Prevalence of HIV infection and bacteriologically confirmed tuberculosis among individuals found at bars in Kampala slums, Uganda

Individuals found at bars in slums have several risk factors for HIV and tuberculosis (TB). To determine the prevalence of HIV and TB among individuals found at bars in slums of Kampala, Uganda, we enrolled adults found at bars that provided written informed consent. Individuals with alcohol intoxication were excluded. We performed HIV testing using immunochromatographic antibody tests (Alere Determine HIV-1/2 and Chembio HIV 1/2 STAT-PAK). TB was confirmed using the Xpert MTB/RIF Ultra assay, performed on single spot sputum samples. We enrolled 272 participants from 42 bars in 5 slums. The prevalence of HIV and TB was 11.4% (95% CI 8.1–15.8) and 15 (95% CI 6–39) per 1,000 population respectively. Predictors of HIV were female sex (aOR 5.87, 95% CI 2.05–16.83), current cigarette smoking (aOR 3.23, 95% CI 1.02–10.26), history of TB treatment (aOR 10.19, 95% CI 3.17–32.82) and CAGE scores of 2–3 (aOR 3.90, 95% CI 1.11–13.70) and 4 (aOR 4.77, 95% CI 1.07–21.35). The prevalence of HIV and TB was twice and four times the national averages respectively. These findings highlight the need for concurrent programmatic screening for both HIV and TB among high risk populations in slums.

www.nature.com/scientificreports/ Similarly, there is a high prevalence of HIV in areas that have a high density of bars 24 . It is likely that the prevalence of HIV among bar customers, employees and neighbours of bars in slums is high due to the high risk sexual behaviour, low risk perception of sex with commercial sex workers and inconsistent condom use observed among individuals at bars in slums 21,25,26 . Further, individuals who attend bars are less likely to visit health facilities and new TB cases and HIV infections among customers, staff and residents at bars in slums are missed 12 . The prevalence of HIV and TB among individuals found at bars in slum settings is not widely reported because studies in this setting have focused on each disease in isolation.
The aim of this study was to determine the prevalence of HIV and TB and predictors of HIV infection among individuals found at bars in Kampala slums in Uganda.

Materials and Methods
Study population and setting. Using a cross sectional study design, we enrolled participants from the 5 divisions of Kampala city, the capital city of Uganda, in December of 2019. Kampala city is sub-divided into 5 administrative divisions: Kawempe, Lubaga, Makindye, Nakawa and Central divisions. A map supplied by the Kampala Capital City Authority designates 29 settlements as slums. We randomly selected one slum from each division of Kampala. Bwaise, Kamwokya, Nalukolongo, Kibuli and Naguru Go-down slums were selected from Kawempe, Central, Lubaga, Makindye and Nakawa divisions respectively. At each slum settlement, we identified bars by snowball sampling method with the aid of a village health team leader or a local council chairperson due to lack of organised housing and registration status of bars in slums. The lack of official licensing of the bars and their operation outside the legally recommended hours makes individuals attending these places a "hardto-reach" and "hidden" population that justifies the use of the snowball sampling method 27 . A bar was defined as any place from which alcoholic beverages were sold and consumed at the premises. At each bar, study participants were sampled and enrolled consecutively until the desired sample size for the given slum was achieved. The study inclusion criteria were any adult (≥ 18 years) who was found at a bar during the study visit and provided written informed consent. We excluded potential participants with alcohol intoxication as this would impair ability to provide informed consent. Alcohol intoxication was defined as Hack's impairment index of ≥ 0.6 28 .

Study measurements.
We administered a pre-tested questionnaire that sought for socio-demographic data, history of alcohol and cigarette use, history of TB exposure and treatment and past medical history. The socio-economic status of participants was measured by the number of durable household items using the equity tool available at: https ://www.equit ytool .org/the-equit y-tool-2/ 29 . Using possession of common household items as a measure of household wealth is used in national demography surveys in Uganda 30 . The socioeconomic status was divided into 3 tertiles: rich, middle class and poor after principal component analysis. We used the CAGE questionnaire to assess the level of alcohol consumption. The CAGE questionnaire is a set of 4 simple questions: "Have you ever: (1) felt the need to cut down your drinking; (2) felt annoyed by criticism of your drinking; (3) had guilty feelings about drinking; and (4) taken a morning eye opener?" 31 . A score of 2-3 is indicative of a high index of suspicion for alcohol abuse while a score of 4 is diagnostic of alcohol dependence 32 . A study nurse drew 1 ml of blood following standard procedures 33 . We performed HIV testing on whole blood samples using a rapid immunochromatographic antibody test (Alere Determine HIV-1/2) and sequentially confirmed positive samples with another test (Chembio HIV 1/2 STAT-PAK) according to the Uganda ministry of health HIV testing guidelines 34 . Participants also provided single spot sputum samples that were tested using the Xpert MTB/RIF Ultra assay for bacteriological confirmation of TB at Mulago Hospital tuberculosis unit laboratory according to World Health Organisation Xpert MTB/RIF user guidelines 35 . Spot sputum samples are equally sensitive as early morning sputum samples for TB diagnosis 36 . Twenty one participants were unable to expectorate sputum spontaneously. For these, we induced sputum by nebulising 0.45% hypertonic saline using an ultrasonic nebuliser following normal post bronchodilator spirometry findings according to standard procedures 37 . To ensure participant privacy and confidentiality, we delivered pre-and post-HIV testing counselling in a designated housing unit in the slum. Newly diagnosed HIV-positive and bacteriologically confirmed TB participants were linked to Makerere University Joint AIDS Program and Mulago Hospital tuberculosis treatment unit for HIV and TB care respectively. Sample size estimation and statistical analysis. Using the prevalence of sputum positive TB among Kampala slum dwellers with a chronic cough of 18% 38 , the sample size was estimated to be 283 considering a 95% confidence interval and possible incomplete data rate of 20%. Similarly, a sample size of 178 individuals was deemed adequate for estimating the prevalence of HIV considering a 10.3% prevalence of HIV among youths in slums of Kampala 39 . The sample size was distributed across the 5 sampled slum settlements proportional to population size of the division using data from the Uganda demographic survey 30 . Data were entered in EpiData 3.1 and analysed with Stata 15 (StataCorp, College Station, TX, USA). The prevalence of HIV was estimated as the proportion of participants with a positive HIV test to the total number of study participants, expressed as a percentage. Similarly, the prevalence of bacteriologically confirmed TB was estimated as the proportion of participants with a positive Xpert MTB/RIF Ultra assay to the total participants, expressed per 1,000 population. As a secondary objective, we performed logistic regression analysis to determine predictors of HIV infection. At univariable analysis, we examined the associations between categorical and continuous variables and HIV infection using chi-square tests and Rank sum test respectively. Prior to multivariate analysis, we assessed the individual associations between independent variables and HIV infection using individual simple logistic regression analysis. Variables that showed associations with a cut off threshold of p value < 0.2 and didn't have sparse data in categories were considered at the multivariable analysis. At multivariable analysis, we used a mul-  Table 2. There were 4 participants with newly diagnosed bacteriologically confirmed TB, corresponding to a prevalence of 15 (95%CI: 6-39) per 1,000 population. One individual had HIV/TB co-infection corresponding to a prevalence of 4 per 1,000 population. Thirty (11.0%) participants reported either history of TB treatment, current TB treatment or had newly diagnosed bacteriologically confirmed TB.

Characteristics of Bacteriologically confirmed TB individuals found at bars in slums of Kampala.
Of the 4 participants with bacteriologically confirmed TB, 3 were customers while 1 was a bar owner.

Discussion
The aim of this study was to determine the prevalence of HIV and bacteriologically confirmed TB among individuals found at bars in slums of Kampala. We further analysed for predictors of HIV infection in this population. We found the prevalence of HIV to be twice the national average of 5.7% among adults 40 . The prevalence of HIV significantly differed by slum settlement and was higher among participants who had never tested for HIV. Female sex, history of tuberculosis treatment, current history of cigarette smoking and CAGE scores of 2-3 and 4 were independent predictors of HIV infection. The prevalence of bacteriologically confirmed TB among individuals found at bars in Kampala slums was 4 times the national estimate of 4 per 1,000 population 41 . Our study findings show a high prevalence of HIV and TB among individuals found at bars in slums. Moreover, all TB patients and more than half of the HIV infections were newly diagnosed. These data indicate that many new HIV infections and TB cases are missed, and programmatic screening for the two infections is warranted among customers, staff, bar owners and other individuals found at bars in slums. The geospatial variation in the prevalence of HIV within the same city suggests a need for slum specific HIV prevention strategies 42 .
Similar to our findings, Lama et al. (2016), found the prevalence of HIV among bar patrons to be 12.7% in the capital of Botswana and 19% of first time testers were HIV positive 43 . In Tanzania, the prevalence of HIV was 19.0% among female bar and hotel workers which is comparable to our finding of 19.7% among females 44 . The prevalence of HIV is known to be higher among women than men in sub-Saharan Africa owing to different responses to socio-demographic risk factors, sexual behaviours and HIV/AIDS awareness 45 www.nature.com/scientificreports/ found at bars. Additionally, an evaluation of the prevalence of other sexually transmitted diseases (STDs) among individuals at bars in slums would be desirable since STDs increase the risk of HIV infection 49 .
Studies that have evaluated the prevalence of TB in bar settings have been contact investigations of an index case and their findings may not compare to our cross sectional evaluation of the TB prevalence. Nonetheless, using DNA fingerprinting, these studies demonstrate that bars are an avenue for TB transmission and could propagate an outbreak among customers, employees and members of their households 15,[50][51][52][53] . Similar to our finding, Godoy et al.(2017), found that the prevalence of active TB was 16 per 1,000 population upon investigation of contacts of an index case (a bar staff) 17 . Conversely, Kline, Hedemark and Davies (1996) found it at 144 per 1,000 population among contacts of an index bar patron, including customers and employees 19 . The prevalence of TB among individuals found at bars in slum settlements is not well reported elsewhere using a cross sectional design other than contact tracing investigations.
Our results underscore the epidemiological and clinical interrelatedness of TB and HIV among individuals found at bars in three ways. First, history of TB treatment was associated with HIV and this could be attributed to HIV-related depletion of CD4 T-lymphocytes which are crucial for immune responses against TB 54 . Secondly, from our multivariable logistic regression model, the odds ratios of HIV infection increased with a higher CAGE score. This suggests a dose dependent relationship between HIV and problematic alcohol use that leads to high sexual risk behaviour and inconsistent condom use 55   www.nature.com/scientificreports/ scores is also reported by a meta-analysis of studies across Africa 56 . This dose dependant relationship has been observed between tuberculosis risk and alcohol use as well 22 . Lastly, cigarette smoking was associated with HIV infection in our study yet it is an established risk factor for tuberculosis 57 . Therefore, these results suggest that people living with HIV found at bars also had TB risk factors. This calls for concurrent screening for HIV and TB by national HIV and TB programs among high risk populations in slums 58 . We did not directly evaluate the association between HIV and TB in the multivariable model population due to the small number of TB outcomes. However, current TB treatment was commoner among HIV-positive individuals than HIV-negative persons (p = 0.035). We have not found reports that have concurrently evaluated the prevalence HIV and tuberculosis among individuals in bars in slum settings and thus our findings fill an important knowledge gap. Our study had some limitations. We did not perform sputum cultures which could have increased our detection of bacteriologically confirmed tuberculosis. Further, by excluding participants with alcohol intoxication, we might have inadvertently excluded individuals with a higher risk for TB and HIV. However, this was necessary to ensure that consent to participate was obtained wilfully. Lastly, the small sample size affected the precision of the effect sizes of predictors of HIV infection. This analysis did not account for other confounders such as sexual behaviour, knowledge and practices regarding HIV and TB prevention and ease of access to HIV and TB diagnostic services, all of which could influence the risk of TB and HIV in this population. Subsequent studies could evaluate these factors among individuals found at bars with a larger sample size.

conclusion
There was a high prevalence of HIV and TB among individuals found at bars in the slums of Kampala, Uganda. Female sex, current cigarette smoking, history of TB treatment, and CAGE scores of 2-3 and 4 predicted HIV infection. These study findings highlight the need for concurrent programmatic screening for HIV and TB among high risk populations in slums to identify the missed infections that drive the two epidemics. We recommend slum specific programs for HIV and TB prevention that focus on women and reduction of cigarette smoking and alcohol use among individuals at bars in slum settings.

Data availability
Datasets generated and/or analysed during this study are available from the corresponding author upon reasonable request.