Low proportion of women who came knowing their HIV status at first antenatal care visit, Uganda, 2012–2016: a descriptive analysis of surveillance data

Uganda is located in East Africa and is composed of 34.6 million people [16]. As of 2018, there were an estimated 9 million women of reproductive age [17]. In 2016, Uganda’s fertility rate was 5.4 children per woman [18].

Among Ugandan adults, the HIV prevalence declined from 7.3% in 2011 [19] to 6.2% in 2016 [7]. In two national surveys conducted 5 years apart, the HIV prevalence was higher among women than men (8.2% versus 6.1%) in 2011 [19] and 7.6% versus 4.7% in 2016 [7]. HIV prevalence is also higher among women living in urban areas (9.8%) than those in rural areas (6.7%) [7].

Administratively, Uganda was divided into 116 districts at the time of the study. These districts were categorized into 10 regions on which HIV programming was and still is based [19]. The HIV prevalence differs in each of the 10 HIV regions, with Central 1 having the highest prevalence at 8.0% and West Nile having the lowest prevalence at 3.1% [7] (Fig. 1).

We conducted secondary data analysis of first ANC visit attendance obtained from the Uganda National Health Management Information System (HMIS). Health facilities collect information on different health variables and summarize them on various standardized HMIS forms. These are submitted to the districts, where the information is entered in the web-based District Health Information Software version 2 (DHIS2) [20]. The DHIS2 is managed by the Ministry of Health and collates data from all health facilities in Uganda.

Our data source was DHIS2 from the HMIS 105 report. The HMIS 105 report is an integrated monthly health facility report that contains, among other variables, monthly attendance figures for the Maternal and Child Health Services (MCH). The MCH section contains sub-sections on antenatal services including first ANC visit attendance and HIV status at first ANC visit. Starting in 2015, the indicator of ‘pregnant women who knew status before first ANC visit’ was further revised to also include ‘pregnant women who came with known HIV-positive status’ [20]. Thus the difference between ‘pregnant women who knew status before first ANC visit’ and ‘pregnant women who came with known HIV-positive status at first ANC visit’ were the pregnant women who came with known HIV-negative status at first ANC visit. Women who attended first ANC visit with documentation of an HIV negative test result obtained within the past 4 weeks were considered to have a ‘known HIV negative status at first ANC visit’ [8]. The rationale for the stringent four-week period is to be able to capture the pregnant women that might have seroconverted at the earliest opportunity. Women who had an HIV care card or a documented HIV-positive test result from a test done at any point in the past were considered to have a ‘known HIV positive status at first ANC visit’ [8].

We extracted data on total first ANC visit attendance and pregnant women who knew their HIV status at first ANC visit for all 116 districts in Uganda during 2012–2016. For known HIV positive status at first ANC visit, we could only extract data for 2015 and 2016 for the reason above. We also extracted data on pregnant women who newly tested HIV-positive for the first time at the time of their first ANC visit during 2015 and 2016.

We used Microsoft Excel, Epi info, and STATA 14 for analyses. Prior to analysis, we categorized the districts into the 10 regions used for HIV programming in Uganda, adapted from the AIDS Indicator Survey of 2010–2011 (Fig. 1) [19]. We used frequencies and proportions to report sample characteristics at national level. We calculated proportions of women with known HIV status at first ANC visit at national and regional levels and used line graphs to describe the trend of known HIV status at first ANC visit for the period 2012–2016.

We determined the annual incidence of women with known HIV status by calculating the proportion of women that came with known HIV status at first ANC visit of the total first ANC visit attendance for that year. We used the improved Poisson with generalized linear models to examine whether known HIV status at first ANC visit increased over the period under study. The known HIV status at first ANC visit was the outcome of interest and year was the independent variable. We opted for the modified Poisson regression because the outcome variable was a count - that is the number of women that came with a known HIV status at the first antenatal care visit, and we were able to obtain incident rate ratios which were a closer association between dependent and independent variables than odds ratios from logistic regression models.

We interpreted the resulting incident rate ratio (IRR) as the average change in the proportion of women who came with a known HIV status at first ANC visit and used the 95% confidence interval to ascertain significance.

We also calculated proportions of women with known HIV positive status at national and regional levels and compared these for 2015 and 2016. However, we could not describe their trends because of the very short time frame of available data of only two years.

We calculated and compared the proportions of pregnant women that tested HIV-positive for the first time at their current pregnancy at national level for 2015–2016. This was for comparability with known HIV positive status at first ANC visit in the 2 years.

The number of women that attended first ANC visit ranged from 1,431,418 in 2012 to 1,715,377 in 2016. There was no statistically significant difference in the number of women that attended first ANC visit visits over the years 2012 to 2016. HIV prevalence at first ANC visit was 2.9% in 2015 and rose to 5.9% in 2016. The proportion of women that came with known HIV status was highest at 6.9% in 2016. The proportion of women that came with a known HIV positive status at first ANC visit in 2015 and 2016 was slightly higher than that of women that were newly tested HIV positive at first ANC visit in 2015 and 2016 (Table 1).

The proportion of women that came with known HIV status ranged from 63,270 (4.4% of first ANC visit attendance) in 2012 to 119,082 (6.9% of first ANC visit attendance) in 2016 (Table 1). This increase in proportion was significant (IRR = 1.14, 95% CI 1.14–1.14) (Fig. 2).

During 2012 to 2016, the proportions of women that came with known HIV status varied in the different regions and were less than 10% in all the regions except Kampala for the years 2015 and 2016. In 2015, the proportion of women that came with a known HIV status in Kampala was 10.0% and it rose to 11.1% in 2016. Most of the regions had an increase in trend of known HIV status over the years 2012 to 2016 except the West Nile and Mid-Eastern that had a declining trend (Fig. 3). The proportion of women with known HIV status in the West Nile region stagnated at 2.7% in 2012 and 2016, while in the Mid-Eastern region ranged from 3.4% in 2012 to 5.2% in 2016.

Our findings should be interpreted with the following limitations. We used DHIS2 data which is aggregate data and so we could not look out for individual effects such as repeat pregnancies in the same woman during the study period. Also, some variables were new and could not be assessed over the whole study period. Relatedly, the new variables (data elements) are initially not very accurate because the health workers that often double as data entrants take some time getting accustomed to looking out for and reporting them.

Our estimate of the proportion of women who knew their current HIV status at first ANC visit was likely an underestimate due to the documentation required to determine a known HIV status at first ANC visit. However, the underestimation was most likely skewed to the HIV negative women who had to have had a test within 4 weeks. The HIV positive women are less likely to be underestimated because they are more likely to report for ANC with their HIV care card.

Our analysis was only a bivariate trend. The effect of other characteristics that could have been considered in the final model such as the woman’s parity and age or whether she was a rural or urban dweller could not be assessed. This is because the data we used was aggregate and so could not account for individual characteristics. Other covariates such as government spending and donor funding would have been important to analyze, however, the donor funding by region differs from the national programming HIV regions considered in this analysis.

In addition, ANC data in DHIS2 have potential selection biases such as: distribution of public and private ANC services, misrepresentation since not all women attend ANC in DHIS2 reporting facilities and a small proportion opts not to attend professional ANC at all [23‐25]. Nevertheless, a large proportion of Uganda’s population attends public health facilities [26] and so the results can be generalized to the entire country.

Finally in countries with a mature and generalized HIV epidemic such as Uganda, ANC indicators are important sources of data in HIV surveillance and provide good data on epidemic trends over time [23‐25]. Our findings therefore can be used as proxy indicator of adult Ugandan women’s seeking behavior to know their HIV status, thus reflecting the national and sub-national trends of women of reproductive age who know their HIV-positive status in Uganda.