AIDS Behav (2009) 13:246–257 DOI 10.1007/s10461-007-9341-5
ORIGINAL PAPER
What You Think You Know Can Hurt You: Perceptual Biases About HIV Risk in Intimate Relationships Jennifer J. Harman Æ Megan A. O’Grady Æ Kristina Wilson
Published online: 13 December 2007 Ó Springer Science+Business Media, LLC 2007
Abstract The use of heuristic biases and the false consensus effect can lead individuals to misperceive risk of HIV infection. The current paper presents the results of two studies which sought to examine whether individuals, (a) weigh risk relevant information accurately in their assessments of HIV risk, and (b) are susceptible to the false consensus effect in their assessments of actual intimate partner risk behaviors. The results of the first study support the hypothesis that individuals do not use objective risk information based on probability statistics in their assessments of HIV risk. In addition, the results of the second study find that female partners exhibit the false consensus effect regarding sexual risk behaviors, whereas male partners did not. Discussion centers on the role of perceptual biases in HIV risk behaviors among individuals in intimate relationships. Keywords False consensus HIV/AIDS Risk assessment Intimate relationships Gender differences
Introduction Individuals in intimate relationships see the best in their partners and may even distort information in order to view them in a positive light (Murray et al. 1996). Sadly, focusing on positive and ignoring negative information can lead to inaccurate perceptions that can put individuals at risk for contracting communicable diseases such as HIV J. J. Harman (&) M. A. O’Grady K. Wilson Department of Psychology, Colorado State University, Fort Collins, CO 80523-1876, USA e-mail:
[email protected]
123
infection. Indeed, beliefs regarding the likelihood of contracting a disease are important predictors of preventative behaviors (Rogers 1975). Unfortunately, little research has specifically examined how individuals assess and utilize their partner’s HIV risk relevant information.
Risk Perception While perceptions of risk have long been identified as important factors in theories of health behavior (e.g., Hachbaum 1958), relatively little is known about how risk relevant information is calculated, particularly in the context of perceptions of HIV risk. Individuals have difficulties using the odds and percentages that form probability statistics to make judgments of risk (Weinstein 1998), and even if they are aware of such statistics, they may not be used when making risk decisions. People often use cognitive heuristics based on what information is available (Slovic 2001), which helps to reduce complex mental tasks into simpler ones (e.g., Tversky and Kahnemen 1974). Although heuristics are useful and adaptive cognitive strategies, they can also lead to persistent biases in perception which has implications for the accuracy of risk assessments (Slovic et al. 1979). Behaviors related to HIV infection can be weighted in terms of risk severity. While some research has suggested that there is little to no relationship between risk perception and using protection during sexual behavior (e.g., Gerrard et al. 1996; O’Sullivan et al. 2006), other research on HIV preventative behaviors (e.g., consistent condom use) suggests that people will only carry out appropriate preventative behaviors if they accurately perceive the behaviors they engage in to be risky (e.g., Stall and Catania 1994). Although there is some contradictory evidence in
AIDS Behav (2009) 13:246–257
this area, if people are indeed using risk perceptions in their decision to engage in safe sexual behavior, it is still important to understand what information they are using to calculate their risk. Therefore, it is critical to determine if people make judgments about the likelihood that an individual is infected with HIV based on objective risk assessments or whether they use a subjective calculation system.
Perception of Partner HIV-specific Risk Behaviors Many individuals in heterosexual intimate relationships feel invulnerable to HIV infection (Exner et al. 1997; Harman et al. 2007), and do not engage in effective HIV prevention practices with their primary partners (Bowen et al. 2001). If risk status based on both individual and partner factors is not accurately perceived, it is unlikely that an individual will believe that they are at risk for HIV infection or engage in preventative behaviors. With 32% of new HIV infections within the United States being transmitted through heterosexual contact (CDC 2005) and heterosexual sexual contact being the largest infection category for women and second largest for men (CDC 2007a), examining how risk information is perceived and utilized in heterosexual intimate relationships is important to understand. The false consensus effect (FCE) can provide a framework for understanding intimate partners’ perceptions of each other. The FCE (Ross et al. 1997) refers to the tendency to overestimate the amount other people share one’s own attitudes and beliefs. The robustness of this effect has been demonstrated across a wide variety of attitudinal and behavioral perceptions (for a thorough review, see Mullen et al. 1985). While the FCE has traditionally examined how the generalized other is misperceived, recent research has found this effect with specific others, such as peers (Prinstein and Wang 2005). Indeed, FCE effects tend to be stronger when the ‘‘other’’ is perceived to be similar (Sherman et al. 1985) or less socially distant to the individual (Jones 2004). Given previous research, it is likely that the FCE will also be found within intimate relationships. Increased acquaintance has not been found to lead to greater accuracy in person perception (Kenny 2004). Assumed similarity (Higgins et al. 1982) is thought to be one explanation for this finding, whereby perceivers tend to see others as they see themselves, and these effects are particularly strong in close relationships (Marks and Miller 1987). As intimate partners become part of one’s self-concept (Aron et al. 1991), partner perception may be biased in a manner that serves to keep the relationship intact. Moreover, perceptions of close others are likely to match some idealized
247
prototype, whereby individuals may view their partners as very similar to themselves or as a good person (Kenny and Acitelli 2001). These findings suggest that perceptions of partners are often inaccurate and give reason to believe that the FCE may also be found in intimate relationships.
The FCE and Health Risk Behaviors Nonnormative health behaviors such as drug use and sexual risk behaviors are particularly prone to the FCE (Sherman et al. 1983; Suls et al. 1988; Whitley 1998). For example, sexually experienced young adults tend to overestimate the amount of sexual activity in which their peers engage as compared to their sexually inexperienced peers (e.g., Whitley 1998). Although there have been numerous explanations for the FCE effect (e.g., salience of the behavior/attitude), selective-exposure, has been strongly supported within research on sexual risk behaviors (Whitley 1998). Selective-exposure presumes that because people tend to expose themselves to people with similar attitudes and patterns of behavior, they are likely to believe that others think and act like they do. Due to the FCE, partners in intimate relationships may have inaccurate perceptions of each other’s HIV risk behavior and therefore may inaccurately consider their partners to be sexually ‘‘safe’’ (Hammer et al. 1996). Close relationships function as a buffer against events that contribute to both mental and physical health problems (Coyne and Downey 1991; Prager 1995), but feelings of trust and security and the desire to maintain such feelings serve as an overlooked source of HIV infection (Misovich et al. 1997). In addition, individuals tend to believe that it is highly unlikely that their partner engages in behaviors that would put them at risk for HIV infection (O’Donnell et al. 1994).
The Current Study The purpose of the two studies presented in this article is to examine how information about sexual and drug risk factors is used to form perceptions of HIV risk in heterosexual relationships. The first study was conducted in order to determine how individuals generally weigh HIV risk information (e.g., sexual and drug risk behaviors) about romantic partners in their judgments of likelihood of HIV infection. The second study was conducted to examine whether perceptions of an intimate partner’s HIV risk information is biased or accurate. If perceptions about risk are biased, this would indicate that individual’s see their partners risk as being similar to their own, illustrating the FCE.
123
248
Study 1 The purpose of the first experiment is to examine how HIV risk related information is utilized to make assessments and form beliefs about the likelihood of HIV risk infection. It was hypothesized that the judgments regarding the likelihood that a hypothetical partner is infected with HIV will not be based on the use of probability statistics.
Method Participants Participants were 103 Colorado State University students. In order to preserve anonymity of participants, no demographic or identifying information was collected, so the actual numbers of men and women were not recorded. Based on class rosters, the sample was approximately 54.3% female and the majority were upper division students (juniors and seniors; 95.7%). Participants received extra course credit for their participation.
Procedure Questionnaires were administered in two psychology classes. Participants read a brief description of the study, and then completed a 10–15-min survey. After finishing, participants placed their survey into a box at the front of the classroom to protect their anonymity.
Measures The questionnaire was designed to examine how information regarding sexual and drug use risk behaviors is utilized to form beliefs about a hypothetical partner’s HIV status. The questionnaire included two sections, which are discussed in detail below.
Primary Partner and HIV Risk Perception Participants were first asked to imagine a scenario in which they discuss past drug use and sexual behaviors with their primary partner. In order to allow participants to respond to questions as though their actual relationship partner had disclosed this information, the scenario was intentionally vague in nature. All participants were presented with the following scenario: Imagine the following scenario. You are having a discussion with your primary partner about their past
123
AIDS Behav (2009) 13:246–257
sexual and drug use behaviors. In this conversation, information is disclosed regarding your partner’s other sexual partners, their use of condoms and their use of non-injection and injection drugs. If you currently are not in a relationship, please answer the questions as if you are in a relationship. Following this scenario, participants were presented with 24 specific risk behaviors that their partner had engaged in. For example, they were asked to think about a situation in which their primary partner discloses information regarding their frequency of condom use in an extra-relationship situation (e.g., When having vaginal sex with people other than you, your partner used condoms about half the time). They were given several additional situations in the same scenario, but varying both condom use frequencies (e.g., Almost every time, Almost never) and type of sexual activity (e.g., vaginal versus anal sex) during sexual encounters with other partners. In addition, participants were asked to think about a situation in which their partner discloses information regarding prior drug use (e.g., Your partner has injected cocaine in their lifetime). Participants then rated additional situations that varied both in terms of type of drug used (e.g., amphetamines, crack, heroin, speedball) and route of administration (e.g., injection versus noninjection). Participants rated the likelihood that given each individual piece of risk information ‘‘disclosed’’ by their partner in the scenario, their primary partner would be infected with HIV. Therefore they only considered one situation/behavior at a time when making judgments of likelihood of HIV infection. Responses were rated on a 7-point scale, ranging from 1 (very unlikely) to 7 (very likely). Only general risk behaviors (e.g., anal sex) were assessed in this study and therefore more specific sexual behaviors (e.g., receptive versus penetrative anal sex) were not included in the survey.
HIV Knowledge Knowledge of HIV transmission and preventative behaviors was then assessed by administering a selection of items from the Health and Relationships Survey (Misovich et al. 1998). Participants rated how much they agreed or disagreed with statements on a 5-point scale, ranging from 1 (Strongly disagree) to 5 (Strongly agree). The selected items were specific to knowledge of sexual and IV-drug use transmission of HIV. Internal reliability for this scale was lower (a = .63) than alphas previously reported (a = .75; Fisher et al. 1996), however the mean number of questions correctly answered was similar to other studies, with 77% of questions answered correctly.
AIDS Behav (2009) 13:246–257
249
Risk Weightings and Ratios
as their use tends to occur with greater frequency (NIDA 2005b), so the use of both injected and non-injected heroin was weighted as less risky than the use of stimulant drugs. The objective risk associated with drug use was also weighted in terms of the high-risk sexual behavior associated with different classes of drugs. Sexual behaviors that put one at risk for HIV infection are higher among individuals who use crack cocaine (e.g., Hoffman et al. 2000) or methamphetamines (Somlai et al. 2003) than other drugs such as heroin (e.g., Cohen et al. 1994), resulting in higher risk severity weightings. Club drugs are very similar to methamphetamines (NIDA 2005c), and were therefore weighted the same. Finally, risk of HIV infection is higher for injection drug users (e.g., Kim et al. 1993), so injection drug use was weighted heavier for risk than non-injection drug use. Furthermore, the risk of HIV infection is higher for individuals who inject cocaine (Somali et al. 2003) or speedball (heroin mixed with cocaine; Johnson et al. 2002) than heroin, so these injection drugs were weighted as riskier.
Objective Risk Weightings Behaviors related to HIV infection were weighted in terms of risk severity based on scientific data (e.g., National Institute of Allergy and Infectious Diseases 2000; see Table 1). Risk associated with general sexual behaviors was weighted based on frequency of condom use and type of sex act (e.g., vaginal or anal sex). For example, due to the possibility of leakage, breakage and improper use, condoms are not 100% effective in preventing HIV transmission. Also, anal sex carries a greater risk of exposure to HIV than vaginal sex (Levin 2002). For this reason, anal sex was weighted as twice as risky as vaginal sex, and infrequent condom use was also weighted as more risky during anal sex (for a more thorough description of the risk calculations, see Harman 2005). Because of the exploratory nature of the current study, more specific behaviors related to HIV sexual risk transmission were not included (e.g., penetrative versus receptive anal intercourse) in these risk calculations, but would be useful to assess in future research. Weight calculations for HIV risk severity were also made for the use of non-injection drugs. Frequency of use needed to maintain a high is one factor that can be used to calculate the objective risk associated with specific drugs. Crack use was weighted as more risky than cocaine, because the high from crack use only lasts 5–10 min in comparison to a 15–30 min high from non-injected cocaine (National Institute on Drug Abuse (NIDA) 2005a). Likewise, other stimulant drugs were weighted more risky than depressants, Table 1 Ratios of risk severity for HIV risk factors Risk factor
Subjective ratio
Objective ratio
Risk Ratios Risk ratios were calculated for both the objective risk and for participant’s subjective risk behavior assessments. For example, objective risk ratios for condom use were calculated by dividing the risk weighting of vaginal sex without a condom by the risk associated with almost always using condoms. Risk ratios were then calculated for participant’s subjective assessments by dividing the mean ratings of likelihood of HIV infection for each risk behavior by the mean ratings for absence of risk behavior. In addition to presenting the objective risk ratios, Table 1 also presents the subjective risk ratios.
Condom use—Vaginal sexa Sometimes
1.39
5.00
Almost Never
1.70
10.00
Sometimes
1.28
7.50
Almost never
1.52
10.00
Heroin
4.57
3.00
Cocaine
4.51
4.00
Speedball
4.47
4.00
Heroin
2.24
0.50
Cocaine
2.04
1.00
Amphetamines
2.14
1.00
Club drugs (e.g., ecstasy)
2.38
1.00
Condom use—Anal sexa
Results HIV Knowledge
Drug use—Injectedb
Drug use—Non-injectedb
a
Ratios in relation to always using condoms
b
Ratios in relation to not using drugs
Mean HIV knowledge scores were 10.83 (SD = 1.97) items correct out of 14 total items, reflecting a high HIV knowledge score. HIV knowledge was entered as a between-subject’s factor in all of the following analyses, but did not significantly interact with any of the perception variables, and will therefore not be discussed further. Perceptions of Partner HIV Drug Risk Ratios A within-participants ANOVA was conducted to determine whether there were significant differences between subjective ratings of risk with different types of drug usage.
123
250
Results indicated that ratings of the likelihood that a person is infected with HIV differed significantly based on the usage of specific non-injection drugs, F(3, 101) = 2.80, P \ .05. In comparing the subjective risk ratios with the objective risk ratios, perceptions of the use of non-injection cocaine, amphetamines and club drugs were weighted about twice as high as the objective risk ratios. The subjective ratio for heroin use also did not reflect the objective risk; heroin was perceived to be riskier than all of other drugs except club drugs. Next, subjective risks associated with injection drug use were tested. These results were not significant, F(2, 101) = 1.35, n.s., indicating that participants did not rate the likelihood that a person is infected with HIV differently based on the class of drug injected. Furthermore, participant’s subjective assessments of these behaviors were not similar to the objective risks. For example, the objective risk severity weight assigned to use of injection heroin was 3.0, as compared to a severity weight of 4.0 for use of cocaine or speedball. The subjective risk ratios were comparatively larger. In comparison to no drug use, use of injection heroin was weighted as 4.57 times as risky and injection cocaine use was weighted as 4.51 times as risky. Therefore, participants perceived injection drug use to be riskier than the objective risk would imply.
Perceptions of Partner HIV Sexual Risk Ratios Frequency of condom use during vaginal and anal sex with someone other than a primary partner was then analyzed. Significant differences were found based on frequency of condom use during vaginal sex with non-primary partners, F(2, 101) = 103.20, P \ .05. Although participants perceived some level of risk, their estimations were not based on objective risk statistics. For example, rare condom use was weighted as 10 times the objective risk severity as almost always using condoms. Participant’s subjective perceptions did not reflect this ratio; rare condom use was perceived to be only 1.70 times more risky than almost always using condoms. As with condom use frequency for vaginal sex, there were significant differences on risk perceptions of condom use frequency for anal sex with non-primary partners, F(2, 101) = 70.57, P \ .05. For example, never using a condom during anal sex was calculated to be 1.33 times as risky as using condoms about half of the time, and 10 times as risky as using condoms all of the time based on the objective risk ratios. Although participant’s perceived risk ratios were in the same direction as actual risk ratios, they were considerably lower overall (see Table 1). Participants rated rare condom usage during anal sex as only 1.52 times as risky as almost always using condoms.
123
AIDS Behav (2009) 13:246–257
Study 1 Discussion Results of the first study confirmed the hypothesis that when assessing the HIV risk status of a sexual partner, individuals in this sample did not use objective probability statistics. These findings are consistent with previous research suggesting that people have difficulty using realistic probability statistics to make judgment decisions (Weinstein 1998; Slovic 2001). Even after controlling for knowledge about HIV transmission, participants overestimated the risk of heroin use in comparison to other injected and non-injected drugs. It is possible that although participants may know how HIV is transmitted, they may not be as knowledgeable about how specific behaviors or drugs may put them at greater or lesser risk for transmission. Unfortunately, familiarity with injection and non-injection drugs was not assessed in the current study, so the subjective ratings reflect the general knowledge that participants in this study have about drug risk. Because injection drug use has historically been a primary transmission category (CDC 2007a), and heroin is one of the most commonly injected drugs (CDC 2007b), it appears that participants may be basing their perceptions on these facts. Unfortunately, risk of HIV transmission via drug use has changed considerably, and is influenced by a great numbers of factors, such as frequency of injection and riskier sexual behaviors while under the influence of the drug, which is higher for stimulant drugs like cocaine. Similar results were found for sexual behaviors, whereby participants were not accurate in their assessments of the risk severity associated with specific sexual risk behaviors. The results suggest that participants may have either underestimated the effectiveness of consistent condom use in preventing HIV, or underestimated the risks associated with unprotected sex. It is possible that a heuristic bias may be interfering in this risk estimation process (Kahneman et al. 1982). It is interesting to note that although participant’s subjective risk ratios were lower than actual risk ratios, they were in the same direction. It is possible that as subjective and actual risk ratios are in the same direction, this alone may be enough to protect individuals from HIV infection. However, given evidence suggesting that participant’s subjective risk ratios were considerably lower overall, it is possible that this misperception of risk could lead to decisions to engage in unprotected sexual activity, thereby putting the individual at risk for HIV infection. Future research should examine this possibility further. One limitation of this first study is that participants were asked to think about hypothetical intimate partners. Individuals may be less likely to make accurate judgments about HIV risk factors with their actual sexual partners. Unfortunately, even in such relationships, HIV risk
AIDS Behav (2009) 13:246–257
relevant information is not often disclosed prior to a sexual encounter (Bowen and Michal-Johnson 1989; Stebleton and Rothberger 1993), so individuals may not have relevant information to assess a partner’s HIV-risk status. In summary, the results of the first study demonstrate that individuals often inaccurately perceive the HIV risk involved with various sexual and drug risk behaviors. It is still not clear if individuals in intimate relationships are biased in their perceptions of each other. False consensus about risk behavior may serve as a heuristic or mental shortcut in the assessment of the HIV risk posed by one’s sexual partner. The second study was designed to test the FCE among partners in intimate relationships using the risk ratios obtained in the first study.
Study 2 The purpose of the second study was to examine whether perceptions of an intimate partner’s HIV risk behaviors are accurate or biased. The hypothesis of the second study is that the FCE will be evident in perceptions of partner’s HIV risk behaviors; they will estimate their partner’s HIV risk behaviors to be more similar to their own.
Method Participants Survey data were collected on the Internet from both male and female members of heterosexual couples as part of another study (Harman 2005). The survey was posted onto PsychData (http://www.psychdata.com), an on-line research service, and participants completed the survey within 24 h of their partner. After indicating consent, participants entered their first and last initials, the initials of their intimate partner, and each of their birthdays in order to match their responses. At the end of the survey, participants were given the opportunity to enter a lottery for a $100 gift certificate from http://www.Amazon.com as an incentive. Multiple strategies were used to recruit couples for this study. First, links to the survey were posted on two websites that advertise psychological studies: the Social Psychology Network (http://www.socialpsychology.org/ expts.htm) and Psychology Research on the Net (http:// www.psych.hanover.edu/research/exponnet.html). Interest group moderators listed under ‘‘Relationships and Romance’’ and ‘‘Family and Home’’ categories on Yahoo (http://www.groups.yahoo.com) and Google (http://www. groups-beta.google.com) were also solicited to post information about the study for their group members. The study
251
was also posted onto networking groups from the Ryze Business Networking website (e.g., Retired Moms; http://www.ryze.com) and the Friendster network (http:// www.friendster.com), which are free, on-line networking sites. In addition, information about the study was posted onto the volunteer section of an on-line classified ads website titled Craigslist (http://www.craigslist.org) at 13 randomly selected cities throughout the United States. Brochures about the study were also distributed in locations such as day care centers, retail stores, New York City airports (John F. Kennedy and LaGuardia), local bars and restaurants in New Haven, CT, and at food vendors at the University of Connecticut. Finally, information about the study was emailed to the listservs of several professional societies (e.g., Society for Personality and Social Psychology), and instructors at the University of Connecticut and Quinnipiac University. Seventy-five heterosexual couples completed the survey in it’s entirety within 24-h of each other. The mean age of participants was 29.33 (SD = 9.29), and the majority were White (89.3%). The average length of time together with their partner was 4.46 years (SD = 3.15). Most couples lived together (59.3%), and the remaining participants lived with roommates (11.3%), family members (14.7%), or alone (12.7%). Participants were either married to their partner (38.7%) or dating their partner exclusively (42.7%). The remaining participants were engaged to their partner (13.3%), dating their partner casually, or in some other form of relationship (5.3%). Participants provided information about their HIV/AIDS status, HIV/AIDS testing history, and sexual communication with their partner. Of the 150 total participants, 43% reported that they had never been tested for HIV/AIDS. Of those that had been tested, most had only been tested one or two times (39%). None of the participants reported that they were HIV-positive, 77% reported that they were HIVnegative and 23% reported that they did not know their HIV status. Most participants reported that they felt comfortable talking to their partner about HIV and how they could protect themselves against infection (M = 4.15 SD = 1.08 on a 5-point scale). Measures Each couple answered questions about their sexual and drug use behavior as well as their perceptions of their partners sexual and drug use behavior. Sex Risk Participants responded to questions asking about the number of partners they have had vaginal and anal sex with
123
252
other than their primary partner within the last 3 months. A small amount (8.7%) of the sample indicated that they had vaginal sex, and 2.7% had anal sex with someone other than their primary partner. Of those having extra-relationship sex, most individuals indicated that they had sex with only one partner (79.9% for vaginal and 75% for anal sex) and none indicated that they had sex with more than three partners. Participants also indicated the percentage of time that a condom was used in those extra-relationship sexual encounters, using a 5-point Likert-type scale (1 = Every time, 5 = Never). For vaginal sex, most participants indicated that they used condoms every time with their other partners (61.5%). Half of the participants indicated that they used condoms every time when they had anal sex with another partner and the other half never used them. These sexual behaviors were assigned a risk ratio based upon results from the first study, and were multiplied by the number of partners that the individual reported. For example, if a participant indicated having vaginal sex with one other person besides their partner and used condoms 75% of the time (Sometimes), a weight of 1.39 was assigned (See Table 1). These weighted risk ratios were then summed for an overall sex risk score. Participants also answered the same questions about their perceptions of their partner’s extra-relationship vaginal and anal sexual behaviors in the last 3 months. Items were measured the same as the self-rating items, however participants could indicate I don’t know for condom use items. In order to be conservative in our statistical analyses, I don’t know was assigned a Sometimes value, as it was assumed that not knowing might lead people to overestimate risk. As with the self ratings, the overall sex risk score was derived by summing the vaginal and anal sex risks, which were weighted by risk ratios and number of partners. Drug Risk Participants indicated whether they used or tried a series of injected (heroin, cocaine, speedball and/or other) and noninjected drugs (heroin, cocaine, crack, amphetamines, club drugs, and/or other) that have been related to HIV risk behaviors. Risk ratios from the first study were then inserted as replacements for each drug that the participants indicated using. The total drug risk ratios were then added together to create an overall drug risk. As with sex risk, participants also rated their partners as to whether they have used or tried the same injected and non-injected drugs. Results The Actor Partner Independence Model (APIM; Kashy and Kenny 2000) was utilized to test our hypothesis using
123
AIDS Behav (2009) 13:246–257
MPlus (Muthe´n and Muthe´n 2005), a structural equation modeling statistical software package. APIM is a statistical method which conceptualizes and measures the interdependence that exists within intimate relationships (Cook and Kenny 2005). In testing the APIM models, the residual variances of the endogenous variables and the two exogenous variables were allowed to be correlated for all analyses. Each couple was the unit of analysis, and chisquare statistics were used to determine the goodness-of-fit for each model. A non-significant P value in the chi-square statistic suggests that the hypothesized model fits the data well and the null hypothesis should not be rejected. The first APIM model tested partner perceptions of sexual risk behaviors. This model provided for a good fit for the data, v2(5) = 7.37, P \ .19, CFI = 1.00, RMSEA \ .001. Correlations between partners on actual (r = -.03) and perceptions of partner risk (r = -.05) were not significant. Both male and female partners were inaccurate about each other’s behaviors, as the diagonal paths were non-significant, however only female partners were biased in their assessments of their partner’s sexual risk behaviors (b = .29, P = .01, see Fig. 1). Therefore, females, but not males, demonstrated the FCE. Women assumed that their partner’s HIV sexual risk behaviors were similar to their own. Another APIM model was tested examining drug risk. Unfortunately, reports of drug use besides alcohol were quite low in this sample, with 67% indicating they had never used non-injected drugs (besides alcohol) and 98.7% never injecting any drugs. As such, the model did not provide for a very good fit for the data when using a Chi square fit index, v2(5) = 151.01, P \ .001. However, the CFI (1.00) and RMSEA fit indices (\.01), which are much less influenced by sample size, indicated a fairly good fit for the data. The paths from actual to perceived risk behaviors were all significant in this model (see Fig. 2). As indicated by the significant diagonal paths, both male (b = .66, P \ .01) and female (b = .64, P \ .01) partners were accurate in their assessments of each other’s drug use behaviors. Interestingly, there was also some bias in participant’s perceptions, as the horizontal paths were also significant for both male and female partners. Correlations between male and female partner’s actual drug use was significant (r = .60, P \ .01), so it is possible that basing partner’s behaviors on one’s own past drug use behavior, in this sample, leads to accurate assessments. An additional analysis was conducted to determine whether perceptions of partner sex risk predicted frequency of condom use. This analysis did not provide for a good fit for the data, v2(5) = 34.50, P \ .001, and none of the paths in this model were significant. Therefore, perceptions of a partner’s sexual risk did not appear to influence whether condoms were ultimately used in these intimate relationships.
AIDS Behav (2009) 13:246–257
253
Fig. 1 Accuracy and bias of male and female partner’s sexual risk
.29* Female actual Female actualsex sex risk
Female perceived Female perceived partner sex partner sexrisk risk -.04 -
-.05
-.03
-.04
Male actualsex actual sexrisk risk
--.07
Male perceived perceived partnersex partner sexrisk risk
χ2(5) =7.37, p= .19, CFI 1.00, RMSEA < .01. *p
