Background

Approximately 1.3 million infants are exposed to human immunodeficiency virus (HIV) each year through gestation, childbirth, and breastfeeding [1]. Despite tremendous global progress in expanding prevention of mother-to-child transmission (PMTCT) services, an estimated 150,000 children were newly infected with HIV in 2020 [2]. Approximately half of new infections occur during gestation and childbirth [3]. Disease progression among infants with HIV is rapid with mortality peaking in the first 2–3 months of life [4] and reaching 50% after 2 years [5]. Early diagnosis and antiretroviral treatment (ART) can significantly improve survival [6,7,8] and thus are critical to continue global pediatric HIV progress.

Conventional early infant diagnosis (EID), typically performed by centralized laboratories, is logistically complex. It requires caregivers to return to the health facility with their infants several times after delivery to initiate testing, receive results, complete follow-up testing, and initiate care. Despite substantial recent investment in diagnostic networks and centralized laboratory capacity, only 63% of HIV-exposed infants received an EID test by the recommended 4–8 weeks of age in 2020 [9, 10]. Further, nearly 40% are no longer in care by 18 months of age, with most loss to follow-up occurring in the first 6 months [11]. While conventional, central laboratory-based EID programs can reduce costs through economies of scale, this approach results in frequent diagnostic delays and loss to follow-up, limiting access to ART. Only 54% of children living with HIV received ART in 2020 [12].

Several strategies have been assessed to improve existing EID services and thus the health and survival of HIV-exposed infants. Point-of-care (PoC) testing improves turnaround times from sample collection to communication of results and ART initiation [13,14,15,16,17] and is recommended by the World Health Organization (WHO) [18]. Other interventions aimed at reducing turnaround time of conventional, laboratory-based testing, such as SMS printers, mobile/electronic health solutions, more efficient sample transport, and the use of hub-and-spoke models for EID have been evaluated on a limited basis in LMICs [19,20,21,22,23]. Adding HIV testing at birth offers potential to improve EID coverage and reduce pre-ART mortality through earlier identification and treatment of infants with HIV [24, 25]. Expanding access to EID beyond PMTCT programs offers the opportunity to identify infants who may be missed by conventional EID programs, especially in settings with high maternal HIV prevalence and low coverage of PMTCT services [26]. Further integrating HIV care for mothers and infants by providing combined interventions from the continuum of health and social services (e.g., adherence support, assisted disclosure of HIV status) as well as engaging the community in the delivery of health services (e.g., mentor mothers) can increase coverage, engagement in care, cost-effectiveness, and sustainability [27, 28].

Evidence of success of EID interventions identifying infants with HIV, improving linkage to care, demonstrating operational feasibility, and improving overall patient outcomes is accumulating [13,14,15, 17, 29]. However, limited evidence on the economic implications of these interventions is available. To inform decisions about EID program design and implementation, costs and cost-effectiveness estimates of EID are needed, particularly for high HIV burden, resource-poor settings. In this scoping review, we systematically summarize the available literature on the costs and cost-effectiveness of EID in low- and middle-income countries (LMICs). We also discuss practical implications and key limitations of existing studies.

Methods

We conducted a scoping review, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) checklist [30] as well as general related guidance [31]. A study protocol was made publicly available on the Open Science Framework on June 8, 2021 [27]. In line with PRISMA-ScR recommendations, we did not perform a quality appraisal of the included studies.

Information sources and search strategy

We searched the bibliographic databases Embase and MEDLINE (via Ovid) and EconLit (via EBSCOhost) for eligible pre-print and peer-reviewed records published in English between January 1, 2008 and June 8, 2021. We restricted our search to records published since 2008 based on 2008 WHO guidance recommending all HIV-exposed infants be tested by 2 months of age followed by immediate ART initiation for infants with HIV [32]. We also searched the archives of major HIV conferences (International AIDS Society conferences including AIDS and the Conference on Retroviruses and Opportunistic Infections) and Google Scholar (stopping screening after 50 irrelevant hits). The search strategy was based on four search terms clusters: HIV, infants, EID, and costs/cost-effectiveness (Additional file 1: Table S1).

Inclusion and exclusion criteria

We included studies of HIV-exposed infants in LMICs (defined by the World Bank classification [33]) exposed to interventions/programs aimed at improving access to EID and/or completion of the EID cascade [34] and reporting costs or cost-effectiveness outcomes. The EID cascade was defined as (1) identification of the HIV-exposed infant (known HIV exposure or symptomatic infant), (2) HIV testing, (3) communication of results, (4) linkage to care, and (5) ART initiation. If applicable, relevant comparators were alternative interventions or the local standard-of-care. We excluded commentaries, correspondence articles, and reviews, but screened the references of reviews to identify additional original articles for inclusion.

Outcomes

Our primary extracted outcomes were reported costs, cost savings, incremental cost-effectiveness ratios (ICERs), and net health or monetary benefit, as defined by Drummond et al. [35]. Secondary outcomes related to health or the EID care process (e.g., turnaround time, proportions of infants initiating treatment) were extracted as alternative disease-specific effects to enrich our discussion of the economic evidence for EID within the context of LMIC infant populations where health utilities are typically unavailable. For articles that did not report the reference year for costs, we assumed it to be 2 years prior to publication. All costs were converted to 2021 USD using the International Monetary Fund Gross Domestic Product (GDP) annual deflator for the United States [36].

Study screening, data extraction and analysis

Three review authors (KE, KEF, BPG) screened titles and abstracts of retrieved records after the removal of duplicates using Covidence [37]. Full-text review was conducted by KE, AE, and VO. Two review authors (AE and VO) extracted outcome data from the included studies to Microsoft Excel 16.60 (Microsoft Corporation, Redmond, USA), and a third review author (KE) cross-checked the data. Discrepancies were discussed among the authors and resolved by consensus. Extracted data were summarized in narrative and tabular formats. Descriptive statistics including frequencies and percentages and ranges of costs for comparable tests were compiled.

Results and discussion

Characteristics and data sources of included studies

We identified 1786 studies including 1011 studies from database searches and 775 studies from the references of reviews. After removing 508 duplicates, we screened titles and abstracts of 1278 studies and reviewed the full text of 99 studies. We included 29 studies on the costs and cost-effectiveness of EID. Reasons for exclusion were lack of cost data (51%), article type (e.g., review or opinion article) (17%), unavailability of abstract (10%) or abstract for which the full results were later published (7.1%), interventions not related to EID (7.1%), population not HIV-exposed infants (5.7%), or study setting outside of LMICs (1.4%) (Fig. 1).

Fig. 1
figure 1

PRISMA flow diagram

Full size image

Among the included studies, there were 12 model-based cost-effectiveness analyses (11 full texts and one abstract) and 17 cost analyses (14 full texts and three abstracts) published between 2008 and 2021 (Additional file 1: Table S3). All studies were conducted in sub-Saharan Africa except for one study from Thailand [38]. Nine cost analyses included primary cost data collection [39,40,41,42,43,44,45,46,47]. Cost estimates for other studies were derived from programmatic data, published estimates (e.g., Clinton Health Access Initiative, Global Fund [48, 49]), or the literature. Effectiveness data used in cost-effectiveness analyses were collected from Joint United Nations Programme on HIV/AIDS pooled analyses, WHO and UNICEF estimates, programmatic data, and the literature.

Costs per HIV early infant diagnosis test

We categorized EID tests into four groups: PoC-nucleic acid testing (NAT; e.g., Abbott m-PIMA, Cepheid GeneXpert®), laboratory-based NAT, rapid antigen- or antibody-based tests, and unspecified NAT. Currently, only NAT are recommended for EID [18]. Unit costs per test are reported in Table 1. Seven studies reported unit costs for PoC assays [13, 40, 50,51,52,53,54], 13 for laboratory testing [13, 24, 38, 39, 41, 44, 46, 50,51,52,53,54,55], two for rapid testing [41, 56], and one for unspecified NAT [57]. All unit costs are expressed in 2021 USD unless otherwise specified. Reported PoC-NAT cost per test were 21.46–51.80 USD. Costs for commercially available laboratory-based NAT were 16.21–42.73 USD.

Table 1 Unit cost per test for HIV early infant diagnosis
Full size table

Variation in unit costs for PoC assays may be explained by inclusion of capital costs. Costs for m-PIMA that included equipment costs were > 20 USD higher than those that did not include equipment costs. GeneXpert® costs per test were less sensitive to variation due to inclusion of equipment costs, with studies excluding equipment costs reporting 21.46–26.20 USD and those including equipment costs reporting 29.96–33.74 USD (Table 1). This may be due to the incorporation of utilization, including the ability to run multiple tests simultaneously, in the unit cost. m-PIMA can run one test at a time, whereas GeneXpert® analyzers support two or four tests run simultaneously. However, utilization is only relevant when equipment costs are included and only one study including equipment costs from Zimbabwe specified that they considered utilization [40]. The types of services that were included in the unit costs for centralized laboratory testing were more varied and less commonly detailed.

One study, conducted in Zambia, evaluated PoC p24 assays which may be more affordable than PoC-NAT tests (< 15 USD per test) and do not require specialized equipment [56]. Despite low sensitivity in very young infants, PoC p24 assays could play a role in diagnosing infants > 4 weeks of age at rural sites where the significant capital investment in PoC-NAT testing platforms is not feasible [56]. An assay that is 80% sensitive and links 99% of positive infants to care achieves the same level of ART coverage as an assay that is 95% sensitive and only links 85% of positive infants to care [58]. However, PoC p24 assays are currently not approved.

A study from Uganda reported that rapid antibody screening before EID testing of infants with a positive serology result was a cost-saving measure at 10–30 USD per test [41]. This is no longer recommended in the context of declining MTCT rates as well as wider availability of NAT and inferior sensitivity of antibody tests compared to NAT [18].

HIV early infant diagnosis program costs

Among 24 studies reporting costs of an EID intervention or program, these were reported as lifetime cost per HIV-exposed infant, average cost per HIV-positive diagnosis, cost per HIV-exposed infant person-year, or total price of the intervention/program (Table 2). Few studies assessed the same intervention or reported costs in the same way, making comparison of costs across studies challenging. Most studies evaluated costs or cost-effectiveness of EID testing approaches including four studies on PoC EID [45, 50, 52, 53], three on birth testing [24, 44, 59], one study that assessed both PoC and improvements to centralized EID [51], one that reported costs of confirmatory testing in EID programs [54], one of added screening of mothers at 6-week infant immunization visits with referral to EID for infants at risk of acquiring HIV [57], and one of rapid antibody screening to rule out negative infants before NAT [41].

Table 2 Costs of early infant HIV diagnosis interventions and programs
Full size table

Lifetime PoC EID testing costs were estimated at 264 and 470 USD per infant in Zimbabwe [51, 52] and 1.2–4.7 million USD total program costs [50, 53] in representative sub-Saharan African countries. Modelled total PoC EID program costs were slightly higher for m-PIMA compared to GeneXpert® but similar for settings with low and high PMTCT coverage [50, 53]. While unit costs for PoC EID are generally higher than laboratory-based testing, PoC testing addresses well-recognized challenges of conventional laboratory-based EID including improving turnaround times, increasing the proportion of infants with HIV initiating ART, and leading to earlier ART initiation [13, 15, 17, 29]. As initial investment in PoC-NAT platforms and infrastructure to support decentralized testing is significant [45], costs are highly impacted by throughput. Average throughput across eight sub-Saharan African countries in a 2019 study was 0.7–3 tests/day/health facility with an associated additional cost of 10 USD/test compared with optimal throughput (defined as 70% of platform capacity) in the same setting [13]. Integrating capital costs across programs (e.g., HIV viral load and tuberculosis testing) and/or health facilities via hub-and-spoke models and thereby increasing throughput can reduce costs [50]. Similarly, personnel sharing across services may increase efficiency without lowering the quality of services [46].

The discounted cost of birth testing from a modelling study in South Africa was 1012 USD per HIV-exposed infant with an in-utero infection rate of 1.8% [24]. The incremental cost of testing infants exposed to HIV at birth in Lesotho was 9114 USD per infant identified as infected at birth with an in-utero infection rate of 0.5%. This decreased to 2289 USD with an in-utero infection rate of 2%, similar to the undiscounted cost of 2140 USD per infant in the previous study. In countries with low coverage of PMTCT programs and higher in-utero infection rates (e.g., Nigeria [1, 2]) birth testing may be cost-effective compared to birth plus 6-week testing [44]. Targeted testing at birth only for infants at elevated risk of HIV acquisition (e.g., mother started ART late in pregnancy or has a high viral load around the time of delivery) reduces the burden on an already strained health workforce and therefore may be more appropriate for settings with low in-utero transmission rates [44].

Studies of other service delivery interventions, including co-located post-partum maternal and child health services in South Africa [27], sample transport in Uganda and Nigeria [22, 60], consolidation of EID testing in a single lab in Uganda [61], electronic communication systems in Uganda, Mozambique, and Kenya [20, 23, 62, 63], and a quality assurance system modelled in five sub-Saharan African countries [43], were also identified (Table 2). One study reported costs of immediate versus delayed ART initiation following EID testing in Thailand [38]. Three studies focused on cost variations across region or type of health facility within existing programs [42, 46, 47]. These studies reported wide variation of cost estimates across settings and therefore recommended context-specific cost estimates to inform budgeting and planning [46].

Cost-effectiveness of HIV early infant diagnosis

Table 3 summarizes the results of the 12 cost-effectiveness analyses. All studies stated at least one of the interventions evaluated to be cost-effective or cost-saving. ICERs were expressed as incremental costs per year-of-life saved (YLS)/per life-years gained (LYG), per death averted, or per additional infant initiating ART within 60 days. One study modelled costs and effects separately and did not report an ICER [59] and one study only reported an ICER for mother-infant pairs [27], and these were not included in the table, however costs were included in Tables 1 and 2.

Table 3 Cost-effectiveness analysis of HIV early infant diagnosis results of included studies
Full size table

Included studies used the Cost-Effectiveness of Preventing AIDS Complications Pediatric model [24, 27, 51, 54, 57, 64] (i.e., a validated state transition model simulating individual costs and HIV disease outcomes [65, 66]), decision tree models [41, 43, 53, 59], and cohort state transition simulation models [38, 50]. Seven studies used a lifetime horizon for the model [24, 27, 38, 51, 52, 54, 57], while the remaining used time horizons of 5 years [53], 2 years [59], 1 year [43], and 18 months [41, 50]. Nine studies used a discount rate of 3% per year for both costs and health benefits [24, 27, 38, 41, 43, 51, 52, 54, 57], two studies reported only undiscounted costs and benefits [50, 53], and one study, an abstract, did not specify whether discounting was applied [59, 67].

Out of four cost-effectiveness studies comparing PoC-NAT to centralized testing, only two reported a willingness-to-pay threshold. Willingness-to-pay thresholds are vital for decision-makers to be able to assess whether resource allocation for an intervention is worth the investment and are often oriented at the country-specific per-capita GDP, particularly in LMIC settings (WHO CHOICE). ICERs per YLS for PoC EID were 52% [51] and 67% [64] of the country-specific (Zimbabwe) per-capita GDP. ICERs for studies that did not report a willingness-to-pay threshold ranged from 23 to 1554 USD per additional child initiating ART within 60 days and 90–5976 per death averted (2018 USD) and were lower for GeneXpert® compared to m-PIMA [50, 53]. Several models assumed 100% EID uptake [51, 52] which excludes the potential costs and benefits of improving access to EID. This assumption favors PoC testing because it is more likely to increase access to EID compared to laboratory-based programs.

While decentralized testing increases access and linkage to ART, it often comes with increased challenges of supply chain management and maintenance. A system-level quality assurance system added to PoC EID programs and aimed at reducing screening interruptions and the misdiagnosis rate was found to be cost-saving in four of five countries modelled [43]. The modelled quality assurance system included external proficiency testing, reports, and corrective action including supervisory visits, equipment maintenance, and refresher trainings. Quality assurance systems can easily be extended to other PoC testing applications and may improve the overall level of service at primary health facilities.

Confirmatory testing was also demonstrated to be cost-saving in South Africa [54], and two cost-effectiveness analyses of PoC testing included scenarios with PoC and laboratory-based confirmatory testing [50, 53]. Without confirmatory testing, more than 10% of infants initiating ART may not actually be HIV-infected in settings with similar MTCT rates to South Africa [54]. ICERs for confirmatory testing at the PoC versus laboratory were slightly more favorable [50, 53], and the WHO now supports PoC testing to confirm positive results [18].

Two cost-effectiveness studies comparing birth plus 6-week testing to 6-week testing only, conducted in South Africa and Lesotho, concluded that cost-effectiveness of birth plus 6-week testing was dependent on prompt ART initiation and the degree to which ART reduces mortality [24, 44]. Birth plus 6-week testing exceeded the willingness-to-pay threshold of 50% of per-capita GDP in South Africa when the added cost was > 7 USD or NAT costs exceeded ~ 36 (2021 USD) [24]. Several estimates included in this review of both PoC and laboratory-based NAT costs in real-world settings exceeded this value [13, 40, 44, 53].

Tracking of infants testing negative at birth to ensure they complete 6-week testing is crucial to detect intrapartum and early breastfeeding transmission. With loss to follow-up rates > 37% between birth and 6-week testing, 1-year survival for infants with HIV in South Africa was lower compared to testing only at 6 weeks of age [24]. Thus, targeted birth testing of infants at high risk of HIV acquisition may be more appropriate given the significant resource investment in testing and tracking of infants to ensure they complete follow-up testing and are linked to care.

ICERs for HIV exposure screening and referral to EID at infant immunization visits compared to standard 6-week NAT ranged from 10 to 78% of country-specific per-capita GDP in three sub-Saharan African countries [57]. Initial rapid HIV testing to screen out uninfected infants before NAT was stated to be cost-effective in Uganda, however, a willingness-to-pay threshold was not specified [41]. The latter is no longer recommended in the context of declining MTCT rates and inferior sensitivity of rapid diagnostic tests compared to NAT, as well as the wider availability and similar cost of PoC-NAT for EID. Rapid diagnostic tests for HIV serology are recommended for diagnosing HIV in children > 18 months [18].

Knowledge gaps and practical implications

Several gaps in the literature on the cost-effectiveness of EID were identified here. Compared with effectiveness studies, sources of heterogeneity across economic evaluations are more numerous, limiting generalizability of cost-effectiveness results [68]. Cost-effectiveness analyses in this scoping review most commonly compared costs and health benefits of an intervention with current best practice or standard-of-care. Comparison of results across studies is complicated by the fact that standard-of-care is typically not well defined, differs greatly across settings, and is changing rapidly in many countries. Future cost-effectiveness studies will need to carefully consider further changes to these standard-of-care comparisons to accurately guide decision-making.

Lack of data availability in resource-poor settings, both for costs and long-term outcomes of children living with HIV, means model parameters are informed by few estimates from the literature, and it is often necessary to combine data from multiple sources (Table 1 and Additional file 1: Table S2). Cost-effectiveness analyses included in this review made efforts to use the best available data at the time of the study and used sensitivity analyses to compensate for uncertainty, however, the resulting long-term model predictions are still subject to considerable uncertainty. Considering that resource use and opportunity costs are highly context-dependent, decision-makers should focus on the most applicable studies to their settings to effectively distribute resources rather than attempting to synthesize less applicable results from multiple studies. Where generalizable results are unavailable, conducting further economic evaluations could be considered, incorporating local data on costs and where possible, outcomes of children with HIV [64].

Intervention scenarios discussed here generally assume that existing human resources would be sufficient to cover scale-up of EID interventions including task-shifting testing from laboratories to health facilities with PoC EID. This assumption may be unrealistic in settings where uptake of EID is expected to increase. Future economic analyses could include health system constraints by limiting the feasible coverage of interventions to align with current capacity or account for increased human resource costs related to expanding services.

In the absence of available data, the cost-effectiveness modelling of EID presented here does not incorporate additional activities designed to increase uptake, retention in care, and adherence to treatment. This may include traditional service delivery in healthcare settings as well as community health workers and/or mentor mothers. As a result, there remains a limited understanding of the impact of a comprehensive package of services for EID. With many countries moving towards widespread PoC EID, there is an opportunity for economic evaluations to inform priority setting and support the design of optimal service delivery models, but empirical cost data is needed. Evaluations of EID interventions and programs could therefore consider including data collection of real-world implementation costs. Further, full costs of program delivery including outreach should be represented.

Lastly, there were no studies evaluating the costs or cost-effectiveness of routinely offered facility-based testing. As EID is mostly delivered as part of PMTCT services, infants born to mothers receiving inadequate or no PMTCT interventions who are at higher risk of vertical HIV acquisition are also the most likely not to receive a diagnostic test within the first 2 months of life. In settings with high maternal HIV prevalence and poor PMTCT coverage, facility-based testing of infants with unknown HIV status in a range of clinical settings can help close the gap in EID coverage. The yield of positive test results was found to be high for inpatient care and malnutrition clinics in a systematic review of EID testing outside of PMTCT services [69]. More data on the costs and cost-effectiveness of testing infants in specific healthcare settings as a strategy to reduce HIV-related mortality are needed.

Strengths and limitations

To our knowledge, this is the first review to broadly describe the economic evidence on multiple EID interventions and/or programs. We conducted a broad search of the literature including peer-reviewed and grey literature and extracted extensive information to summarize EID unit costs, intervention costs and cost-effectiveness findings. We also converted findings to a common currency to increase comparability. Limitations of our study include restricting our search to studies published since the 2008 WHO recommendation to test HIV-exposed infants for HIV by 2 months of age, which is, however, the period in which major developments in EID started. Additionally, while we used a broad coverage GDP deflator rather than a consumer price index, it is unclear how the relevant costs in the respective settings have changed since the studies were conducted. Comparison of economic evidence across studies was limited due to heterogeneity of studies in interventions and comparators evaluated, the scope of costs included, as well as assumptions made in terms of model design. Finally, we did not systematically assess the quality of the included studies and potential resulting biases, as is common for scoping reviews.

Conclusions

The available cost and cost-effectiveness evidence for EID of HIV covers a broad range of interventions and suggests most EID interventions are indeed cost-effective. Few studies reported cost or cost-effectiveness estimates for the same intervention in comparable settings, and resources included in the cost estimates vary widely. Thus, comparison of costs across studies is challenging. Relatively few studies included primary cost data collection, and several report a lack of context- and setting-specific cost data as a limitation. Similarly, cost-effectiveness modelling studies must make assumptions based on limited data both for costs and outcomes of children exposed to HIV.

Increasing uptake and coverage of EID will likely be achieved through a package of services supporting EID service delivery and engagement in care. The scope of studies in this review did not cover the additional costs and benefits outside of EID programs that such comprehensive service delivery would provide. Future cost and cost-effectiveness studies capturing costs and benefits of EID services as they are delivered in real-world settings are needed to support the needs of decision-makers.