General self-efficacy in relation to unprotected sexual encounters among persons living with HIV

Abstract

This study examined general self-efficacy in relation to sexual risk behavior among persons living with HIV and evaluated psychometric properties of the Positive Self Questionnaire, a novel measure of general self-efficacy. The Positive Self Questionnaire showed high internal consistency, a factor analysis supported by a single factor structure, and convergent validity supported by significant correlations in predicted directions with indicators of mental health. The Positive Self Questionnaire was related to unprotected sexual encounters, even after controlling for other factors. Results suggest that general self-efficacy is important to examine when assessing sexual risk behavior; an internally consistent measure is available for such endeavors.

Across the globe, approximately 33.4 million individuals are living with HIV/AIDS, while over 2 million new HIV infections occur each year (UNAIDS, 2009). Reducing the number of new cases of HIV is a major global health priority. The vast majority of new HIV infections are the result of sexual risk behavior, specifically unprotected vaginal and anal intercourse (e.g. Centers for Disease Control and Prevention (CDC), 2011). Many people living with HIV (PLH) still engage in unprotected sexual encounters after learning their status, despite the threat of additional transmission of infection (Crepaz and Marks, 2002; Wenger et al., 1994). These unprotected sexual behaviors put PLH at risk of contracting additional sexually transmitted infections and put their sexual partners at risk for contracting HIV. To better understand how to reduce HIV transmission rates and how to improve the health of PLH, it is imperative that factors related to sexual risk be identified.

General self-efficacy (GSE) is one factor that may be predictive of HIV transmission risk behavior. GSE encapsulates a general sense of mastery over novel situations and an ability to successfully perform tasks (Sherer, 1982). GSE has been associated with positive health outcomes, such as less depression, decreased physical disability, better somatic health, and greater adherence to medication regimens (Luszczynska et al., 2009); lower perceived stress and fewer symptoms of posttraumatic stress disorder (PTSD) (Ebstrup et al., 2011); and healthier coping styles, including more active and acceptance-based coping (Mystakidou et al., 2010). Within the population of PLH specifically, GSE has been found to be predictive of medication adherence and better physical functioning (Luszczynska et al., 2007). While this research has demonstrated that GSE predicts both physical and mental health functioning, less is known about how GSE relates to HIV transmission and sexual risk behavior.

Sexual risk behavior has been found to be associated with condom self-efficacy (CSE), a specific type of self-efficacy in which one feels competent obtaining and using condoms (Wulfert and Wan, 1993). Self-reported CSE has been found to be related to actual condom use in studies targeting both adults (e.g. Farmer and Meston, 2006; Teng and Mak, 2011) and adolescents (Zimmerman et al., 2007). However, across studies, CSE appears to be a weak predictor of actual condom use. Meta-analysis has found CSE to have only a small-to-medium correlation with condom use in heterosexual adults (Sheeran et al., 1999), and a measure assessing sexual risk behaviors in the context of HIV/AIDS risk failed to find significant correlations between CSE and condom use among adolescents (Faryna and Morales, 2000). Furthermore, interventions to increase the levels of CSE have had only modest effects on actual condom use (Harvey et al., 2009; O’Leary et al., 2008; Weinhardt et al., 2007). This may be attributed in part to the complexity of the process of actual condom use, which draws on multiple specific self-efficacies (Sheeran et al., 1999; Teng and Mak, 2011).

As indicated by these mixed results, much variance in sexual risk behavior is not explained by CSE (Basen-Engquist and Parcel, 1992). Whereas CSE measures a situation-specific perceived competency, GSE has been shown to be related to a variety of specific self-efficacies and is a more stable and, therefore, generalizable characteristic than CSE (Chen et al., 2001). Hence, GSE may be a more robust predictor of actual condom use than CSE. Many studies have attempted to determine the factors that contribute to risky sexual behavior in HIV-positive individuals, but few have examined the relationship between that population’s GSE scores and their sexual risk behavior. In one of these few studies, Harrison-Genus (2009) found that HIV-positive Jamaican women with higher levels of GSE were more likely to negotiate safe sexual encounters with their partners.

Current and standard measures of GSE assess feelings of competence and control in performing tasks, as well as perceived ability to cope with daily hassles and stressful life events (Schwarzer and Jerusalem, 1995). However, research into personality factors has indicated that the construct of GSE may also include components related to self-esteem, locus of control, and positive affect (Judge et al., 2002). Utilizing this broader definition of GSE may enhance examinations of the association between GSE and sexual risk behavior. Research has indicated that positive views of the self may be linked to disclosure of serostatus to sexual partners and decreased HIV-related sexual risk (Moskowitz and Seal, 2011; Stokes and Peterson, 1998). Thus, developing and administering a brief measure of GSE that incorporates the constructs of positive views of the self, self-esteem, and positive affect may lead to improved modeling of factors leading to HIV risk behavior, as well as providing additional targets for HIV prevention interventions.

This study examined the relationship between GSE and risky sexual behavior using a brief, novel measure of GSE, the Positive Self Questionnaire (PSQ), which is based on the General Self-Efficacy Scale (GSES; Schwarzer and Jerusalem, 1995). Although the existing measures of GSE, such as the GSES, have adequate psychometrics (Chen et al., 2001; Scherbaum et al., 2006), the PSQ was designed to augment these measures by improving readability, decreasing length, and using positive items that would be well-received by participants. It also defined GSE more broadly, so as to assess self-esteem, optimism about the future, and positive thoughts about the world and oneself. We predicted that this brief, novel measure of GSE would provide preliminary evidence of reliability and validity in a sample of adults with HIV/AIDS. Furthermore, we hypothesized that greater GSE would be associated with less self-reported sexual risk behavior, predicting variance in this outcome above and beyond that predicted by CSE.

Method

Participants

The sample consisted of seropositive men and women living in the Milwaukee area (N = 71). Participants were recruited from community-based clinics for a pilot study of an intervention designed to reduce trauma-related stress symptoms and HIV transmission risk behavior. Data from baseline visits (i.e. before randomization into an intervention condition) were analyzed. Inclusion criteria were as follows: HIV-positive, reported at least one trauma-related symptom during the past 3 months, reported HIV transmission risk behavior in the past 3 months, English speaking, and at least 18 years of age. The sample comprised 62% men (n = 44), 36.6% women (n = 26), and 1 transgendered (male to female) individual. The overall mean age for the sample was 41 (range = 27–56). Study participants identified as 84.5% White (n = 60), 9.9% African American (n = 7), 4.2% Latino/Hispanic (n = 3), and 1.4% Native American (n = 1). A total of 61.9% of the sample (n = 44) had completed a high school degree or equivalent, while 87.3% (n = 62) reported being unemployed. Most participants (63.3%; n = 45) were currently involved in a sexual relationship.

All participants provided informed consent to participate in the study; the informed consent document was read aloud and consent was obtained by a trained research assistant. All study procedures were approved by the Institutional Review Board of the Medical College of Wisconsin.

Measures

The PSQ was used to assess GSE. Several measures were used to help evaluate the convergent validity of PSQ, including measures of trauma symptoms, response to acute stress, depression, coping, and CSE. Sexual risk behavior was also assessed. Demographic information (gender, age, ethnicity, employment, and relationship status) was assessed by self-report items.

GSE.

The PSQ comprises nine statements about personal attitudes and behaviors that respondents answer on a 4-point scale from “not at all” to “exactly true,” with higher scores indicating greater agreement with the items and greater self-efficacy. Psychometric properties were calculated for the PSQ, as described below.

Trauma symptoms.

Trauma symptoms were assessed with the Impact of Events Scale–Revised (IES-R; Weiss and Marmar, 1995), a psychometrically valid and reliable measure that asks participants to rate their distress related to specific symptoms of PTSD on a 0 (none) to 4 (extremely) scale. Subscales measure each of the primary symptom clusters related to PTSD, namely intrusions, hyperarousal, and avoidance (Creamer et al., 2003). The IES-R total score demonstrated strong internal consistency in the current study (α = .95).

Response to acute stress.

The Stanford Acute Stress Reaction Questionnaire (SASRQ; Cardeña et al., 2000) is a 30-item measure including subscales of Dissociation and Functional Impairment. The Dissociation subscale includes 10 items, rated on a 0 (low) to 5 (high) scale, measuring dissociative symptoms of numbness, lack of awareness of surroundings, derealization, depersonalization, and amnesia. The Functional Impairment subscale includes two items, rated on a 0 (low) to 5 (high) scale, measuring difficulties in social and occupational functioning due to response to stress. This questionnaire has demonstrated good psychometric properties with populations exposed to a variety of stressful events (Cardeña et al., 2000). The SASRQ subscales demonstrated satisfactory to excellent internal consistency in the current study (Dissociation α = .90; Functional Impairment α = .75).

Depressive symptoms.

Depressive symptoms were measured with the Center for Epidemiologic Studies Depression (CES-D) Scale (Radloff, 1977), which consists of 20 items assessing mood, somatic symptoms, and interpersonal relationships during the past 7 days on a 4-point scale based on frequency of occurrence, ranging from 0 (rarely or none of the time (less than 1 day)) to 3 (most or all of the time (5–7 days)). Following the recommendation of Kalichman et al. (2000), we looked only at those items measuring the cognitive and affective symptoms of depression, as somatic symptoms tend to be highly correlated with HIV symptoms and may lead to inflated diagnoses of depression among PLH. This cognitive-affective subscale ranged from 0 to 48 in the current sample. The CES-D cognitive–affective subscale demonstrated good internal consistency in the current study (α = .85).

Coping strategies.

Coping strategies were assessed with the Brief COPE (Carver, 1997), which asks participants to rate how often they used different coping strategies on a 1 (not at all) to 4 (a lot) scale. For the current study, we selected a priori scales which we would expect to be related to GSE, namely the Denial, Active, Growth, and Acceptance subscales, with the Denial scale expected to be negatively associated with GSE and the others expected to be positively associated. Total scores for both scales ranged from 2 to 8, and all four subscales displayed satisfactory to good internal consistency in the current study (α = .76–.85).

CSE.

The CSE Scale (Rotheram-Borus et al., 1997) is a 10-item measure of ability and willingness to use condoms. It consists of a single scale, with items asking participants to rate their confidence in performing condom-related activities on a scale from 0 (not confident at all) to 4 (extremely confident). Total scores for the measure in the current sample ranged from 2 to 40, and the scale displayed excellent internal consistency (α = .90).

Sexual risk behavior.

Sexual risk behavior was assessed via a Sexual Risk Behavior Assessment Schedule (SERBAS) that has been used in previous studies of sexual risk and injection drug risk behavior (Weinhardt et al., 2004). This measure asks participants to report on sexual behavior during the last 3 months in general and with the last five partners specifically. To define sexual risk behavior in the current study, we looked at the total number of unprotected anal or vaginal intercourse acts during the last 3 months. As the distribution of responses was heavily skewed, we dichotomized the scores as “any unprotected sexual encounters” (coded as 1) and “no unprotected sexual encounters” (coded as 0).

Analytic strategy

Cronbach’s alpha was examined as an estimate of internal consistency for the items in the PSQ. To establish the construct validity of the PSQ, factor analysis was conducted using principal component analysis, to evaluate whether the items in the measure loaded on a single factor. Convergent validity was also examined; bivariate correlations were computed between the PSQ and mental health scales assessing theoretically related constructs that were expected to be negatively correlated with PSQ scores, including trauma symptoms, response to acute stressors, and depression. The use of denial was expected to be negatively correlated with the PSQ, while active, acceptance-based, and growth coping were expected to be positively correlated. Furthermore, CSE was expected to be moderately and positively correlated with the PSQ. Finally, we included the PSQ in a hierarchical logistic regression equation predicting engagement in unprotected sexual encounters. Participants’ dichotomous sexual risk scores were used as the outcome in this analysis. We controlled for demographic characteristics by entering as covariates, age, ethnicity, and gender in the first block of the regression analysis. Age was treated as a continuous variable, ethnicity was dichotomized to represent White/Caucasian and non-White/Caucasian participants, based on the techniques used in the previous research (e.g. Sikkema et al., 2009), and gender was dichotomized to represent male and female participants. We excluded the single transgendered participant from this analysis, as we did not have sufficient numbers of transgendered individuals to look at transgendered status as a control variable. We then entered CSE scores as a continuous predictor in the second block. Finally, we entered PSQ scores as a continuous predictor in the third block, to test whether GSE predicted additional variance in sexual risk behavior above and beyond CSE. To control for any covariance between the measures of CSE and GSE, both CSE and PSQ scores were centered before being entered into the equation. All statistical analyses were performed using SPSS 17.0 (SPSS Inc., Chicago, IL).

Results

Psychometric properties of the PSQ

Internal consistency.

The PSQ showed high internal consistency. Cronbach’s α was .92.

Factor structure and construct validity.

Using principal component analysis with an unrotated solution and retaining factors with an eigenvalue > 1, a single factor was extracted, explaining 61.83% of the total variance. As shown in Table 1, all nine items were retained in the final factor structure as their factor loadings were greater than .4. Factor loadings ranged from .62 (“I spend time to identify long-range goals for myself”) to .85 (“When I think about tomorrow, I usually feel good about what I’ve done today”).

Table 1.

Items and factor loadings for the general self-efficacy factor of the Positive Self Questionnaire (N = 71).

Factor 1. General self-efficacy (9 items, eigenvalue = 5.56)		Loading
1	I spend time to identify long-range goals for myself	.62
2	I feel in charge to make things happen	.77
3	I feel responsible for my own life	.76
4	I try to see the positive in anything that happens to me	.79
5	When I think about tomorrow, I usually feel good about what I have done today	.85
6	I am driven by a sense of purpose	.84
7	I am able to choose my own actions	.83
8	I find time to laugh	.80
9	There are abundant opportunities that await me	.79

Convergent validity.

As presented in Table 2, most of the scales showed evidence for the convergent validity of the PSQ by correlating moderately in the predicted directions. The PSQ scores were significantly (p < .05) negatively associated with scores on the measures of intrusive symptoms and hyperarousal symptoms of trauma, dissociation, functional impairment due to acute stress, depression, and coping through denial. Scores on the PSQ were significantly and positively associated with coping through acceptance and CSE. Participants’ responses on the PSQ, however, were not significantly related to their responses on the measures of overall trauma symptoms, avoidance symptoms, active coping, or growth coping.

Table 2.

Correlations between the Positive Self Questionnaire, mental health variables, coping strategies, and CSE (N = 71).

Measure	Correlation with Positive Self Questionnaire total score
Trauma symptoms total score (IES-R)	−.20
Intrusion symptoms (IES-R)	−.24*
Avoidance symptoms (IES-R)	−.06
Hyperarousal symptoms (IES-R)	−.26*
Dissociation symptoms (SASRQ)	−.28*
Functional impairment (SASRQ)	−.30*
Depression CES-D	−.35**
Denial coping (Brief COPE)	−.31**
Active coping (Brief COPE)	.18
Growth coping (Brief COPE)	.07
Acceptance coping (Brief COPE)	.44**
CSE	.32**

IES-R: Impact of Events Scale–Revised; SASRQ: Stanford Acute Stress Reaction Questionnaire; CSE: condom self-efficacy.

Note:

^*p < .05,

^**p < .01.

CES-D: Center for Epidemiologic Studies - Depression.

Predicting sexual risk behavior from GSE

The results of the logistic multiple regression are presented in Table 3. In Block 1, neither age, ethnicity (dichotomized as White/non-White), nor gender predicted variance in sexual risk behavior. In Block 2, CSE scores entered the model as a significant predictor of sexual risk behavior (β = −.10, odds ratio (OR) = .91, p <.01) and significantly improved model fit (χ² = 9.45, p < .01). In Block 3, scores on the PSQ predicted additional, unique variance in engagement in unprotected sexual encounters (β = −.15, OR = .86, p <.05) above and beyond the effect of CSE, resulting in a better fitting model (χ² = 7.80, p < .01). The overall model predicted an estimated 34% of the variance in sexual risk behavior (final χ² = 19.99, p < .001).

Table 3.

Results of a hierarchical logistic regression model predicting unprotected sexual intercourse from age, ethnicity, gender, condom self-efficacy, and general self-efficacy (N = 70).

Variable	OR, Step 1	(95% CI)	OR, Step 2	(95% CI)	OR, Step 3	(95% CI)
Step 1: χ2 = 2.94
Age	1.04	(.97–1.12)	1.04	(.96–1.12)	1.03	(.96–1.09)
White/Non-White	2.92	(.71–12.06)	3.71	(.79–17.51)	3.90	(.69–22.10)
Gender	.92	(.33–2.53)	1.39	(.46–4.20)	1.58	(.47–5.29)
Step 2: χ2 = 9.45**
Condom self-efficacy			.91**	(.84–.97)	.92*	(.85–.99)
Step 3: χ2 = 7.60**
Positive Self total score					.86*	(.77–.97)

OR: odds ratio; CI: confidence interval.

Note: model χ² = 19.99, Nagelkerke R² = .34,

p < .001;

^*p < .05,

^**p < .01.

Discussion

The current study indicates that the novel measure, the PSQ, demonstrates excellent psychometric properties, with strong internal consistency and a single factor structure, as well as good evidence of convergent validity. Furthermore, the PSQ is brief, easily administered, and may produce a more useful model for predicting health-related outcomes than more specific efficacy measures such as CSE. Future research involving interventions that target reduction of HIV transmission risk behavior may benefit from including measures such as the PSQ to investigate whether improving GSE leads to the reduction of sexual risk behaviors that can lead to HIV transmission.

Consistent with previous research, we found that CSE was a significant predictor of sexual risk behavior, but that it accounted for only a modest amount of variance. Including GSE, as measured by the PSQ, provided significantly heightened predictive power, consistent with the possibility that GSE plays an important role in actual condom use. This result may indicate that GSE is more generalizable than CSE to the myriad situations in which an individual’s intention to use a condom is tested. Researchers examining sexual risk behavior may consider including a measure of GSE, such as the PSQ, in their HIV prevention questionnaire batteries. Clinically, interventions to reduce sexual risk behavior by developing CSE (e.g. teaching proper condom use and role-playing assertive communication of the intention to use a condom) have had modest effects (Harvey et al., 2009; O’Leary et al., 2008; Weinhardt et al., 2007). Interventions that also foster GSE in PLH might produce more robust outcomes. Such interventions might include behavioral activation, problem solving, and exploration of past successes and perceived competencies. Further research is needed to establish the generalizability of our findings to other populations and to identify effective methods of increasing GSE.

An additional benefit of including the PSQ in research may be to mitigate a phenomenon in which participation in survey research results in increased levels of stress (Flagel et al., 2007); this may be problematic in research wherein people with HIV are asked to report on their negative or pathological experiences. Inclusion of a measure that encourages positive self-reflection, such as the PSQ, in research protocols may improve participants’ experience of the survey, which could in turn reduce attrition and increase participation in other research. This phenomenon could be examined in future research with the PSQ. Future psychometric paradigms should also extend and replicate our findings by establishing its test–retest reliability and convergent validity with other measures of GSE, so as to ensure that the PSQ is truly a valid measure of self-efficacy.

Finally, it is important to note that the PSQ used an expanded definition of GSE, consistent with previous research highlighting the overlap between self-efficacy, self-esteem, and positive affect (Judge et al., 2002). Despite utilizing this broad definition in the development of the questionnaire items, the items on the PSQ clustered as a single, internally consistent factor. This provides some additional evidence of the overlap between self-efficacy and positive views of the self. It also suggests that utilizing this expanded definition of self-efficacy and focusing on building self-esteem and self-respect may add additional power to interventions aimed at reducing sexual risk. However, self-efficacy is likely a broad construct influenced by multiple social and psychological factors; these factors may in turn influence sexual risk directly. Future studies could usefully examine the role of social norms and socially determined health beliefs on both GSE and sexual risk behavior.

Findings from the current study should be interpreted in light of several methodological limitations. The relatively small sample size may have contributed to sampling errors in the results; however, the sample size was large enough to detect reasonable outcomes, and the moderate effect sizes suggest meaningful clinical implications. Moreover, while the cross-sectional design of the study allows for the examination of associations between GSE and sexual risk behavior, it does not allow conclusions about clear temporal or causal relationships to be drawn. The cross-sectional design also prevented establishing test–retest reliability for the PSQ. In addition, the current study did not include a preexisting, more-researched measure of GSE, so the evidence of convergent and divergent validity of the PSQ is limited to associations with measures of other theoretically related constructs. Finally, the sample, recruited primarily from community clinics in the Milwaukee area and all reporting at least one symptom of trauma, represents a specific subset of the population of men and women diagnosed with HIV in the United States. It would be desirable therefore to extend the findings of the present study by replication with samples representing additional geographic areas and ethnic groups, as well as individuals with and without trauma histories.

As previously noted, HIV is a prominent public health concern, and it is important to identify the factors that lead to risky sexual activity in PLH in order to try to reduce HIV transmission risk. Our study showed a significant relationship between GSE and unprotected sexual intercourse, even accounting for CSE. It may be useful to develop an intervention that seeks to enhance the qualities measured by the PSQ in PLH; in addition, the reliability of the PSQ should be further examined, as well as its generalizability to the broader population of PLH.