A cross-sectional community-level survey assessing knowledge and uptake of PMTCT services among women of child-bearing age was performed March – June, 2011 in the Kenya Medical Research Institute (KEMRI) and US Centers for Disease Control (CDC) Health and Demographic Surveillance System (HDSS) area in Nyanza Province, Kenya.

The HDSS covers 385 villages with a population of approximately 220,000. Three regions situated in Siaya County make up this area: Karemo, Gem and Asembo. [25] All three regions are rural. The HDSS was launched in September 2001 by the US Centers for Disease Control and Prevention (CDC) in collaboration with the Kenya Medical Research Institute (KEMRI) and serves as a community-based platform. The HDSS provides demographic and health information as well as disease- or intervention-specific information. Since its inception, the HDSS has collected 13 years of population-level data. Home-based HIV counseling and testing (HBCT) of the HDSS population has been conducted since 2008, but had not been implemented in all areas at the time of the study. The HDSS area includes 41 health facilities that provide care to pregnant women.

Women, maternal age 14 years and older (mothers 14–17 are considered emancipated minors), who were residents in the HDSS area, and had delivered a baby within the previous year (January to December 2010) were recruited. Residency was defined as having lived in the area for at least 4 consecutive months. We leveraged existing HDSS program data to identify and recruit two groups representing two target populations for prevention services: a random community sample to assess factors influencing uptake of general services (access to ANC and maternal HIV testing); and a second sample of HIV infected women, known via previous home-based testing and counseling in the region, to assess factors influencing uptake of HIV-specific services (use of ARVs and infant HIV testing). The same inclusion criteria applied to both general community and HIV-positive groups, with the addition that women in the HIV-positive group must have completed HIV-testing prior to the delivery of the infant.

Sampling was limited by the number of women meeting recruitment criteria in the HDSS database. We were able to generate a random list of 523 women from non-HBCT areas to represent a general community assessment of ANC and HIV-testing uptake. In HBCT areas, only 275 women were HIV positive, thus all HIV positive mothers who were diagnosed prior to delivery were approached.

Trained fieldworkers were assigned a list of names and locations for all selected participants. Village reporters assisted fieldworkers to locate the mother participant and introduce the study. Fieldworkers administered surveys and recorded GPS location on hand-held PDAs using electronic forms (Pendragon Software Corporation, Buffalo Grove, IL); paper forms were used as a back-up. Mothers' surveys were adapted from clinic-based surveys used previously in this region [24]. Women were surveyed regarding their knowledge, opinions and use of ANC and PMTCT services at last pregnancy.

Outcomes of interest included self-report of uptake of ANC, HIV testing, maternal ARVs for PMTCT, infant testing, and infant ARVs. Potential cofactors assessed included demographics, educational achievement, and marital status, as well as knowledge about HIV and disease transmission. Due to limitations in determining income through household surveys, asset-based indicators (ownership of goods, including mobile phones, cattle, television or refrigerator, and roof type) were used as measures of socioeconomic status. [26], [27] Fieldworkers were not informed of HIV status of participants, thus participant self-reported HIV status was used. In 2010, Kenya national PMTCT guidelines adopted WHO PMTCT Option A (zidovudine during pregnancy with infant nevirapine during breastfeeding for women without advanced HIV, or triple-drug antiretroviral therapy for women with advanced disease) with a provision to implement Option B (triple-drug antiretroviral therapy for all women, stopping after breastfeeding for those without advanced disease) in higher resource systems. As national adoption of new guidelines was slow, and only half of national facilities offered PMTCT, [12] we chose to assess self-report of any maternal ARV uptake at each of three time-points: antenatal, perinatal, and postpartum.

Descriptive proportions of those accessing services in the PMTCT cascade were generated. Data analysis utilized chi-square analyses with Fisher's exact tests for comparison of proportions using STATA SE version 11 (STATACorp, College Station, Texas). Wilcoxon Mann-Whitney tests were used for comparisons where continuous data were not normally distributed. Multivariate analysis using generalized linear models further assessed adjusted prevalence ratios of uptake. A priori covariates of age and education level were included in the model with correlates identified in univariate analyses. Variables were retained in the model if they were significantly associated with the outcome and/or if their inclusion substantially changed the estimates by 10%.

Analysis was modeled to assess key steps in the PMTCT cascade: correlates of ANC attendance were assessed among women in the community sample (n = 405); correlates of HIV testing were assessed among women attending ANC with unknown or previously negative HIV status (n = 362); and correlates of maternal and infant ARV uptake were assessed among self-identified HIV-positive women (n = 216).

Prior to study start, community engagement activities were held targeting the area senior health officials, the local community advisory board, the village reporters, chiefs and assistant chiefs. Written informed consent was obtained from all study participants, both to be interviewed and also to have their data from these surveys linked to their HDSS record. In Kenya, women with pregnancy are considered emancipated and were therefore able to consent to study participation without parental assent. All study procedures, including enrollment of emancipated minors, were approved by the University of Washington Institutional Review Board (#36022) and the Kenya Medical Research Institute Ethical Review Committee (#1714). Written permission was also received from provincial medical and public health offices.

A random sample of 523 women who delivered in the previous year was identified, 437 (83.6%) were located, and 405 (92.7% of those located) agreed to participate (Figure 1). Among 275 women known to be HIV infected through prior HBCT, 247 (89.8%) consented to participate. However, only 173 (70%) of those enrolled reported to fieldworkers that they were HIV-positive and subsequently answered questions about PMTCT. Thus, the HIV-infected sample includes 43 women from the random community sample and 173 women from the HBCT sample to total 216 women.

Among the 405 women in the general community sample, most were married (79% monogamous and 15% polygamous) (Table 1). The majority were of poor socioeconomic status; 42% held any job or small-business (less than 2% were salaried workers), and less than half of the women had completed a primary education. The median number of mobile phones per household was 1 (IQR 0-2), with 71% of women reporting one or more phones in the household. Under half of households reported cattle ownership (median 0, IQR 0-2), and >90% of households did not own a television or refrigerator. Approximately one third (29%) had better than a grass roof (a marker of higher socioeconomic status), and all but one of these were iron sheet.

Women reported a median of 4 (IQR 3-6) pregnancies. Although uptake of ANC was high (94%), most women started ANC late (median 5 months gestation at first ANC visit, IQR 4-6) and completed fewer than the recommended 4 ANC visits (median 3 visits, IQR 2-4). Uptake of skilled delivery, defined as a doctor, nurse or midwife, was 44%, with 20% reporting no assistance at all. Less than half of women reported delivering in a health facility (n = 170, 42.0%); 210 (52%) delivered at home and 14 (3.5%) delivered on the roadside while attempting to reach a health facility.

Most (94%) women from the random community sample reported attending ANC during the last pregnancy, among whom 89% reported that they were offered HIV testing and 87% reported being tested. Among women in the random sample, regardless of ANC attendance, 324 (80%) reported uptake of HIV testing, 4% were known positive, and 16% did not accept testing during the last pregnancy (Figure 2). Among women in the random sample who ever agreed to HIV testing, prevalence of HIV by self-report was 11% positive, 85% negative and 4% declined to answer.

Among 216 HIV-positive women, 82% reported receipt of maternal ARVs at any time-point (prenatal, labor and delivery, or postpartum), however uptake at all three time-points was considerably lower (54%). This was primarily driven by lower uptake during labor/delivery time-points (62% compared to 72% antenatal and 72% postpartum).

Seventy-nine percent of HIV-positive women reported administering infant ARVs for prevention of HIV transmission. Among the 81% (176/216) of infants tested for HIV, 11% of mothers reported that their infant was HIV-positive and 76% reported that the infant tested negative, 6% declined to report their child's HIV status, and 7% stated that they did not remember the test results. Timing of infant HIV diagnosis was not reported, though a national early infant diagnosis program with testing recommendations at six weeks was initiated in the previous year, December 2009. Among the 19 positive infants, 15 (79%) were reported to be receiving care and treatment for HIV, 1 child died prior to receiving treatment, and another 3 did not receive treatment because of fear of others finding out (n = 1), denial (n = 1), and being unaware of treatment (n = 2).

Factors associated with report of ANC uptake at most recent pregnancy among women in the community sample are detailed in Table 2. In bivariate analyses, completion of primary education was associated with uptake of ANC. Among women who did not access ANC, 12% had completed primary education compared to 48% of women who accessed ANC (p<0.001) [Adjusted Prevalence Ratio (aPR):1.09; 95% Confidence Interval (CI): 1.04–1.15]. Women who did not access ANC reported a higher number of pregnancies than those who accessed ANC (median 5 vs. 4, p<0.001) (aPR: 0.94; 95%CI: 0.89–0.99).

Of 362 women attending ANC whose prior HIV status was negative or unknown, 87% reported being tested for HIV, and most HIV testing (89%) occurred at the first ANC visit (Table 2). Cofactors associated with HIV testing were occupation (31% versus 15% of those not accessing vs. accessing ANC, p = 0.03), mobile phone at home (50% vs. 75% of those not accessing vs. accessing ANC, p<0.001), partner having been tested for HIV (36% vs. 60% of those not accessing vs. accessing ANC, p = 0.003), and knowledge of PMTCT (34% vs. 55% of those not accessing vs. accessing ANC, p = 0.01). In adjusted models, socioeconomic indicators (cattle, phone ownership, and roof type) and report of partner HIV testing were significantly associated with an increased likelihood of acceptance of HIV testing.

Factors associated with uptake of maternal ARVs are described in Table 3. Women who did not access any ARVs reported fewer ANC visits than those who used ARVs (median 2 vs. 4, p = 0.002), and they were less likely to have a partner who was tested for HIV (53% vs. 76% of those without vs. with ARV uptake). Among those who knew their partner's HIV status, having an HIV-positive partner was associated with uptake of maternal ARVs. Correlates of uptake of a complete course of maternal ARVs included more education (41% vs. 61% of those who had an incomplete vs. complete ARV course had finished primary school, p = 0.006). Women who accessed a skilled provider at delivery had increased uptake of ARVs specifically during the labor/delivery time point (p = 0.02). In multivariate analyses, adjusted for number of ANC visits, partner HIV testing, and access to a skilled provider at delivery, older age and a higher number of ANC visits during the last pregnancy were associated with uptake of any antiretrovirals for PMTCT. A full course of PMTCT was associated with completion of primary education. Adjusted prevalence ratios for each of the primary outcomes are detailed in Table 4.

Administration of infant ARVs for PMTCT was highly correlated with maternal ARV uptake. While 160 of 174 (92%) women who took ARVs also administered them to their infant, 10 out of 39 (26%) women who did not take ARVs gave their infants ARVs (p<0.001).

Among women known to be HIV infected through previous home-based counseling and testing, 14 (5.7%) declined to answer questions about their HIV status and 60 (24.3%) reported to our field workers that they were HIV-negative. In a sensitivity analysis attempting to account for the low self-report among known HIV-positive women, in which we considered the most liberal (that all known HIV-positive women who did not self-report HIV-positive status accessed PMTCT services) and conservative (that all known HIV-positive women who did not self-report HIV-positive status did not access PMTCT) estimates, the expected range of any PMTCT ARV uptake was 61–87%, ARVs at all three time points was 35–66%, and infant ARV administration was 59–84%. In the HBCT sample, we further compared characteristics of women who reported or denied HIV-positive status in both univariate and multivariate models, and did not find differences in education, marital status, economic indicators, delivery provider, or HIV knowledge. HIV-positive women who did not report their status were younger than those who did (median 25.5 vs. 29 years) (Table 5).