Correlates and Effects of Alcohol Use Expectancies Among Persons Living with HIV in Uganda

Abstract

Unhealthy alcohol use fuels difficulties with HIV disease management and potentiates secondary transmission of HIV but less is known about how these alcohol use expectancies may shape alcohol use behaviors, particularly in the presence of depressive symptomatology. In this paper, we utilize data from a prospective study of 208 people living with HIV in Southwest Uganda, to examine the correlates of alcohol use expectancies and their association with unhealthy alcohol use. Affective depressive symptoms were positively associated with alcohol use expectancies. Gender moderation was observed such that depression was more strongly associated with alcohol use expectancies among women. In unadjusted analyses, alcohol use expectancies were marginally associated with unhealthy alcohol use and this association was not significant in adjusted analyses. Findings underscore the need to strengthen screening for depression and alcohol use within HIV care services, particularly among women.

Introduction

Uganda has one of the highest rates of alcohol consumption in Africa, with a total per capita alcohol consumption of 9.5 L [1]. Alcohol plays a major role in the HIV epidemic in Sub-Saharan Africa: it has been associated with high-risk sexual behavior, leading to HIV infections and poor HIV care outcomes, e.g., poor adherence to anti-retroviral therapy and mortality [2–7]. Psychological distress following an HIV diagnosis [8, 9] and mental health problems such as depression, dysthymia and suicidality are common among persons living with HIV (PLWH) [10–14]. Mental health problems among PLWH have been attributed to HIV-related stress and negative life events such as death worry, HIV stigma, declining physical functioning, and poverty [10–13]. Indeed, many studies have documented high rates of mental health problems among PLWH [10–12, 14]. Of the myriad of biological, psychological, cognitive and socio-ecological factors that have been implicated in the etiology of unhealthy alcohol use [2, 15–17], evaluative cognitions such as alcohol expectancies- defined as beliefs about the effects of alcohol [18, 19]—have emerged as potent correlates of quantity and frequency of alcohol consumption.

Theoretical [20, 21] and empirical studies [22–24] suggest that individuals are more likely to consume alcohol in response to stressful situations, particularly if they lack effective coping responses. Specifically, motivational [19] and tension-reduction [20] theories of alcohol use assert that stressful situations may induce alcohol consumption as a way of relieving stress. The relief from stress associated with alcohol consumption is the result of both pharmacological and expectancy effects. Alcohol expectancy theory [25] asserts that an individual’s beliefs about the expected effects of alcohol (i.e., alcohol use expectancies) influence the frequency and quantity of alcohol consumption. Alcohol use expectancies can be formulated before initiation of alcohol consumption [19, 26]. Positive expectancies can increase the likelihood of problem drinking in anticipation of positive expected effects (e.g. social facilitation, coping, sexual enhancement, behavioral disinhibition and conformity), while negative expectancies can reduce the likelihood of problem drinking in anticipation of negative expected effects (e.g., negative affective change, loss of control, health effects) [27–31]. Although much of the past research has focused on positive alcohol expectancies [31, 32], there is evidence that both positive and negative alcohol expectancies are important for understanding patterns of alcohol consumption, albeit with differential effects [26, 33].

Alcohol use expectancies, and specifically, sex-related alcohol expectancies [34–36], release of inhibitions such as self-confidence [36, 37], and negative alcohol expectancies [37], have been found to be strongly associated with unhealthy alcohol use. These associations have also been documented among PLWH in SSA [32, 38] and North America [7, 39], particularly in the context of HIV sexual risk behaviors. For example, men living with HIV who have higher sex-related alcohol use expectancies were more likely to consume alcohol during unprotected sex [39]. In South Africa, studies have documented positive associations between sex-related expectancies and alcohol use before sex [32, 38].

Gender plays an important role in the etiology of problematic alcohol use. The gendered patterning of both unhealthy alcohol use and mental health problems such as depression is well documented. Globally, men are consistently more likely to consume alcohol and to have a higher likelihood of unhealthy alcohol use compared to women [40–42] and this pattern is attributed to both biological factors (e.g., women have a lower genetic risk for alcohol use disorders and more negative biological consequences from drinking compared with men) and socio-cultural factors (e.g., greater social sanctions against female drinking, and adherence to gender norms about masculinity) [43–47].

Gender differences in alcohol use expectancies have also been documented: men have higher positive alcohol expectancies than women [22, 48], largely due to differences in how males and females cope with stress. Women are more likely to internalize stressful stimuli and have lower expectancies for alcohol-induced stress reduction, while men are more likely to externalize stress and have higher expectancies for alcohol-induced stress reduction [22]. Indeed, previous studies have found that gender moderates the relationship between depression and alcohol use but the direction of these moderating effects has been inconsistent. For example, some studies report stronger associations between depressive symptoms and unhealthy alcohol use among men [22, 49] while other studies reported greater unhealthy alcohol use among females with depressive symptoms [50]. Sample population characteristics (e.g., age) and differences in outcome measures (e.g., daily alcohol use, unhealthy alcohol use, alcohol use disorder) may contribute to these variations in the study findings.

Despite the preponderance of research on the important contributory role of alcohol use expectancies in the etiology of problematic alcohol use, less is known about the correlates of alcohol use expectancies in Africa, and how these outcome expectancies may shape alcohol use behaviors, particularly in the presence of mental health problems such as depression. Such understanding is necessary to inform ongoing efforts to develop or adapt existing alcohol use interventions such as alcohol expectancy challenges, which have demonstrated efficacy in the United States [25, 51] but have not been tested in the African context. The aims of this paper were to examine: (1) the correlates of alcohol use expectancies, and (2) their association with unhealthy alcohol use among persons living with HIV in Uganda participating in a research study examining alcohol use in the first year of HIV care. Specifically, we examined the influence of socio-demographic, psychosocial, and HIV disease correlates on alcohol use expectancies, and the relationship between alcohol use expectancies and unhealthy alcohol use among PLWH in Uganda. We hypothesized that: (1) positive alcohol expectancies (release of inhibition and sex-related expectancy) would be positively, and significantly, correlated with unhealthy alcohol use, while negative alcohol expectancies would be negatively, and significantly, correlated with unhealthy alcohol use; (2) the association between positive expectancies and unhealthy alcohol use would be stronger for individuals with higher depressive symptoms due to positive expectancies related to tension reduction/stress dampening; and (3) the association between alcohol use and positive alcohol use expectancies will be stronger for men with depressive symptoms compared to women, as men are more likely to utilize external coping strategies such as alcohol use.

Methods

Setting

This paper utilized baseline data from BREATH (Biomarker Research on Ethanol Among Those with HIV)—a one year prospective cohort study of HIV-infected adults entering HIV care in Mbarara, Southwest Uganda [52]. Participants were recruited from the Mbarara Regional Referral Hospital Immune Suppression Syndrome (ISS) clinic. ISS clinic patients were eligible to participate in the study if they were ages 18 years or older, their primary language was Runyankole (the local language) or English, they had not previously received HIV care, they resided within 60 km of the clinic and either they reported any alcohol consumption in the prior year to HIV clinic counselors, or the counselors suspected that they were alcohol consumers.

Study Procedures

BREATH cohort participants were administered a structured interview consisting of numerous measures including socio-demographics, alcohol use, mental health (e.g., depression) and psychosocial variables (e.g., social support, spirituality, alcohol use expectancies). Interviews were conducted in a private room in either English or Runyankole – the dominant regional language. Our selection of variables for these analyses was guided by both theoretical models [19, 20] and empirical studies of alcohol use expectancies [22–24].

Protection of Human Subjects

The study was approved by Institutional Review Boards at the University of California San Francisco, Mbarara University of Science and Technology (MUST) and the Uganda National Council for Science and Technology.

Variables

We focused on alcohol use expectancies and alcohol use as the main outcome variables for aims 1 and 2. Alcohol use expectancies were the primary predictor for Aim 2. We also assessed frequency and quantity of alcohol use as secondary outcomes in Aim 2 analyses.

Alcohol Use Expectancies

We used the 11-item East Africa Alcohol Expectancy Scale (AFEXS) that was developed and validated in 181 HIV-infected adults recruited from the BREATH and URBAN ARCH Uganda Study- a cohort study of 450 people living with HIV that examines the effects of heavy alcohol consumption on disease progression [53]. The global scale and sub-scales demonstrated strong reliability [53]: global expectancy scale (Cronbach’s α = 0.90; M = 32, SD = 14); sex-related expectancies (Cronbach’s α = 0.93; M = 17.8, SD = 4.4), release of inhibition (Cronbach’s α = 0.73; M = 8.6, SD = 4.4), and negative expectancies (Cronbach’s α = 0.74; M = 5.6, SD = 2.9). For our analyses, we computed the global summative score and three sub-scale scores (i.e., sex-related expectancies, release of inhibition and negative expectancies). For Aim 1, we recoded the global and sub-scale scores into binary variables (high/low) based on a median split because these scores were not normally distributed and the AFEXS scale does not have a cut-off value for subject categorization.

Alcohol Use

We examined three indicators of alcohol use. A composite measure of unhealthy alcohol use as primary outcome. Frequency of alcohol use and quantity of alcohol use as secondary outcomes.

Unhealthy Alcohol Use

To improve measurement sensitivity, we created a combined measure of unhealthy alcohol use using the Alcohol Use Disorders Identification Test-Consumption (AUDIT-C) and phosphatidylethanol (PEth), an alcohol biomarker [54]. The combined measure was based on a positive AUDIT-C and/or PEth ≥ 50 ng/ml. We modified the AUDIT-C to elicit responses for drinking in the prior 3 months and used cut-offs of ≥ 3 for women and ≥ 4 for men [52] to create a binary variable for positive for self-reported unhealthy drinking. PEth is a biomarker for unhealthy alcohol use in the past 2–3 weeks, which is assessed using liquid chromatography – tandem mass spectrometry (LC-MS/MS)- following extraction into methanol on Dried blood spots (DBS) by the United States Drug Testing Laboratory [52]. The lower limit of quantification was 8 ng/ml, and the most common PEth homologue (16:0/18:1) was detected [54–57]. Based on previous studies, we considered individuals to have unhealthy alcohol consumption if their PEth level was ≥ 50 ng/ml [54, 56].

Frequency of Alcohol Use

We used the first question of the AUDIT-C to measure frequency of alcohol use, i.e., how often do you have a drink containing alcohol in the past year? We categorized respondents reporting two or more times per week as frequent alcohol consumers.

Quantity of Alcohol Use

We used the second question of the AUDIT-C to assess quantity of alcohol use per drinking occasion, i.e., how many drinks containing alcohol did you have on a typical day when you were drinking in the past year? We categorized respondents reporting three or more drinks per occasion as high alcohol consumers.

Predictor Variables

Depressive Symptoms

We used the depression section of the Hopkins Symptoms Checklist (DSHSCL), a 15-item Likert-type scale that measures symptoms of depression according to the criteria set by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV). We added an additional item, “feeling like I don’t care what happens to my health” [58, 59]. Given the evidence suggesting that somatic measures of depression may inflate depression scores among HIV-infected persons [60, 61], we followed Bolton’s suggestion [58] and excluded somatic symptoms of depression. We excluded the following scale items: “feeling low in energy, slowed down”, “feeling fidgety”, “poor appetite” and “having difficulty falling or staying asleep”. The reliability of the scale was excellent: Cronbach’s α = 0.90 (M = 1.7, SD = 0.7). We computed the mean depressive score, which we centered to create standardized z scores.

Low Social Support

We used the modified version of the Duke-UNC Functional Support Questionnaire [62] to assess social support. This 10-item questionnaire measures respondent’s emotional/affective support and material/instrumental support. The scale reliability in this sample was good: Cronbach’s α = 0.89 (M = 35.0; SD = 6.2). We chose to dichotomize responses into a binary variable – yes or no, using a mean score of less than three as the cut-off.

Coping and Self-Efficacy

We used the Coping self-efficacy (CSE) scale, a 26-item measure of one’s confidence in performing coping behaviors when faced with life challenges [63]. This scale demonstrated strong reliability: Cronbach’s α = 0.89 (M = 88.6, SD = 25.3). Total scores were mean centered to create standardized z scores.

Spirituality

We used the short version of the Ironson-Woods Religiosity Index, which has been associated with long survival, health behaviors, less distress, and low cortisol in people living with HIV [64]. This 22-item scale has strong reliability: Cronbach’s α = 0.94 (M = 91.9, SD = 9.4). Total scores were mean centered to create standardized z scores.

Gender

We defined gender as a binary variable – males and females.

Covariates

We included the following socio-demographic variables: age (mean centered), educational attainment, marital status, employment status, household socio-economic status, partner HIV status and HIV symptom burden. We categorized educational attainment into a binary variable – no secondary school or any secondary school education. We categorized marital status into a binary variable – married (including co-habiting or living with partner as if married) and unmarried (single or widowed). We used principal components analysis to create a household asset index from measures of housing quality and available energy sources [65, 66]. We grouped participants into three categories: low (bottom 40%), middle (middle 40%) and high (top 20%). Taking into consideration polygamous marriages, we conceptualized partner HIV status as having any HIV-infected partner. We elicited HIV-related severe symptoms lasting at least 4 weeks, from a list of nine of the most prevalent symptoms chosen by the AIDS Clinical Trials Group Index [67]. We computed a standardized score of HIV symptom burden.

Analyses

Aim 1

We conducted four sets of unadjusted and adjusted logistic regression analyses to assess the associations of several covariates with each of the four alcohol use expectancies variables i.e., global score and three subscales. The covariates included psychosocial factors (i.e., depressive symptoms, spirituality, social support, and coping self-efficacy); and socio-demographic factors (i.e., age, gender, education, marital status, household wealth, employment, HIV symptoms and partner HIV status). We conducted adjusted analyses to examine the main effects of the factors listed above (step 1), as well as interactive effects between gender and depressive symptoms (step 2) on alcohol use expectancies. The psychosocial factors, age, and gender were included in all adjusted models, as well as any other covariate that was statistically significant (p < 0.05) in unadjusted analysis.

Aim 2

We conducted unadjusted and adjusted logistic regression analyses to examine the relationship between alcohol use expectancies (global score and subscales) and unhealthy alcohol use. Adjusted analyses included socio-demographics (i.e., age, gender, education, marital status, household wealth, employment, HIV symptoms and partner HIV status), psychosocial variables (i.e., spirituality, social support, coping self-efficacy and depression), and gender by depression interaction terms as potential covariates. Only statistically significant socio-demographic and psychosocial variables were included in the adjusted analyses.

Lastly for Aim 1 and Aim 2, we conducted multiple imputation via chained equations in Stata (using 25 imputed datasets), to account for missing data. All variables listed above were included in the imputation model. The results obtained using the imputed data were similar to those using the observed data. So, the results we present here are based on the observed data.

Secondary Analyses

We also examined the relationship between alcohol use expectancies and two other alcohol use measures: frequency of alcohol use and quantity of alcohol use per drinking occasion. We conducted separate logistic regression analyses for the global expectancies scale and each sub-scale and these alcohol use measures, adjusting for the same variables as the analysis of unhealthy alcohol use in Aim 2.

Results

Participant Characteristics

From July 2011 to July 2013, 3747 new patients were screened and 621 were considered eligible for the BREATH study. Sixty-one per cent (n = 381) agreed to enroll in the study. Similar proportions of those who enrolled compared to those who declined participation were female (45 versus 47%, respectively, P = 0.59). Of the 381 people enrolled, 213 were randomized to the BREATH cohort described here and 168 were randomized to later study participation, for an examination of assessment reactivity [68] and were not included here. Of the 213, five were found ineligible after enrollment and three were excluded because PEth testing was not performed, leaving a sample size of 205 participants.

Of the 205 participants (Table 1), 57.6% were male, the median age was 30 years (inter-quartile range (IQR): 25–38 years), and 51.7% were married or cohabiting. Almost two-thirds (64.4%) of participants did not have a secondary school education and the majority (94.2%) were currently employed. Almost two-thirds (64.7%) of respondents had unhealthy alcohol use. Approximately one-third (38.5%) reported frequent alcohol use and 40.2% reported a high number of drinks per drinking occasion in the prior three months. The median scores for the alcohol expectancies were as follows: Global alcohol expectancies: 31 (IQR: 22–43); Release of inhibition sub-scale: 8 (IQR: 5–12); Sex-related expectancies: 15 (IQR: 11–27) and Negative expectancies: 5 (IQR: 3–8). The median depression score was 1.5 (IQR: 1.2–2.1) and 17.7% of participants reported low social support.

Table 1.

Descriptive statistics of participant socio-demographic characteristics, BREATH study participants at baseline (n = 205)

	N (%) or Median (IQR)	Mean (SD)
Age	30 (25–38)	31.9 (9.4)
Gender
Male	118 (57.6)
Female	87 (42.4)
Educational attainment
No secondary school education	132 (64.4)
Any secondary education	73 (35.6)
Marital status
Not married	99 (48.3)
Married	106 (51.7)
Any HIV-infected spouse(s)
No/don’t know	152 (74.2)
Yes	53 (25.9)
Household asset index
Low	80 (39.0)
Medium	83 (40.5)
High	42 (20.5)
Employment status
Currently employed	193 (94.2)
Currently not employed	12 (5.9)
Months since HIV diagnosis	0.3 (0.1–1.3)	3.9 (12.3)
Depression score	1.5 (1.2–2.1)	1.7 (0.7)
HIV symptom burden (9 symptoms)	2 (0–4)	2.3 (2.2)
Low social support
No	168 (82.4)
Yes	36 (17.7)
Spirituality score	88 (88–97)	91.9 (9.4)
Coping and self-efficacy	90 (71–107.5)	88.6 (25.3)
Unhealthy alcohol use, prior 3 months (AUDIT-C positive or PEth ≥ 50)
No	72 (35.3)
Yes	132 (64.7)
Frequent alcohol use, prior 3 months
No	126 (61.5)
Yes	79 (38.5)
Quantity of alcohol per drinking occasion, prior 3 months
No	122 (59.8)
Yes	82 (40.2)
Alcohol use expectancies
AFEXS	31 (22–43)	32.0 (14.0)
AFEXS sub-scale: Release of inhibition expectancies	8 (5–12)	8.6 (4.4)
AFEXS sub-scale: Sex-related expectancies	15 (11–27)	17.8 (9.3)
AFEXS sub-scale: Negative expectancies	5 (3–8)	5.6 (2.9)

Aim 1: Correlates of Alcohol Use Expectancies

Global Alcohol Expectancies

In the unadjusted analyses (Table 2), the number of affective depressive symptoms were not significantly associated with global alcohol use expectancies. However, in the adjusted logistic regression analyses (Table 2), the number of affective depressive symptoms was significantly associated with high global alcohol use expectancies (adjusted odds ratio (AOR) = 1.56, 95% CI 1.08, 2.25; z = 2.39; p = 0.02). In the adjusted analyses, unmarried respondents had significantly lower odds of reporting high global alcohol use expectancies, relative to married respondents (AOR = 0.42, 95 CI 0.20 to 0.90; z = − 2.23; p = 0.03).

Table 2.

Correlates of alcohol use expectancies: unadjusted and adjusted odds ratios (OR) and 95% confidence intervals (CI, BREATH study participants at baseline)

	AFEXS				Release of inhibition expectancies				Sex-related expectancies				Negative expectancies
	Unadjusted		Adjusted (n = 155)		Unadjusted		Adjusted (n = 155)		Unadjusted		Adjusted (n = 155)		Unadjusted		Adjusted (n = 156)
	OR (95% CI)	z (p-value)	OR (95% CI)	z (p-value)	OR (95% CI)	z (p-value)	OR (95% CI)	z (p-value)	OR (95% CI)	z (p-value)	OR (95% CI)	z (p-value)	OR (95% CI)	z (p-value)	OR (95% CI)	z (p-value)
Agea	1.06 (0.80, 1.40)	0.41 (0.68)	0.93 (0.65, 1.35)	− 0.37 (0.71)	0.93 (0.71, 1.23)	− 0.50 (0.62)	0.87 (0.59, 1.27)	− 0.73 (0.47)	1.24 (0.93, 1.64)	1.47 (0.14)	1.10 (0.76, 1.59)	0.51 (0.61)	0.97 (0.74, 1.28)	− 0.19 (0.85)	1.10 (0.78, 1.57)	0.55 (0.58)
Gender		− 1.36 (0.18)		− 0.72 (0.47)		− 0.63 (0.53)		− 0.84 (0.40)		− 2.64 (<0.01)		− 1.94 (0.05)		0.12 (0.91)		− 0.31 (0.76)
Male (ref)	1.00		1.00		1.00		1.00		1.00		1.00		1.00		1.00
Female	0.68 (0.39, 1.19)		0.76 (0.36, 1.61)		0.84 (0.48, 1.46)		0.72 (0.34, 1.55)		0.47*** (0.26, 0.82)		0.47 (0.22, 1.01)		1.03 (0.59, 1.80)		0.89 (0.42, 1.87)
Educational attainment		− 0.17 (0.86)				0.60 (0.55)				− 0.36 (0.72)				− 0.85 (0.40)
Less than secondary school (ref)	1.00		–		1.00				1.00				1.00
Secondary school or higher	0.95 (0.53, 1.69)		–		1.19 (0.67, 2.12)		–		0.90 (0.51, 1.60)		–		0.78 (0.44, 1.39)		–
Marital status		− 2.11 (0.04)		− 2.23 (0.03)		− 1.13 (0.26)	–			− 1.98 (0.05)		− 2.16 (0.03)		− 0.37 (0.71)
Not married	0.55** (0.32, 0.96)		0.42** (0.20, 0.90)		0.73 (0.42, 1.26)		–		0.57** (0.33, 0.99)		0.43** (0.20, 0.92)		0.90 (0.52, 1.56)		–
Married (ref)	1.00		1.00		1.00				1.00		1.00		1.00
Any HIV-infected spouse(s)		− 0.97 (0.33)				− 1.21 (0.23)				− 0.73 (0.47)				− 0.21 (0.83)
No	0.73 (0.39, 1.38)		–		0.68 (0.36, 1.28)		–		0.79 (0.42, 1.49)		–		0.93 (0.50, 1.75)		–
Yes (ref)	1.00		–		1.00		–		1.00		–		1.00		–
Household asset index
Low	0.50** (0.27, 0.93)	− 2.19 (0.03)	0.62 (0.29, 1.34)	− 1.21 (0.23)	0.52** (0.28, 0.97)	− 2.04 (0.04)	0.39** (0.17, 0.87)	− 2.30 (0.02)	0.55* (0.29, 1.02)	− 1.88 (0.06)	–		0.76 (0.41, 1.41)	− 0.88 (0.38)	–
Medium (ref)	1.00		1.00		1.00		1.00		1.00		–		1.00		–
High	0.62 (0.29, 1.30)	− 1.27 (0.20)	0.79 (0.31, 2.02)	− 0.49 (0.63)	0.78 (0.37, 1.65)	− 0.64 (0.52)	0.61 (0.24, 1.58)	− 1.02 (0.31)	0.59 (0.28, 1.24)	− 1.39 (0.16)	–		0.98 (0.46, 2.05)	− 0.06 (0.95)	–
Current employment status		− 1.04 (0.30)				− 1.07 (0.29)				− 1.55 (0.12)				− 1.02 (0.31)
Employed (ref)	1.00		–		1.00		–		1.00		–		1.00		–
Unemployed	0.52 (0.15, 1.79)		–		0.51 (0.15, 1.75)		–		0.35 (0.09, 1.32)		–		0.53 (0.15, 1.81)		–
Affective depressive symptomsa	1.27 (0.96, 1.68)	1.67 (0.10)	1.56** (1.08, 2.25)	2.39 (0.02)	1.57*** (1.17, 2.12)	2.97 (<0.01)	1.96*** (1.33, 2.90)	3.39 (<0.01)	1.10 (0.84, 1.45)	0.70 (0.48)	1.10 (0.77, 1.58)	0.53 (0.59)	1.75*** (1.28, 2.38)	3.54 (<0.01)	1.88*** (1.29, 2.74)	3.26 (<0.01)
HIV symptom burdena	0.82 (0.62, 1.09)	− 1.38 (0.17)	–		0.84 (0.64, 1.11)	− 1.22 (0.22)	–		0.78* (0.59, 1.03)	− 1.73 (0.08)	–		1.09 (0.83, 1.44)	0.63 (0.53)	–
Low social support		0.23 (0.82)		0.96 (0.34)		0.90 (0.37)		0.58 (0.56)		1.39 (0.17)		2.47 (0.01)		0.69 (0.49)		− 0.34 (0.73)
No (ref)	1.00		1.00		1.00		1.00		1.00		1.00		1.00		1.00
Yes	1.09 (0.53, 2.24)		1.66 (0.59, 4.67)		1.39 (0.67, 2.87)		1.36 (0.48, 3.82)		1.68 (0.81, 3.48)		3.79** (1.32, 10.90)		1.29 (0.63, 2.65)		0.84 (0.31, 2.27)
Spiritualitya	0.94 (0.69, 1.28)	− 0.38 (0.70)	0.99 (0.70, 1.39)	− 0.07 (0.95)	0.72* (0.51, 1.01)	− 1.91 (0.06)	0.81 (0.56, 1.16)	− 1.15 (0.25)	0.89 (0.65, 1.21)	− 0.73 (0.47)	0.87 (0.62, 1.23)	− 0.77 (0.44)	0.83 (0.60, 1.14)	− 1.16 (0.25)	0.94 (0.67, 1.31)	− 0.38 (0.70)
Coping self-efficacya	0.79 (0.59, 1.06)	− 1.59 (0.11)	0.77 (0.54, 1.09)	− 1.50 (0.14)	0.84 (0.63, 1.12)	− 1.17 (0.24)	0.87 (0.61, 1.25)	− 0.75 (0.45)	0.77* (0.57, 1.03)	− 1.76 (0.08)	0.69** (0.49, 0.98)	− 2.06 (0.04)	0.69** (0.51, 0.93)	− 2.42 (0.02)	0.83 (0.59, 1.16)	− 1.10 (0.27)
Interaction
Gender × depressive symptoms
Males: depressive symptomsa	0.99 (0.66, 1.49)	− 0.05 (0.96)	–		1.33 (0.87, 2.02)	1.31 (0.19)	–		0.93 (0.62, 1.40)	− 0.34 (0.73)	–		1.33 (0.87, 2.02)	1.31 (0.19)
Females: depressive symptomsa	1.72** (1.13, 2.61)	2.55 (<0.01)	–		1.94*** (1.25, 3.00)	2.95 (0.01)	–		1.44* (0.97, 2.15)	1.81 (0.07)	–		1.94*** (1.25, 3.00)	2.95 (<0.01)

^***p < .01;

^**p < .05;

^*p < .10

^az-scores

Release of Inhibition

In the unadjusted analyses (Table 2), the number of affective depressive symptoms were positively associated with an increased odds of high release of inhibition alcohol use expectancies (unadjusted odds ratio (UOR) = 1.57, 95% CI 1.17, 2.21; z = 2.97; p < 0.01). In the adjusted analyses (Table 2), the number of affective depressive symptoms remained significantly associated with an increased odds of high release of inhibition alcohol expectancies (AOR = 1.96, 95% CI 1.33, 2.90; z = 3.39; p < 0.01). In the adjusted analyses, respondents residing in low asset households had lower odds of reporting high release of inhibition expectancies, relative to respondents residing in medium asset households (AOR = 0.39, 95% CI 0.17, 0.87; z = − 2.30; p = 0.01).

Sex-Related Alcohol Expectancies

In the unadjusted and adjusted analyses (Table 2), the number of affective depressive symptoms were not significantly associated with sex-related alcohol expectancies. In the unadjusted analyses (Table 2), women had significantly lower odds of reporting high sex-related expectancies than men (UOR = 0.47, 95% CI 0.26, 0.82; z = − 2.64; p < 0.01) but this relationship did not remain statistically significant in the adjusted analyses. In the adjusted analyses, unmarried respondents had lower odds of reporting high sex-related expectancies compared to married respondents (AOR = 0.43, 95% CI 0.20, 0.92; z = − 2.16; p = 0.03). Low social support was significantly associated with higher odds of reporting high sex-related alcohol expectancies, (AOR = 3.79, 95% CI 1.32, 10.90; z = 2.47; p = 0.01), and increasing levels of coping self-efficacy were significantly associated with lower odds of reporting high sex-related alcohol use expectancies (AOR = 0.69, 95% CI 0.49, 0.98; z = − 2.06; p = 0.04).

Negative Alcohol Expectancies

In the unadjusted analyses (Table 2), the number of affective depressive symptoms was positively associated with high negative alcohol expectancies, while coping self-efficacy was negatively associated with high negative alcohol expectancies. In the adjusted analyses (Table 2), the number of affective depressive symptoms was positively associated with high negative alcohol expectancies (AOR = 1.88, 95% CI 1.29, 2.74; z = 3.26; p < 0.01).

Gender and Depression Interactions

As shown in Table 2, females with depressive symptoms had significantly higher odds of reporting high global alcohol use expectancies, release of inhibition expectancies, sex-related alcohol expectancies and negative alcohol expectancies. The associations between depression and the expectancies were weaker for the males (Fig. 1).

Fig. 1.

Fitted plots of Gender*Depression interactions stratified by Alcohol use expectancies (AFEXS) scale and sub-scales

Aim 2: Alcohol Use Expectancies and Unhealthy Alcohol Use

In the unadjusted logistic regression analyses (Table 3), higher global alcohol use expectancies were associated with higher odds of unhealthy alcohol use but this association did not reach statistical significance (UOR = 1.32; 95% CI 0.98, 1.77; z = 1.82; p = 0.07). We observed a similar pattern with the sub-scale analyses. Depression was not significantly associated with unhealthy alcohol use and the gender*depression interaction was also not statistically significant. In the adjusted analyses (Table 3), alcohol use expectancies were not significantly associated with unhealthy alcohol use. Being female and having a HIV-negative partner remained significantly associated with lower odds for unhealthy alcohol use.

Table 3.

Unadjusted (UOR) and adjusted odds ratios (AOR) and 95% confidence intervals (CI) from multivariable logistic regression models examining the relationship between alcohol use expectancies and unhealthy alcohol use among BREATH study participants at baseline

	Model 1 (n = 203)				Model 2 (n = 203)				Model 3 (n = 203)				Model 4 (n = 204)
	UOR (95% CI)	z (p-value)	AOR (95% CI)	z (p-value)	UOR (95% CI)	z (p-value)	AOR (95% CI)	z (p-value)	UOR (95% CI)	z (p-value)	AOR (95% CI)	z (p-value)	UOR (95% CI)	z (p-value)	AOR (95% CI)	z (p-value)
Alcohol use expectancies
AFEXSa	1.32* (0.98, 1.77)	1.82 (0.07)	1.28 (0.94, 1.73)	1.57 (0.12)	–		–	–	–		–		–		–
Release of inhibition expectanciesa	–		–		1.26 (0.94, 1.69)	1.52 (0.13)	1.27 (0.94, 1.73)	1.54 (0.12)
Sex-related expectanciesa	–		–		–		–	–	1.27 (0.94, 1.71)	1.58 (0.12)	1.18 (0.87, 1.61)	1.08 (0.28)	–		–
Negative expectanciesa	–		–		–		–	–	–		–		1.23 (0.92, 1.66)	1.40 (0.16)	1.31* (0.97, 1.78)	1.76 (0.08)
Covariates
Gender		− 3.30 (<0.01)		− 2.89 (<0.01)		− 3.30 (<0.01)		− 3.04 (<0.01)		− 3.30 (<0.01)		− 2.84 (<0.01)		− 3.30 (<0.01)		− 3.12 (<0.01)
Male (ref)	1.00		1.00		1.00		1.00		1.00		1.00		1.00		1.00
Female	0.37*** (0.20, 0.67)		0.41 *** (0.22, 0.75)		0.37*** (0.20, 0.67)		0.39*** (0.21, 0.71)		0.37*** (0.20, 0.67)		0.41 *** (0.22, 0.76)		0.37*** (0.20, 0.67)		0.38*** (0.21, 0.70)
Any HIV- infected spouse(s)		− 2.11 (0.04)		− 1.82 (0.07)		− 2.11 (0.04)		− 1.82 (0.07)		− 2.11 (0.04)		− 1.80 (0.07)		− 2.11 (0.04)		− 1.65 (0.10)
No	0.46** (0.22, 0.95)		0.49* (0.23, 1.06)		0.46** (0.22, 0.95)		0.49* (0.23, 1.06)		0.46** (0.22, 0.95)		0.50* (0.23, 1.06)		0.46** (0.22, 0.95)		0.53 (0.25, 1.13)
Yes (ref)	1.00		1.00		1.00		1.00		1.00		1.00		1.00		1.00

^***p < .01;

^**p < .05;

^*p < .10

^az-scores

Gender and Depression Interaction

There was no evidence of an interaction between gender and depressive symptoms on unhealthy alcohol use.

Secondary Analyses

Quantity of Alcohol Use

In both unadjusted and adjusted analyses, alcohol use expectancies were significantly associated with consuming three or more drinks per drinking occasion (Table 4). This positive association was observed for global alcohol use expectancies (AOR = 1.50; 95% CI 1.12, 2.02; z = 2.87; p < 0.01), and each of the sub-scales: release of inhibition expectancies (AOR = 1.44; 95% CI 1.08, 1.94; z = 2.44; p = 0.01); sex-related alcohol use expectancies (AOR = 1.38; 95% CI 1.03, 1.85; z = 2.16; p = 0.03); and negative alcohol use expectancies (AOR = 1.42; 95% CI 1.06, 1.91; z = 2.33; p = 0.02).

Table 4.

Unadjusted (UOR) and adjusted odds ratios (AOR) and 95% confidence intervals (CI) from multivariable logistic regression models examining the relationship between alcohol use expectancies and quantity of alcohol use (i.e., high number of drinks (≥3) per drinking occasion) among BREATH study participants at baseline

	Model 1 (n = 203)				Model 2 (n = 203)				Model 3 (n = 203)				Model 4 (n = 204)
	UOR (95% Cl)	z (p-value)	AOR (95% Cl)	z (p-value)	UOR (95% Cl)	z (p-value)	AOR (95% Cl)	z (p-value)	UOR (95% Cl)	z (p-value)	AOR (95% Cl)	z (p-value)	UOR (95% Cl)	z (p-value)	AOR (95% Cl)	z (p-value)
Alcohol use expectancies
AFEXSa	1.53 *** (1.14, 2.05)	2.87 (<0.01)	1.50 *** (1.12, 2.02)	2.69 (<0.01)	–		–		–		–		–		–
Release of inhibition expectanciesa	–		–		–		–		1.43** (1.08, 1.91)	2.46 (0.01)	1.44** (1.08, 1.94)	2.44 (0.01)	–		–
Sex-related expectanciesa	–		–		1.44** (1.08, 1.91)	2.50 (0.01)	1.38** (1.03, 1.85)	2.16 (0.03)	–		–		–		–
Negative expectanciesa	–		–		–		–		–		–		1.35** (1.02, 1.80)	2.08 (0.04)	1.42** (1.06, 1.91)	2.33 (0.02)
Covariates
Gender				− 2.38 (0.02)				− 2.28 (0.02)				− 2.59 (0.01)				− 2.74 (<0.01)
Male (ref)	–		1.00		–		1.00		–		1.00		–		1.00
Female	–		0.48** (0.26, 0.88)		–		0.49** (0.27, 0.91)		–		0.45** (0.24, 0.82)		–		0.42*** (0.23, 0.78)
Any HIV-infected spouse(s)				− 0.94 (0.35)				− 0.92 (0.36)				− 0.86 (0.39)				− 0.75 (0.46)
No	–		0.73 (0.38, 1.41)		–		0.73 (0.38, 1.42)		–		0.75 (0.38, 1.45)		–		0.78 (0.40, 1.51)
Yes (ref)	–		1.00		–		1.00		–		1.00		–		1.00

^***p < .01;

^**p < .05;

^*p < .10

^az-scores

Frequency of Alcohol Use

Alcohol use expectancies were not significantly associated with frequent alcohol use in unadjusted or adjusted analyses (data not shown).

Discussion

Drawing on baseline data from a study of HIV-infected adults in Southwest Uganda, we examined the socio-demographic and psychosocial correlates of alcohol use expectancies and the relationship between depression, alcohol use expectancies and alcohol use among PLWH who were past year consumers of alcohol and enrolling in HIV care in Uganda. Similar to prior studies, we found that affective depressive symptoms were associated with higher alcohol use expectancies on the global AFEXS scale [29, 31, 51, 69]. Affective depressive symptoms were also significantly associated with release of inhibition and negative alcohol use expectancies sub-scales but not the sex-related expectancies sub-scale. While the unadjusted models showed positive associations between alcohol use expectancies and unhealthy alcohol use, these associations were not statistically significant in adjusted models. In secondary analyses, we found that higher alcohol use expectancies were significantly associated with increased odds of consuming three or more drinks per drinking occasion. The positive association between alcohol use expectancies and quantity of alcohol consumption is consistent with the alcohol use expectancy theory and other motivational models of alcohol use, which assert that individuals with high alcohol use expectancies are more likely to have severe alcohol use [22, 27, 30, 70]. The absence of a strong association between alcohol use expectancies and our primary outcome (unhealthy alcohol use) but a significant association between alcohol use expectancies and quantity of alcohol consumed may suggest that the patterns of alcohol use such as frequency and quantity of alcohol use (and not merely presence of alcohol) may be more relevant to understanding the relationship between alcohol use expectancies and depression among PLWH. These patterns of alcohol use are not captured in our primary measure of unhealthy alcohol use. As such, future studies should explore a range of measures capturing patterns of alcohol use (e.g., quantity of use, binge drinking) in order to fully explicate these relationships.

Depression as a Vulnerability Factor

Negative affective states (e.g., depression, dysthymia, and suicidality), which have been widely documented among PLWH [10–12, 14], have been shown to play an important role in the relationship between alcohol use expectancies and unhealthy alcohol use. Our analyses indicate that PLWH with affective depressive symptoms have higher alcohol use expectancies. Additionally, we found that higher alcohol use expectancies increased the odds of consuming three or more drinks per drinking occasion. This finding is consistent with both theoretical (i.e. motivational) models of alcohol use and empirical studies, which suggest that individuals may consume alcohol with the expectation that it will relieve tension, stress and negative affective states [20, 22–25]. However, because we did not find an association between alcohol use expectancies and unhealthy alcohol use, the ultimate impact of the relationship of depression and expectancies is unclear. This finding may suggest that other unmeasured behavioral factors may be more salient in shaping unhealthy alcohol use among HIV-infected persons in Uganda.

Gender as a Vulnerability Factor

Our findings revealed unique patterns of vulnerability: females with affective depressive symptoms had higher alcohol use expectancies. Biological and socio-cultural factors contribute to gendered patterns in alcohol use disorders. Generally, men tend to have higher rates of alcohol use disorders compared to women [43, 46, 47, 71, 72]. Consistent with the literature, we found that unhealthy alcohol use was higher among males compared to females [43, 46, 47, 71, 72]. However, alcohol use expectancies may be higher among women with depressive symptomatology. This finding suggests women with depressive symptomatology may be more vulnerable to unhealthy alcohol use due to higher alcohol use expectancies. Biological factors (e.g., alcohol pharmacokinetics, genetic susceptibility, impulsivity and risk-taking propensity) and socio-cultural factors (e.g., gender norms that promote male alcohol use; social sanctions against female alcohol use) play an important role in gender differences in alcohol use [43, 46, 47]. However, changing patterns in women’s alcohol use behavior have eroded the gender gap in alcohol use with a younger age at onset, and more rapid progression towards alcohol use disorders [40, 73–75]. Given that women are almost twice as likely as men to have depression [76], the association between depressive symptomatology and alcohol use expectancies highlights is a potential pathway for higher alcohol-related adverse outcomes among women with co-morbidities such as depression.

Conclusion

We examined the socio-demographic and psychosocial correlates of alcohol use expectancies and the relationship between depression, alcohol use expectancies and unhealthy alcohol use among PLWH in Uganda. We found that affective depressive symptoms were positively associated with alcohol use expectancies and depression was more strongly associated with alcohol use expectancies among women. While alcohol use expectancies were not associated with an overall measure of unhealthy alcohol use (i.e., composite measure of PEth and AUDIT-C), we found that higher alcohol use expectancies increased the odds of consuming three or more drinks per drinking occasion. These findings are consistent with alcohol use expectancies theories and other motivational models of alcohol use. Our findings are not without limitations. First, the study was limited to persons who self-disclosed alcohol use or were suspected alcohol consumers in the prior year at their initial clinic visit; therefore, alcohol users who denied alcohol use were omitted, which could introduce bias in the study estimates. Additionally, our findings are based on patients enrolling into HIV care and may not be generalizable across time because as these patients initiate anti-retroviral therapy, they also receive psychosocial support and health education on potential impact of alcohol on HIV care outcomes. Nonetheless, these supports could be explored as intervention support to mitigate the impact of alcohol use on anti-retroviral therapy.

All together, these findings provide useful insights on potential drivers of unhealthy alcohol use among HIV-infected persons – specifically, the potential role of mental health problems such as depression in shaping alcohol use expectancies. From a preventive and clinical perspective, the findings suggest the importance of addressing mental well-being in the context of HIV treatment programs. More specifically, there is a need to strengthen screening for depression and other mental health problems, increase access to mental health services, and to develop psychosocial interventions to address mental health problems and alcohol use among PLWH. In Uganda, HIV treatment clinical guidelines have strengthened screening and referral for mental health problems among PLWH. However, these efforts are curtailed by the limited access to mental health professionals, cultural attributions of mental health problems, and the stigma associated with mental health [77]. Psychosocial interventions delivered by lay health workers in with HIV clinics can expand access to mental health services and target alcohol use among PLWH [78–81]. These interventions are also less likely to be hampered by the stigma attached to formal mental health institutions, could reduce other health service access barriers such as time and cost of seeking mental health services outside of HIV clinics, and allow for better follow-up of patients as reports are directly integrated into patient’s clinic records. Lastly, these findings underscore the need for longitudinal studies to examine the relationships between depression, alcohol use expectancies and unhealthy alcohol use among PLWH. Such understanding is necessary to develop effective interventions to mitigate the impact of unhealthy alcohol use on HIV care outcomes in Uganda.

Data Availability

Data are available from the Principal investigator (JAH) upon request.