Compliance to Viral Load Monitoring Schedules Among Women Attending Prevention of Vertical HIV Transmission Services Before and During the COVID-19 Pandemic in Ehlanzeni District, Mpumalanga, South Africa

Human immunodeficiency virus (HIV) viral load (VL) monitoring was likely interrupted during the Coronavirus disease 2019 (COVID-19) pandemic. We used routine data on repeat VL testing among 667 prevention of vertical HIV transmission (PVT) clients in Ehlanzeni district, to determine compliance to VL testing recommendations and associated factors during different time periods: pre-COVID-19, transition, and COVID-19. Descriptive and multivariable Poisson regression analyses were conducted, with and without including revised PVT-guidelines rolled out in January-2020. Among 405 women with ≥ 2 VL tests, the overall median age was 30 years (interquartile range: 26–35 years). Compliance to recommended VL testing guidelines ranged between 81.5% (172/211) and 92.3% (191/207) at different time periods. Across all three periods and when revised PVT-guidelines were used, being compliant was significantly reduced among those with earliest VL = 50–999 copies/ml (incidence rate ratio (IRR) = 0.71 [95% confidence interval (CI) 0.61–0.82], p value < 0.001) and VL ≥ 1000 copies/ml (IRR = 0.18 [95% CI 0.09–0.36], p value < 0.001). When guideline revisions were excluded, compliance was only significantly reduced among those with VL ≥ 1000 copies/ml (IRR = 0.14 [95% CI 0.06–0.32], p value < 0.001) and increased during the COVID-19 period versus pre-COVID-19 (IRR = 1.10 [95% CI 1.05–1.15], p value < 0.001). Similar significant associations between compliance and VL level were observed when the COVID-19 period was analyzed separately. Significantly increased compliance to VL testing among the 25–34 years age-group versus younger women was also observed across all periods. These results highlight the importance of strengthening strategies such as short message service reminders and educational messaging, reaching all age-groups, to fast-track implementation targets for VL monitoring. Supplementary Information The online version contains supplementary material available at 10.1007/s10461-023-04192-z.


Introduction
The World Health Organization (WHO) recommended viral load (VL) testing for monitoring the effectiveness of antiretroviral treatment (ART) in human immunodeficiency virus (HIV) positive persons in 2013 [1].VL monitoring is vital to tracking individual responses to ART, population-level effectiveness of ART, and progress towards United Nations Acquired Immunodeficiency Syndrome (UNAIDS) global targets for combating the HIV epidemic [2].The UNAIDS's third target in its 95-95-95 goals is to ensure that 95% of people on ART are virally suppressed by 2030 [3,4].However, progress towards the third target has been slow.In 2022, about 71% of people with HIV (PWHIV) globally were virally suppressed [5].New interim targets for 2025 were developed to accelerate progress towards the 2030 goals.One of the main 2025 targets is to have 95% of pregnant and breastfeeding women with HIV (PBWHIV) having suppressed VL for elimination of vertical HIV transmission (EVT) [5][6][7][8][9][10].The 2015 South African guidelines for Prevention of vertical HIV transmission (PVT) recommended a repeat VL in 1 month if VL ≥ 1000 copies/ml and within 6 months when VL < 1000 copies/ml, while the 2019 guidelines reduced the window period for repeat VL testing when VL = 50-999 copies/ml to 8-10 weeks and recommended the use of PVT codes on the National Health Laboratory Service (NHLS) VL requisition forms to electronically identify and permit frequent VL performed in PBWHIV at antenatal care (ANC), delivery, and postpartum, for rapid follow-up and care [11,12].
Progress towards EVT has been slow and new infections continue to occur due to existing gaps in PVT; (i) new HIV infections during pregnancy, (ii) late HIV diagnosis and initiation of ART, and (iii) sub-optimal retention in care and viral suppression [13][14][15][16][17][18][19][20][21].VL monitoring in this population has been sub-optimal, and compliance to repeat VL testing guidelines is not documented in South Africa [22][23][24].Documenting compliance to VL testing in lowmiddle income countries is not yet common practice, yet it is crucial for understanding progress towards EVT.One example in Kenya reported that between 2016 and 2018 only 6% of PBWHIV who used public hospitals achieved a repeat VL test within the recommended schedule [25].Additionally, PVT program activities are expected to have been impacted by the Coronavirus disease 2019 (COVID- 19) pandemic and the restrictions on mobility and trade that were imposed to curb it.
In many countries, HIV services were interrupted, resources shifted towards the pandemic, and healthcare workers were allocated to COVID-19 tasks.Additionally, lockdown restrictions and fear of contracting COVID-19 made it difficult for people including pregnant and breastfeeding women to access healthcare services, and conducting face-to-face encounters was near impossible [10,26,27].This may explain the decline of 25% or more in testing for HIV in six countries (Jamaica, Nigeria, Rwanda, Mozambique, Cameroon, Sierra Leone, and Botswana) among pregnant women [10].In the 2022 UNICEF report, pregnant women from the Republic of Moldova missed and delayed HIV-related laboratories tests during the COVID-19 pandemic [28].Reports of delays in VL monitoring among PWHIV when clinics and laboratories were functioning sub-optimally during COVID-19 have also been noted [29,30].Considering the existing gaps and effects of COVID-19, VL monitoring for timely management of HIV infection among PBWHIV was likely to be negatively affected, increasing the potential for vertical transmission.For these reasons, we conducted an exploratory study aimed at determining compliance (i.e., timely tests within guideline recommended number of weeks/ months) to VL testing guidelines, and associated factors among PVT clients, before and during the COVID-19 pandemic in Ehlanzeni District, South Africa, within a context of changing PVT-guidelines.

Study Design and Setting
This was a retrospective cohort analysis of pregnant women and postpartum mothers, who were previously included in a facility-based cross-sectional study (the PHANGISA study).The PHANGISA study took place in the eight largest community healthcare centers of the Ehlanzeni District, located in rural Mpumalanga province of South Africa, from September to December 2019.Ehlanzeni's HIV prevalence during pregnancy is the highest in the province and is one of the districts with the highest HIV prevalence (20%) in the general population nationally [31][32][33].The PHANGISA study was designed to measure the prevalence of maternal VL non-suppression among pregnant or postpartum women with HIV and is described in detail in Ngandu et al. [34].
The 667 HIV-positive mothers who were enrolled in the PHANGISA study were retrospectively and virtually followed up using routine VL testing data from the NHLS' Corporate Data Warehouse (NHLS-DW).Study participant names, surnames and dates of birth were used to link study participants with their respective HIV VL test results entered in the NHLS-DW from 1-April-2019 until 30-September-2021 (the study period).Once test results were linked on demographic data, the NHLS-DW unique identifier-created to assign multiple tests to a single person-was used to extract additional VL tests performed on the PHANGISA study participants within the study district and nationally.

Sampling and Sample Size Determination
Sample precision was calculated assuming that all participants (N = 667) enrolled during the PHANGISA study would be included in the analysis.Based on this, the study had 100% power to detect assumed VL testing compliance of 38 ± 5% (based on the Mpumalanga province's known compliance) OR 90.2% power to detect compliance of 38 ± 3% OR 58.7% power to detect compliance of 38 ± 2% with 5% level of significance [35].

Inclusion Criteria and Exclusion Criteria
The inclusion criteria for this current study were: all HIVpositive pregnant and postpartum mothers (i) aged 15 years and above enrolled in the primary study, (ii) with at least two VL test results in the NHLS-DW during the study period, and (iii) who reached 24 months postpartum only after the start of COVID-19 lockdown period (i.e., after 25-March-2020).Mothers with no NHLS-DW data or insufficient NHLS-DW data i.e., only one VL test throughout the study period, and those who reached 24 months postpartum before the COVID-19 period were excluded.

Primary Outcome
The primary outcome measure was compliance to repeat VL testing within each period (i.e., pre-COVID-19, transition, or COVID-19), and was defined as the proportion of participants with a timely test (i.e., within guideline recommended number of weeks/months) out of the sample expected to have a repeat test during that period.The samples for each period were determined as follows: pre-COVID-19 sample (Fig. 1    delayed or missed), within each of the three COVID-19 periods using 2015 and 2019 guidelines [10,11].Two sets of calculations were done for the primary outcome: one accounted for the rollout of the revised 2019 guidelines from January-2020 onwards wherein the time to a repeat VL test when earliest VL = 50-999 copies/ml was shortened from 6-monthly to 8-10 weekly (referred to hereafter as the 2015/2019-PVT analysis); the other set assumed a potential delay in the rollout of the revised guidelines and thus used the 2015 recommendations throughout the study period (i.e., the 2015-PVT analysis).

Main Exposures
The first main exposure variable was the time periods: pre-COVID-19, transition, and COVID-19, defined by when VL tests were performed (Fig. 1).The second main exposure variable was the COVID-19 lockdown levels (1 through to 5) at the repeat VL test which changed differentially over the course of the COVID-19 period.This second exposure was only applied to the COVID-19 period.

Other Independent (Exposure) Variables
Sociodemographic factors collected during the PHANGISA study recruitment, perceived to influence healthcare uptake and HIV indicators were evaluated [34].These included the participant's PVT stage at recruitment (pregnant or postpartum), maternal age in years and further categorized (15-24, 25-34, 35-49 years) in secondary analysis to understand outcomes among the historically known HIV high-risk adolescent and young women age-group, level of education (ranging from no formal schooling to higher learning), marital status (never married or never lived with a partner, married or living together, widowed, divorced or separated), source of income (employed, spouse/partner, depend on others and government grant/no income), household monthly income categorized according to the national poverty line household cut-off (≤ R3200 or > R3200) [36], body mass index (BMI) (obese, overweight, normal, underweight) [37], partner's HIV status (positive, negative, unknown), condom use frequency (never, sometimes, always), planned pregnancy (yes or no), gestation at first ANC visit (0-12 weeks, 13-20 weeks, 21-40 weeks) and number of ANC visits (0-4 visits, 5-12 visits) were categorized according to the World Health Organization four-visit focused ANC model [38], timing of HIV diagnosis and ART initiation (before recent pregnancy, before 28 weeks gestation, after 28 weeks gestation, around labour/delivery, at postnatal before or after 6 months), current ART regimen (first-, second-, third-line, unknown), missed an ART dose last 7 days (yes or no), and facing any ART adherence challenges (yes or no).

Data Management and Analysis
The primary outcome variable and exposure variables were generated using definitions explained earlier.Data were analyzed in Stata/BE-basic edition 17.0 [StataCorp, College Station, Texas USA] and R version 4.2.0 [2022-04-22 ucrt].The socio-demographic data and the longitudinal VL tests data were merged.Data were cleaned to exclude participants not meeting the inclusion criteria for this analysis.Proportions were used to describe population characteristics and to report compliance to repeat VL testing.The outcome variable was also presented as a binary of on-time versus delayed or missed for the rest of the analyses.The distribution of the outcome by various exposures was assessed using a Chi-square test, for the pre-COVID-19, transition, and COVID-19 periods, separately.A partially overlapping samples z-test which compares proportions for two dichotomous samples was used to compare compliance to repeat VL testing between pre-COVID-19 and transition periods, pre-COVID-19 and COVID-19 periods, and transition and COVID-19 periods.Data in each comparison comprised a

Ethical Considerations
Ethics approval for the study was obtained from the South African Medical Research Council (SAMRC) ethics committee (EC021-6/2020).All PHANGISA participants gave consent for their routine data to be accessed and used for other related research in the future.Permission from the NHLS-DW gatekeeper was obtained to provide VL results matched to PHANGISA study participants.

Sample Size Realization
Out of the target sample of 667, data for 526 (78.9%) participants were successfully extracted from the NHLS-DW using personal identifiers.A further 23% (121/526) of participants were excluded from the study using the study exclusion criteria, resulting in a final sample of 405 participants (Fig. 2).There were insufficient VL data for 243/667 (36.4%) participants.With the achieved final sample size (n = 405), the study had 89% power to detect VL testing compliance of 86 ± 3% with 5% level of significance.The number of participants included in the pre-COVID-

Socio-demographics
Of the 405 included in the analysis, the overall median age of study participants was 30 years (interquartile range: 26-35) (Table 2).Over 90% of participants had at least grade 8 education level.About 60% of the participants were receiving  less than R3200 of income monthly and approximately 35% did not know the HIV status of their male partner.

PVT-Related Factors
A majority (70%) of the participants were in the postpartum period at recruitment into the PHANGISA study (Table 3).Most (63.2%) attended their first ANC visit within 0-12 weeks of pregnancy, and around 75% were diagnosed with HIV before their current or recent pregnancy and had initiated ART.
When age was categorized to investigate the differences between the historically known high-risk age group (15-24 years) and older age groups, all results remained the same except for age, across the three stages and irrespective of PVT guidelines.There was significant overall increased compliance among the 25-34 years agegroup versus younger (15-24 years) women (IRR = 1.12 [95% CI 1.01-1.23],Poisson p value = 0.027) in the 2015/2019-PVT analysis-Table S3; and (IRR = 1.10 [95% CI 1.00-1.20],Poisson p value = 0.035) in the 2015-PVT analysis-Table S4).

Discussion
This study documented compliance to VL testing in PBWHIV before and during the COVID-19 pandemic in a rural district in South Africa, during April-2019 to September-2021.Given the study period overlapped with the rollout of the 2019-PVT guidelines in January-2020, three months before the COVID-19 hard lockdown in South Africa, we considered a scenario with and without the use of the revised PVT guidelines to delineate the effect of the pandemic in the context of changing guideline practices.Our results indicated that compliance to VL testing was largely affected by the frequency of recommended repeat tests irrespective of the COVID-19 pandemic status and that the COVID-19 pandemic lockdown delayed the implementation of the revised PVT-guidelines.The increased likelihood of being compliant during the COVID-19 period compared to the pre-COVID-19 period, observed only in the 2015-PVT analyses where frequent visits for low-level viraemia were not applied, supports the delayed implementation of the revised guidelines.The higher proportion of delayed/missed VL testing visits among those on first-line ART regimen during the transition period when frequent visits were applied, also supports delayed implementation of the revised guidelines, given the majority of those with increasing VL would be expected to be on first-line regimen.This likely reflected the sub-group with low-level viraemia who might have been prescribed the 6-monthly repeat visit instead of the 8-10 weekly visit.The association between first-line ART regimen and high VL was also confirmed in the baseline cross-sectional point prevalence study, supporting a delay in treatment intervention for those with detectable viraemia [34].These observations confirm that the COVID-19 pandemic interrupted the training of healthcare workers for the implementation of the 2019-PVT guidelines.Training should have been resumed with urgency after the pandemic restrictions were eased.
The second major finding of this study raises a conundrum over the implementation of the revised and previous guidelines alike.According to this study, compliance to more frequent VL testing visits, which are prescribed for those at highest risk of VHT, was sub-optimal even before the COVID-19 pandemic.In both PVT-guideline scenarios and across the entire study period, compliance to VL testing was significantly reduced among those with high-level viraemia VL ≥ 1000 copies/ml, who were routinely prescribed 4-6 weekly clinic visits.Compliance was significantly reduced when VL = 50-999 copies/ml (i.e., low-level viraemia), only when the revised guidelines were used, wherein low-level viraemia clients are prescribed 8-10 weekly clinic visits instead of the previous 6-monthly visits.Shortening time to repeat clinic visits for those with low-level viraemia was motivated by growing evidence of high risk for VHT [40].The revised guidelines imply an increase in the population of those requiring frequent visits amid sub-optimal prior implementation.Implementation interventions need to be prioritized urgently.
To improve VL monitoring compliance, continuous evaluation and strengthening of the existing interventions, such as short message service (SMS) appointment reminders sent to PVT clients as part of Accelerating Programme Achievements to Control the Epidemic program is vital [41].SMS reminders have previously been shown to improve retention and viral suppression among PVT clients, early infant diagnosis follow-up testing, and uptake of repeat HIV testing [42][43][44].Recommendations for improving outcomes include (i) eLABS, that deliver VL results to healthcare workers digitally as soon as the results are ready and are piloting SMS reminders for patients [45] and (ii) Results for Action (RfA) reports, that collate HIV VL results weekly and identify high VL belonging to PBWHIV for clinical action, using codes introduced in the 2019-PVT guidelines [11,12,46,47].Furthermore, some previously proposed interventions that have shown potential to improve and encourage VL testing include educational efforts on the benefits of VL testing, using point-of-care VL testing for quick result turnaround times and enhanced access to VL testing in resource-limited settings, enhanced adherence counselling, compensating patients, and services aimed at young PBWHIV as they are less likely to receive VL testing [33,[48][49][50][51]. Periodic evaluation will also help in understanding the gaps that exist around implementation of guidelines and assist in addressing the impact of natural disasters such as COVID-19 pandemic on PWHIV.
The better compliance to VL testing observed among older compared to younger women, observed when age was categorized in a secondary analysis, is not unexpected, given known challenges in achieving viral suppression, uptake of care and VHT risk in younger women [34,[50][51][52][53].The insignificance of age in the COVID-19 period sub-analyses, regardless of PVT-guidelines used, could be due to the universally enforced COVID-19 movement restrictions.Other participant sociodemographic factors such as education and income were not risk factors unlike in other related PVT indicators [50,51,54,55].
This study had limitations and strengths.Although the message of delayed implementation of guidelines is undeniable, the observed levels of compliance to VL testing are best interpreted as an over-estimation.This study had a higher (~ 82%) compliance to repeat VL testing among PVT clients during pre-COVID-  [35], and what was seen among PVT clients in Kenya (only 6% of repeat VL testing within recommended schedule) [25] and Republic of Moldova during the COVID-19 pandemic among PWHIV (28.6%) [28], but comparable to observations of a similarly small PWHIV sub-population in southwestern Uganda (80.3%) [55].A high proportion of women were excluded because of only one VL test observed (102/667) in the NHLS-DW, despite searching nation-wide routine data to account for women's mobility.Deaths of study participants due to COVID-19 or loss-to-follow-up may have contributed to the exclusions.However, the realized study sample was highly powered for the analyses of repeated measures and sufficiently portrayed changes (or lack thereof) over time.In contrast, the subsamples used for reporting point prevalence of compliance to VL testing at each time-period were insufficient for external validity.However, the distribution of age in the remaining sample is similar to that in the baseline sample as well as the national VL monitoring survey, indicating that the sample that could not be traced on the NHLS-DW (mostly due to inconsistent use of patient demographics and the absence of a unique health identifier) was random [22,34].This qualifies the appropriateness and reliability of the rest of the results attained from the repeated measures regression analyses.
The sociodemographic factors were collected at one time-point, thus the effect of time-changing factors was not measured.
Although there are a few limitations, the study managed to combine observations of routine VL testing among PBWHIV over a period of 30 months, including 12 months before the onset of the pandemic in South Africa, with an assessment of changes in PVT-guidelines within that period, and raised an alarm for implementation priorities.Given that there have been extremely limited assessments of compliance to VL testing for PBWHIV in the past, the implementation and evaluation of the 2019 recommendation to use PVT codes on the NHLS laboratory requisition forms to identify and track VL performed in PBWHIV at ANC, delivery and postpartum, needs to be prioritized.

Conclusions
Although the point prevalence of compliance to repeat VL testing were over-estimated, the study had high power to accurately monitor VL testing over a 30-month period and delineate the effect of the COVID-19 pandemic within a context of rolling out revised PVT-guidelines.Two major observations were made: Compliance to repeat VL testing was significantly reduced among those with viraemia and -a step-2)-included all participants with the earliest VL test (pre-COVID-19 earliest) and expected repeat VL test falling before the start of lockdown (i.e., before 25-March-2020); Transition sample (Fig. 1-b step-2)-included all participants with the latest observed VL test (transition earliest) in the pre-COVID-19 period and the expected repeat VL test falling during the COVID-19 period; COVID-19 period sample (Fig. 1-c step-2)-included all participants with the earliest observed VL test (COVID-19 earliest) and the expected repeat VL test both observed during the COVID-19 period.The numerators for timing of repeat tests are defined in Fig. 1-step-3 and

Fig. 2
Fig. 2 Final sample realization after exclusion and denominator for each period

STEP 1: COVID- 19 periods STEP 2: Denominator defining each sample period STEP 3: Timing of repeat test
. The difference between the calculated expected date and the actual observed VL test date was used to determine the timing of the VL test (on-time,

Table 1
Calculation of compliance to repeat VL tests during COVID-19 periods using 2015 and 2019 South African PVT-guidelines Univariable and multivariable Poisson regression (with robust error variance) analyses were conducted to identify exposure factors associated with compliance to repeat VL testing across the three periods (including all repeat VL tests) and during the COVID-19 period separately (sub-analysis).Exposure variables with an overall univariable Poisson p value < 0.200 were selected for each multivariable model.Factors-age, education level, and income were kept in both overall multivariable analysis and sub-analysis a priori.A significance level of 0.050 was used in the multivariable analysis to indicate a significant association with compliance to VL testing.All descriptive and inferential statistics involving the outcome variable were conducted twice, i.e., for the 2015/2019-PVT analysis and the 2015-PVT analysis.

Table 3 PVT
-related factors n = number of cases per category; ART Antiretroviral treatment; PVT Prevention of vertical HIV transmission; ANC antenatal care

Table 4
Factors associated with compliance to VL testing 2015 or 2019 guidelines among PVT clients across the three time periods (N = 700 repeat tests)

Table 4 (
Variables with missing data, Factors with overall univariate p-value < 0.2 bolded, Bolded multivariate significant p-value < 0.05.IRR Incidence rate ratio; CI Confidence Interval; n = number of cases per category; BMI Body mass index; ART Antiretroviral treatment; ANC Antenatal Care; PVT Prevention of vertical HIV transmission *

Table 5
Factors associated with compliance to VL testing 2015 or 2019 guidelines among PVT clients during COVID-19 period (N = 207)

Table 5 (
19 compared to what was *Variables with missing data, Factors with overall univariate p-value < 0.2 bolded, Bolded multivariate significant p-value < 0.05.IRR Incidence rate ratio; CI Confidence Interval; n = number of cases per category; BMI Body mass index; ART , Antiretroviral treatment; ANC Antenatal Care; PVT Prevention of vertical HIV transmission

Table 6
Factors associated with compliance to VL testing 2015 guidelines among PVT clients across the three time periods (N = 695 repeat tests)

Table 6 (
Variables with missing data, Factors with overall univariate p-value < 0.2 bolded, Bolded multivariate significant p-value < 0.05.IRR Incidence rate ratio; CI Confidence Interval; n = number of cases per category; BMI Body mass index; ART Antiretroviral treatment; ANC Antenatal Care; PVT Prevention of vertical HIV transmission *

Table 7
Factors associated with compliance to VL testing under 2015 guidelines among PVT clients during COVID-19 period (N = 207) , both before and during the COVID-19 pandemic; and the rollout of the revised PVT-guidelines for shortening time to the next clinic visit for a low-level viraemia, was delayed due to the COVID-19 pandemic restrictions initiated barely three months into the rollout.To improve compliance to repeat VL testing, especially among those with detectable VL and young women, there is a need to strengthen interventions by (i) training healthcare workers on the new guidelines, (ii) educational messaging on VL testing benefits, (iii) SMS appointment reminders, and (iv) adolescent-PBWHIV friendly services, all which have shown a potential to improve uptake and timeous linkage for appropriate treatment and care.