The HIV care cascade in sub‐Saharan Africa: systematic review of published criteria and definitions

Abstract Introduction The HIV care cascade examines the attrition of people living with HIV from diagnosis to the use of antiretroviral therapy (ART) and suppression of viral replication. We reviewed the literature from sub‐Saharan Africa to assess the definitions used for the different steps in the HIV care cascade. Methods We searched PubMed, Embase and CINAHL for articles published from January 2004 to December 2020. Longitudinal and cross‐sectional studies were included if they reported on at least one step of the UNAIDS 90‐90‐90 cascade or two steps of an extended 7‐step cascade. A step was clearly defined if authors reported definitions for numerator and denominator, including the description of the eligible population and methods of assessment or measurement. The review protocol has been published and registered in Prospero. Results and discussion Overall, 3364 articles were screened, and 82 studies from 19 countries met the inclusion criteria. Most studies were from Southern (38 studies, 34 from South Africa) and East Africa (29 studies). Fifty‐eight studies (71.6%) were longitudinal, with a median follow‐up of three years. The medium number of steps covered out of 7 steps was 3 (interquartile range [IQR] 2 to 4); the median year of publication was 2015 (IQR 2013 to 2019). The number of different definitions for the numerators ranged from four definitions (for step “People living with HIV”) to 21 (step “Viral suppression”). For the denominators, it ranged from three definitions (“Diagnosed and aware of HIV status”) to 14 (“Viral suppression”). Only 12 studies assessed all three of the 90‐90‐90 steps. Most studies used longitudinal data, but denominator–denominator or denominator–numerator linkages over several steps were rare. Also, cascade data are lacking for many countries. Our review covers the academic literature but did not consider other data, such as government reports on the HIV care cascade. Also, it did not examine disengagement and reengagement in care. Conclusions The proportions of patients retained at each step of the HIV care cascade cannot be compared between studies, countries and time periods, nor meta‐analysed, due to the many different definitions used for numerators and denominators. There is a need for standardization of methods and definitions.


| INTRODUCTION
The Joint United Nations Programme on HIV/AIDS (UNAIDS) adopted the 90-90-90 targets in 2014 to track progress towards ending the HIV epidemic. Targets to be reached by 2020 include that 90% of people living with HIV (PLWH) are aware of their status, 90% of those diagnosed initiate antiretroviral therapy (ART) and 90% of those on ART achieve undetectable viral loads to end the HIV epidemic by 2030 [1]. The HIV care cascade examines the attrition of PLWH from diagnosis of the infection to starting ART and achieving suppression of viral replication. It is used to monitor HIV programme performance and to identify gaps and opportunities for specific interventions to improve retention and outcomes [2].
The number of studies examining the HIV care cascade has increased steeply, from no such study before 2011 to over 161 studies in 2020 (based on a PubMed search combining free-text words "HIV", "care" and "cascade"). These studies found that in sub-Saharan Africa and elsewhere, significant gaps remain. For example in South Africa's North West Province, awareness of HIV status among PLWH remained similar at around 70% in 2014 and 2016 and was considerably lower in men than in women [3]. Among HIV-positive individuals presenting for initiation of ART in Dakar, Senegal, 16% were lost to follow-up within one year [4]. However, studies differ concerning the definitions, methods and calculations used to construct the care cascade [5][6][7]. It is, therefore, difficult to compare published cascade research across regions and calendar periods.
Only a few studies have evaluated the methods used to define the HIV care cascade in low-and middle-income countries or sub-Saharan Africa [2,[5][6][7][8]. To fill this gap, we performed a systematic review of studies published in the academic literature from sub-Saharan Africa to assess the different methodological approaches used to define the steps in the HIV care cascade.

| METHODS
We examined guidance on systematic reviews of observational studies [9,10] when planning this review and report our review according to the PRISMA statement [11]. The review protocol has been published [12] and was registered in Prospero (PROSPERO registration number CRD42017055863) [13].

| Search strategy and inclusion criteria
We searched PubMed, Embase and CINAHL for articles published in English from 1 January 2004 up to 3 December 2020, the date of database search. We used Medical Subject Headings (MeSH terms) and free-text search. The MeSH terms for HIV and AIDS, and terms "cascade", "continuum", "linkage to care", "retention in care" and "ART initiation" were cross-referenced with terms for 62 African countries. The detailed search strategy has been published elsewhere [12]. Studies were included if they reported on at least one step of the 90-90-90 cascade or on at least two steps of the extended cascade. Pairs of reviewers (AG and FV, CM and DK) screened titles and abstracts. Two reviewers (CM, DK) independently screened the full text of potentially eligible studies using a standardized eligibility checklist. Disagreements were resolved through discussion. Figure 1 illustrates the populations involved in the cascade (bubbles) and the numerators and denominators (connectors linking populations) proposed by WHO and others for crosssectional and longitudinal cascades [6,14,15]. Seven populations (steps) are involved: (i) PLWH (diagnosed or undiagnosed); (ii) PLWH who have been diagnosed with HIV infection (the first 90); (iii) PLWH who have been diagnosed and linked to care; (iv) PLWH who have been diagnosed and who are retained in pre-ART care; (v) PLWH who have been diagnosed and started ART; (vi) PLWH who have been diagnosed and are retained on ART (the second 90); (vii) PLWH who have been diagnosed, are retained on ART and are virologically suppressed (the third 90). A study reported on a step if it provided the number of patients in the numerator and denominator or a percentage of patients completing the step. We considered the step as clearly defined if the authors reported definitions for the numerator and denominator, such as the description of the population studied and methods of assessment or measurement.

| Data extraction
We developed a standardized data extraction sheet, which was piloted by four reviewers (AG, CM, DK and ME) and revised. Data extraction was performed independently in pairs by three reviewers (CM, DK, Leona Hulbert), with disagreement resolved by discussion. Data were then entered into an EpiData database [16]. We classified study design into cross-sectional studies, longitudinal studies and studies that combine both designs (mixed design). We defined regions of sub-Saharan Africa according to the classification used by the International epidemiological Databases to Evaluate AIDS (IeDEA) [17].

| Selection of eligible studies
Overall, 3810 articles were identified ( Figure 2). After removal of duplicates, 3364 articles were screened by title and abstract, of which 412 were considered potentially eligible and underwent full-text screening. Five studies reported on data from the Kenya AIDS Indicator Survey 2012 (KAIS) [19]. To avoid double-counting of data from the KAIS, we included the publication reporting on the largest number of steps in the cascade [20]. A total of 82 studies met the inclusion criteria.

| Study characteristics
The median year of publication of studies was 2015 ( Table 1)

| Data collection and management
In 61 (74.4%) of the studies, data collection was performed by the clinic or hospital where participants were in care. Sixtyone (74.4%) studies used data collected through routine clinical records, 35 (42.7%) used questionnaires and 33 (40.2%) studies collected samples specifically for the study. Twentyeight studies (45.9%) reported using data captured in an electronic database only, and most of the studies (56, 68.3%) had a personal identifier assigned to study participants, which allowed for tracking of the participants along the cascade (Table 1).   The number of different definitions used in the 82 studies for the numerators defining the different steps ranged from four definitions (for step "People living with HIV") to 21 (step "Viral suppression"). Similarly, the number of different denominators ranged from three (step "Diagnosed and aware of HIV status") to 14 (step "Viral suppression") ( Table 2). The median number of definitions used across the seven steps was 8 for the numerator and 8 for the denominator.

| Numerators and denominators along the cascade
The most frequently studied cascade populations were PLWH on ART with viral suppression (the third 90), followed by PLWH who have been diagnosed and are retained on ART (the second 90) and PLWH who have been diagnosed and started ART (Figure 3). In cross-sectional studies, PLWH who had been diagnosed with HIV infection was the most common denominator used to describe retention at a subsequent step. The number of patients retained on ART was another commonly used denominator for viral suppression.
As expected, in longitudinal studies, it was common to use the previous steps as the denominator for the subsequent step ( Figure 3). For example the proportion of PLWH on ART with suppression of HIV replication was based on the number who were retained on ART. The latter, in turn, was based on the number of PLWH starting ART and so forth. In recent years (2018 to 2020) the denominator used for the step "Start of ART" changed, with more studies using "Diagnosed with HIV, " rather than "Retained in care" ( Figure S1).

| People living with HIV
Fifteen studies reported numbers or percentages of PLWH [8,[22][23][24][26][27][28][29][31][32][33][34][35][36]: 10 were cross-sectional, three were cohort studies and two of mixed design. All but one reported on the setting in which the testing took place ( Table 2) and used the number who tested HIV positive as their definition of the numerator. One study included self-reported HIV status among people who declined testing. Denominators were also consistent, with all studies using the number of people tested for HIV ( Figure 3). Target populations tested differed between studies, ranging from adult residents of a high prevalence community to a random sample of the general population (Table 2).

| Diagnosed and aware of HIV infection
Twenty-four [20-24,26-31,33-35,37-46] studies reported on people diagnosed with HIV and aware of their status, with 19 providing definitions for the denominator and numerator. Ten were cross-sectional, 10 longitudinal and 4 included longitudinal and cross-sectional components (mixed design). Awareness of HIV infection either relied on self-reported information (questionnaires or interviews) or was assessed through clinical records; four studies did not specify the source of information. In most studies, denominators included the PLWH testing positive for HIV, but people attending HIV testing or the whole population were also used ( Table 2). Only for eight studies, HIV diagnosis was the starting point of the cascade.

| Retention in pre-ART care
Nineteen studies reported on retention in pre-ART care [  ART and 34 (74%) provided clear definitions of numerators and denominators. The majority of studies were longitudinal (29 studies, 63%), 13 were cross-sectional and four of mixed design. Studies not reporting a clear definition described this step as "currently on ART" or similar. Numerators were either self-reported or ascertained through clinical records, pharmacy records or pill counts (Table 2). Time points for assessing numerators ranged from retention for six months to ten years of ART. The most frequently used denominator was based on the patients starting ART (26 studies, 56%) ( Figure 3).

| DISCUSSION
This systematic review of published studies from sub-Saharan Africa examined the approaches used to define the steps in the HIV care cascade and the steps covered in studies. Its results show that the definitions of the steps along the HIV care cascade are highly heterogeneous. Only 12 studies assessed all three 90-90-90 steps. Many studies focused on a subset of steps, with most studies assessing the start of ART, retention on ART, and achieving viral suppression. These steps are easier to assess within a typical clinic-based cohort and are also part of the 90-90-90 goals proposed by UNAIDS, which might explain the numerous studies reporting on these steps. Fewer studies examined the earlier steps, including the proportion of PLWH who have been diagnosed with HIV infection (the first 90) or the step of linkage to care.
The various definitions used for numerators and denominators mean that the proportions of patients retained at each step cannot easily be compared between studies, countries and time periods. Also, the statistical combination of data in meta-analyses, for example, for one country or region, would be problematic in this situation because "apples are combined with oranges" [100]. The heterogeneous definitions may be explained by the focus of studies, which may have been on retention in general, not specifically on the 90-90-90 targets. Of note, for the step of viral suppression, where definitions could be expected to be more consistent than for other steps, we identified a total of 21 different numerator definitions. Changing viral load thresholds over time and different timepoints of measuring contributed to the many different numerators, reflecting changing recommendations for defining viral suppression. The denominators also varied widely, ranging from all PLWH to patients starting ART. In recent years, PLWH diagnosed with HIV became a more common denominator for the proportion of patients on ART, reflecting the introduction of the "treat-all" guidelines. "Retention in pre-ART" care does no longer apply with "treat-all".
Previous studies from high-income and lower-income settings have highlighted the lack of uniformity in definitions [5][6][7]101,102]. To the best of our knowledge, this is the first comprehensive review of the methods used to assess the full HIV care cascade in sub-Saharan Africa, including the definitions of numerators and denominators from cross-sectional and longitudinal studies. Most previous studies focused on one part of the HIV care cascade, for example highlighting the lack of uniformity following specific community-based testing strategies [103,104], or on other settings, for example highincome countries [7,105]. Several studies from high-income settings have previously highlighted the heterogeneity and lack of standardization [101,102]. Haber and colleagues recommended investments in population-based, longitudinal cohorts so that data can be linked at the individual level across steps (denominator-denominator linkage) and within steps (denominator-numerator linkage) [6]. We found that the majority of studies reporting on the cascade use longitudinal rather than cross-sectional data. However, denominator-denominator or denominator-numerator linkages over several steps were rare. Globally agreed-upon measures to quantify the HIV care cascade are needed to evaluate the scope of and factors that are associated with attrition along the cascade [6,14,105,106].
A strength of our study is that we included smaller surveys and cohorts that are not necessarily representative of the national level but highly relevant to many settings in sub-Saharan Africa. Some weaknesses need to be considered. Our review was based on a systematic search of the academic literature and did not consider other reports on the HIV care cascade. For example we did not include reports published by governments describing national care continua. Of note, a systematic review of national data found that only a few complete national continua of care were available, and there was much heterogeneity in the methods for determining progress towards the 90-90-90 target [5]. A previous systematic review identified 13 cascades from the USA, Canada, Denmark, Georgia and Australia and recommended the use of population-based data sets to improve comparability [7]. Another study examined the linkage steps to care and ART initiation following communitybased detection of HIV [107]. Similar to our research, the authors found that definitions of numerators and denominators and observation periods were heterogeneous and that a metaanalysis was inappropriate. We did not examine disengagement and reengagement in care. Such "churning in and out of HIV care" was beyond the scope of this review [108].
We did not systematically search for reports from the Population-based HIV Impact Assessment (PHIA) surveys and did not include them [109,110]. The PHIA project performed nationally representative cross-sectional surveys for 13 different countries, measuring the three 90-90-90 steps of the HIV care cascade. In contrast to the academic literature, multinational studies, which use a standardized methodology, can provide data that are comparable across countries. Where a population-based nationwide cohort is missing, combined information from cross-sectional studies, cohort studies and from health surveillance is often used to evaluate the HIV cascade of care. A weakness that applies both to the PHIA surveys and the academic literature is the limited coverage of data on the care cascade. In our review, we identified studies from 19 countries in sub-Saharan Africa. Therefore, data on the care cascade may still be lacking for some countries, including countries profoundly affected by the HIV epidemic. The IeDEA collaboration is an extensive network of clinical cohorts, which can address the later steps of the HIV cascade in a standardized way [17,111].

| CONCLUSIONS
In conclusion, the many different numerators and denominators used along the HIV care cascade limit the comparability between studies. It is not clear whether observed differences in results are due to real differences or different approaches to the calculation of numerators and denominators. Future studies assessing the HIV care cascade should provide clear definitions of numerators and denominators used. To identify gaps and opportunities for specific interventions to improve the cascade of care, we need standard definitions, such as those definitions proposed by the WHO [14]. AG, BR, CM, FV and ME conceived the study and wrote the protocol. AG, FV, CM and DK performed title and abstract screening. AG, CM and DK performed full-text screening. CM and DK performed data extraction and analysis. CM and ME wrote the first draft of the paper. All authors reviewed and approved the final manuscript.

A C K N O W L E D G E M E N T
We are grateful to Leona Hulbert for her help with data extraction and to Lea Sophie Heinrichsmeier for help with proof checking.

F U N D I N G
Funding of the International epidemiology Databases to Evaluate AIDS Southern Africa (IeDEA-SA) collaboration was provided by 10 institutes, centres and programmes of the US National Institutes of Health (NIH): the U.S. National Institutes of Health's National Institute of Allergy and Infectious Diseases (NIAID) (https://www.niaid.nih.gov), the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) (https://www.nichd.nih.gov), the National Cancer Institute (NCI) (https://www.cancer.gov), the National Institute of Mental Health (NIMH) (https://www.nimh.nih.gov), the National Institute on Drug Abuse (NIDA) (https://www.drugabuse.gov), the National Heart, Lung, and Blood Institute (NHLBI) (https://www.nhlbi.nih.gov), the National Institute on Alcohol Abuse and Alcoholism (NIAAA) (https://www.niaaa.nih.gov), the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (https://www. niddk.nih.gov), the Fogarty International Center (FIC) (https://www.fic.nih.gov) and the National Library of Medicine (NLM) (https://www.nlm.nih.gov/) under Award Number U01AI069924. The research reported is solely the responsibility of the authors and does not necessarily represent the official views of the U.S. National Institutes of Health. ME was supported by special project funding (Grant No. 189498) from the Swiss National Science Foundation (www.snf.ch). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the U.S. National Institutes of Health or any governments or institutions mentioned earlier.