Background

HIV testing is the entry point into the HIV service cascade for the general population and is important for identifying infections and stopping the spread of the disease [1]. Data from Zimbabwe, Malawi, and Zambia reveal that the major barrier to reaching UNAIDS targets of 90% of people living with HIV diagnosed, 90% of them on treatment, and 90% of them having suppressed viral load is diagnosing HIV infections [2]. Globally, by the end of 2017, 75% of persons living with HIV were aware of their status [3].

Some countries have used extensive intake forms to collect data on HIV testing and counseling (HTC) clients from routine HIV testing facilities. However, given the volume of data, few have been able to analyze and use the data [4,5,6]. In the absence of national population-based data, which are usually only conducted every five years, few countries understand who is being tested for HIV beyond basic demographics such as sex and age [7]. In 2012, Uganda’s National HIV Testing Services Policy recommended follow-up testing after 3 months for individuals thought to be in the window period. The 2016 update revised this to 14 days to 3 months later, depending on the population [8]. New guidelines are under development. The annual number of people tested for HIV in Uganda increased from 7.0 million in 2013 to 8.6 million in 2014 [9]. This translates to approximately half of adults being tested each year and on the surface bodes well for efforts to diagnose 90% of HIV infections. However, testing data must be carefully scrutinized because it most often represents the number of tests conducted rather than unique people tested. The proportion of testers who are retesters, that is, people who have been diagnosed with HIV but who test again, is often unknown and could artificially increase the yield of HIV positivity. In one study in Mozambique, for instance, 13.0% of HIV-positive voluntary counseling and testing clients and 29.4% of those testing through provider-initiated counseling and testing were retesters [10]. Efforts to diagnose 90% of HIV infections require that testing targets the right people, especially those most at risk. For example, data from surveys conducted in Kampala, the capital of Uganda, from 2012 to 2013 indicate that only 20.2% of men who have sex with men [11], and 46% of female sex workers living with HIV are aware of their status [12].

With a better understanding of the demographic characteristics of who is being tested and who needs to be tested, countries can better target outreach and testing interventions. Additional information about the risk behaviors of testers can further allow HIV testing counselors to tailor messages to encourage shorter repeat testing intervals for those at greatest risk [13].

Much of the literature on individuals testing for HIV multiple times focuses on those who previously tested negative for HIV [14,15,16]. Little is known about the sub-group of people living with HIV who return to testing facilities for retesting, ostensibly in hopes that they have been cured of the disease. Such individuals may inflate positivity rates and also put additional strain on resources, both human and otherwise. We implemented a computer-assisted self-interview system (CASI) at Mildmay Uganda’s clinic in greater Kampala, a large HIV testing and treatment facility, to collect biobehavioral information about individuals testing for HIV. Here we assess correlates of living with undiagnosed HIV and compare testers who are diagnosed for the first time to retesters.

Methods

Study Design and Setting

We analyzed cross-sectional first-visit data from Mildmay Clinic in Uganda, located just south of Kampala, collected between January 2011 and October 2013. All clients aged 13 or older attending Mildmay for the purpose of client-initiated HIV testing and counseling at a voluntary counseling and testing center during this period were consented to have their routinely collected data analyzed.

Data Collection

Clients attending Mildmay underwent routine registration, followed by Mildmay counseling staff providing group pre-test counseling. Mildmay laboratory staff then drew venous blood and tested specimens for HIV in real time using Uganda’s national HIV serial rapid testing algorithm which includes Determine (screening, Alere, Waltham, MA, US), Stat Pak (confirmatory, Chembio, Medford, NY, US), and Unigold (tiebreaker, Trinity Biotech, Bray, Ireland). Interviews were conducted after blood draw.

Interview data were collected primarily through CASI (without audio) or audio CASI (ACASI, using Questionnaire Design System™, Nova Research, Silver Spring, MD), either in Luganda, the main language spoken in greater Kampala, or English on either palm-held devices or netbooks. CASI was used by literate HTC clients who did not want to use the audio feature. Staff were present while participants underwent a tutorial explaining how to navigate the A/CASI environment and remained nearby during the subsequent self-administered interview in order to respond to questions or switch to a face-to-face interview setting if participants had trouble with the self-interview. Interview domains included demographics, HIV testing characteristics, mental health (Patient Health Questionnaire-2 (PHQ-2) screening tool, with a score of 3 + defined as screening positive for possible depression), drug and alcohol use (Alcohol Use Disorders Identification Test-10 (AUDIT-10) scale, cut-off of 8 for hazardous drinking), as well as HIV related risk behaviors, focusing mostly on sexual risk behavior [17, 18]. All interview data were self-reported. Depending on respondents’ answers and ease of computer use, the interview lasted approximately 30-45 min. Test results were returned after the interview, and counseling and referrals provided for HIV treatment.

Data Management and Analysis

Following the interview, staff printed auto-generated summaries of key risk behaviors and made them available to Mildmay’s counselors to inform the ensuing routine post testing counseling sessions taking place in separate, private rooms. Later, participants’ interview data were merged with the corresponding Mildmay’s HIV testing results for analysis. Data cleaning and analysis were conducted using SAS software, Version 9.3 (Cary, NC). We examined participants’ sociodemographic and biobehavioral characteristics, and compared participants by HIV status and testing history using 2 tests.

We confined our analysis to the first test during the study period for each participant. Two separate bivariate logistic regression analyses were conducted to examine factors associated with our outcomes of interest: (1) HIV infection, and (2) retesting for HIV. Adjusted multivariable logistic regression models were then developed for each analysis that included variables found to be statistically significant in the unadjusted models (p < 0.10) and a P value < 0.05 was considered statistically significant. Unadjusted and adjusted odds ratios and corresponding 95% confidence intervals (95% CI) were reported. Missing values were excluded from the analysis.

Human Subjects Considerations

We obtained written informed consent to retain interview and biomarker data for data analysis. Children ages 13 and above are allowed to obtain HIV testing without parental permission in Uganda. As such, permission for analyzing the self-interview data was not obtained from guardians of children because doing so would require informing the guardian that a child sought testing and possibly discourage testing, something that national guidelines sought to prevent.

HTC clients who did not provide consent continued to receive Mildmay’s usual HTC services but their data were not included in this analysis. Personal identifiers, including Mildmay ID numbers, were collected as per Mildmay’s routine procedures. All personal identifiers were removed prior to data analysis. We collected alphanumeric fingerprint codes without storing fingerprint images in order to distinguish new from returning HTC clients. No monetary or material incentives were given. The protocol was approved by the relevant institutional review boards.

Results

A total of 12,233 individuals participated in this study. All individuals aged ≥ 13 years seeking HTC services at Mildmay during the analysis period agreed to an interview to facilitate counseling, and 99.8% agreed to have their data analyzed. Our analysis of user experience of the ACASI interview format revealed that 41.5% of testers would give different answers to a person than to a computer and 62.8% appreciated the privacy aspect of ACASI even though they knew that their responses would be viewed by a counselor (Table 1).

Table 1 Perspectives on using computer-assisted self-interview
Full size table

Characteristics of People Accessing Voluntary Counseling and Testing

Females comprised 57.2% of the clients evaluated (Table 2). The majority of clients (65.9%) came from Wakiso District (where Mildmay Clinic is located) and 27.2% from Kampala. HIV-negative testers were young compared to people with HIV. Among those testing HIV-negative, 35.8% were less than 25 years old compared to 18.6% of those newly diagnosed with HIV and 18.9% among those aware of their HIV infection (p < 0.0001). HIV-negative testers were also better educated, with 23.6% having at least 14 years of education versus 9.3% of those unware of their infection and 16.0% of those who were aware (p < 0.0001). Whereas more than half of males with HIV (52.2% of unaware and 53.9% of aware) lived with a sex partner, only 36.6% of HIV-negative males did (p < 0.0001). Cohabitation was less common among female testers (p = 0.0002), among whom 39.9% of HIV-negative females, 35.7% of unaware females, and 37.4% of aware females lived with a sex partner.

Table 2 Socio-demographic characteristics of individuals seeking testing at Mildmay Clinic, Uganda
Full size table

Both depression and harmful drinking were most common in testers with HIV. Prevalence of depression was similar among those unaware and aware of their infection (40.1% vs. 38.8%, p = 0.4008). Harmful drinking was more prevalent among those unaware of their infection compared to aware (29.2% vs. 22.6%, P < 0.0001). Over one-quarter of all females had ever been forced to have sex (Table 3). Among males, the estimate was similar among HIV-negative participants (8.0%) and those previously diagnosed with HIV (12.2%). More than one in 10 males (14.1%) bought sex in the last 6 months, and the proportion of people reportedly selling sex was approximately twice as high in males than females, with males also having a higher HIV prevalence among both the HIV-aware and unaware groups compared to females (p < 0.0001). Condom use at last sex was 28.6% among those without HIV, 22.4% among the unaware, and 28.3% among the aware (p < 0.0001).

Table 3 Biobehavioral characteristics of HIV testing clients at Mildmay Clinic
Full size table

Prior HIV testing was much more common among HIV-negative participants compared to those newly diagnosed with HIV (67.0% and 26.7%, respectively, p < 0.0001). While 38.1% of HIV-negative participants tested because they felt ill or feared they had AIDS, 66.5% of newly diagnosed participants tested for this reason (p < 0.0001). Similarly, 32.1% of HIV-negative participants believed they were likely to get HIV in the next year compared to 63.7% of unaware HIV-infected individuals (p < 0.0001). More than half of each group did not know the HIV status of their last sex partner, and among HIV-unaware individuals, 78.2% did not know (p < 0.0001). The CD4 count of males was lower than for females (p < 0.0001); 61.8% of males and 50.5% of females unaware of their status had a CD4 count less than 350, the treatment eligibility threshold in Uganda at the time of data collection.

HIV prevalence was 39.0%; however, 37.2% of people diagnosed with HIV already knew their sero-status. HIV prevalence among those who had not previously tested HIV positive was 23.4% among males and 32.8% among females, for a combined prevalence of 28.5%.

Factors Associated with Undiagnosed HIV

In multivariate analysis, the odds of having undiagnosed HIV infection were lower for males compared to females (adjusted Odds Ratio (aOR) 0.60, 95% CI 0.53–0.67) and for people living in Wakiso District compared to Kampala (aOR 0.65, 95% CI 0.58–0.73) (Table 4). While the adjusted odds of undiagnosed HIV were higher among those aged ≥ 19 years than for those aged 18 years or lower, the odds did not increase with age; and for those aged > 49 years, the odds of undiagnosed HIV were significantly lower than among the age groups 25–34 years and 35–49 years. All categories of people who have ever been married were more likely to have undiagnosed HIV than those who had never been married, with widowers having the highest odds (aOR 3.34, 95% CI 2.58–4.33).

Table 4 Factors associated with HIV positivity among HIV− and unaware HIV+ testers
Full size table

The odds of having undiagnosed HIV were also higher among individuals screening positive for depression (aOR 1.16, 95% CI 1.04–1.28) and those screening for harmful drinking behavior (aOR 1.23, 95% CI 1.10–1.39) according to the PHQ-2 and AUDIT-10 scales, respectively. Those who had never tested for HIV were more likely to be undiagnosed (aOR 5.72, 95% CI 5.13–6.37) than those who had, as were those who perceived themselves as being extremely likely to acquire HIV in the next year (aOR 3.53, 95% CI 3.03–4.12). Neither selling sex in the last 6 months nor the number of sexual partners in the last 6 months were significantly associated with undiagnosed HIV.

Factors Associated with Retesting for HIV

The odds of retesting among individuals testing positive for HIV were lower for males compared to females (aOR 0.80, 95% CI 0.70–0.92) and those screening positive for harmful drinking behavior compared to those who did not (aOR 0.77, 95% CI 0.66–0.88) (Table 5). Having been forced to have sex was associated with a greater odds of retesting (aOR 1.26, 95% CI 1.10–1.46). Retesting was also associated with higher education level and perceived social status below ‘better off’. Retesting testing was not associated with age, district of residence, marital status, having sold or bought sex in the last 6 months, or the number of partners in the same period.

Table 5 Factors associated with retesting among PLHIV
Full size table

Discussion

Our examination of people testing for HIV in one of Uganda’s largest HIV service facilities reveals important information about who is getting tested and the characteristics of those testing for HIV. Perhaps most importantly, it has revealed the large proportion of PLHIV who seek retesting after having already been diagnosed. In the Mildmay context, as the clinic is a well-known treatment provider, it is possible that some HIV-positive clients may have opted for testing at Mildmay with the intention of seeking care or treatment at this facility. In our sample, individuals already diagnosed with HIV accounted for an excess 1776 people tested for HIV. At a national level, with more than 8.6 million people testing, and assuming prevalence of 7.1%, this could result in an additional 227,753 people being tested unnecessarily [9]. It overestimates the “positive yield” among tested clients and could also inflate testing coverage. It further contributes to an underestimation of linkage to care rates. Population-based surveys with viral load testing and testing for the presence of antiretroviral medications provide the best portrait of progress toward 90-90-90 goals and can help calibrate service data [19, 20].

Though adult HIV prevalence in Uganda is estimated at 7.1%, prevalence among testers at this facility who had not been previously diagnosed with HIV was four times higher (28.5%) [21]. Though unlikely, there may be a high concentration of people in Mildmay’s catchment area who are unaware of their infection status. That a large proportion of people living with HIV (PLHIV) come from Kampala suggests the need to assess the reach and quality of testing services in the city. It may be that people feel more comfortable testing farther from home to maintain a sense of anonymity. Simultaneously, the greater odds of undiagnosed HIV among people from Kampala suggest a greater need for prevention, testing, and treatment services in the city.

The odds of undiagnosed HIV infection were also higher among females than males, and ever married females in particular, especially those who are no longer married, emphasizing the need for testing campaigns to focus on these populations. Consistent with findings from other studies, those testing for HIV for the first time were more likely to be HIV-infected than people who had tested before [4, 5, 16]. This is to be expected as repeat testers had less exposure time since their last negative test than first-time testers. Repeat testing among the “worried well”, that is, people who are not at high risk for HIV but who nonetheless test regularly, may also account for the lower prevalence among repeat testers. Regular testing can decrease time from infection to diagnosis and the start of treatment with the result of decreased morbidity, mortality and risk for onward transmission [1, 22, 23]. However, focusing on reducing the time between tests may come at the expense of testing others for the first time. Testing people more often with the result of identifying more new infections may result in fewer PLHIV being identified as such infections are relatively rare, but as they may be in the acute phase of infection their viral load might be higher and the transmission risk higher [24, 25]. Data on community viral load among testers at Mildmay could facilitate prioritization of groups for testing. To decrease the use of health resources by the worried well, HIV counselors could advise people with no or little risk behavior of what constitutes risk behavior and when testing may be warranted.

The increased odds of undiagnosed HIV infection among those with depression and harmful alcohol behaviors highlights an opportunity for better linkages between mental health and substance abuse services with HIV services. Individuals with depression or who abuse alcohol could be tested for HIV and referred for services as needed [26]. HIV testing, self-tests, or referrals for testing could be co-located at or near liquor stores, bars, barbers, or other venues where men can be found [27].

In our study the odds of undiagnosed HIV infection increased significantly with each level of increased self-perceived risk of acquiring HIV. Outreach workers should be encouraged to screen for HIV risk behaviors and take extra efforts to connect individuals with higher self-perceived risk to HIV testing, including self-testing. Where HIV test kits are limited asking people their self-perceived risk of getting HIV may be a useful way to triage who to test and who to request to return for testing at a later time.

Over one-third (37.2%) of people diagnosed with HIV already knew they were infected. Such retesting of individuals puts a strain on supplies and human resources. Given the high positivity, it is possible that some of these people were silent transfers who had tested positive and possibly even started treatment elsewhere [28, 29]. Other possible explanations include being informed that they have been cured by a traditional or faith healer, believe that they may be cured because they are virally suppressed, or may have challenges accepting their HIV status. Populations with a higher odds of retesting may benefit from additional counseling after diagnosis and while on treatment to reinforce that HIV is a chronic and incurable disease. Qualitative research should be undertaken to determine why people with known infection are restesting and inform how to decrease retesting.

Our findings are limited by the cross-sectional nature of this survey, that we included only one testing facility, and that non-biological results rely on self-reported data. Nevertheless, our assessment of participant experience with ACASI suggests that the self-interview format promoted the provision of more accurate self-reported responses by participants, consequently increasing the validity of our findings. ACASI facilitates the collection of sensitive data and should be exploited to its fullest [30,31,32,33,34,35]. Due to the small number of individuals who tested multiple times during the study period, we are unable to describe behavior changes over time, or before and after a diagnosis with HIV. The lack of viral load data also hinders our understanding of transmission risk of people testing for HIV at Mildmay Clinic. Funding constraints prevented the testing of HIV viral load and the presence of antiretroviral medications. Such testing would facilitate an understanding of why so many who had already been diagnosed with HIV visited Mildmay for HIV testing. Conducting qualitative interviews of a sample of people testing at Mildmay could provide further insight. In the meantime, testing programs may wish to consider having HIV counselors ask clients whether they have already been diagnosed with HIV.

Conclusion

Understanding who is and who is not testing for HIV and who is aware of their infection is key to identifying people with undiagnosed infection. To reach these people testing strategies should focus on those who have never tested before. To better understand the proportion of HIV testing clients who have already been diagnosed with HIV, testing providers should ask clients for their status rather than assuming that all clients are undiagnosed. Positivity data from HIV testing facilities should be used with caution and adjusted to account for repeat testers who have already been diagnosed with HIV. Including an assessment of prior HIV diagnosis during pre-test counseling may reduce the number of PLHIV retested.