Expansion and scale-up of HIV care and treatment services in four countries over ten years

Background Scale-up and expansion of antiretroviral therapy (ART) for people living with HIV (PLHIV) have been a global priority for more than 15 years. Methods We describe PLHIV at enrollment in care and ART initiation in Ethiopia, Kenya, Mozambique and Tanzania from 2005–2014 and report on enrollment location, CD4 count and loss to follow-up (LTF), death, and combined attrition (LTF and death) pre- and post-ART initiation over time. Pre-ART outcomes were estimated using competing risk and post-ART using Kaplan-Meier estimators; LTF defined as no visit within six months pre-ART and 12 months after ART start. Results From 2005–2014, 884,328 PLHIV enrolled in care at 350 health facilities, median age was 32.0 years (interquartile range [IQR] 26.0–42.0), and majority were female (66.5%). The proportion of PLHIV enrolled at primary and rural facilities increased from 12.9% and 15.3% in 2005–2006 to 43.5% and 41.7% in 2013–2014 (p<0.0001). Median CD4+ cell count at enrollment increased from 171 cell/mm3 in 2005–2006 (IQR 71–339) to 289 cell/mm3 in 2013–2014 (IQR 133–485) (p<0.0001). A total of 460,758 (57.4%) PLHIV initiated treatment. Cumulative risk of LTF for PLHIV prior to ART initiation 12 months after enrollment was 33.5% (95%CI 33.36–33.58) and 21.98% (95%CI 21.9–22.1) after ART initiation. Pregnant women and the youngest PLHIV group had the highest attrition after ART initiation, at 24 months 40.8% (95%CI 40.1–41.6) of pregnant women and 47.4% (95%CI 46.4–48.4) of PLHIV 15–19 years were not retained. Attrition at 12 months after enrollment among PLHIV regardless of ART status was 38.5% (95%CI 38.4–38.6). Conclusion Over 10 years of HIV scale-up in four sub-Saharan African countries, close to a million PLHIV were enrolled in care increasingly at rural and primary facilities with increasing CD4 count. Loss to follow-up from HIV care remains alarmingly high, particularly among pregnant women and younger PLHIV.


Methods
We examined routinely collected patient-level data from health facilities in Ethiopia, Kenya, Mozambique, and Tanzania as part of the Identifying Optimal Models of HIV Care in Africa study, funded by President's Emergency Plan for AIDS Relief (PEPFAR) through the US Centers for Disease Control and Prevention (CDC). In partnership with the Ministries of Health in each country, all health facilities received support from ICAP at Columbia University and offered a standard package of services, including HIV testing, pre-ART and ART care, and prevention and treatment for opportunistic infections, per each country's national guidelines. National ART eligibility guidelines in each country changed over the period of observation and are summarized in S1 Table. Ethics and administrative approvals were obtained from the Columbia University Medical Center institutional review board (IRB) and the Associate Director of Science Office at the CDC, as well as from review boards in each country.
The study population included all adult PLHIV �15 years of age who enrolled at the health facilities from January 1, 2005 through December 31, 2014. Patients who enrolled in care <12 months prior to the last date of data collection at each health facility were excluded. At all facilities, medical record data from routine care visits were entered into on-site electronic databases by trained data capturers with data quality support from ICAP. Loss to follow-up (LTF) was defined as not having a recorded outcome of death or transfer out and before ART initiation (pre-ART) as not having a clinic visit for >12 months and >6 months after ART initiation. Data on deaths and documented transfer to other facilities were ascertained from facility records. Time to LTF or death was calculated from the date of ART initiation to the date of death (if available) or the last visit date.
We describe characteristics of PLHIV at enrollment in HIV care and at time of ART initiation based on year of enrollment and country including age, sex, point of entry, CD4 cell count (CD4+), and WHO HIV disease stage (measured up to 90 days prior and 30 days after). Cochran-Armitage tests for proportions and Kruskal-Wallis tests for medians were used to compare characteristics of PLHIV at enrollment and ART initiation in the periods [2005][2006] and 2013-2014. We report cumulative incidence of LTF, death, and combined attrition (LTF and death) before (pre-ART) and following ART initiation among patients who were ARTnaïve at enrollment (patients reporting prior ART or current ART at time of enrollment were excluded from retention analyses). We also analyzed LTF, death, and combined attrition for all enrolled patients regardless of ART status in order to measure overall lack of retention at enrollment facilities among all PLHIV entering care. Survival analyses were conducted using competing risk estimators for pre-ART outcomes (treating ART initiation as a competing risk for pre-ART death and combined attrition, and ART initiation and death as competing risks for pre-ART LTF). For outcomes after ART initiation and for the analysis of all enrolled patients, Kaplan-Meier estimators were used. Unadjusted sub-distributional hazards (pre-ART) and Cox proportional hazards models (following ART initiation) were used to compare cumulative incidence by groups. Statistical analyses were performed in SAS 9.3 and Stata 12.

Results
A total of 884,328 patients enrolled in care at the 350 health facilities in the four countries between 2005 and 2014 ( Table 1). The median duration of data collection at the individual facilities included in the analysis was 8 years (interquartile range (IQR) 7-9). Over time, more patients enrolled at primary health facilities; among all PLHIV 12.9% enrolled at primary health facilities in 2005-2006 compared to 43.5% in 2013-2014 (p<0.0001) (Fig 1). The proportion of PLHIV enrolled at rural health facilities also increased from 15 (Table 4). Almost two-thirds (64.0%, 95%CI 63.5-64.6) of PLHIV 15-19 years of age and more than half (54.5%, 95%CI 54.3-54.7) of PLHIV 20-29 years were not retained at 36 months.

Conclusions
This analysis of almost one million PLHIV enrolled in care at 350 health facilities over ten years across four countries in Eastern and Southern Africa reflects the evolution of the HIV response over that decade, the progress made in expanding access to more populations with HIV, as well as the remaining challenges. By the end of the first decade of rapid ART expansion, a higher proportion of PLHIV entered care and started ART at primary health facilities and in rural areas, demonstrating successful decentralization of HIV services. Women continued to represent the majority of patients enrolled in care and starting ART, with a large increase in pregnant women starting ART over time. Median CD4+ at enrollment, while increasing over time, remained low at 289 cell/mm 3 in 2013-2014. Attrition was high-roughly 20% of all PLHIV enrolled in care had no further recorded visits and almost 35% of PLHIV not yet on ART were LTF or had died within 12 months. While early attrition was lower among PLHIV after starting ART compared to those not on treatment, by 36 months estimates for attrition for patients not yet on ART and those on treatment were around 40%. In the latest period of observation, 2013-2014, pre-ART attrition had declined to 26% at 36 months but for PLHIV on ART, it had increased from 34% to 41%. Overall, regardless of ART status, almost half (48%) of PLHIV across the four countries were not retained at the facilities where they enrolled at 3 years after entry into HIV care with the highest attrition observed among pregnant women and young PLHIV.
A key finding from our analysis is the enormous increase in access to HIV care and treatment over time. Decentralization efforts have led to greater availability of HIV services at primary health clinics and in rural areas. Previous studies have demonstrated comparable health outcomes among PLHIV who receive care at lower level health facilities and lower rates of lost to follow-up among PLHIV initiating treatment at primary health clinics. [9][10][11] A study from Malawi found that in one district, expanding access to ART from the district hospital to primary health clinics decreased average travel distance from 7.3 to 4.7 kilometers and led to a 10% increase in visit attendance. [12] Decentralization has also been accompanied by task shifting, including expansion of nurse initiated ART (NIMART) and development of differentiated care models, such as community ART groups, that are continuing to expand access and make care more patient-oriented. [13][14][15] Over the first decade of ART scale-up, women constituted the majority of PLHIV entering care and starting treatment. Our data reflect a dramatic increase in pregnant women initiating ART which could in part be due to improved reporting of pregnancy status at enrollment but mirrors changes in guidelines moving towards Option B+ (ART for all pregnant women). [16] We also found a high proportion of women enrolling in care who did not return after the first visit and alarmingly high LTF among pregnant women. Overall, 48% of pregnant women were lost before starting treatment and 31% of women who were pregnant at the time of ART initiation were lost to follow-up by 12 months. These data are consistent with previous reports of high loss to follow-up among pregnant women. [17,18] The follow-up period for this analysis ends at the time when many countries were expanding Option B+, however a recent systematic review of retention of pregnant and postpartum women in sub-Saharan Africa in the era of Option B+ by Knettel et al [19] found that only 76.4% (95%CI 69.0-83.1) were still in care at 12 months after enrollment. These results along with our findings from earlier periods underscore the urgent need to identify strategies to ensure that women remain in care and on treatment through pregnancy, breastfeeding period and thereafter. [20,21] Perhaps one of the most alarming findings is the large number of patients LTF which is consistent with other studies, including our finding of the highest attrition among pregnant women and younger PLHIV. [8,[22][23][24][25] We also report a large proportion (20%) of patients who did not return after their first visit to the health facilities which few other retention analyses have documented. Tracing studies have provided important estimates of outcomes among PLHIV recorded as LTF in routine care settings. In a recent systematic review of outcomes of PLHIV on ART who were lost to follow-up in Africa, it was noted that 34% of patients successfully traced had died and 23.9% had transferred to another health facility, and that over time, in later cohorts, deaths appear to have declined while silent transfer have increased. [8] In a meta-analysis of data from nine tracing studies, it was noted that among PLHIV LTF after ART initiation, approximately 22% had died, 22% were alive but not on ART, and 15% had transferred to another clinic. [22] The authors also found that women and PLHIV with less advanced disease at ART initiation were more likely to have undocumented transfers. It is likely that some of the PLHIV identified as LTF in our analysis had undocumented transfer and may still be in care, nonetheless, it is still concerning that close to half of all patients enrolled in HIV care were not retained at three years. It is also possible that the introduction of "treat all" approaches (ART initiation at HIV diagnosis) as recommended by the WHO starting in 2015 will improve attrition; however there are few reports of long-term patient outcomes following the introduction of new guidelines from resource limited settings. While the data used for this analysis are several years old and may not reflect outcomes in the "treat all" era, our findings underscore the urgent need to identify drivers of LTF and to develop differentiated service delivery models that meet the needs and preferences of high-risk PLHIV. [14,15] Encouragingly, our analysis showed improvement in disease stage among PLHIV at entry to HIV care over time reflecting expanded testing efforts and earlier engagement in care. While we saw fewer clinically advanced patients (WHO stage III/IV) at enrollment and increasing CD4 + counts at both entry and ART initiation, nonetheless, in 2013-2014, 36% of PLHIV enrolling in care across the four countries had CD4+ <200 cells/mm 3 . PHIA data highlight the continued challenges around early HIV diagnosis; in Tanzania (2016-2017) only 60% of PLHIV were aware of their HIV infection. [26] Even as attention has shifted toward 'treat all' with a focus on immediate ART initiation, it remains critical to identify and provide services targeted to the many PLHIV who continue to enter care and initiate ART with HIV. [27,28] Strengths of this analysis include the very large sample size, multi-country cohorts, long duration of follow-up, which allowed for three-year retention estimates, and the timespan over the first decade of HIV treatment scale-up. In addition, our data cover the landscape of care settings from tertiary hospitals in urban areas to rural primary health clinics and are from multiple sub-national regions across four countries. These data are highly representative of where most PLHIV receive care in RLS, particularly in sub Saharan Africa, and reflect typical outcomes observed in real world settings. A key weakness is that our data are limited to information recorded in medical charts and are subject to significant amounts of documentation gaps. We also cannot distinguish between gaps in care and missing data, for instance we are unable to determine whether half of all PLHIV enrolled did not have a CD4+ test or whether their results were not recorded in the clinical chart. We also have limited death data drawn only from medical charts.
This analysis of almost one million PLHIV enrolled in HIV care and treatment during a critical decade in the HIV response offers insights from programmatic level supporting the enormous successes of the scale-up of HIV services as well as highlighting the challenges that still remain to be overcome.