What patient and health facility attributes predict retention and virologic suppression in private for-pro t health centers? A retrospective cohort analysis of data from a large private sector antiretroviral treatment program in Nigeria

Muyi Aina (  muyi.solina@gmail.com ) Solina Center for International Development and Research https://orcid.org/0000-0001-9686-7685 Zeena Yesufu Solina Center for International Development and Research Abdulateef Salisu Solina Center for International Development and Research Echezona Ezeanolue Solina Center for International Development and Research Charles Mensah Solina Center for International Development and Research Patrick Dakum Solina Center for International Development and Research


Introduction
The 2018 Nigerian National AIDS Indicator and Impact Survey (NAIIS), the largest ever population-based HIV impact assessment (PHIA) globally, showed a prevalence of 1.5% among people aged 15-69 years [1]. This represents a remarkable decline in HIV disease burden since the country began using sentinel surveys to track its epidemic. Prior to this, the national prevalence rose from 1.8% in 1991 to a peak of 5.8% in 2001 before declining to 5.0% in 2003 and 2.3% in 2015 [2] [3][4] [5]. This trend suggests that the massive efforts of the government and its development and donor partners to control the HIV epidemic during the last three decades have achieved results [3] [6] [7][8] [9] [10].
With its estimated 1.9 million people living with HIV, Nigeria remains the 4th largest contributor to the global burden of HIV [1]. In 2017, the country accounted solely for more than half of the 260,000 new HIV infections in West and Central Africa, of which 36,000 were in children [11]. It also contributed over half of deaths due to AIDS related illnesses in the region [11]. Nigeria's HIV epidemic is mixed, with persisting pockets of high prevalence [12]. The national prevalence of 1.5% masks subnational variations that range from 0.3% in Kebbi, Katsina and Jigawa States in the northwest region, to 5.5% in Akwa-Ibom state in the south-south region [1] [6]. Only 7 out of the 36 states in the country account for more than 50% of the national HIV burden, and 80% of the combined burden is attributable to only 13 of the 37 states[1] [13]. These disparities re ect wide inequities in poverty, education and access to HIV prevention services, as well as geographical differences in high-risk behaviors and harmful socio-cultural practises, stigma and discrimination [2][12] [14].
The HIV response in Nigeria is government-led but largely donor-funded, with many programs initially focusing on supporting the public sector's capacity to provide comprehensive HIV prevention, care and treatment services in secondary and tertiary health facilities [7] [8] [15] [16]. However, coverage of services remained limited by the inadequate number of service delivery points [17]. To bridge the coverage gaps, the government and its partners rolled out a strategy to decentralize ART to primary health centers and implemented a task shifting policy that allowed nurses and allied health workers to be trained and empowered to provide varying levels of services in the HIV prevention, care and treatment continuum [18][19] [20]. Currently, Nigeria still falls far short of the UNAIDS 90-90-90 targets set for 2020, with only about 1 million people, approximately 52% of PLHIV, currently on ART [11][21] [22]. Despite this huge gap in service coverage, the potential for Nigeria to expand access to HIV services through private providers remains largely under-tapped [20] [23]. Up to 65% of all citizens, and 72% of the poorest quintile, access basic health services from private sector outlets [24] [25]. As such, private providers can not only help accelerate progress towards the revised 95-95-95 targets by 2030, but also reduce inequities in access to HIV prevention, care and treatment services. There is currently little or no data on performance of private facilities in provision of HIV services in Nigeria.
Solina Center for International Development and Research (SCIDaR), with sub-grant funding from the Institute of Human Virology, Nigeria (IHVN), designed and began implementing a model to support private for-pro t facilities to provide HIV prevention, care and treatment services in Nigeria in 2013. This paper presents patient outcomes and predictors of those outcomes in adult patients receiving ART in private health facilities supported by the SCIDaR-IHVN program in Benue and Nasarawa states, and the Federal Capital Territory (FCT), all high burden regions in northcentral Nigeria between 2013 and 2019.

Methods
This study is a retrospective analysis of routinely collected longitudinal data for patients enrolled in the SCIDaR-IHVN private sector HIV care and treatment program in Benue, Nasarawa, and the FCT between January 2013 and November 2019. All adult patients (aged 18 years and above), who received ART for at least one month in any of the 214 private health facilities supported by the program were included in this study. Routinely collected patient-level demographic and clinical data was extracted from the PEPFAR Retention and Audit Determination Tool (RADET) database. The data was cleaned, organized and analyzed using Stata® 13 SE [26].

Data analysis
The outcomes of interest were viral suppression (< 1,000 copies/ml) in most recent viral load test [27], all-cause exit, death, loss to follow up (LTFU)), and regimen switch to 2nd line. Demographic and clinical care characteristics were compared across states using Pearson's chi squared tests of proportions for categorical variables, including gender, pregnancy status, year enrolled in ART, duration on ART (years), starting ART regimen base (Tenofovir, Zidovudine or other), current ART treatment line (1st or 2nd line), ART re ll appointment duration (1, 2 and 3 or more months), and host health facility support type (model facilities received more intensive technical assistance than standard support facilities). Means of continuous variables (age at ART start, baseline CD4 + count) were compared across states with one-way ANOVA.
Treating time on ART, in months, at point of exit as the analysis time, overall and state-speci c incidence rates with 95% con dence intervals were calculated for all-cause exit, mortality, LTFU, treatment stoppage and transfers out. In addition, state-speci c and overall incidence rates of regimen switch from 1st to 2nd line was determined. Cox proportional hazards models were used to identify demographic, clinical and patient care characteristics that were associated with all-cause exit, mortality, and LTFU in the study population. In addition, a logistic regression model was used to identify patient characteristics that were predicted regimen switch to 2nd line. Viral suppression rates (proportion of patients with most recent viral load < 1,000 c/ml) were calculated for each state and for the overall study population. The state-speci c rates were compared using a Pearson chi squared test. Predictors of viral suppression in the study population were determined using multivariate logistic regression models. Patient characteristics assessed for relationship with viral suppression include gender, age at ART start (years, categorized into quartiles), state of residence, pregnancy status, host health facility support type, ART enrolment year, duration on ART in years, months of ARV re ll (1 month, 2 months, or ≥ 3 months), baseline CD4 + count (categorized into quartiles), current regimen base, and contact testing status. Adjusted ORs for viral suppression by ART enrolment year and by duration on ART in years from the best t multivariate logistic regression model were charted to identify any trends over time.

Results
A total of 22,010 patients meeting the inclusion criteria were included in this analysis. This number included 442 patients (2.0%) who had commenced ART prior to joining the program in 2013. As shown in Table 1, 42.7%, 22.2% and 35.1% of study subjects were in Benue, FCT and Nasarawa respectively. Majority (70.9%) were female, with a signi cantly higher proportion of females in Benue than FCT and Nasarawa. Among the female study subjects, 3.9% were pregnant and 3.6% breastfeeding at time of data recording. Mean age at start of ART was 33.8 years (SD 9.9 years). Overall, almost 1 in 6 (17.3%) patients commenced ART in 2013 or earlier, with Benue contributing more to this group (19.9%) compared to FCT (12.7%) or Nasarawa (16.9%). Overall annual enrolment increased gradually to peak at 24.0% in 2016, after which enrolment began to decline. As at time of exit from the program or censoring, 31.8% of subjects had been on ART for a year or less and 12.9% had received ART for 4 years or more. Mean baseline CD4 + count was slightly higher for Benue-based patients at 385.2 cells/ml, compared to the overall mean of 374.5 cells/ml. The majority (81.5%) of patients were started on tenofovir (TDF) -based treatment regimens and few (1.0%) had switched to second line regimens as of time of analysis. As table 1 shows, majority of patients in FCT (54.7%) and Nasarawa (52.4%) re lled their medications monthly, while most patients in Benue (50.7%) re lled every two months. Overall contact testing rate was 85.9% on the program. A slightly lower proportion of clients in Benue (78.0%) had their contacts tested. 44.5% of patients in FCT, 33.7% in Nasarawa and none in Benue received their care in health facilities categorized as model sites.

FCT -Federal Capital Territory
All-cause exit, mortality, loss to follow up (LTFU) and treatment stoppage       Table 4, odds of regimen switch increased progressively as time on ART increased up till 3-4 years. Odds of switch was not statistically different between those patients treated for 4 years or more when compared to those treated for less than one year. Patients whose starting regimen was AZT-based were less likely to be viral suppressed compared to those started on TDF-based regimen (OR 0.72, 95% CI 0.61-0.86). As shown in Table 5, decreasing age at ART start was associated with progressively worse viral outcomes, with patients in the bottom age quartile having 29% decreased odds (OR 0.71, 95% CI 0.57-0.88) compared to those in the highest age quartile. Viral outcomes showed signi cant positive trends over time (ART enrolment year) as well as with increased duration on ART (Fig. 1)

Discussion
Few reports on patient level HIV treatment outcomes from Nigeria have been published, and there is no large-scale outcomes data from private for-pro t health facilities. The potential to improve equity and access to health services in low and middle income countries (LMICs) through better private sector engagement is widely recognized in the literature, yet very little empirical information on their effectiveness and quality of service exists [19,[28][29][30][31][32][33][34][35]. This study provides the rst private sector data on patient outcomes in a large-scale HIV treatment program in Nigeria. Benue, Nasarawa and FCT are all in the north-central zone of the country, with signi cant HIV burdens -5.3%, 2.0% and 1.6% respectively, compared to a national average of 1.7% [1].
Population estimates for viral suppression among PLHAs in Nigeria vary widely between geo-political zones, ranging from 33.7% in the south-south zone to 65.7% in the north-central [1]. We found an overall viral suppression rate of 75.6% in our study, strengthening the case for expanding the role of private for-pro t facilities in HIV care and treatment. Although the viral suppression rate compares favorably with other available data, it falls short of the last 90 of the UNAIDS 95-95-95 targets, pointing to the need to intensify programmatic support and capacity building for these facilities to enhance their quality of care and patient outcomes. This study found better retention in care, lower mortality and higher viral suppression rates in Benue, likely attributable to a longer history of large-scale HIV care and treatment services in the state, as it initially was the target of much programming support, being the rst epicenter of Nigeria's HIV epidemic.
Females had better outcomes, including retention in care and viral suppression, which is in keeping with other studies from Nigeria and elsewhere [36][37][38][39]. Better health seeking behaviour of women and earlier detection through Prevention of Mother To Child Transmission (PMTCT) interventions may contribute to the observed better outcomes. Low baseline CD4 + counts, use of AZT-based regimens and shorter duration on ART were associated with worse virologic outcomes in this study, consistent with ndings of other researchers in Africa and elsewhere [40][41][42][43][44][45][46][47].
Despite the potential that private for-pro t health facilities hold, engaging them in public and social services like HIV prevention, care and treatment services is fraught with challenges [29-31, 33, 35]. Private providers are poorly regulated with unclear standards of care that often translates into variable quality of care [29]. In many countries, effective engagement, oversight and regulation of the various private health-care providers may be constrained by imperfect strategic intelligence, limited nancial in uence and weak institutional capacity [31]. Private facilities are generally small in size, which, coupled with the nancial burden associated with private healthcare, results in fewer patients per facility, making program support less e cient [30]. In addition, the core business model of for-pro t by the private sector may limit transparency in operations and impose user fees which often makes it di cult for donors to reconcile their goals of minimizing out of pocket expenses [29,35].
Nonetheless, our study showed that intensi ed technical support for these facilities (model facility support) yielded signi cant improvements in quality of care and patient outcomes. Furthermore, performance improved with experience, and outcomes have continued on an upward trajectory (supplemental material). The scale, reach and acceptability of private health facilities make them especially critical to achieving UNAIDS 95-95-95 by 2030 targets.
This study was limited by the non-availability of some data for analysis. Trends in repeat viral load and CD4 + testing records could be informative in understanding determinants of disease progression in the study population but repeat test measures were unavailable for analysis. Also, because viral load testing was not universally accessible during the earlier years of program implementation, testing was prioritized for patients in whom it was speci cally indicated, such as those with signs of clinical failure or evidence of poor adherence. This may have resulted in an under-estimate of viral suppression rates and may affect the overall generalizability of our ndings. Declarations Ethics approval and consent to participate No patient identi ers were extracted from the RADET database for this analysis. This study was reviewed and approved by Nigeria's National Health Research Ethics Committee (NHREC).

Consent for publication Not Applicable
Availability of data and materials The datasets generated and analyzed for this study are not publicly available due to ethical reasons, but can be accessed from the corresponding author upon reasonable request.

Competing interests
The authors have no competing interest to disclose.

Funding
The private for-pro t HIV care and treatment program on which the study was carried out was funded through a sub-grant from the Institute of Human Virology, Nigeria (IHVN)'s Action to Control HIV Epidemic through Evidence (ACHIEVE) grant number 1U2G GH002099-03 from US Centers for Disease Control and Prevention.
Authors' contributions MA conceptualized the study, MA and ZY analyzed the data and prepared initial draft and revisions of manuscript and results tables.
AS, EE, CM and PD reviewed the manuscript and made critical inputs into its nalization. All authors approved the nal manuscript for publication. Figure 1 Trends in viral suppression on the private for-pro t HIV program