Abstract
Introduction:
Self-report is widely used to assess alcohol use in research and clinical practice, but may be subject to social desirabilitybias. We aimed to determine if social desirability impacts self-reported alcohol use.
Methods:
Among 751 HIV-infected patients from a clinic in Southwestern Uganda,we measured social desirability using the Marlowe-Crowne Social Desirability Scale (SDS) Short Form C, self-reported alcohol use (prior 3 months) AUDIT-C, and phosphatidylethanol (PEth), a biomarker of prior 3 weeks’ drinking. We conducted multiple regression analyses to assess the relationship between SDS score (low, medium, and high levels) and (1) any self-reported recent alcohol use, among those who were PEth-positive (≥8ng/ml), and (2) continuous AUDIT-C score, among those reporting any recent alcohol use. We controlled for PEth level, age, gender, education, economic assets, marital status, religion, spirituality/religiosity, social support, and study cohort.
Results:
Of 751 participants, 59% were women; the median age was 31 years (interquartile range [IQR]: 26–39). Median SDS score was 9 (IQR: 4–10). Two-thirds (62%) self-reported any recent alcohol use; median AUDIT-C was 1(IQR: 0–4). Among those who were PEth-positive (57%), 13% reported no recent alcohol use. Those with the highest SDS tertile had decreased odds of reporting any recent alcohol use compared to the lowest tertile, but the association did not reach statistical significance in multivariable analyses (adjusted odds ratio 0.55 [95% Confidence Interval (CI): 0.25,1.23]). Among participants self-reporting recent alcohol use, SDS level was negatively associated with AUDIT-C scores (adjusted β: −0.70 [95% CI: −1.19, −0.21] for medium versus low SDS and −1.42 [95% CI: −2.05, −0.78] for high versus low SDS).
Conclusions:
While use of objective measures (e.g. alcohol biomarkers) is desirable for measuring alcohol use, SDS scores may be used to adjust self-reported drinking levels by participants’ level of social desirability in HIV research studies.
Keywords: Phosphatidylethanol, self-report, social desirability, alcohol biomarker
Introduction:
HIV/AIDS and alcohol use are significant causes of morbidity and mortality in Sub-Saharan Africa (Williams et al., 2016).The deleterious relationship between alcohol use and HIV has been demonstrated in many studies (Hendershot et al., 2009; Fatch et al., 2013; Kalichman et al., 2013; Vagenas et al., 2015; Scott-Sheldon et al., 2016). Alcohol use not only affects HIV disease progression through poor antiretroviral therapy (ART) adherence (Kahler et al., 2017), but is also related to the transmission of HIV/AIDS (Woolf-King et al., 2013; Williams et al., 2016).However, alcohol use isa potentially modifiable behaviour, and several interventions exist to reduce such use. The early detection of unhealthy alcohol use is important in order to be able to initiate alcohol use reduction interventions. Early detection and accurate quantification of alcohol use is also useful in order to mitigate some of the adverse outcomes of unhealthy alcohol use in HIV care (e.g., alcohol use combined with hepatotoxic drugs, or poor ART adherence).
HIV positive persons in sub-Saharan Africa have been shown to under-report how much alcohol they consume (Bajunirwe et al., 2014; Muyindike et al., 2017). In Uganda, HIV treatment guidelines advise against concomitant alcohol use and ART (Ministry of Health UGANDA, 2016).Therefore, some HIV clinics in Uganda conduct group health education sessions that focus on the negative effects of alcohol use on HIV disease and its interaction with ART (Wandera et al., 2017); such sessions may lead to the perceived need to minimize reporting of alcohol use. However, despite this, HIV clinics, as well as a broad range of medical settings, and most health research studies, rely on self-report to determine whether patients consume alcohol, and if so, how much.
Self-report is rapid and inexpensive; however, it may be subject to several biases, including social desirability bias (Davis, Thake and Vilhena, 2010; Vu et al., 2011). Social desirability bias is the tendency to respond to questions in amanner that is likely to be most socially sanctioned,especially when reporting socially or culturally ‘unacceptable’ habits or behaviours. This may lead to under-reporting of such behaviours, which in turn may ultimately translate into missed opportunities for intervention and missed contraindications for medications (Copeland et al., 1977; Richman et al., 1999).
To account for the existence of social desirabilitybias in research, the Marlowe-Crowne Social Desirability Scale (SDS) and its variations have been used to measure social desirability worldwide, including in sub-Saharan Africa (Vu et al., 2011). Using versions of this scale, prior studies have shown associations between the level of social desirability and self-reported alcohol use (Davis, Thake and Vilhena, 2010; Latkin et al., 2017). However, these studies did not include an objective measure of alcohol use, and thus associations are difficult to interpret due to the possibility of inaccurate alcohol use measurement by self-report.
Phosphatidylethanol (PEth) is a highly sensitive and specific biomarker that is correlated with the amount of alcohol consumed during the previous 2–3 weeks (Wurst et al., 2015). PEth has highlighted differences between self-report and alcohol metabolites in several studies assessing alcohol use amongst HIV positive persons in Africa (Hahn et al., 2015; Papas et al., 2016; Muyindike et al., 2017; Magidson et al., 2018) and has also been correlated with level of alcohol use (Hahn, Anton and Javors, 2016). Thus it is a useful tool for objectively measuring any and level of alcohol use, and can serve to help determine whether social desirability is a factor in alcohol use reporting.
This study sought to investigate the relationship between the SDS level and 1) any self-reported alcohol use (yes/no), and 2) level of self-reported alcohol use, while controlling for PEth level, among persons with HIV in Uganda. We also explored whether age and gender modified the relationship between social desirability and self-reported alcohol use, because previous studies have shown that age (Muyindike et al., 2017) and gender (Bajunirwe et al., 2014) may affect self-reported alcohol use.
Material and Methods:
For this analysis, we analysed baseline data from two longitudinal cohorts of adults living with HIV recruited from the Mbarara Regional Referral Hospital’s (MRRH) Immune Suppression Syndrome (ISS) clinic in south-western Uganda. The studies are part of the Uganda-Russia-Boston Alcohol Network for Alcohol Research Collaboration on HIV/AIDS (URBAN ARCH) Consortium and are described in detail elsewhere (Hahn et al., 2015, 2018). ISS clinic patients were eligible for the Alcohol Drinking Effects Prior to Treatment (ADEPT) study from 2011 through 2014 if they were ≥18 years, fluent in English or Runyankole (the local language), lived within 60 kilometres of the clinic, and were not yet eligible to initiate HAART. Recruitment was targeted to include approximately 50% unhealthy drinkers (defined after study enrolment as Alcohol Use Disorders Identification Test – Consumption (AUDIT-C) score of ≥3 for women and ≥4 for men, or PEth value ≥50 ng/ml) (Hahn et al., 2018). The Biomarker Research on Ethanol Among Those with HIV (BREATH) study was a cohort study examining changes in alcohol use during the first year of HIV/AIDS care at the MRRH ISS Clinic, conducted from 2011 to 2014. BREATH enrolment criteria were similar to those in ADEPT, but additionally included: new to HIV care, self-reported alcohol use in the previous year at clinic enrolment, and no restrictions on HAART eligibility. BREATH participants were randomly assigned to participate in a main cohort study arm (followed for one year, with quarterly study visits), or a minimally assessed study arm. Those in the minimally assessed arm were interviewed only once at 6 months after enrolment; the purpose of this arm was to examine assessment reactivity (Emenyonu et al., 2017).
Approvals from the institutional review boards of the Mbarara University of Science and Technology (MUST), the University of California, San Francisco (UCSF), and Boston University were obtained, as well as the Uganda National Council for Science and Technology. Participants provided written informed consent prior to recruitment into both studies. Participants were informed at enrolment during the consenting process that their blood would be used to test for alcohol that was consumed in the last 2–3 weeks and also that a breathalyser test would be performed as part of the study procedures.
Measurements
At the study visit, participants completed an interviewer-administered structured questionnaire in English or Runyankole, and breath alcohol tests and blood draws for the following laboratory tests were performed.
Laboratory testing
Venous blood samples were collected; CD4 cell count was tested at the MUST Clinical Research Laboratory, and viral load was tested at the UCSF Virology Core Laboratory in San Francisco, CA. Dried blood spot (DBS) cards were also prepared from the venous blood draw. DBS testing for phosphatidylethanol (PEth), a biomarker of alcohol use, was conducted using liquid chromatography and tandem mass spectrometry (LC-MS/MS) (Jones et al., 2011) at the United States Drug Testing Laboratory in Des Plaines, Illinois. A PEth result of ≥ 8ng/ml was taken as positive.
Dependent/outcomevariables
We examined two measures of self-reported alcohol use in the prior three months as the primary outcome variables.These measures were any self-reported alcohol use (yes/no) in the prior three months (outcome 1), and level of alcohol use, as a continuous measure, using the Alcohol Use Disorders Identification Test – Consumption (AUDIT-C) (Bush et al., 1998) (outcome 2). We modified the AUDIT-C to ask about a reference period of the prior 3 months.
Independent variable
The primary explanatory variable of interest was social desirability, measured using the Marlowe-Crowne Social Desirability Scale Short Form C. The Marlowe-Crowne Social Desirability Scale Short Form C is a 13-item instrument used to assess a participant’s need for social approval (Reynolds, 1978). The higher the score, the more the participants’ demonstrated need for social approval. Some of the items in the scale include: “I’m always willing to admit it when I make a mistake”; “I am always courteous, even to people who are disagreeable”; “I have never been annoyed when people expressed ideas very different from my own”; “I have never deliberately said something that hurt someone’s feelings”. A 28-item version of the scale has been tested in the Ugandan setting with good reliability, α=0.7 (Vu et al., 2011). The Kuder-Richardson reliability coefficient (Kuder and Richardson, 1937) of the SDS Short Form C in our sample was 0.44. We graphically examined the relationship between SDS score and continuous AUDIT-C; the relationship did not appear linear. Therefore, SDS was analysed as a 3-level variable based on tertiles for all analyses.
Covariates
Demographic characteristics collected during the baseline study interview included participant gender, age, education, marital status, and religion. We created a household asset index using principal components analysis to assess socioeconomic status (Filmer, D ; Pritchett, 2001). The asset index was based on ownership of durable goods, household quality, and available energy sources; the bottom 40% was considered low, 41to 80% middle, and the top 20% high. We measured social support using a modified 11 item version of the Duke University-University of North Carolina social support scale (Broadhead et al., 1988), with a mean score of <3 indicating low perceived levels of social support (Antelman et al., 2001). We measured spirituality and religiosity using the short version of the Ironson-Woods Spirituality and Religiosity Index (SRI) (Ironson et al., 2002) and included it as a continuous variable, with higher scores indicating higher levels of spirituality and religiousness.
Statistical analysis
We described the characteristics of the overall sample at baseline using proportions for categorical variables, and medians and inter-quartile ranges (IQR) for continuous variables. We limited our analyses of outcome 1 to those who were PEth-positive (≥8 ng/ml) because we were reasonably sure that they were consuming alcohol, given the high specificity of PEth (Hahn, Anton and Javors, 2016). We conducted logistic regression models controlling for potential confounders to assess the association between SDS level and any self-reported alcohol use. For the outcome, level of self-reported alcohol use (outcome 2), we limited the analysesto participants who were self-reported current drinkers (AUDIT-C >0). We limited our analyses of outcome 2 to those who self-reported any alcohol use because we hypothesized that social desirability may be related to the level of reported alcohol use, among those who report engaging in any alcohol use. We conducted linear regression models controlling for potential confounders to assess the association between the SDS tertiles and AUDIT-C scores, among self-reported drinkers. Lastly, we conducted exploratory regression analyses assessing whether age (dichotomized as <35 and ≥35; we chose a cut off of 35 as it was roughly the median age, gender, and PEth level (<50 ng/ml versus ≥50 ng/ml) were possible effect modifiers. This was evaluated by testing separate 2-way interactions between SDS level and each potential effect modifier. We selected the following covariates a priori for inclusion in the above multivariable models: gender, age, education, household asset index, marital status, religion, social support, and spirituality/religiosity index score; these variables have been frequently associated with alcohol use (Crum, Helzer and Anthony, 1993; Ironson et al., 2002; Mavandadi et al., 2015; Adong et al., 2018). We included an indicator variable for whether the subject was a participant in ADEPT, BREATH main cohort arm or BREATH minimally assessed arm. We also included PEth as a continuous covariate in the models, to allow us to examine the relationships between SDS tertile and self-reported alcohol use, controlling for the objective level of alcohol consumption. Approximately ten percent of observations were missing data for at least one variable of interest. To account for missing data, we used multiple imputation via chained equations to impute missing data (Sterne et al., 2009). All analyses were conducted using Stata version 14.2.
Results
The analysis included 751 participants; 59% were women with a median age of 31 years (IQR: 26–39). The median SDSscore was 9 (out of a possible 13, IQR: 4–10). Forty-two percent of participants were in the low (score 1–8) SDS group, 39% were in the medium (score 9–10) group, and 20% were in the high (score 11–13) group. Sixty-two percent (n=462) self-reported any alcohol use in the past 3 months.The sample median AUDIT-C score was 1 (IQR: 0–4.). Fifty-seven percent (n=430) of the participants were PEth-positive (PEth≥8 ng/ml).
Among those who were PEth-positive (n=430), 13% self-reported no alcohol use in the last 3 months; 9% (17/186) of those in the lowest SDS tertile reported no alcohol use in the last 3 months, 14% (21/149) of those in the middle SDS tertile reported no alcohol use,in the last 3 months and 19% (17/89) of those in the highest SDS tertile reported no alcohol use in the last 3 months. In bivariate analysis among participants who were PEth-positive, those with the highest SDS tertile had decreased odds of reporting any recent alcohol use, compared to the lowest tertile[odds ratio (OR): 0.43 (95% confidence interval [CI]: 0.21 to 0.88] (Table 2). However, the association was attenuated and not statistically significant in the adjusted analyses (adjusted OR [aOR] 0.72 [95% CI: 0.35 to 1.48] and 0.55 [95%CI: 0.25 to 1.23] for the medium and high SDS tertiles compared to the lowest, respectively).
Table 2.
Unadjusted and adjusted odds ratios (OR) and 95% confidence intervals (CI) for any self-reported alcohol use in the prior 3 months at baseline, among participants who are PEth-positive ([≥8 ng/mL] - “PEth-confirmed drinkers”). (n = 430).
| Unadjusted OR (95% CI) | p-value | Adjusted OR (95% CI) | p-value | |
|---|---|---|---|---|
| Social desirability score | 0.07 | 0.34 | ||
| Low (1–8) | 1.00 | 1.00 | ||
| Medium (9–10) | 0.60 (0.31, 1.19) | 0.72 (0.35, 1.48) | ||
| High (11–13) | 0.43 (0.21, 0.88) | 0.55 (0.25, 1.23) | ||
| PEth (per 50 ng/ml) | 1.03 (0.99, 1.08) | 0.18 | 1.05 (1.00, 1.11) | 0.06 |
| Gender | 0.43 | 0.96 | ||
| Women | 1.00 | 1.00 | ||
| Men | 0.79 (0.45, 1.41) | 0.98 (0.50, 1.94) | ||
| Age (years) | 0.98 (0.95, 1.01) | 0.13 | 0.98 (0.94, 1.01) | 0.18 |
| Education | 0.13 | 0.23 | ||
| Less than secondary | 1.00 | 1.00 | ||
| Secondary or more | 1.68 (0.85, 3.30) | 1.63 (0.73, 3.63) | ||
| Household asset index | 0.06 | 0.15 | ||
| Low | 1.00 | 1.00 | ||
| Middle | 2.18 (1.12, 4.21) | 2.02 (1.00, 4.11) | ||
| High | 1.65 (0.77, 3.55) | 1.32 (0.52, 3.31) | ||
| Marital status | 0.05 | 0.06 | ||
| Not married | 1.00 | 1.00 | ||
| Married | 0.57 (0.32, 1.01) | 0.51 (0.25, 1.03) | ||
| Religion | 0.51 | 0.24 | ||
| Protestant | 1.00 | 1.00 | ||
| Catholic | 0.81 (0.45, 1.47) | 0.71 (0.38, 1.34) | ||
| Muslim | 0.42 (0.13, 1.37) | 0.27 (0.07, 1.00) | ||
| Other | 0.61 (0.07, 5.53) | 0.50 (0.05, 5.23) | ||
| Low social support? | 0.88 | 0.57 | ||
| No | 1.00 | 1.00 | ||
| Yes | 0.95 (0.48, 1.88) | 0.79 (0.35, 1.77) | ||
| Overall SRI score | 0.99 (0.96, 1.02) | 0.46 | 1.00 (0.96, 1.03) | 0.78 |
| Study | <0.01 | <0.01 | ||
| BREATH cohort | 1.00 | 1.00 | ||
| BREATH comparison | 0.27 (0.12, 0.60) | 0.27 (0.11, 0.62) | ||
| ADEPT cohort | 0.53 (0.24, 1.15) | 0.56 (0.24, 1.31) |
Among participants self-reporting any recent alcohol use, higher SDS tertile was significantly associatedwith lower levels of self-reported alcohol use in both bivariate and multivariable analyses. The regression coefficients for the multivariable model of AUDIT-C score were as follows: adjusted β: −0.70 (95%CI: −1.19 to −0.21) and −1.42 (95%CI: −2.05 to −0.78) for those with medium and high versus low SDS tertiles, respectively (Table 3).
Table 3.
Unadjusted and adjusted linear regression coefficients (β) and 95% confidence intervals (CI) for AUDIT-C score for the prior 3 months, among participants who self-reported any alcohol use in the prior 3 months (n = 462).
| Unadjusted β (95% CI) | p-value | Adjusted β (95% CI) | p-value | |
|---|---|---|---|---|
| Social desirability score | <0.01 | <0.01 | ||
| Low (1–8) | ref | ref | ||
| Medium (9–10) | −0.78 (−1.30, −0.25) | −0.70 (−1.19, −0.21) | ||
| High (11–13) | −1.23 (−1.88, −0.57) | −1.42 (−2.05, −0.78) | ||
| PEth (per 50 ng/ml) | 0.10 (0.08, 0.13) | <0.01 | 0.09 (0.06, 0.11) | <0.01 |
| Gender | <0.01 | <0.01 | ||
| Women | ref | ref | ||
| Men | 1.16 (0.70, 1.62) | 1.40 (0.89, 1.90) | ||
| Age (years) | 0.03 (0.01, 0.06) | 0.01 | 0.01 (−0.02, 0.03) | 0.67 |
| Education | 0.83 | 0.54 | ||
| Less than secondary | ref | ref | ||
| Secondary or more | −0.05 (−0.55, 0.44) | −0.16 (−0.66, 0.35) | ||
| Household asset index | 0.50 | 0.30 | ||
| Low | ref | ref | ||
| Middle | 0.24 (−0.30, 0.78) | 0.35 (−0.15, 0.85) | ||
| High | 0.36 (−0.29, 1.01) | 0.43 (−0.24, 1.09) | ||
| Marital status | 0.27 | <0.01 | ||
| Not married | ref | ref | ||
| Married | −0.27 (−0.74, 0.21) | −0.82 (−1.30, −0.34) | ||
| Religion | 0.34 | 0.30 | ||
| Protestant | ref | ref | ||
| Catholic | −0.39 (−0.88, 0.10) | −0.41 (−0.85, 0.04) | ||
| Muslim | 0.43 (−1.02, 1.89) | 0.06 (−1.26, 1.37) | ||
| Other | 0.37 (−1.74, 2.48) | 0.36 (−1.55, 2.27) | ||
| Low social support? | 0.77 | 0.95 | ||
| No | ref | ref | ||
| Yes | 0.09 (−0.49, 0.66) | −0.02 (−0.58, 0.54) | ||
| Overall SRI score | −0.00 (−0.03, 0.02) | 0.77 | −0.01 (−0.03, 0.02) | 0.61 |
| Study | 0.01 | 0.05 | ||
| BREATH cohort | ref | ref | ||
| BREATH comparison | −0.70 (−1.38, −0.03) | −0.50 (−1.12, 0.12) | ||
| ADEPT | 0.27 (−0.26, 0.80) | 0.33 (−0.21, 0.87) |
The overall effect modification by genderand age of the relationship between social desirability and either any alcohol use (p-values 0.16 and 0.26) or AUDIT-C (p-values 0.10 and 0.82) did not reach statistical significance (Tables 4 and 5). However, stratified analyses conducted for descriptive purposes,suggested that high social desirability may be associated with lower odds of any self-reported alcohol use among women but not men. Also, both medium and high social desirability may be associated with lower odds of any self-reported alcohol use among those <35 but not those ≥35 (Table 4).The analyses of effect modification by PEth level did not reach statistical significance for either outcome (outcome 1, p=0.75; outcome 2, p=0.76, data not shown).
Table 4.
Adjusted odds ratios (AOR) and 95% confidence intervals (CI) for any self-reported alcohol use in the prior 3 months at baseline (outcome 1), among participants who are PEth-positive (≥8 ng/mL, “PEth-confirmed drinkers”). Models stratified by gender and age.
| Stratified by gender* p-value for interaction=0.16 |
Stratified by age** p-value for interaction=0.26 |
|||||||
|---|---|---|---|---|---|---|---|---|
| Women | Men | Age < 35 | Age >=35 | |||||
| AOR (95% CI) | p-value | AOR (95% CI) | p-value | AOR (95% CI) | p-value | AOR (95% CI) | p-value | |
| Social desirability score | 0.09 | 0.76 | 0.09 | 0.98 | ||||
| Low (1–8) | 1.00 | 1.00 | 1.00 | 1.00 | ||||
| Medium (9–10) | 0.69 (0.21, 2.30) | 0.72 (0.28, 0.85) | 0.33 (0.10, 1.06) | 1.10 (0.39, 3.12) | ||||
| High (11–13) | 0.23 (0.06, 0.87) | 0.95 (0.32, 2.81) | 0.27 (0.07, 0.95) | 1.04 (0.32, 3.32) | ||||
stratified models adjusted for: PEth, age, education, economic assets, marital status, religion, spirituality/religiosity, social support, and study.
stratified models adjusted for: PEth, gender, education, economic assets, marital status, religion, spirituality/religiosity, social support, and study.
Table 5.
Adjusted linear regression coefficients (β) and 95% confidence intervals (CI) for self-reported recent alcohol use (AUDIT-C, outcome 2), among participants who self-reported any alcohol use in the prior 3 months. Models stratified by gender and age.
| Stratified by gender* p-value for interaction=0.10 |
Stratified by age** p-value for interaction=0.82 |
|||||||
|---|---|---|---|---|---|---|---|---|
| Women | Men | Age < 35 | Age >=35 | |||||
| Adjusted β (95% CI) | p-value | Adjusted β (95% CI) | p-value | Adjusted β (95% CI) | p-value | Adjusted β (95% CI) | p-value | |
| Social desirability score | <0.01 | <0.01 | <0.01 | <0.01 | ||||
| Low (1–8) | ref | ref | ref | ref | ||||
| Medium (9–10) | −1.11 (−1.70, −0.52) | −0.23 (−1.04, 0.58) | −0.61 (−1.22, 0.00) | −0.93 (−1.77, −0.10) | ||||
| High (11–13) | −1.15 (−2.06, −0.24) | −1.44 (−2.39, −0.49) | −1.38 (−2.18, −0.58) | −1.65 (−2.72, −0.58) | ||||
stratified models adjusted for: PEth, age, education, economic assets, marital status, religion, spirituality/religiosity, social support, and study.
stratified models adjusted for: PEth, gender, education, economic assets, marital status, religion, spirituality/religiosity, social support, and study.
Discussion:
We evaluated the associations of social desirability with self-reported alcohol use among HIV positive participants in Uganda.The results of the analysis that were limited to those who were PEth-positive (i.e. those we were reasonably sure were consuming alcohol) showed that those with the highest SDS scores had decreased odds of reporting any alcohol use compared to those with the lowest SDS scores, but this association was not statistically significant in analyses adjusting for several variables, including the PEth biomarker. The results revealed a more robust association for social desirability with levels of drinking. Among participants who reported any alcohol use in the last 3 months, higher SDS scores were significantly associated with lower levels of self-reported alcohol using the AUDIT-C, by an average of 0.7 AUDIT-C points for the middle tertile of SDS scores, and 1.4 AUDIT-C points for the highest tertile of SDS scores (adjusted analysis). These findings together suggest that social desirability may be associated with how much alcohol useis reported among those who admit to drinking alcohol, but in our study where participants were aware of biomarker measurement, social desirability was not associated with whether any alcohol use was reported. Thus, for future studies measuring level of alcohol use by self-report, additionally adjusting for social desirability analytically may improve the validity of research conclusions. Also, measures that reduce social desirability during self-report may be useful in order to improve alcohol use quantification. However, measuring and/or reducing social desirability may not have an impact on determining whether any recent alcohol use has occurred. Thus objective measures and/or alcohol biomarkers or other interventions that encourage reporting,suchas a bogus pipeline, in which the threat of biologically validating a reported behaviour encourages truthful reporting (Adams et al., 2008), may be needed.
We conducted exploratory analyses of effect modification by gender and by age group. While we found no statistically significant effect modification overall, stratified analyses suggested that women and those <35 years old who had the highest tertile SDS scores had lower odds of self-reporting any alcohol use compared to those in the lower tertile of SDS scores. The direction of this association was similar (although attenuated and not significant) among men. Among those >=35 years, no associations between SDS tertiles and self-reporting any alcohol use were detected.These findings among women and younger participants are consistent with the higher level of stigma associated with alcohol use for women compared to men in Uganda (Kabwama et al., 2016), and possible increased impression management among younger persons (Davis, Thake and Vilhena, 2010). The study findings suggest that effects of social desirability on alcohol use reporting may vary by subgroup; however, further research is needed to explore gender and age differences in how alcohol use is reported.
Several study limitations merit discussion. First, these findings may be specific to the setting of HIV care, and/or the setting of semi-rural Uganda. Social desirability may be more salient in this setting. Alcohol use among persons living with HIVin Uganda is advised against (Ministry of Health UGANDA, 2016), and clinic-based group health education includes focus on the negative effects of alcohol use on HIV disease and its interaction with ART, which may lead to the perceived need to minimize reporting of alcohol use. Previous studies using PEth have shown that many persons with HIV in Uganda and elsewhere under-report alcohol consumption (Bajunirwe et al., 2014; Asiimwe et al., 2015; Papas et al., 2016; Magidson et al., 2018), particularly to clinic health care providers (Muyindike et al., 2017). These data were collected prior to 2016, when ART initiation became universal upon clinic entry in Uganda, after the implementation of the World Health Organization HIV treatment guidelines (Ministry of Health Uganda, 2016). Thus, individuals presenting to the clinic could have perceived that self-reported alcohol use might impact the clinical recommendation for ART prescription.
Participants were informed during the consent process that a breathalyser test would be performed and also that blood samples collected would be used to test for prior 2–3 weeks of alcohol use. This may have resulted in higher levels of reporting of alcohol use than would have been found in settings without breathalysers or biomarker specimen collection (Hahn et al., 2012). This may explain the relatively low levels of denial of any alcohol use (outcome 1) among those testing PEth-positive (13%), and the lack of significant association of social desirability with reporting any alcohol use.
A further limitation was the poor internal consistency of the 13-item social desirability scale we used in our study (Cronbach’s alpha = 0.44). Although we conducted cognitive interviewing (Gordon B, 2005) for the scale prior to launching the study, which confirmed good understanding of the scale items, internal consistency analysis could only be conducted after data collection. The low reliability coefficient suggests we may not have been adequately measuring the general construct of social desirability. A longer 28-item version of the scale has been used in this setting with better reliability (alpha=0.70) (Vu et al., 2011), suggesting that a scale with more items may be needed in future studies. Psychometric validation of self-report scales is an on-going research need in this setting, and a more extensive evaluation of the SDS may be required to confirm the findings we report here.
A strength of the study was its use of a sensitive and specific biomarker of alcohol use to include an objective measure of alcohol consumption, PEth. This enabled us to compareself-reported alcohol use among participants with likely similar alcohol consumption levels.
In conclusion, social desirability was not significantly associated with any recent self-reported alcohol use overall, although stratified analyses suggested that it might be a predictor of denying alcohol use among women and younger patients, with the magnitude of these associations appearing strongest among these groups in descriptive, exploratory analyses. Among those who self-reported any recent alcohol use, higher social desirability was associated with lower levels of self-reported alcohol use. Thus, social desirability may have a role in minimizing reporting theamount of alcohol consumed. This suggests some possible pathways for improving measurement of alcohol use. When possible, objective measures of alcohol use should be used. In situations where biomarkers or other objective measures are not feasible, methods are needed to improve the accuracy of alcohol self-report. The design of alcohol assessment tools or interview methods that minimise social desirability may be helpful. These may include the use of self-administered surveys, which have been shown to increase the rate of reporting of stigmatized behaviours (Anton, 1985); these surveys may possibly be embedded within larger health surveys to reduce socially desirable reporting. Bogus pipelines such as the use of breathalysers may also be considered. In the HIV research setting, implementing social desirability scales may enable examination of their use in analyses, serving to adjust drinking levels among those who do report any recent drinking by their level of social desirability.
Table 1.
Participant characteristics in two studies of alcohol use by persons with HIV in southwestern Uganda (n = 751).
| Median(Interquartile range) | ||
|---|---|---|
| N (%) | - | |
| Gender | ||
| Men | 309 (41.2) | - |
| Women | 442 (58.9) | - |
| Age(years) | - | 31(26–39) |
| Education | ||
| Less than secondary | 504 (67.1) | - |
| Secondary or more | 247 (32.9) | - |
| Household asset index | ||
| Low | 295 (39.3) | - |
| Middle | 303 (40.4) | - |
| High | 152 (20.3) | - |
| Marital status | ||
| Married | 383 (51.0) | - |
| Not married | 368 (49.0) | - |
| Religion | ||
| Protestant | 393 (52.3) | - |
| Catholic | 276 (36.8) | - |
| Muslim | 48 (6.4) | |
| Other | 34 (4.5) | - |
| Social desirability scale | - | 9(4–10) |
| Social desirability scale | ||
| Low (1–8) | 309 (41.6) | - |
| Medium (9–10) | 288 (38.8) | - |
| High (11–13) | 146 (19.7) | - |
| Low social support | ||
| Yes | 173 (23.2) | - |
| No | 573 (76.8) | - |
| Any alcohol use, prior 3 months (self-report) | ||
| Yes | 462 (61.8) | - |
| No | 286 (38.2) | - |
| AUDIT-C score (prior 3 months) | - | 1(0–4) |
| PEth ≥ 8 ng/ml | ||
| Yes | 430 (57.4) | |
| No | 319 (42.6) | |
| PEth result (ng/ml) among all | - | 18.8(1–141.8) |
| SRI score (overall) | - | 97(88–107) |
| Study | ||
| BREATH cohort | 205 (27.3) | - |
| BREATH comparison | 141 (18.8) | - |
| ADEPT cohort | 405 (53.9) | - |
Acknowledgments
Funding for this project was provided by the National Institutes for Health (NIH): NIH U01 AA20776, R01 AA018631, U24 AA020778, and U24 AA020779 and K24 AA022586
Footnotes
Conflicts of interest: None declared.
References:
- Adams J, Parkinson L, Sanson-Fisher R. and Walsh R. (2008) ‘Enhancing self-report of adolescent smoking: The effects of bogus pipeline and anonymity.’, Addictive Behaviors, 33(10), pp. 1291–1296. [DOI] [PubMed] [Google Scholar]
- Adong J, Lindan C, Fatch R, Emenyonu N, Muyindike W, Ngabirano C, Winter M, Lloyd-Travaglini C, Samet J, Cheng D. and Hahn J. (2018) ‘The Relationship Between Spirituality/Religiousness and Unhealthy Alcohol Use Among HIV-Infected Adults in Southwestern Uganda’, AIDS and Behavior. Springer US, 22(6), pp. 1802–1813. doi: 10.1007/s10461-017-1805-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Antelman G, Smith Fawzi M, Kaaya S, Mbwambo J, Msamanga G, Hunter D. and Fawzi W. (2001) ‘Predictors of HIV-1 serostatus disclosure: a prospective study among HIV-infected pregnant women in Dar es Salaam, Tanzania’, AIDS, 15(14), pp. 1865–1874. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Anton NJ (1985) ‘Methods of coping with social desirability bias: A review’, European Journal of Social Psychology., 15(3). [Google Scholar]
- Asiimwe SB, Fatch R, Emenyonu NI, Muyindike WR, Kekibiina A, Santos GM, Greenfield TK and Hahn JA (2015) ‘Comparison of Traditional and Novel Self-Report Measures to an Alcohol Biomarker for Quantifying Alcohol Consumption Among HIV-Infected Adults in Sub-Saharan Africa’, Alcoholism: Clinical and Experimental Research, 39(8), pp. 1518–1527. doi: 10.1111/acer.12781. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bajunirwe F, Haberer JE, Boum YI, Hunt P, Mocelo R, Martin J. and Hahn J. (2014) ‘Comparison of Self-Reported Alcohol Consumption to Phosphatidylethanol Measurement among HIV-Infected Patients Initiating Antiretroviral Treatment in Southwestern Uganda’, PLos ONe, 9(12), p. e113152. doi: 10.1371/journal.pone.0113152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Broadhead W, Gehlbach S, de Gruy F. and Kaplan B. (1988) ‘The Duke-UNC Functional Social Support Questionnaire. Measurement of social support in family medicine patients.’, Med Care, 26(7), pp. 709–23. [DOI] [PubMed] [Google Scholar]
- Bush K, Kivlahan DR; McDonell M, Fihn S. and Bradley K. (1998) ‘The AUDIT Alcohol Consumption Questions (AUDIT-C) An effective brief screening test for problem drinking.’, Archives of Internal Medicine, 158(16), pp. 1789–1795. doi: 10.1001/archinte.158.16.1789. [DOI] [PubMed] [Google Scholar]
- Copeland K, Checkoway H, McMichael A. and Holbrook R. (1977) ‘Bias due to the misclassification in the estimation of relative risk’, Am J Epidemiology, 105(5), pp. 488–95. [DOI] [PubMed] [Google Scholar]
- Crum R, Helzer J. and Anthony J. (1993) ‘Level of education and alcohol abuse and dependence in adulthood: A further inquiry’, American Journal of Public Health, 83(6), pp. 830–837. doi: 10.2105/AJPH.83.6.830. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Davis CG, Thake J. and Vilhena N. (2010) ‘Social desirability biases in self-reported alcohol consumption and harms’, Addictive Behaviors, 35(4), pp. 302–311. doi: 10.1016/j.addbeh.2009.11.001. [DOI] [PubMed] [Google Scholar]
- Emenyonu N, Fatch R, Muyindike WR; Kekibiina A. and Woolf-king SE; Hahn J. (2017) ‘Randomized Study of Assessment Effects on Alcohol Use by Persons With HIV in Rural Uganda’, Journal of Studies on Alcohol and Drugs, 78(2), pp. 296–305. doi: 10.15288/jsad.2017.78.296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fatch R, Bellow B, Bagenda F, Mulogo E, Weiser S. and Hahn J. (2013) ‘Alcohol Consumption as a Barrier to Prior HIV Testing in a Population-Based Study in Rural Uganda’, AIDS and Behavior, 17(5), pp. 1713–1723. doi: 10.1007/s10461-012-0282-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Filmer D ; Pritchett L. (2001) ‘Estimating wealth effects without expenditure data or tears: an application to education enrollments in states of India.’, Demography, 38(1), pp. 115–32. [DOI] [PubMed] [Google Scholar]
- Gordon B, W. (2005) Cognitive Interviewing A Tool for improving Questionniare design. [Google Scholar]
- Hahn JA, Emenyonu NI, Fatch R, Muyindike WR, Kekiibina A, Carrico AW, Woolf-king S. and Shiboski S. (2015) ‘Declining and rebounding unhealthy alcohol consumption during the first year of HIV care in rural Uganda, using phosphatidylethanol to augment self-report’, Addiction, 111(2), pp. 272–279. doi: 10.1111/add.13173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hahn JA, Fatch R, Kabami J, Mayanja B, Emenyonu NI, Martin J. and Bangsberg DR (2012) ‘Self-report of alcohol use increases when specimens for alcohol biomarkers are collected in persons with HIV in Uganda’, Journal of Acquired Immune Deficiency Syndromes, 61(4), pp. 63–64. doi: 10.1097/QAI.0b013e318267c0f1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hahn J, Anton R. and Javors M. (2016) ‘The Formation, Elimination, Interpretation, and Future Research Needs of Phosphatidylethanol for Research Studies and Clinical Practice’, Alcoholism: Clinical and Experimental Research, 40(11), pp. 2292–2295. doi: 10.1111/acer.13213. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hahn J, Cheng D, Emenyonu N, Lloyd-Travaglini C, Fatch R, Shade S, Ngabirano C, Adong J, Bryant K, Muyindike W. and Samet J. (2018) ‘Alcohol Use and HIV Disease Progression in an Antiretroviral Naive Cohort’, Journal of Acquired Immune Deficiency Syndromes, 77(5), pp. 492–501. doi: 10.1097/QAI.0000000000001624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hendershot CS, Stoner SA, Pantalone DW and Simoni JM (2009) ‘Alcohol use and antiretroviral adherence: Review and metaanalysis’, Journal of acquired immune deficiency syndromes (2009), 52(2)(October 2009), pp. 180–202. doi: 10.1097/QAI.0b013e3181b18b6e.Alcohol. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ironson G, Solomon G, Balbin E, O’Cleirigh C, George A, Kumar M, Larson D. and Woods T. (2002) ‘The Ironson – Woods Spirituality / Religiousness Index Is Associated With Long Survival, Health Behaviors, Less Distress, and Low Cortisol in People With HIV / AIDS’, Ann Behav Med, 24(1), pp. 34–48. [DOI] [PubMed] [Google Scholar]
- Jones J, Jones M, Plate C. and Lewis D. (2011) ‘The detection of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanol in human dried blood spots’, Analytical Methods, 3(5), pp. 1101–1106. doi: 10.1039/c0ay00636j. [DOI] [Google Scholar]
- Kabwama SN, Ndyanabangi S, Mutungi G, Wesonga R, Bahendeka SK and Guwatudde D. (2016) ‘Alcohol use among adults in Uganda: Findings from the countrywide non-communicable diseases risk factor cross-sectional survey’, Global Health Action, 9(1), p. 31302. doi: 10.3402/gha.v9.31302. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kahler C, Lui T, Cioe P, Bryant V, MM P, Kojic E, Onen N, Baker J, Hammer J, Brooks J. and Patel P. (2017) ‘Direct and Indirect Effects of Heavy Alcohol Use on Clinical Outcomes in a Longitudinal Study of HIV Patients on ART.’, AIDS and Behavior, 21(7), pp. 1825–1835. doi: 10.1007/s10461-016-1474-y.Direct. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kalichman SC, Grebler T, Amaral CM, McNerey M, White D, Kalichman MO, Cherry C. and Eaton L. (2013) ‘Intentional Non-Adherence to Medications among HIV Positive Alcohol Drinkers: Prospective Study of Interactive Toxicity Beliefs’, Journal of General Internal Medicine, 28(3), pp. 399–405. doi: 10.1007/s11606-012-2231-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kuder G. and Richardson M. (1937) ‘The theory of the estimation of test reliability.’, Psychometrika, 2(3), pp. 151–160. [Google Scholar]
- Latkin CA, Edwards C, Davey-Rothwell MA and Tobin KE (2017) ‘The relationship between social desirability bias and self-reports of health, substance use, and social network factors among urban substance users in Baltimore, Maryland’, Addictive Behaviors. Elsevier, 73, pp. 133–136. doi: 10.1016/j.addbeh.2017.05.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Magidson J, Fatch R, Orrell C, Amanyire G, Haberer J. and Hahn J. (2018) ‘Biomarker-Measured Unhealthy Alcohol Use in Relation to CD4 Count Among Individuals Starting ART in Sub-Saharan Africa’, AIDS and Behavior. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mavandadi S, Helstrom A, Sayers S. and Oslin D. (2015) ‘The Moderating Role of Perceived Social Support on Alcohol Treatment Outcomes’, Journal of Studies on Alcohol and Drugs, 76(5), pp. 818–822. doi: 10.15288/jsad.2015.76.818. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ministry of Health Uganda (2016) ‘Consolidated Guidelines for Prevention and treatment of HIV in Uganda 2016’, (December). [Google Scholar]
- Muyindike WR, Lloyd-Travaglini C, Fatch R, Emenyonu NI, Adong J, Ngabirano C, Cheng DM, Winter MR, Samet JH and Hahn JA (2017) ‘Phosphatidylethanol confirmed alcohol use among ART-naïve HIV-infected persons who denied consumption in rural Uganda’, AIDS Care - Psychological and Socio-Medical Aspects of AIDS/HIV, 29(11), pp. 1442–1447. doi: 10.1080/09540121.2017.1290209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Papas R, Gakinya B, Mwaniki M, Keter A, Lee H, Loxley M, Klein D, Sidle J, Martino S, Baliddawa J, Schlaudt K. and Maisto S. (2016) ‘Associations Between the Phosphatidylethanol Alcohol Biomarker and Self-reported Alcohol Use in a Sample of HIV Infected Outpatient Drinkers in Western Kenya.’, Alcohol Clin Exp Res., 40(8), pp. 1779–1787. doi: 10.1111/acer.13132.Associations. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Reynolds WM (1978) ‘Development of Reliable and Valid Short Forms of the Marlowe-Crowne Social Desirability Scale’, (1972), pp. 119–126. [Google Scholar]
- Richman WL, Weisband S, Kiesler S. and Drasgow F. (1999) ‘A meta-analytic study of social desirability distortion in computer-administered questionnaires, traditional questionnaires, and interviews’, Journal of Applied Psychology, 84(5), pp. 754–775. doi: 10.1037/0021-9010.84.5.754. [DOI] [Google Scholar]
- Scott-Sheldon L, Carey K, Cunningham K, Johnson B, Carey M. and MASH RT (2016) ‘Alcohol use predicts sexual decision-making: A systematic review and meta-analysis of the experimental literature’, AIDS and Behavior, 20, pp. 19–39. doi: 10.1016/j.biomaterials.2015.05.042.Shifts. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sterne J, White I, Carlin J, Spratt M, Royston P, Kenward M, Wood A. and Carpenter J. (2009) ‘Multiple imputation for missing data in epidemilogical and clinical research: potential and pitfalls’, BMJ, 338(b2393). [DOI] [PMC free article] [PubMed] [Google Scholar]
- Vagenas P, Azar MM, Copenhaver MM, Springer SA, Molina PE and Altice FL (2015) ‘The Impact of Alcohol Use and Related Disorders on the HIV Continuum of Care : a Systematic Review Alcohol and the HIV Continuum of Care’, Curr HIV/AIDS Rep., pp. 421–436. doi: 10.1007/s11904-015-0285-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Vu A, Tran N, Pham K. and Ahmed S. (2011) ‘Reliability of the Marlowe-Crowne social desirability scale in Ethiopia, Kenya, Mozambique, and Uganda’, BMC Medical Research Methodology. BioMed Central Ltd, 11(1), p. 162. doi: 10.1186/1471-2288-11-162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wandera B, Tumwesigye N, Nankabirwa J, Mafigiri D, Parkes-Rantashi R, Kapiga S, Hahn J. and Sethi A. (2017) ‘Efficacy of a single, brief alcohol reduction intervention among men and women living with HIV/AIDS and using alcohol in Kampala, Uganda; a randomized trial.’, J Int Assoc Provid AIDS Care, 16(3), pp. 276–285. doi: 10.1177/2325957416649669.Efficacy. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Williams EC, Hahn JA, Saitz R, Bryant K, Lira MC and Samet JH (2016) ‘Alcohol Use and Human Immunodeficiency Virus (HIV) Infection: Current Knowledge, Implications, and Future Directions’, Alcoholism: Clinical and Experimental Research, 40(10), pp. 2056–2072. doi: 10.1111/acer.13204. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Woolf-King S, Steinmaus C, Reingold A. and Hahn J. (2013) ‘An update on alcohol use and risk of HIV infection in sub-Saharan Africa: Meta-analysis and future research directions’, International journal of alcohol and drug research, 2(1), pp. 99–110. doi: 10.7895/ijadr.v2i1.45. [DOI] [Google Scholar]
- Wurst F, Thon N, Yegles M, Schruck A, Preuss U. and Weinmann W. (2015) ‘Ethanol metabolites: their role in the assessment of alcohol intake.’, Alcohol Clin Exp Res., 39(11), pp. 2060–72. [DOI] [PubMed] [Google Scholar]