Study setting

This study was embedded in an observational study of the prevalence of tuberculosis, HIV and syphilis sposnsored by the US Centers of Disease control. We studied consenting prisoners in Kality Prison, a high-security federal prison, located 15 kilometers from central Addis Ababa, the capital of Ethiopia. It provides long-term incarceration, shorter-term detention before and during trials, and temporary holding of prisoners before their transfer to other federal prisons. It consists of: a) a central high-security area divided into eight zones, b) an adjacent lower security area containing medical clinics that manage HIV, syphilis, and TB, c) two inpatient wards with a total of 58 beds, d) a 40-bed TB and general medical ward, and e) a well-equipped clinical laboratory and digital x-ray facility. Study personnel had access only to the latter areas.

During our study an average of about 60 new prisoners entered the prison daily and the resident prisoner census averaged about 4,500. However, we had no access to more detailed census data. Full-time medical staff at Kality consists of 3 physicians, 9 health officers (physician’s assistants), 40 nurses, 6 laboratory technicians, 3 radiography technicians who both obtained and read radiographs and a part-time consulting radiologist. Clinics specific for TB, HIV and other STIs, manage all prisoners with these diagnoses according to Ethiopian national guidelines.

Recruitment and enrollment

Potential participants were recruited by prison medical staff either during health screening at entry (entrants) or at ad hoc group meetings within each prison zone (residents). Full-time, prison health care workers who were trained in ethical conduct of research obtained written informed consent by providing information to groups of 5–30 prisoners followed by individual interviews to answer additional questions. Prisoners were informed that they could accept, refuse or discontinue participating in the study, without affecting their health care, imprisonment or parole.

Study design

We initiated two independent, cross-sectional waves of screening of residents (R) and continuous, but not universal, screening of newly-admitted, entrant (E) prisoners for TB and HIV from August 2014 through November 2016. We were unable to recruit or interview entrants who arrived when research staff were unavailable. Residents, excluding those who had been examined as entrants, were screened in two waves: (R1 = 8/2014 to 4/2015) and (R2 = 11/2015 to 11/2016), Entrants screened during these waves were designated E1 and E2 and those screened in the gap between waves were designated entrants screened in the gap period (EG).

Data collection

Demographic data including information on risk factors for HIV and syphilis were collected using a structured questionnaire developed in English, translated to the Ethiopian national language (Amharic) and validated through back-translation into English. Data was collected by trained health care workers and recorded on paper case report forms (CRFs) with a unique participant identification number.

Data collected using the paper CRFs was directly entered into the University of California, San Diego (UCSD)-based database OCCAMS (Open-source Clinical Content Analysis and Management System) by trained data entry clerks at Kality and verified by a full-time data manager. Data quality was completely audited (100%) by the study data manager. The investigators reviewed regular tracking reports of internally accumulating data generated by the UCSD-based statistical team. The data was also assessed by the Centers for Disease Control and Prevention (CDC) three times and by an independent auditor twice during the study. Formal reporting of audits identified deficiencies and errors to which the study team responded with appropriate corrections and/or methodological adjustments.

Variables collected as potential risk factors for HIV before or after imprisonment were:

Socio-demographic. Age, gender, self-identified ethnicity, education, literacy, marital status, occupation, and religion.

Medical. TB diagnosis ever, diabetes diagnosis ever, cancer diagnosis ever, malnutrition, urinary pain (dysuria), scrotal swelling, genital discharge, genital ulcer, syphilis infection (at Kality and ICL labs), HIV infection (at Kality lab),

Behavioral. Current smoking, alcohol use prior prison, current use of Khat, and number of lifetime sex partners.

Prison-related conditions. Residence prior to prison (urban/rural), prison location (entrant/resident), prior imprisonment, prior prison duration, and current prison duration.

Housing conditions before prison. Number of people in the house and number of rooms in the house.

Serological testing

All subjects who consented for HIV had a rapid HIV test performed at the Kality laboratory. HIV rapid testing followed the standard prison algorithm of sequential testing following the national guidelines [8]. I Initial testing used KHB (Shanghai Kehua Bioengineering, Ltd, Shanghai, China) kit [9], subsequent confirmation with HIV1/2 STAT-PAK (CHEMBIO Diagnostic systems, Inc., Medford, NY, USA) [10] and resolution of discordance with Uni-Gold (Trinity Biotech, Jamestown, NY, USA). All tests used an immunochromatographic assay method with various levels of sensitivity and specificity [11] (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Published sensitivity and specificity of standard test kits used at Kality Prison Laboatory with performance references from package inserts.

https://doi.org/10.1371/journal.pone.0271666.t001

Clinical care and treatment

HIV positive cases were referred and linked to the ART clinic in Kality for treatment, care and follow-up per the national guidelines [12].

Ethical and regulatory considerations

This study was designed to assess screening procedures that directly benefit prisoners by diagnosing treatable diseases common in Ethiopian prisons. It was conducted after protocol review by the US Centers for Disease Control and the Ethiopian Federal Prison Administration and by the Institutional Review Boards of the National Ministry of Science and Technology of Ethiopia, the Ethiopian Public Health Association and the University of California, San Diego. Informed written consent was obtained from all study participants. Those who were unwilling to undergo some of the procedures (eg, blood drawing or HIV testing) were included, but excused from those procedures.

Data analysis

Data entered into OCCAMS were exported to the SPSS version 20.0 software program for analysis. Descriptive summaries were provided by study groups (E1, E2, EG, R1 and R2) and overall. Prevalence rates of HIV were compared between entrants and residents (E1 vs R1, E2 vs R2) using Chi-square and Fisher’s exact test. Sensitivity and specificity were calculated to assess accuracy or diagnostic performance. Multivariable logistic regression models were conducted to study the risk factors associated with HIV infection.

Socio-demographic characteristics of participants

Of a total 13,803 prisoners enrolled in a study of screening for tuberculosis and HIV, 10,714 (78%) consented to HIV testing. Participants were 83.9% men and 16.1% women, young (median age: 26 years, IQR: 21–33), single (61%), literate (89%), and urban residents (91%) without prior incarceration (96%) (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Associations of HIV prevalence with socio-demographic and medical characteristics and history of incarceration and sexual behavior for all the participants and in entrants and residents separately.

https://doi.org/10.1371/journal.pone.0271666.t002

Sero-prevalence of and risk factors for HIV infection in all participants and separately in entrants and residents

The prevalence of HIV based on the rapid testing algorithm at Kality lab in all participants was 3.4% (360/10,714) (Table 2). We compared all 360 HIV-positive (HIV+) to the 10,354 HIV-negative (HIV-) prisoners for risk factors in bivariate analysis (Table 2). HIV prevalence increased as the number of lifetime sexual partners increased from 0 to ≥11 partners (1.6% to 5.0%, p<0.001) and risk was especially high in those having 2–10 sex partners (AOR = 1.6, p<0.001). Prevalence of HIV was also significantly higher among participants reporting symptoms of other STIs: genital ulcer (7.6% vs 3.2%, p<0.001), genital discharge (6.1% vs 3.2%, p = 0.002), and scrotal swelling (6.9% vs 3.1%, p = 0.012. (Table 2). Contrary to expectations based on dysuria being a symptom of gonococcal or chlamydial urethritis, it was associated with reduced risk of HIV infection (p = 0.007).

HIV prevalence in all entrants was 2.8% (170/6019), and in all residents was 4.0% (190/4695), a 1.4 fold difference (p<0.001). We compared differences in percentage to To see if patient characteristics HIV prevalence in entrants (E) and residents (R) differed from those in all participants we compared their rates by sociodemographic categories (Table 2). Prevalence of HIV was greater in resident than entrant men (E = 2.5% < R = 4.3, p < 0.001), but in women HIV prevalence was greater in entrants than residents (E = 5.5 > R = 3.2, p = 0.02). Differences in HIV prevalence were found in several ethnic groups, but was significant only in the largest group, the Amhara (E = 2.9 < R = 4.9, p = 0.001).

HIV prevalence differed in only two educational strata: the largest group with 1–8 years of education (E = 2.6 < R = 2.8, p = 0.022) and those with a university diploma (E = 3.0 < R = 7.3, p = 0.010). Of major occupational categories, only the largest category (“other,” not further defined) had a significantly increased prevalence in residents (R = 4.4 > E = 2.8, p = 0.001. Two other occupational groups differed with borderline significance: housewives (E = 11.5 > R = 1.8, p = 0.064) and government employees (E = 2.0 < R = 4.4, p = 0.071) and in all occupational categories residents had more prevalence HIV than entrants. Of groups classified ed by marital status, only the largest group, “never married”, differed significantly (E = 1.9 < R = 2.7, p = .039).

Resident prisoners from both urban and rural places of residence before prison had higher HIV prevalence that entrants: urban E = 2.9 < R = 3.8, p = 0.024, and rural E = 1.0 < R = 5.8, p <0.001. Prisoners who came from homes with either 1 or 2–5 rooms, had the expected higher rates in residents and then entrants: 1 room (E = 3.2 < R = 4.5, p = 0.025) and 2–5 rooms (E = 2.2 < R = 3.8, p = 0.03). Other characteristics with an expected, statistically significantly greater HIV prevalence in residents than entrants included history of prior imprisonment and current number of years in prison.

The prevalence of HIV during and in the gap (= G) between each of the two waves screening (1 or 2) by prisoner status (entrant = E or resident = R) were 2.2% in E1, 2.8% in E2, 2.7% in EG, 2.4% in R1, and 5.4% in R2 (Table 3). In the first wave, HIV prevalence was similar in residents (R1) and entrants (E1) (2.4% vs 2.2%, p = 0.80), but in the second wave was higher in residents (R2) than entrants (E2) (5.4% vs 2.7%, p<0.001) (Table 3). Thus, HIV prevalence was borderline higher in E2 than in E1 (2.7%/2.2% or 1.3-fold, p = 0.07),but strikingly higher (2.3 fold) in residents higher in R2 than R1 (5.4%/2.4% p<0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Associations of HIV prevalence with socio-demographic and medical characteristics and history of incarceration and sexual behavior for all the participants and separately in men and women.

https://doi.org/10.1371/journal.pone.0271666.t003

Multivariable logistic regression model of risk factors for HIV infection by socio-demographic, incarceration, sexual behavior, and medical characteristics in all participants

In a multivariable logistic regression model (Table 4), multiple factors were significantly and independently associated with HIV infection as messured by adjusted odds ratios (AOR). These included: a) increasing number of lifetime sexual partners (p<0.001), b) age (AOR per age category ranging from 1.5–6.4, p<0.001), c) female sex (AOR = 2.5, p<0.001), d) marital status (p = 0.016) especially being divorced/separated (AOR = 2.7, p = 0.001), e) lifetime history of TB diagnosis (AOR = 5.9, p<0.001), f) syphilis (AOR = 4.6, p<0.001), and g) history of genital ulcer (AOR = 2.2, p = 0.037). Unexpectedly, dysuria, a common symptom of some sexually-transmitted infections, was associated with reduced risk of HIV infection (AOR = 0.49, p = 0.007) (Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Prevalence of HIV in entrants and residents by screening period.

https://doi.org/10.1371/journal.pone.0271666.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Multivariable analysis of risk factors for all HIV infected participants.

https://doi.org/10.1371/journal.pone.0271666.t005

Gender differences in HIV prevalence by categories of socio-demographic, incarceration, sexual behavior, and medical characteristics

Comparisons in HIV prevalence rates by multiple categories identified many that differed between male and female prisoners (Table 3). Women were significantly more likely than men to be HIV+ in 4 of 6 age categories under 40 years, but not for the two categories >40 years (p = 0.001). Women were more likely to be HIV+ than men in ethnicity categories; Oromo and the “other.” HIV prevalence in women was higher than in men across all levels of education (p = 0.046), but rates were not significantly higher at any single educational level. The sexes did not differ overall in HIV prevalence by occupational categories. In the occupational category of “none”, women had the highest HIV prevalence and men the lowest (W 9.4 vs M 1.2%, p = 0.003. an 8-fold difference. However, numbers of HIV+ prisoners in this “none” category were small (3 men and 6 women).

No marital status category had a statistical significance diference in HIV prevalence by sex. However, a greater proportion of women who were unmarried (W 3.0 vs M 2.0%) or who were divorced / separated (W 20 vs M 11.5%) were HIV+ compared to their male counterparts. Women residing in urban settings before imprisonment were more likely to be HIV+ than urban men (4.3 vs 3.1, p = 0.013). Women living prior to imprisonment in either a one-room residence (6.0 vs 3.3%, p = 0.001) or a 6–10 room residence (6.1 vs 2.8%, p = 0.038) were more likely to be HIV+ than equivalently-housed men.

Women entering prison were more likely to be HIV+ than entering men (W 5.5% vs M 2.5%, p< 0.001). For residents the difference was reversed, but not significant so (3.2%in women vs 4.3% in men, p = 0.125). Previously imprisoned women tended to have higher HIV+ prevalence than similar men, but this difference was of only borderline significance (4.1 vs 3.1, p = 0.059). Most Kaliti prisoners had been imprisoned for <5 years (92.6%) and in this group women were significantly more likely to be infected (4.2 vs 3.1%, p = 0.021). The number of lifetime sexual partners was a greater risk factor for HIV infection in women than men for the largest two categories: 2–10 (9.4 vs 4.2%, p<0.001) and >10 (20.0 vs 4.5%, p = 0.033). For symptoms of current genital infections (discharge, ulcers, or dysuria), HIV infection was more common in women, but not in men, with these symptoms.

Prevalence of HIV infection in all prisoners and in entrants and residents

As expected, this study found that the overall HIV prevalence in a large central prison (3.4%) was higher than in the general population of Ethiopia (2.0%). HIV prevalence varies markedly among Ethiopian and other African prisons. For example, HIV prevalence at Kality (3.4%) was 12-fold lower than that in Swaziland correctional services (40%) [13] and 1.7-fold lower than that in Ghanaian prisons (5.9%) [14]. It was also lower than in two regional prisons in Ethiopia from a northern region (Woldiya = (6.3%) (15) and southern region (Tabor/Hawassa) prison = 6.0%) [15]. In contrast, prevalence of HIV in a south central prison (Jimma Regional) was lower than Kaliti’s (2.6%) [16]. These differences within Ethiopia may reflect the differing prevalences of HIV in the general populations of various regions of Ethiopia: for example, 4.1% in Amhara (north-central), 1.8% in the SNNPR (south), and 3.0% in Oromia (south-central) [17].

We hypothesized that HIV prevalence rates in the first screening wave would be higher in residents (R1) than entrants (E1) because of intramural HIV transmission. However, differences in HIV prevalence at our first screening (i.e., between E1 and R1) were small and not statistically significant. In contrast, in the second wave, resident prisoners (R2) had significantly higher rates than entrants (E2). Thus, findings only in the second wave were consistence with our hypothesis that intramural transmission of STIs would result in a higher prevelance in residents than entrants.

The reason for this unexpected increase in HIV prevalence in residents between the two screening periods is unclear. A possible explanation is that a subgroup of inmates with either known HIV infection or high risk of HIV did not participate in the R1. They may have feared that a diagnosis of HIV would lead to adverse consequences such as shunning by other prisoners or mistreatment by prison authorities. Some of them may have volunteered for R2 after being reassured by the lack of such consequences for prisoners participating in R1. Less likely explanations are a substantial increase in sexual activity within the prison between waves or errors in laboratory testing. These seem unlikely because conditions in the prison and its laboratory did not change substantially during this period.

Our second hypothesis was that HIV prevalence would decline in residents between the two waves of screening because treatment of HIV-infected residents and entrants would reduce transmission within the prison by reducing HIV in their bood and genital fluids. However, this hypothesis was not supported by our findings. In fact, HIV prevalence was greater in R2 than in R1 by 2.3-fold. Inaccuracy in prevalence estimates in residents for reasons discussed above may have obscured any beneficial effects of screening.

Risk factors for HIV infection

Risk factors that were independently associated with HIV infection in all prisoners in a multivariable analysis were a) increasing age, b) female gender, c) smaller number of rooms in their pre-prison residence, d) marital status (being a widow/widower, divorced or separated, p = 0.016), e) TB diagnosis ever, f) number of lifetime sex partners (especially having 2–10), g) history of genital ulcer(s), and h) reaginic antibody evidence of latent syphilis. Many of these factors reflect either lower socio-economic status or greater exposure to sexually transmitted diseases.

Risk factors for HIV in Kality prisoners were similar to those in the general Ethiopian population [18]. HIV prevalence was 1.3-fold higher in prisoners who resided in urban than rural areas. This difference is less than the 7-fold greater prevalence of HIV in urban than rural dwellers in the general Ethiopian population [19]. Large differences in HIV prevalence were found in several categories of marital status: divorced/separated (13.1% vs 2.9%), married (4.3% vs 0.8%) and never married (2.1% vs 1.0%), except widowers (10.2% vs 11.5%). Prevalence of HIV increased with the number of lifetime sex partners and the current duration of imprisonment. Neither prior imprisonment nor its duration influenced the prevalence of HIV positivity.

Housewives, divorced / separated persons or widows/widowers and the better educated (graduates of preparatory school or more) had higher HIV prevalence than other occupations, marital groups, and educational categories. Consistent with this pattern, prevalence of HIV was highest at the extremes of education and occupational categories i.e. higher in the more educated (merchants and government employees) and the least educated (peasants and those with no formal education) than those with intermediate levels of education. The reasons for this non-linear (U-shaped) distribution are unclear, but may be related to greater access to multiple sexual partners by the more affluent and by the poor than by other Ethiopians, but no direct evidence supports this hypotheses. We found that a history of symptoms of sexually transmitted infections such as scrotal swelling, genital discharge, and genital ulcer or of syphilis were associated with HIV infection in men, but not in women.

Risk factors for HIV infections in Ethiopian prisoners in Jimma Regional Prison were similar to those that we found [19]. They included: a) male sex, b) age > 50 years, c) primary or less school education, d) history of a blood transfusion (a factor we did not address), and e) serologic diagnosis of syphilis

Gender differences in HIV prevalence by categories of socio-demographic, incarceration, sexual behavior, and medical characteristics

Prevalence of HIV in the female Kaliti prisoners in this study (72/1728 = 4.2%) was higher than that for men (3.2%).(Table 2) A review of HIV epidemiology in Addis Ababa emphasized the role of female sex workers and “hot spots in transmission [20]. Thus, female prisoners might be at higher risk because they were more likely to have been sex workers than other Ethiopian woman. However, we could not examine that possibility because we were not permitted to collect data on that risk factor. Unexpectedly, HIV prevalence in the imprisoned women was slightly lower than that estimated for contemporaneously studied (2014), mainly urban, pregnant, Ethiopian women (5.5%) [3]. Thus, they appear no more likely to have been HIV infected than other women.

Many characteristics differed between male and female prisoners. The prevalence of HIV was strikingly 2.2-fold higher in women than men entering Kality Prison (W = 5.5% > M = 2.5%, p<0.001). The ratio was reversed in residents (W = 3.2% < M = 4.3,% p = 0.125). Reasons for this large difference between entrants and residents in the ratio of HIV prevalence in men and women is unclear, but include differences between the sexes in participation by HIV+ prisoners in the study or their management such as release or transfer to other prisons.

HIV+ was more prevalent in female than male prisoners in multiple demographic, behavioral, and serological categories consistent with increased exposure to HIV and other sexually transmitted diseases. Risk factors indicating a higher risk for HIV infection in women than in men included: a) age < 40 years, b) endorsed Oromo and the “other” categories of Ethiopian ethnicities, c) urban origins, d) unmarried divorced or separated, e) had been previously imprisoned, and f) had more than two lifetime sexual partners. They were more likely than men to be HIV+ across all levels of education, but not by occupational categories.

A studies 1679 workers in two Ethiopian factories earlier period of the HIV epidemic (1997) found very higher HIV prevalence in both men (~ 10%) and women (~12%) [21]. Risk factors for HIV were similar to those in Kaliti prisoners 17 years later. In this earlier study the major self-reported risk factors were indicators of high-risk sexual behaviors such as a) higher lifetime sexual partners in men, b) low income history of rape and casual sex in women and c) serologic evidence of sexually transmitted diseases (syphilis and herpes simplex virus, type 2) in both sexes.

The prevalence of HIV was significantly higher in this prison than in the general population confirming that prisoners are a high-risk group for HIV in Ethiopia as they are globally. The higher prevalence of HIV in prisoners than the general population confirms that HIV screening and treatment of prisoners should be a priority. Prevalence rates were higher in resident than entrant prisoners in only one of two periods that we surveyed not consistently supporting or falsifying our hypothesis that intramural transmission of HIV was occurring. Moreover, HIV prevalence increased in residents after screening and treating both entrants and residents for over a year. This finding does not support our second hypotheses that screening for and treating HIV infection would reduce its prevalence in this prison.

HIV prevalence was higher in female than male prison entrants, consistent with the pattern in the general population requiring that they be similarly screened and treated of HIV infection. HIV prevalence in residents of both sexes were similar, an unexplained difference in pattern from that in entrants.

Strengths and limitations

The strengths of this study included: a) inclusion of large numbers of both residents and entrants with wide demographic diversity and b) two cross-sectional waves of screenings of residents and continuous screening of entrants. Our study had the following limitations.

First, participation was voluntary and some entrants were not even approached because they arrived and were processed on weekends or at times the study staff were not available. Second, no unique identifying number was available to us that would allow us to link data on prisoners seen multiple times in the study. Moreover, prisoners were frequently transferred to other prisons or released between the two cross-sectional waves of screening. Thus, we were unable to assess prisoners longitudinally and settled for two cross-sectional waves of assessment of residents.

Third, we did not collect data on blood-related activities associated with transmission (e.g. tattooing or transfusions) and were not allowed to ask about sexual risk activities during imprisonment. Fourth, for security reasons study investigators and research staff who were not prison employees were denied access to high-security areas of the prison where the residents were screened. Thus, they could not directly evaluate the quality of recruitment, interviewing, or assessment methods employed in residents by the prison medical staff. However, we were allowed to observe and assess the screening of entrants. Fifth, we had no access to data on the daily numbers or rates of new entrants, census of residents, or transfers to or from the prison.