Risk and rates of hospitalisation in young children: a prospective study of a South African birth cohort

Introduction Children in sub-Saharan Africa (SSA) are disproportionately affected by morbidity and mortality; there is also a growing vulnerable population of children who are HIV-exposed uninfected (HEU). Understanding reasons and risk factors for early-life child hospitalisation will help optimise interventions to improve health outcomes. We investigated hospitalisations from birth to two years in a South African birth cohort. Methods Mother-child pairs in the Drakenstein Child Health Study were followed from birth to two years with active surveillance for hospital admission and investigation of aetiology and outcome. Incidence, duration, cause, and factors associated with child hospitalisation were investigated, and compared between HEU and HIV-unexposed uninfected (HUU) children Results Of 1136 children (247 HEU; 889 HUU), 314 (28%) children were hospitalised in 430 episodes despite >98% childhood vaccination coverage. The highest hospitalisation rate was from 0–6 months, decreasing thereafter; 20% (84/430) of hospitalisations occurred in neonates at birth. Amongst hospitalisations subsequent to discharge after birth, 83% (288/346) had an infectious cause; lower respiratory tract infection (LRTI) was the most common cause (49%;169/346) with respiratory syncytial virus (RSV) responsible for 31% of LRTIs; from 0–6 months, RSV-LRTI accounted for 22% (36/164) of all-cause hospitalisations. HIV exposure was a risk factor for hospitalisation in infants (IRR 1.63 [95% CI 1.29–2.05]) and longer hospital admission (p=0.004). Prematurity (HR 2.82 [95% CI 2.28–3.49]), delayed infant vaccinations (1.43 [1.12–1.82]), or raised maternal HIV viral load in HEU infants were risk factors; breastfeeding was protective (0.69 [0.53–0.90]). Conclusion Children in SSA continue to experience high rates of hospitalisation in early life. Infectious causes, especially RSV-LRTI, underly most hospital admissions. HEU children are at particular risk in infancy. Available strategies such as promoting breastfeeding, timely vaccination, and optimising antenatal maternal HIV care should be strengthened. New interventions to prevent RSV may have a large additional impact in reducing hospitalisation.


KEY QUESTIONS
What is already known?
• The Sustainable Development Goals highlight the need to prevent child morbidity and mortality. However, there are limited recent data on hospitalisation rates and determinants in sub-Saharan Africa (SSA), including in children who are HIV-exposed and uninfected (HEU), despite this region having the highest under-5 mortality rate.
• Children who are HEU represent a growing vulnerable population in SSA.

What are the new findings?
• Early-life hospitalisation occurred in 28% of children in our cohort, most frequently in the first 6 months of life, despite high vaccination coverage including the 13-valent pneumococcal conjugate vaccine (PCV) and excluding paediatric HIV infection.
• Infectious causes represented 83% of hospital admissions after birth; lower respiratory tract infection (LRTI) caused most hospitalisations (49%), with RSV accounting for 31% of episodes. In the first 6 months of life RSV-LRTI was responsible for 22% of all-cause hospitalisations and for 41% of LRTI hospitalisations.
• Children who are HEU had higher rates of hospitalisation in infancy through 12 months, compared to HIV-unexposed and uninfected children (HUU), as well as longer hospital stay.
• Prematurity or delayed vaccination were associated with higher risk of hospitalisation; detectable maternal HIV viral load was associated with higher hospitalisation rates in HEU children; breastfeeding was protective against hospitalisation, particularly from gastrointestinal causes.

What do the new findings imply?
• Young children in SSA continue to experience high rates of hospitalisation and infectious causes underly most hospital admissions.
• Renewed focus on measures to prevent hospitalisation in young children is needed, including interventions to encourage breastfeeding practices, timely vaccinations, optimal maternal health, and targeting at-risk children including those who are HEU or premature.
• New interventions against RSV-LRTI may substantially reduce hospitalisation in LMICs.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

INTRODUCTION
The Sustainable Development Goals highlight the need to prevent child morbidity and mortality, with the aim of reducing under-five mortality to less than 25 deaths per 1,000 live births by 2030. 1 Children in low and middle-income countries (LMIC) are disproportionately affected by morbidity and mortality in early life, particularly in sub-Saharan Africa (SSA), and hospital admission represents a key marker of child health worldwide. However, there are limited recent data on rates and causes of hospitalisations in LMICs, especially in the context of HIV exposure. Quantifying hospitalisation burden, identifying causes and determinants of hospital admission, as well as populations who are most at-risk, will allow development of targeted prevention and intervention strategies.
In SSA, HIV is a key risk factor for morbidity and mortality. Prevalence of HIV infection in pregnant women continues to remain high in SSA, however, access to antiretroviral therapy (ART) in pregnancy has expanded and vertical transmission of HIV has decreased substantially. 2 Therefore, numbers of children who are HIV-exposed and uninfected (HEU) are growing, with this population representing over 20% of live births in some countries. 2,3 Several studies have shown that, when compared to HIV-unexposed and uninfected children (HUU), children who are HEU are at higher risk of mortality, 4,5 impaired growth, 6,7 neurocognitive impairment, 8 or severe infections. [9][10][11] Studies have also reported that early-life hospitalisation may be increased in children who are HEU compared to HUU. [12][13][14] However, more evidence is needed to understand the causes and determinants of hospitalisation to improve child outcomes.
Investigating the factors that impact child morbidity and hospitalisation in SSA is critical to inform and optimise interventions, 15 and accelerate progress towards the Sustainable Development Goals. We prospectively investigated hospitalisation rates and determinants amongst children from birth through two years of life in the Drakenstein Child Health Study (DCHS), a large population-based birth cohort outside Cape Town, South Africa, in an area of high HIV prevalence.

Study Population
Pregnant women were enrolled in the DCHS during their second trimester from two primary public healthcare facilities in a poor, peri-urban setting. All women who were over the age of 18 years and intended to be resident in the area for at least one year were eligible. 16 Women gave informed consent and received standard antenatal care. All births occurred at the single central hospital, Paarl Hospital, serving this area. Mother-child pairs were followed from birth through two years. 16,17 Children received vaccination according to the South African Extended Programme on Immunisation schedule which included the 13-valent pneumococcal vaccine, rotavirus vaccine, 18 and vitamin A . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 12, 2023. ; https://doi.org/10.1101/2023.06.08.23289961 doi: medRxiv preprint supplementation per national policy. Delay in vaccination was defined at 14 weeks and 9 months as any vaccinations delayed by more than 2 weeks from the due date.

Hospitalisations
Hospitalisations were tracked by active surveillance at Paarl Hospital, the presenting and only referral public hospital for all children in this area. Study staff conducted daily rounds at the emergency unit and inpatient ward of Paarl Hospital, to ensure real time surveillance. Additionally, hospital staff and mothers enrolled were also sensitized to contact study staff when an admission occurred; there was a 24 hour study phone manned by a member of the study team available to all mothers. 19 Detailed information regarding each admission episode including sociodemographic factors, anthropometry, diagnosis and treatment were obtained by study staff and abstracted from hospital folders, with data recorded on study questionnaires. The main causes of hospitalisations were documented and grouped into lower respiratory tract infection (LRTI); gastroenteritis; other infections; and other causes (for example accident and trauma) based on the treating clinician diagnosis. Where multiple causes were recorded, the most severe cause was used in this analysis. The diagnosis of LRTI was made according to WHO clinical case definitions; all children had a nasopharyngeal sample taken at the time of LRTI for potential pathogens including respiratory syncytial virus (RSV) using nucleic acid amplification testing, as described previously. 20

HIV
Maternal HIV status was established antenatally at enrolment. As per South African prevention of mother-to-child transmission (PMTCT) guidelines, HIV testing of pregnant and breastfeeding women was performed routinely every 3 months during pregnancy and the postnatal period. All mothers with HIV were enrolled in the national PMTCT programme and initiated on ART as recommended by the Western Cape guidelines at the time. Enrolment started in March 2012 and ended in March 2015, and during this time South Africa adopted Option B+, with the recommendation to initiate triple ART for all pregnant women for life (efavirenz, emtricitabine and tenofovir). HIV-exposed children were tested at 6-10 weeks of age using a PCR test, 9 months and 18 months using a rapid test, ELISA or PCR. HIV data were collected by triangulating clinic and hospital folder information and through self-report interviews completed with mothers during the antenatal period, at-birth and postpartum. 21 Clinical measures of maternal CD4 and plasma viral load measurements antenatally were obtained through the National Health Laboratory Service system. Where there was more than one result the highest viral load and lowest CD4 were included.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 12, 2023. ; https://doi.org/10.1101/2023.06.08.23289961 doi: medRxiv preprint

Breastfeeding
Detailed longitudinal data were collected on type of feeding from birth, breastfeeding practices, formula feeding and introduction and use of complimentary foods based on maternal report at each study visit (6-10 weeks, 14 weeks, 6 months, 9 months, 12 months, 18 months and 24 months after birth). Where conflicting responses were given, earlier reports were prioritised; they were considered more accurate due to being closer in time to the event reported, which should minimise recall bias.
Duration of exclusive breastfeeding was derived based on maternal report and defined as occurring until the first maternal report of introduction of solid food or formula for child. 22 Dichotomous variables were also included based on international and national recommendations for duration of exclusive breastfeeding and any breastfeeding.

Sociodemographic variables
Sociodemographic variables were measured using validated questionnaires administered by trained study staff at an antenatal visit at 28 to 32 weeks of gestation. Sociodemographic variables including maternal education, household factors and maternal demographics, were collected using a validated interviewer-administered questionnaire. 17

Birth and Child variables
Prematurity was defined as less than or equal to 37 weeks gestational age at birth. Gestational age was calculated based on a second trimester ultrasound; when this was unavailable maternal report of last menstrual period was used. Child growth measures were conducted at birth and all postnatal study visits. The birth anthropometric measurements were conducted by study staff. Infant length was measured in centimetres to the nearest completed 0.5 cm, where recumbent length was measured up until 18 months using a Seca length-measuring mat (Seca, Hamburg, Germany), and standing length was then measured at 24 months using a wall-mounted stadiometer. Weight was measured in kilograms to the nearest 10g, in light or no clothing, using the Tanita digital platform scale (TAN1584; IL, USA). Infant weight and length measurements were converted to z-scores that adjusted for infant age, sex, and prematurity; where the Fenton growth chart was applied to anthropometry measurements captured from premature children (< 37 weeks' gestation) at birth and up till 50 weeks postmenstrual age. WHO Anthro software was used to adjust for age and sex in measurements taken from full term infants at birth through to 2 years, as well as age-corrected measurements using the Fenton charts for premature infants. 23 Children were classified as ever stunted if they had a height z-score (HAZ) <-2 standard deviations [SD], underweight if they had a weight-for-age z-score (WAZ) <-2 SD, and wasted if they had a weight-for-height z-score (WFH) <-2 SD.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Statistical analysis
The primary outcome was all-cause hospitalisation. All hospitalisations from birth to two years of life were included. The child-time at risk in the cohort was calculated by starting calculations on the day after birth and the time period was censored at two years of age, loss to follow up, or death, whichever occurred earliest. The time hospitalised was excluded from the at-risk period. Hospital admissions due to maternal reasons were not included. Duration of hospitalisation was calculated from date of admission to date of discharge inclusive; where either date was missing, these hospital admissions were excluded from rate calculations but included in other reporting (n=12; 3% of hospitalisations).
We used an age-stratified approach to compare hospitalisations since prior literature highlights higher risks at early ages. We therefore categorised time periods into: (i) 0 -12 months (infancy) with subdivisions of the first year into (ii) 0 -6 months and (iii) 6 -12 months, and (ii) 12 -24 months.
We did not sub-divide later ages due to smaller hospitalisation numbers in the second year. We report incidence rates for each of the time periods and investigated duration of hospitalisation. The main causes of hospitalisations are described and reported as a proportion of all hospitalisations (including those infants hospitalised at birth as a group), and separately as a proportion of hospitalisations subsequent to discharge after birth as a representation of community-acquired morbidity.
Results are presented for all children, as well as HEU and HUU groups separately. Cox Proportional Hazards models were conducted to assess HIV exposure status differences in hospitalisations, where standard errors were computed using the grouped jacknife method in order account for recurrent hospitalisations. Crude incidence risk ratios (IRRs) were reported for unadjusted analyses, whilst adjusted hazard ratios (aHRs) were reported for multivariable analyses. Confounders were selected a priori using a directed acyclic directed acyclic graph (DAG) (Supplementary Figure 1). These included household income and maternal education. Separately we assessed any factors that significantly differed between HEU and HUU groups as potential confounders. Sensitivity analyses were conducted excluding recurrent hospitalisations.
Other potential risk and protective factors for hospitalisation including prematurity, breastfeeding, immunisation status, and HIV-related variables (maternal CD4, viral load and ART) were also assessed using hazard ratios. Finally, association of markers of malnutrition with all-cause hospitalisation rates in both HEU and HUU groups was investigated. All analyses were performed using R software.

Patient and Public Involvement
The public was not involved in the design or conduct of our research.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Sociodemographic characteristics
There were 1143 live births to 1137 mothers; of these, 2 [0.2%] children were diagnosed with HIVinfection and were excluded from this analysis. Five children died within one day of birth and were excluded as they never entered the at-risk period, leaving a total of 1136 children (247 [21.7%] HEU; 889 [78.3%] HUU). Cohort retention was high, with 1000 children in follow up at 2 years (87.5%) with a total of 2075 child-year follow-up time. Baseline clinical and demographic characteristics of mother-child dyads are provided in Table 1. One third of children were born to households with an income of <R1000 (~$75) per month, and almost two thirds of mothers had only primary school education. Only 27% of mothers were employed, and there were high rates of maternal alcohol use (13%) and smoking (28%) during pregnancy. The median birthweight was 3.09 kg (IQR 2.71 -3.42) and gestation was 39 weeks (IQR 38 -40) at delivery; 186 (16%) children were born preterm, the majority (n=129, 69%) late preterm. Vaccination completion rates were high (98% at 9 months), although 42% experienced delayed vaccination at 9 months. HEU and HUU groups were comparable in household income, gestational age, birthweight and immunisation status, however, HEU children were born to mothers of older age with lower educational attainment. A greater proportion of HUU children had mothers who reported smoking in pregnancy compared to HEU (33% versus 12%, p<0.001). Overall, 84% children were ever breastfed, although median duration of exclusive breastfeeding was 1.5 months (IQR 0.7 -3.2) and of any breastfeeding was 6 months (2 -8); fewer HEU children were breastfed and for shorter duration compared to HUU children (p<0.001).

Causes of hospitalisation
Overall, LRTI was the most common cause of hospitalisation (39.3% of cases), with gastroenteritis (17.7%) second, across all ages and in the HEU and HUU groups respectively (Table 3)

Hospitalisation by HIV exposure status
Children who were HEU had a significantly greater probability of hospitalisation (Figure 1a).
Compared to HUU children, the incidence rate of hospitalisation was significantly higher for HEU  (Figure 1b). When birth hospitalisations were excluded, similar associations were found between prematurity, breastfeeding and vaccination delay and hospitalisation (Supplementary Table 5), although hazard ratios were lower for prematurity associations.

Other determinants of hospitalisation
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2023. ; https://doi.org/10.1101/2023.06.08.23289961 doi: medRxiv preprint Of maternal HIV-related factors, a detectable maternal HIV viral load during pregnancy (>40 copies/ml) was associated with hospitalisation in the HEU group between 6-12 months compared to mothers living with HIV who had an undetectable viral load (HR 4.43 [95% CI 1.52-12.87]) but there were no significant associations with maternal CD4 or timing of ART initiation.
In terms of concomitant malnutrition, there was a trend for the HEU group of hospitalized children to be more underweight, stunted, and/or wasted, across all hospitalisations compared to the hospitalised HUU group, particularly in the 12-24 month period. However, this did not reach statistical significance (Supplementary Table 6).

DISCUSSION
In this population-based birth cohort we found high rates of early-life hospitalisation and that more than a quarter of children experienced hospital admission, representing a substantial burden.
Secondly, LRTI was the most common reason for hospitalisation overall, largely driven by RSV-LRTI. Thirdly, HIV exposure was associated with a greater risk of hospitalisation in infancy and with longer duration of admission. Finally, prematurity and delayed vaccination were drivers of hospitalisation for all children, and higher maternal HIV viral load in HEU children, while breastfeeding was protective, particularly from gastroenteritis.
Overall, the incidence of hospitalisation was high, despite over 98% immunization coverage and excluding child HIV infection, ranging from 454/1000 person years in the first 6 months to 82/1000 person years between 12-24 months. This is comparable with previous reported hospitalisation rates in studies from South Africa. 24 The incidence of hospitalisation was greatest between 0-6 months of age, declining substantially into the second year, highlighting the key vulnerability and importance of the first 6 months of life, and the potential for maternal immunisation as an effective preventive strategy.
Infections were the predominant cause of hospitalisation, and our results support LRTI and diarrhoea as key causes of morbidity in children. 15 LRTI was the commonest reason for hospital admission, responsible for almost half of all hospitalisations outside of those occurring at birth. Further, in the first 6 months of life RSV-LRTI was responsible for 22% of all-cause hospitalisations and for 41% of LRTI hospitalisations. These data are consistent with our prior finding that RSV is the leading cause of LRTI in this setting, 25 especially in infants, 20,26 and with global data of the importance of RSV-LRTI in LMICs. 27 Importantly these findings highlight the potential large impact of new RSV preventive interventions (either monoclonal antibody or maternal vaccination), to reduce LRTI and associated hospitalisations in LMIC settings. 28,29 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2023. ; https://doi.org/10.1101/2023.06.08.23289961 doi: medRxiv preprint Our findings add to data from other studies that suggest an impact of HIV exposure on hospitalisation incidence and duration in infants, in both LMIC 13,14 and high income countries, 30 despite the scale up of maternal ART. However, our results show a longer at-risk period, until 12 months, compared to another South African study, which found infection-related hospitalisation risk lasted up to 3 months of age. 13 The increased hospitalisation risk may be due to multiple factors associated with early-life infection, 7 including higher exposure to potential infectious agents in HIV-households, immune dysregulation or reduced transplacental antibody transfer, and low breastfeeding rates. Studies have found that children who are HEU have lower transplacental transfer of maternal antibodies to specific pathogens and some infants have altered responses to vaccination. 31,32 However, another USA study reported HEU infants have increased hospitalisation despite robust antibody responses to vaccines, 33 suggesting cellular immune defects may underlie the susceptibility to infection. Detectable maternal HIV viral load in the peripartum period was a risk factor for HEU child hospitalisation, as reported by others, 13,34 indicative of maternal disease severity and immune compromise, and suggesting viral load monitoring and ART adherence as areas for prevention. We did not show an effect of timing of ART initiation (prior to pregnancy versus during pregnancy) which differs from a European study, where initiation of maternal ART prior to pregnancy reduced child hospitalisation from infectious causes. 35 In addition to HIV exposure, we identified several other factors associated with hospitalisation.
Prematurity was a risk factor for hospitalisation for all children adding to evidence of the association of prematurity with hospitalisation due to infectious causes. 36 Further, in HEU children, prematurity was more strongly associated with hospitalisation compared to HUU children, despite no HEU/HUU group differences in preterm birth or birthweight, suggesting an additive impact of multiple risk factors. Another study also found prematurity contributed to increased infection-related hospitalisation in HEU children, 14 however, in this study timely infant vaccination was poor and there were more preterm and low birthweight HEU than HUU children. Overall vaccination completion in out cohort was very high. However, vaccination delay was associated with hospitalisation, particularly for LRTI, consistent with evidence that optimising vaccination reduces hospitalisation risk. 13 These data highlight the importance of timely vaccination delivery, 15 an issue that has become even more important during COVID which negatively impacted on childhood immunisation programs.
Breastfeeding was protective against hospitalisation, although we had reduced power to detect an effect due to low breastfeeding rates across the cohort. WHO recommends exclusive breastfeeding from birth until six months of age, with continued breastfeeding alongside appropriate complementary foods thereafter, however, only 13% of the cohort breastfed to six months. Prior to 2013, breastfeeding was not recommended for women living with HIV which may have affected feeding practices and reduced our power to show a greater effect in HEU children. Our results identifying . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2023. ; https://doi.org/10.1101/2023.06.08.23289961 doi: medRxiv preprint breastfeeding as a protective factor are consistent with substantial evidence demonstrating the importance of optimal breastfeeding in preventing morbidity. 13,37 However, in the comparison of HEU and HUU groups, our hazard ratios held on adjusting for breastfeeding, fitting with other studies that have found breastfeeding reduces, but does not completely ameliorate, morbidity in children who are HEU. [38][39][40] We found a strong association between breastfeeding and reduced hospitalisation for gastroenteritis, which may explain the higher rates of gastroenteritis-associated hospitalisation in HEU children from 0-6 months. Both LRTI and gastroenteritis have previously been reported to be higher in HEU children than HUU, 5,11,30,[41][42][43] and early, exclusive, and prolonged breastfeeding are known to be protective. 34 Strengths of this study include the well-characterised cohort with detailed prospective data collection, high vaccination coverage, no paediatric HIV, and high cohort retention in a low socioeconomic status African setting. The longitudinal design with comparable controls allowed the assessment of HIV exposure status with heterogeneity in antiretroviral combinations and maternal immune status across the cohort which is representative of child populations in other LMIC settings across sub-Saharan Africa. 10 Limitations include the potential for missing hospitalisations; however, all children were hospitalised at a single central hospital as this is the only public hospital serving the area, active surveillance was carefully done, and there were low rates of loss to follow up across the cohort. The low breastfeeding rates across the cohort may have reduced power to detect differences, and the results need to be replicated in populations with high breastfeeding rates. Finally, the results may not be generalisable to high malaria prevalence settings or settings of high child HIV infection.

Conclusion
Despite high vaccination rates and no child HIV infection, there was a high incidence of hospitalisation, especially during infancy in HEU children, also associated with prematurity, delayed vaccinations, and lack of breastfeeding. LRTI was the most common reason for hospitalisation, and RSV was a predominant pathogen. Our findings indicate the potential for simple, affordable interventions to mitigate hospitalisation risk including encouraging prolonged breastfeeding and timely vaccination as per global guidelines. 44 We also demonstrate the importance of targeting caregivers and families of at-risk groups, including those who are premature and HIV-exposed, with prevention and intervention strategies. 15 In children who are HEU, optimising antenatal maternal HIV treatment and monitoring viral load may decrease morbidity. Finally, the high prevalence of RSV-LRTI as a contributor to hospitalisation especially in the first 6 months of life, suggest that new interventions to prevent RSV may have a major impact in reducing hospitalisation.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2023. ;

SA-MRC. MPN is funded by an Australian National Health and Medical Research Council
Investigator Award APP1174455.

DATA AVAILABILITY STATEMENT
The de-identified data that support the findings of this study are available from the authors upon reasonable request as per DCHS cohort guidelines.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 12, 2023 Vaccination delay defined as any of the vaccinations given at 14 weeks (or 9 months) and younger delayed by more than 2 weeks of due date. Abbreviations: ART = antiretroviral therapy; AZT = zidovudine; HEU = HIVexposed uninfected; HUU = HIV-unexposed uninfected; NRTI = nucleoside reverse transcriptase inhibitor; NNRTI = non-nucleoside reverse transcriptase inhibitor; ZAR = South African rand . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.