has been especially useful in demonstrating the impact of psychosocial variables on HIV disease progression (e.g., (1–2,7)). Defined as the "failure to publicly express any subjectively significant private experience, including, but not limited to, emotional, social, and behavioral impulses" ((13), p. 243), psychological inhibition appears to be a discrete stressor that affects immune function and other markers of mental and physical health (e.g., (14,15)). For example, Pennebaker and colleagues (16) found that writing about significant thoughts and feelings related to a past trauma for 4 consecutive days led to fewer visits to the doctor over the next 6 months compared with writing about a trivial topic. Although the mechanism is not fully understood, Cole and colleagues (2) note that psychological inhibition (e.g., inhibiting the expression of certain thoughts and emotions) can increase activity in the sympathetic division of the autonomic nervous system. Such increased autonomic activity, in turn, appears to be associated with suppressing some aspects of immune function. It is important to note, however, that the research done to support the former point (i.e., about the relationship between psychological inhibition and sympathetic autonomic nervous system [ANS] activity) and that done to support the latter point (i.e., immunosuppression) were not the same. In other words, there is an untested inference in making the full psychological inhibition–ANS activity–immunosuppression claim.
In the realm of HIV, psychological inhibition has been largely operationalized as concealment of sexual orientation (SO). In other words, gay men who do not disclose their SO to others are considered to be psychologically inhibited. Such inhibition, in turn, appears to be associated not only with the HIV-related outcomes outlined previously (2,7), but also with significantly higher incidence of cancer and infectious disease (e.g., pneumonia, bronchitis) among HIV-negative gay men (1). SO, however, is not the only potential source of concealment among persons with HIV/AIDS. HIV infection itself is something that different people choose to conceal or disclose at different points in the course of disease progression for different reasons (17,18).
Like sexual orientation, HIV status is a potentially complex stressor. As mentioned previously, the general theory of psychological inhibition as a mediator of physical health suggests that concealing personally important thoughts, behaviors, and emotions is a stressor that may produce negative immune function effects. However, disclosing HIV status can be an acute and recurring stressor because of stigma, prejudice, and loss of important interpersonal relationships (17,18). Even so, psychological inhibition researchers would hypothesize that concealment is a chronic stressor (as opposed to acute but recurring) and that disclosure should relieve some, if not all, of the effects of that stressor. Holt et al. (17), for example, showed that concealment was the typical strategy for gay men immediately post-HIV diagnosis as those men "came to terms" with their diagnosis. After that initial period, however, disclosure tended to increase in proportion to time since initial diagnosis and symptomatic expression of HIV. Over time, then, disclosure became a form of coping that was associated with increased social support, shared responsibility for sexual activities, and greater self-acceptance.
Given the hypothesized connection between disclosure of SO and improved immune function, along with a substantial literature on the reasons for and effects of HIV status disclosure, we explored in this study whether concealment of HIV status has an effect on immune function similar to that of SO. In addition to adding disclosure of HIV status into the research equation, we also had the opportunity to test whether the previously reported results would be obtained among a sample of persons already stressed by poverty and psychiatric illness. Given that much of the research in this area has been done with relatively affluent, self-selected research volunteers, we thought it was important to take advantage of a real-world sample representing a diverse group of patients with HIV.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
All participants had been referred by their primary care physicians at Madison for psychiatric evaluation and treatment and had completed a visit with a psychiatric provider between January 1, 2000, and December 31, 2004 (see Table 1 for psychiatric diagnosis information). The final sample consisted of male and female patients who had clearly indicated their level of HIV and sexual orientation disclosure at least once (N = 457) and who had two or more valid hematopathology labs, including absolute CD4 cell counts done during the study period (N = 373). CD4 cell counts were included based on the following criteria:
|
These criteria generated a final sample of 373 persons.
Measures
Concealment of Sexual Orientation and HIV Status
Concealment of SO and HIV status was measured using a 5-point Likert-type scale similar to previous research (2,7). In the prior research, the scale anchors referred to various levels of being "in the closet," which is not a phrase commonly used in the HIV context. Therefore, in this research, the scale anchors were changed to 1 = open all the time, 2 = open most of the time, 3 = open some of the time, 4 = not open most of the time, and 5 = open none of the time. Both scales were included on the bottom of the SF-12 questionnaire described subsequently. Because many participants had completed the questionnaire on more than one occasion, the first response during the study period was used in the analyses and a median split was used to create binary concealment conditions as equal in number as possible. For SO, this meant that only those who endorsed being "open all the time" were considered "open" (n = 153) and all others were considered "concealed" (n = 220). For HIV status, participants who were "open all the time" or "open most of the time" became the "open" group (n = 177) and all others became the "concealed" group (n = 196) (see Table 2 for a complete breakdown of responses to the concealment questions).
|
We included people who self-identified as heterosexual in the models despite the common-sense notion that people are not typically motivated to conceal heterosexuality. The problem with automatically excluding those who identify as heterosexual is that (particularly in a racially and ethnically mixed sample) many people who have same-sex relationships do not identify as being gay or bisexual (e.g., (12)). Thus, it may be just as stressful to identify as straight while having same-sex relationships as to identify with, but conceal, homosexuality. To account for possible confounding effects related to sexual identity (gay/bisexual versus straight), however, we included both sexual identity and the interaction between sexual identify and SO disclosure as covariates in the models discussed below (see Table 1 for descriptive data; we refer to that variable as sexual identity to avoid confusing it with SO disclosure). We also included the interaction of sexual identity and HIV status disclosure in the final (p = .89) and full (p = .80) models, but it was not a significant term in either, did not affect the other results, and is not discussed further. Finally, it is worth noting that the 
2 analysis on the disclosure variables (subsequently, Table 1) suggests that the proportion of concealers who identified as gay or bisexual as compared with heterosexual was significantly higher than would have been expected by chance. As would be expected, this was only true for SO and not HIV status.
Absolute CD4 Cell Counts
CD4 cell counts were collected from the clinical records database at Harborview Medical Center. They were measured using flow cytometry using a whole blood assay and are reported as number of CD4 cells/mm3. The first CD4 cell count was used as a baseline covariate in the analyses. Each additional CD4 cell count (from one to four) was used as the longitudinal dependent variable. Because these data were not generated at standard intervals, we calculated the number of days between each hematopathology lab and the original SF-12 date for each participant (CD4 latency; see bottom of Table 1). We then used those data as a time-varying covariate in the analyses. We chose four data points to maximize the number of available data points while allowing for a sufficiently long period of time to analyze the impact of disclosure (4 labs x approximately 90 days between labs = 1 year).
Demographic Data
Most of the demographic data were collected from a form completed by psychiatric providers after each patient visit and, when necessary, confirmed by reference to the patient’s psychiatric intake data and medical records (see Table 1). Monthly income data were taken from Madison Clinic’s Social Work database. To create covariates with reasonable representation at each level, ethnicity, employment, and partner/marital status were transformed into binary variables. For ethnicity, the groups were white (n = 243) versus minority (including black [n = 89], Latino [n = 20], Asian/Pacific Islander [n = 12], Native American [n = 4], and other [n = 5]); for partner/marital status, the groups were single (including single [n = 249], divorced [n = 37], and widowed [n = 5]) versus married/partnered (n = 82); and for employment, the groups were unemployed (n = 281) versus some work or school (including student [n = 16], part-time employment [n = 27], and full-time employment [n = 49]).
Psychiatric Diagnoses and Intravenous Drug Use
Psychiatric diagnoses were generated from each patient’s electronic medical record. The records contain problem lists that are coded using International Classification of Diseases, 9th Revision codes. We created diagnosis-spectrum categories by including related codes in the same categories. For example, the bipolar spectrum included 296.0 to 296.1 and 296.4 to 296.8, whereas the depression spectrum included 296.2 to 296.3. The different subcodes are typically specifiers of such things as recurrence or topography (e.g., whether the last episode was mania or depression in bipolar I disorder). In addition, we included a variable that indicated whether the person had any history of intravenous drug use as noted in the psychiatric intake evaluation. Each patient may have had more than one diagnosis and each spectrum was entered separately (see Table 1).
Physical and Mental Health
We used the Physical Component Summary (PCS) and Mental Component Summary (MCS) scales from the SF-12 (19) to control for general physical and mental health. The SF-12 is a brief self-report instrument with established validity as a health outcomes measure. The PCS and MCS scores were collected at the same time as the concealment measures. Higher SF-12 scores indicate better functioning with U.S. general population means of 50.04 (MCS) and 50.12 (PCS).
Social Support
Because of the potential importance of social support as a moderator of concealment effects, we included the Relation to Self and Others (RSO) subscale from the Behavior and Symptom Identification Scale (BASIS-32 (20)) as a covariate. The BASIS-32 is a psychometrically sound self-report instrument measuring difficulty on 32 symptom and behavior items. Higher scores indicate greater dysfunction. The RSO subscale has a maximum score of 4.
Highly Active Antiretroviral Treatment
To control for potential effects of HIV treatment on CD4 cell counts, we reviewed each patient’s pharmacy records to determine if he or she had ever been treated for HIV infection during the study period. All patients at Madison during the course of this study would have received (if it were clinically indicated) one of several possible antiretroviral combination treatments commonly referred to as highly active antiretroviral treatment (HAART), but also accurately called combined antiretroviral therapy (CART). Because we did not have reliable data describing treatment adherence for each participant, HIV treatment was included in the analyses as a binary covariate. In other words, for the purposes of this research, patients either had, at some point during the study period, been prescribed a HAART regimen at least once or they had not.
Statistical Analyses
We completed all analyses using SPSS for Windows 12.0 (SPSS, Inc., Chicago, IL) except for the mixed-effects random regression modeling, which was performed using MIXREG for Windows 1.2 (21). Statistical significance was set at p 
.05. We used Pearson 2 and independent sample t tests to compare baseline characteristics between different levels of concealment for both SO and HIV status. For the main analyses, we used mixed-effects random regression modeling to determine whether concealment of SO or HIV status was associated with changes in CD4 cell counts over time. We selected mixed-effects random regression modeling for three reasons. The first reason was that they allow the use of correlated longitudinal data with missing observations (assuming the missing data are missing at random); the second was that they allow for the use of covariates; and the third was that there was no requirement that the dependent variable be collected at the same time intervals between or within subjects (i.e., time could be entered as a random effect in the models).
Our main hypothesis was that disclosure of SO and HIV status would independently predict higher CD4 cell counts over time when compared with concealment of SO and HIV status (i.e., a disclosure x time interaction for both SO and HIV status) controlling for baseline CD4 levels and CD4 cell count latencies (i.e., the time between the hematopathology lab and the original assessment of concealment/disclosure). Although it was not an a priori prediction of ours,1 if SO and HIV concealment combined to moderate CD4 cell counts over time, we would have expected a significant three-way interaction (SO disclosure x HIV status disclosure x time). Thus, an initial model was fit containing the three-way interaction, the three two-way interactions (including those representing our main hypotheses), the main effects of time, SO, and HIV status disclosure, baseline CD4 cell counts, CD4 cell count latencies, sexual identity, and the sexual identity x SO disclosure interaction. After this initial model, the nonsignificant three-way interaction term was dropped and the model was refit. Psychiatric diagnosis,2 PCS and MCS scores, ethnicity, marital status, employment, income, years of education, RSO, HAART history, and intravenous drug use history were added as covariates in the full model. The final model was developed by selecting significant predictors—and lower-order terms needed to support significant interaction terms—and dropping nonsignificant predictors with the exception that sexual identity and related terms were maintained despite being nonsignificant to control for that potential confound in the final model. Also, any covariate with a p value less than .15 was maintained for the final analysis based on the recommendation of Tabachnik and Fidel (22).3 In every case, time and intercept were entered into the model as random effects and all other variables were entered as fixed effects.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Outcome Effects
The initial mixed-effects random regression analysis of CD4 cell counts over time (model = intercept, time, SO disclosure, HIV status disclosure, CD4 baseline, CD4 latency, sexual identity, sexual identity x SO disclosure, the three two-way interactions, and the three-way interaction) revealed only time, CD4 baseline, and CD4 latency as significant predictors. However, after the nonsignificant three-way interaction was dropped from the model, the predicted disclosure x time interactions for both SO and HIV status attained significance (see Table 3). As the signs of the interaction estimates suggest, these results were the result of the association between disclosure ("open" was coded as "0" and "concealed" was coded as "1" in the model) and higher CD4 cell counts over time. For the CD4 baseline data, the relationship was that higher baseline CD4 cell counts predicted higher CD4 cell counts over time. For the CD4 latency data, the relationship was that longer latencies predicted lower CD4 cell counts. In other words, as time went on, CD4 cell counts went down. Note that neither the sexual identity covariate nor the sexual identity x SO disclosure interaction term was significant in the model.
|
The statistical significance of these effects was not meaningfully altered in the presence of all potential covariates nor when the nonsignificant covariates were dropped individually or in total. None of the covariates was significant except CD4 baseline and CD4 latency. Only MCS (estimate = –0.00116, standard error = 0.00063, p = .06) had a p value at or below .15 (the recommended threshold mentioned previously). Including all the relevant candidate variables left us with a final model that included intercept, time, CD4 baseline, CD4 latency, SO disclosure, HIV status disclosure, SO disclosure x time, HIV status disclosure x time, SO disclosure x HIV disclosure, sexual identity, sexual identity x SO disclosure, and MCS (see Table 4). We included the sexual identity terms in the final model despite failing to meet the p < .15 criterion to continue to control for possible confounding of those two variables. Note that in the final model, the time main effect was significant but it was qualified by the significant interactions. Figures 1 and 2 show the SPSS-generated means, adjusted for the significant covariates, for SO and HIV status disclosure, respectively.
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
To the best of our knowledge, there have been only four prior studies that investigated the association between psychological inhibition and CD4 cell counts (2,7,10,11). Cole et al. and Ullrich et al. both used SO concealment among gay men as the operational definition of psychological inhibition, whereas Eisenberger et al. and Petrie et al. used a more general concept of emotional expression. Our study is conceptually consistent with the findings from Cole and colleagues in that their results vis-à-vis CD4 cell counts were not altered by inclusion of demographic and biobehavioral (e.g., antiretroviral treatment) variables into the model. However, our sample was different from the Cole et al. sample in a number of ways that might suggest additional generalizability for the finding.
First, our sample was of low-income psychiatric patients at a publicly funded clinic who, it could be argued, might not show the effect of psychological inhibition per se given other stressors that have been linked to disease progression. In other words, the combination of disease progression-related variables (e.g., psychiatric problems, gender, poverty, and ethnicity) might have led (but did not) to a restriction-of-range problem in our data. Second, although our sample was one of convenience, it captured individual patients from a large, urban HIV clinic and not just those who self-selected themselves into a long-term research project. Third, more effective antiretroviral treatments have been developed since the Cole et al. study (2) was completed and those treatments are standard for our population. Finally, our participants were not necessarily physically healthy at the beginning of the study as was the case in the Cole et al. study (the plurality of patients at the Madison Clinic is category C3 according to the taxonomy developed by the Centers for Disease Control and Prevention and 68.2% of the sample meet criteria for AIDS [i.e., A3 or above]).
Our results are both similar to and different from the Ullrich et al. (7) findings in that they found the same kind of relationship between concealment and CD4 cell counts, but they did so using a cross-sectional design. Also, their results were moderated by level of satisfaction with social support. Specifically, Ullrich et al. found that openness about SO was only associated with CD4 cell counts among those who were satisfied with their level of social support. Although we used a measure of interpersonal functioning rather than satisfaction with social support, we did not find any evidence that interpersonal dysfunction moderated the results.
Perhaps more important than those results, however, was our finding that disclosure of HIV status was associated with better CD4 cell counts over time. This lends additional support to the notion that inhibitory processes per se are the key factor in the inhibition/immune suppression link and that the link is not related to a specific kind of inhibition. However, as has been noted in all of the previous research, these data do not, in and of themselves, allow for causal inferences (i.e., that inhibition caused CD4 cell counts to drop or disinhibition caused them to rise). It would also be premature to suggest that individuals should disclose SO and HIV status to be healthier, especially given one study showing concealment can be protective for gay men who are rejection-sensitive (13). How, for whom, and under what conditions to disclose, however, are all interesting questions that warrant additional research.
It is possible that the disclosure results (especially the HIV status results) were the consequence of increased health-related behaviors (e.g., seeking medical care) among those who are open about HIV status and SO and not the consequence of disclosure per se. Although we cannot answer that question definitively, our results were not moderated by HAART prescription during the study period. In addition, our sample consisted of persons who had at least two completed hematopathology labs with a modal number of five (baseline plus four). This suggests that all patients were seeking health care. Although it does not bear directly on our results, Cole et al. (2) had a sophisticated measure of antiretroviral treatment adherence in their model (they used a time-varying treatment covariate that took into account treatment nonadherence) and it did not change the relationship between disclosure and CD4 cell count. They also included a number of health-related behaviors (e.g., exercise, smoking, drinking, and drug use) that did not change the contribution of disclosure to the model. In addition, intravenous drug use (IVDU) was not a significant predictor in our models despite previous research showing that IVDU may be associated with relatively poor antiretroviral medication compliance and response (e.g., (23)).
It is also possible that disclosure of SO, or especially HIV status, may have changed over time. Although it seems implausible that either status would have, on average, changed to "more concealed," to the extent that it is possible it would have implications for interpretation of these results. Changes in the direction of "more open" would be consistent with the hypothesis given the same CD4 cell count results, but there is no way to ensure that the model would stay the same. To address this, we ran a post hoc mixed-effects random regression analysis for ordinal outcome variables (MIXOR (24)) with up to four responses to the disclosure questions per patient as the DV and all of the covariates as predictors. Time was not a significant predictor in either model (SO or HIV status) suggesting that responses to the questions did not change significantly over the study period. For SO, the only significant predictors were ethnicity (being an ethnic minority was associated with less disclosure) and RSO (greater interpersonal dysfunction was associated with less disclosure). For HIV status, the same associations were found for ethnicity and RSO, but three other variables also contributed to the model. History of HAART treatment was associated with more openness as was years of education. Being employed was associated with less disclosure.
An additional limitation to our study was the use of a sample of convenience drawn from an archival database of psychiatry patients. However, this sample extends the findings into a different population and, in combination with prior research, bolsters the generalizability of the finding. Additional prospective research in which disclosure is manipulated rather than measured while controlling for all of the important psychosocial and biobehavioral variables would greatly increase our understanding of the link between psychological inhibition and immune function in the HIV context. For example, a randomized, controlled trial of a psychosocial skills training course focused on building and implementing a disclosure plan for friends, family, and sexual partners might yield interesting results in terms of whether the course actually increases disclosure, influences risk behaviors, and generates the immune results demonstrated in this study. Such research could lend greater clarity to the clinical significance and longer-term trends associated with such disclosure.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
2Diagnoses were entered as yes or no for any depression, anxiety, bipolar disorder, alcohol use, or drug use spectrum diagnoses based on International Classification of Diseases, 9th Revision codes during the study period. Each diagnostic spectrum was entered individually.
3Tabachnik and Fidel also suggest that number of predictors in a multivariate model should be a function of sample size such that N 
8k + 50 where k is the number of variables. Given a sample size of 373 (with over 1200 individual observations), we were comfortable using up to 40 variables but did not use more than 30 in any model.
Received for publication December 8, 2005; revision received July 28, 2006.
DOI:10.1097/01.psy.0000249900.34885.46
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  L. R. Temoshok, R. L. Wald, S. Synowski, and A. Garzino-Demo Coping as a Multisystem Construct Associated With Pathways Mediating HIV-Relevant Immune Function and Disease Progression Psychosom Med, June 1, 2008; 70(5): 555 - 561. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
