AUDIT, AUDIT-C, and AUDIT-3: Drinking Patterns and Screening for Harmful, Hazardous and Dependent Drinking in Katutura, Namibia

Objectives To describe alcohol drinking patterns among participants in Katutura, Namibia, and to evaluate brief versions of the AUDIT against the full AUDIT to determine their effectiveness in detecting harmful drinking. Methods A cross-sectional survey was conducted in four constituencies and 639 participants, 18 years or older, completed a sociodemographic survey and the AUDIT. The effectiveness of the AUDIT-C (first three questions) and the AUDIT-3 (third question) was compared to the full AUDIT. Results Approximately 40% were identified as harmful, hazardous or likely dependent drinkers, with men having a higher likelihood than women (57.2% vs. 31.0%, p<.0001). Approximately 32% reported making and/or selling alcohol from home. The AUDIT-C performed best at a cutoff ≥ 3, better in men (sensitivity: 99.3%, specificity: 77.8%) than women (sensitivity: 91.7%, specificity: 77.4%). The AUDIT-3 performed poorly (maximum sensitivity: < 90%, maximum specificity: <51%). According to AUROC, the AUDIT-C performed better than the AUDIT-3. Conclusions A large proportion of participants met criteria for alcohol misuse, indicating a need for screening and referral for further evaluation and intervention. The AUDIT-C was almost as effective as the full AUDIT and may be easier to implement in clinical settings as a routine screening tool in resource-limited settings because of its brevity.

Although studies have identified the AUDIT-3, which assesses binge drinking (i.e., 6 or more standard drinks in a sitting), as effective in detecting alcohol misuse, the AUDIT-C has performed significantly better than the AUDIT-3 [27,29]. Findings have been less conclusive, particularly among women and when compared to the full AUDIT and the AUDIT-C [25][26][27].
To our knowledge, no studies have evaluated the effectiveness of the AUDIT-C and AUDIT-3 in sub-Saharan Africa. Recent studies on drinking patterns and alcohol screening methods in Namibia are limited and are needed to help mitigate the substantial social and health harms associated with heavy drinking. For this study, participants were selected from the community of Katutura, which is adjacent to Windhoek. Data on the prevalence of alcohol use and abuse in Katutura do not exist. The community has a substantial number of informal drinking venues (i.e., shebeens), where people make and/or sell alcohol from their home. This is one of the first studies to: [1] describe drinking patterns among adults in four constituencies (political jurisdictions) in Katutura, [2] describe the prevalence of making and selling alcohol from home, and [3] compare the AUDIT-C and AUDIT-3 against the full 10-item AUDIT in their effectiveness to detect high-risk drinking (i.e. harmful, hazardous or likely dependent) among this population.

Study Design and Participants
This study was a population-based, cross-sectional survey of adults living in randomly selected households in a convenience sample of four constituencies in Katutura. The four constituencies included: Tobias Hainyeko, Moses Gareb, Samora Machel, and Soweto. Many neighborhoods are informal settlements. These constituencies were selected because a community-based HIV prevention and care outreach program targeting high risk communities had systematically mapped all households (20,863) in 2008, providing a recent and accurate sampling frame. Households were chosen using a 3-stage stratified sampling design. Each stratum was further sub-stratified into 50 distinct geographic areas. Finally, a proportional number of houses were randomly selected within each geographic area. Participant inclusion criteria were 18 years or older, verbal informed consent, and the ability to answer questions in English, Oshiwambo, or Afrikaans. The protocol, including consent procedures, was approved by the Namibian MoHSS and the Institutional Review Board at the Centers for Disease Control and Prevention prior to implementation.

Procedures
Each person approached in the selected households was asked to identify a private place in the home or outdoors to administer the survey. The survey administrator then explained the purpose of the survey, including the AUDIT screen and supplementary questions, the interpretation of results, and resources that would be provided should alcohol misuse be identified. The administrator assured confidentiality. For those willing to participate, verbal but not written consent was obtained and recorded on the screening form to assure anonymity, and the survey was administered individually. Participants were free to not answer any questions that made them uncomfortable and to stop the interview at any time.

Measures
Sociodemographic variables included sex, age, and paid employment. All participants were asked whether they made or sold alcohol from their home. Those who acknowledged drinking alcohol in the last year were asked if they ever drank home-brewed beer, what type of alcohol they drank most of the time (i.e., homebrewed beer, bottled beer, wine, or spirits), and where they typically purchased alcohol.

Data Analyses
Primary analyses described patterns of alcohol consumption among residents of Katutura and examined sex differences. These patterns were analyzed, using survey software, accounting for sample weights, clustering (household), and stratification (constituency), and estimates were population-based, representing the population of the four constituencies of Katutura. Analyses were generated from R 2.15.1 [30,31] and SAS 9.3.
Additional analyses compared the AUDIT-3 and AUDIT-C for sensitivity and specificity against the full AUDIT cut-off score 8 for the overall sample and for men and women separately. Sensitivity, specificity, and 95% confidence intervals account for the sampling design. Sensitivity refers to the "true positive rate," which measures the proportion of persons who are correctly identified as high-risk drinkers. Specificity refers to the "true negative rate," which measures the proportion of persons who are correctly identified as not being high-risk drinkers. The analyses illustrated which cut-off (s) for the AUDIT-3 or AUDIT-C most closely predicted harmful, hazardous, or likely dependent drinking, as defined previously for the original AUDIT. In addition, the area under the receiver-operating curve (AUROC) was used, which compared the areas under the curves with a distribution-free permutation procedure for curves based on data from a paired design [32]. With AUROC, larger areas indicate superior performance, with 1 indicating perfect performance. The anonymized dataset is available upon request from the analysis working group, comprising the corresponding and senior authors, members of the MOHSS, and CDC.

Results
In March 2009, 85% (340) of the 400 approached households participated, which represented 1.6% of households registered across all 4 constituencies (20,863). A total of 639 adults participated; the median number of surveys per household was two, and the range was one to seven. Approximately two-thirds of participants were women (64.6%), and there were no significant differences by sex in the age distribution (p = .57) ( Table 1). Participants' ages ranged from 18 to 80 years, with the largest proportion (43.5%) being 20-29 years. Approximately half (51.3%) reported paid employment, with no significant differences by sex (p = .79).
Approximately one-third (32.2%) of participants reported making or selling alcohol from their homes, including a mix of homebrewed and manufactured alcohol products (Table 1). Among the 467 (73.1%) self-reported current drinkers (i.e., consumed alcohol in the past 12 months), consumption of homebrewed (48.0%) and bottled beer (42.6%) was more prevalent than wine (5.1%) or spirits (2.3%). A significantly higher proportion of men reported consumption of bottled beer than women (50.3% vs. 38.1%) and a higher proportion of women reported consumption of homebrewed beer (54.7% vs. 36.3%) (p = 0.0019).

AUDIT-Drinking Patterns in Katutura
Following AUDIT guidelines, if respondents answered "never" to the first AUDIT question ("How often in the past year did you have a drink containing alcohol?"), then questions 2 through 8 were skipped. AUDIT scores were obtained on 625 participants because of missing data on key questions for 14 participants. The first question was completed by 100% (639) of participants, and 26.9% (172) responded "never." Thus, 73.1% indicated alcohol consumption in the previous 12 months, and the denominator for questions 2 through 8 was 467. In terms of data completeness, the AUDIT-3 was completed by all 467 current drinkers and the AUDIT-C by 466. Table 2 displays participants' drinking patterns and alcohol risk categories. The average full AUDIT score was 7.2 (95% CI: 6.5, 7.9). There were sex differences, with a mean score of 9.5 (95% CI: 8.4, 10.6) for men and 6.1 (95% CI: 5.3, 6.9) for women (p<.0001). When examining drinking categories, 39.5% (95% CI: 35.1%, 43.9%) were classified as harmful, hazardous or likely dependent drinkers by the full AUDIT (AUDIT 8). The breakdown by sex revealed that 57.2% (95% CI: 49.6%, 64.8%) of men, and 31% (95% CI: 26.1%, 36%) of women were categorized as harmful, hazardous drinkers or likely dependent on alcohol (p<.0001).
ROC curves from all cut-off values were constructed for both the AUDIT-C and AUDIT-3 ( Figs. 1-3). Bootstrapped 95% confidence intervals for sensitivity at given specificity points, provided by the analysis in Tables 3 and 4 and vice versa, were overlaid on the ROC curves to provide some estimate of the precision of this analyses. For example, with the AUDIT-C (Fig. 1, solid-lined curve) and a sensitivity of 95.2% (Table 3, cutoff score = 3), the bootstrapped 95% confidence interval for specificity around the ROC curve was estimated to be 69.6%-83.3%, which was wider than the 95% confidence interval estimated using methods for complex samples.
The AUROC illustrated the ability of the AUDIT-C and AUDIT-3 to classify levels of alcohol consumption risk when compared to the full AUDIT. Both AUROCs for the briefer screening tools were significantly superior to the line of identity (x-axis = 1-y-axis) at classifying harmful, hazardous, or likely dependent drinkers. However, for the overall sample and for each sex, the AUROCs indicated that AUDIT-C performed significantly better than the AUDIT-3 in classifying harmful, hazardous, or likely dependent drinkers.

Discussion
This is one of the first studies conducted on drinking patterns and alcohol screening methods in Katutura, Namibia. This population-based study examined alcohol use and abuse patterns among household participants in four constituencies in Katutura. Overall, rates for both were alarming. Over 70% of participants self-identified as current drinkers, with nearly 40% of those reporting harmful, hazardous or likely dependent drinking, and 63.5% reporting binge drinking in the past 12 months. These findings are similar to the study conducted by the Namibian Ministry of Health and Social Services approximately 10 years earlier [13], indicating that harmful drinking behavior may not have changed in Namibia in over a decade. Additionally, although rates were alarming for both men and women, men were more likely to report harmful, hazardous, or especially, likely dependent drinking behaviors than women. Specifically, daily, weekly and monthly binge drinking rates were significantly higher for men. *These statistics are population estimates, which imply stratification by constituency and weighting by household. ** Questions 1-3 represents the AUDIT-C. *** Question 3 represents the AUDIT-3. † Time period for these questions is during the previous 12 months.
doi:10.1371/journal.pone.0120850.t002 AUDIT-C > = 7 This study also is the first to systematically document the prevalence of informal alcohol sales in Katutura. The finding that 32.2% reported making or selling alcohol from their homes indicates a lack of enforcement of national laws governing alcohol sales in Namibia. Both men and women reported making or selling alcohol from their homes, but rates were higher among women (35.3% vs. 25.5%). Further, consumption of homebrewed beer was highly prevalent, particularly among women. Consumption of informally manufactured alcohol complicates accurate assessment of alcohol use and alcohol interventions for high-risk drinking. Further research is needed to address the complexities of alcohol consumption and screening among this population.
Finally, to our knowledge, this study is the first to evaluate the effectiveness of the AUDIT-C and AUDIT-3 in detecting harmful, hazardous or likely dependent drinking in comparison to the full 10-item AUDIT in Namibia. In this resource limited setting, the AUDIT-3 performed poorly. However, the AUDIT-C performed better than the AUDIT-3 among the overall sample and among both sexes, which is consistent with previous findings from other settings and countries [27,29]. Therefore, the brief 3-item version of the AUDIT (AUDIT-C) may be effective in detecting high-risk drinking, whereas the 1-item version (AUDIT-3) may not be as effective. At a cut-off score of 3, essentially all men (99.3%) identified from the full AUDIT as high-risk drinkers would be identified with the AUDIT-C, although some women may be missed as sensitivity was 91.7%. As with the full AUDIT, further clinical assessment of those who screen positive for high-risk drinking is necessary to confirm the finding.
There are limitations to this study. The participants consisted of adults present at the time of a single household visit. For safety reasons, data were collected only during daylight hours and household members working may have been missed. Therefore, more women are represented than men. Previous research has indicated that alcohol use levels may be higher in nonworking populations; therefore, alcohol consumption rates among men may be skewed and overestimated. Other limitations are related to the screening tools. The abbreviated versions of Drinking Patterns and Screening in Namibia the AUDIT were compared only with the full AUDIT and were not administered independently of the full 10-question AUDIT. Results for the briefer versions may have been different, if administered separately. Since 48% of participants reported primary consumption of homebrewed rather than manufactured alcohol, assessment of alcohol content is difficult. The common practice of sharing drinks among a group and drinking from a variety of nonstandard containers presents a challenge in quantifying alcohol consumption among this population. Finally, this study was conducted in one community, Katutura, where sociodemographic characteristics may not be similar to neighboring communities or other parts of Namibia. Therefore, results may not be generalizable, and further research with diverse geographic populations is needed.
Many factors limit alcohol screening in both clinic and community settings, including social acceptability or lack thereof regarding alcohol use, limited time of providers, and lack of training [33,34]. However, given that alcohol is associated with high-risk sexual behavior and Drinking Patterns and Screening in Namibia negative health and social outcomes, it is important for providers in both settings to routinely screen patients for current alcohol use and provide alcohol reduction counselling to those who report harmful, hazardous or likely dependent drinking. Previous research has indicated that healthcare settings are an avenue for brief, effective interventions for alcohol issues [35,36]. For certain populations (e.g., HIV-positive persons, pregnant women), screening and intervention should be offered if any alcohol use is reported. The current findings indicate that providers in resource limited settings may use the AUDIT-C as a very brief, effective alcohol screening tool to identify persons who need further evaluation and intervention. Despite legislation and policies regulating alcohol sales in Namibia, the relatively large prevalence of high-risk drinking and consumption of homebrewed in Namibia indicates potential barriers in enforcing these policies. Programs to promote alternative income generation and enforcement of alcohol sales restrictions should be considered. Additional research and guidance also are needed to help curb high-risk drinking in the context of these unique challenges.