Summary
Background
The HITSystem efficacy trial showed significant improvements in early infant diagnosis retention, return and notification of infant test results, and earlier antiretroviral therapy (ART) initiation compared with standard-of-care early infant diagnosis services in Kenya. This study aimed to analyse data from the HITSystem trial to assess the cost-effectiveness of the intervention in Kenya.
Methods
In this analysis, we extrapolated results from the HITSystem cluster randomised controlled trial to model early infant diagnosis outcomes and cost-effectiveness if the HITSystem was scaled up nationally in Kenya, compared with standard-of-care outcomes. We used a micro-costing method to collect cost data, which were analysed from a health-system perspective, reflecting the investment required to add HITSystem to existing early infant diagnosis services and infrastructure. The base model used to calculate cost-effectiveness was deterministic and calculated the progression of infants through early infant diagnosis. Differences in progression across study arms were used to establish efficacy outcomes. The number of life-years gained per infant successfully initiating ART were based on the Cost Effectiveness of Preventing AIDS Complications model in east Africa. HITSystem cost data were integrated into the model, and the incremental cost-effectiveness ratio was calculated in terms of cost per life-year gained. Sensitivity analyses were done using the deterministic model with triangular stochastic probability functions for key model parameters added. The number of life-years gained was discounted at 3% and costs were adjusted to 2021 values.
Findings
The cost per life-year gained from the HITSystem was US$82⋅72. Total cost for national HITSystem coverage in Kenya was estimated to be around $2⋅6 million; covering 82 230 infants exposed to HIV at a cost of $31⋅38 per infant and a yield of 1133 infants receiving timely ART, which would result in 31189 life-years gained. With sensitivity analyses, the cost per life-year gained varied from $40⋅13 to $215⋅05. 90% of model values across iterations ranged between $55⋅58 (lower 5% threshold) and $132⋅38 (upper 95% threshold).
Interpretation
The HITSystem would be very cost-effective in Kenya and can optimise the return on the existing investment in the national early infant diagnosis programme.
Introduction
The ultimate objective of early infant diagnosis services is to identify and treat infants with HIV to optimise their survival. Each year in Kenya, nearly 54 000 infants are born to birthing parents who are HIV positive and nearly 5200 of those children acquire HIV.1 To prevent peak mortality occurring at age 2–3 months,2 health systems must be able to facilitate early testing, efficient notification of results, and immediate initiation of antiretroviral therapy (ART) among infants with HIV before they are aged 12 weeks. Over US$19 million is invested in HIV testing (for both infants and adults) in Kenya each year; ensuring timely results and action is necessary to maximise this substantial investment.3 The primary driver of cost-effectiveness for early infant diagnosis services is the number of infants diagnosed with HIV infection who are initiated on life-saving ART.
From 2006 to 2016, Kenyan early infant diagnosis guidelines called for infants perinatally exposed to HIV to receive HIV DNA PCR testing at age 6 weeks and then antibody testing at age 9 months and age 18 months with confirmatory PCR testing if the antibody results come back positive. Immediate ART initiation upon a HIVpositive PCR result at any age has been emphasised in practice guidelines since 2014.4 Despite substantial financial investment, systemic and structural barriers cause early infant diagnosis outcomes to fall substantially behind global targets for HIV diagnosis and treatment.5 In 2020, only 76% of infants exposed to HIV presented for early infant diagnosis care.1 Among those who present for care, late presentation for testing (ie, aged older than 7 weeks)6,7 and high rates of loss to follow-up before completion of early infant diagnosis at age 18 months6,8 are persistent challenges. The logistics of HIV DNA PCR testing at age 6 weeks present additional barriers to early identification and treatment of HIV: of the infants who receive testing, only 65% of caregivers are notified of their infant’s test result,6 with a median turnaround time of 55 days from sample collection to caregiver notification of results.9 Among infants and children identified as having HIV, 84% received ART in 2020.1 These delays result in late ART initiation, well after the peak in HIV-related mortality,2 with many infants dying before diagnosis. These gaps and inefficiencies in early infant diagnosis services substantially compromise infant outcomes and reduce the cost-effectiveness of prevention of mother-to-child transmission (PMTCT) and early infant diagnosis programmes.10
Among infants with HIV, initiation of ART can extend life expectancy by 26 years and is very cost-effective (US$800–930 per year of life saved compared with no ART, depending on regimen initiated);11 thus, strategies to increase access to paediatric ART are crucial to maximise the substantial investment in PMTCT and early infant diagnosis services. Similarly, efforts to increase the proportion of infants exposed to HIV that are presented to health facilities for early infant diagnosis would drive the return on existing investment and infrastructure. eHealth interventions present many opportunities to strengthen early infant diagnosis and other HIV services in health facilities.
In 2010, our team developed the HIV Infant Tracking System (HITSystem)—a web-based intervention designed for use in low-resource settings. As a system-level intervention, the HITSystem connects key stakeholders in early infant diagnosis services by sending algorithm-driven alerts to health-care providers at government health facilities and to laboratory technicians at national referral laboratories, and SMSs to mothers and caretakers of infants exposed to HIV. By strengthening time-sensitive communication and accountability for services, this intervention leverages existing technology to mitigate many of the known barriers that create delays in services, which often result in poor health outcomes in low-resource settings.
The cluster randomised trial to evaluate the HITSystem was guided by an implementation science framework.12,13 We collected data on predictors of delayed or missed services and timepoints across the 18-month early infant diagnosis cascade most vulnerable to drop out to inform intervention refinement and health policy.
Between February, 2014, and December, 2015, HITSystem intervention efficacy was assessed in a cluster randomised trial at six Kenyan government hospitals.12 The trial compared early infant diagnosis outcomes among infants enrolled in standard-of-care early infant diagnosis hospitals versus HITSystem-supported hospitals. Hospitals were matched on geographical region, resource level, and volume of patients (high, medium, and low). Matching was done across three standard-of-care hospitals and three hospitals that received the HITSystem intervention. Eligible participants were birthing parents with HIV aged 18 years or older with an infant younger than 24 weeks presenting for their first early infant diagnosis appointment. Median infant age at testing was 6⋅4 weeks (IQR 6⋅1–7⋅3) with 93⋅7% tested before 12 weeks of age. The primary outcome was complete early infant diagnosis retention, defined as completion of all eligible services specifically: (1) initiation of co-trimoxazole for opportunistic infection prophylaxis; (2) PCR testing; (3) return of the PCR result to the hospital; (4) successful notification of the mother of the PCR results; (5) initiation of ART (for infants who were HIV positive); and (6) retesting at 9 months and 18 months (for infants who were HIV negative only).12 Ethics review and approval was obtained from Institutional Review Boards at the Kenya Medical Research Institute (SERU3983) and the University of Kansas Medical Center (STUDY00144753), and all participants provided informed consent.
Compared with the standard of care, the HITSystem increased complete early infant diagnosis retention (60% for standard of care vs 85% for HITSystem; adjusted odds ratio [OR] 3⋅7, 95% CI 2⋅5–5⋅5, p<0⋅0001), including ART initiation among infants who were HIV positive (73% vs 100%). Intention-to-treat analyses for complete early infant diagnosis retention also showed a significant advantage for the HITSystem (54% vs 70%; 2⋅0, 95% CI 1·4–2⋅7; p<0⋅0001). Compared with the standard of care, the HITSystem also showed significant improvements in the secondary outcomes of turnaround time of PCR results (median 38·5 days [IQR 26⋅0–66⋅0] vs 20⋅0 days [15⋅0–32⋅0], p<0⋅0001) and mother notification of PCR results (23⋅0 days [13⋅0–33⋅0] vs 14⋅0 days [7⋅0–24⋅0], p<0⋅0001). Use of the HITSystem also resulted in an overall younger infant age at ART initiation (median age of 25⋅1 weeks [21⋅0–26⋅9] vs 17⋅5 weeks [13⋅7–21⋅8], p=0⋅04).12
Although the efficacy of the intervention was shown in our previous analysis,12 the question remains as to whether the HITSystem intervention is feasible in relation to its cost. Thus, the objective of this analysis was to evaluate the cost-effectiveness of incorporating the HITSystem into routine early infant diagnosis services in Kenya.
Methods
Data analysis
In this analysis, using a health-system perspective, we coupled trial efficacy data with intervention cost data and modelled the cost-effectiveness of the intervention in terms of the incremental cost per life-year gained of the enhanced HITSystem intervention compared with the current standard of care.
We extrapolated results from the HITSystem cluster randomised trial to model early infant diagnosis outcomes and cost-effectiveness if the HITSystem intervention was scaled up to a national level in Kenya, compared with standard-of-care early infant diagnosis. The modelled population includes all infants exposed to HIV who were tested in Kenya during the study intervention period: Feb 16, 2014, to Dec 31, 2015. The population was identified by reviewing the National Laboratory Database for all infants (younger than 2 years) tested for HIV during the targeted window. All PCR tests and results from government health facilities providing early infant diagnosis were recorded in this national repository, which we also used to estimate infant HIV prevalence for the same period: 6⋅5% (range 2⋅0–14⋅0%).
Sensitivity analyses were done by using the deterministic model with the addition of triangular stochastic probability functions for key model parameters (ie, epidemiological data, trial data of early infant diagnosis testing, retention, ART initiation, and intervention cost data; table). Stochastic modelling was done using @Risk software, version 8⋅2. For parameters examined in the sensitivity analyses we identified mostlikely values, low values, and high values. For most parameters we set the point estimate with the study data and varied the high and low values so that they were within 10% of the base-case value. There were multiple exceptions. First, the percentage of infants with HIV used the low and high values from national data. Second, the number of life-years gained per infant initiated on ART used the point estimate and upper and lower 95% CI values from the published trial results. Third, we capped the high value at 100% for parameters where a 10% increase in the base value exceeded 100%. Fourth, for cost data we used observed high and low values across the study sites. Fifth, we set the low value for clinic personnel at US$0 as some non-study sites have implemented the intervention programmatically through task shifting of existing staff without new or additional personnel support and the high value reflects the annual salary of one new mentor mother (ie, a peer staff member with experience as a former PMTCT client) hired to support the HITSystem. Finally, for the sensitivity range for the number of infants enrolled per month, we used the average as the base-case, with high (25 infants), medium (15 infants), and low (eight infants) values based on the maximum and minimum value observed across the six study sites. The average number of staff hours to support HITSystem implementation ranged between 8 h per week for low-volume sites and 20 h per week for high-volume sites.
For the sensitivity analyses, the base-case deterministic model with the stochastic functions added were iterated with a Latin Hypercube sampling method using a Mersenne Twister generator. Initial seeds were chosen randomly, and the same seed was used across all simulations. A model convergence threshold was set to force simulation termination when both the mean and SD on all model outputs varied by less than 1% across cumulative iterations. The model converged after 19 300 iterations. The sensitivity analyses was done using multivariate regression analysis on the resulting dataset generated by the model iterations with the cost per life year gained from the intervention as the dependent variable.
Discounting of 3% was used for the number of life-years gained. This discount value is consistent with the recommendation of the Public Health Service Panel on Cost-Effectiveness in Medicine,16 and the Second Panel on Cost-Effectiveness in Health and Medicine.17 Since the intervention only incurs a cost at the onset of the intervention, discounting of future costs was not required. We also adjusted costs to reflect 2021 values, which were current at the time of analysis. The intervention study ran from 2014 through 2016. All cost data were converted from Kenyan Shillings (KSH) to US$ averaging mid-year exchange rates for 2015 and 2016 (97⋅9 KSH: 1⋅US$). Using historical values of the Consumer Price Index from the Bureau of Labor Statistics we calculated current cost values by adjusting from the midpoint of the intervention study (July, 2015) to the comparable value for September, 2021, with a multiplier of 1⋅14.
Costing
We used a micro-costing method to collect cost data. Cost details were collected at each of the six hospitals on data entry forms provided to project managers. These forms were reviewed by the study team and queries were made as needed to clarify entries and harmonise metrics across the study sites. These data were compiled and summarised. Costs of the HITSystem were analysed from a health-system perspective, reflecting the investment required to add the intervention to existing early infant diagnosis services and infrastructure. Therefore, we did not quantify costs for providing standard early infant diagnosis services in Kenya (eg, personnel, equipment, supplies, other infrastructure). Microcosting of the HITSystem intervention was guided by established guidelines for costing HIV interventions.18 HITSystem implementation costs were estimated under the categories of personnel, technical support, training, equipment, and HITSystem intervention operations. Based on expenditures for HITSystem implementation observed at study and programme sites, low, medium, and high estimates were calculated for each expenditure across an 18-month implementation period, reflecting the duration for a cohort of infants to complete early infant diagnosis services. Furthermore, monthly early infant diagnosis patient volume was averaged for the low-volume, medium-volume, and high-volume health facilities included in the efficacy study. Research costs (eg, administering informed consent, participant remuneration, and research-specific staff salaries) were removed given the goal of nationwide programmatic implementation.
Role of the funding source
The study funder had no role in the study design, data collection, data analysis or interpretation, writing of the manuscript, or the decision to submit for publication.
Results
The cost per life-year gained from the HITSystem intervention was $82⋅72. Total cost of the intervention as modelled, which would be national coverage in Kenya, was estimated to be $2 580 080. This programme would cover 82 230 infants at a cost of $31⋅38 per infant and yield timely administration of ART to 1133 infants diagnosed with HIV (difference in the number of treated infants in the intervention minus standard-of-care arms; using the average perinatal HIV infection rate across low, medium, and high estimates) resulting in 31 189 life-years gained.
The results of the sensitivity analyses using the stochastic model yielded a point estimate for the cost per life-year gained from implementation of the HITSystem intervention of $87⋅07 across the 19 300 iterations of model runs. This estimate varied from a low value of $40⋅13 when model values were set to the most optimistic level, to $215⋅05 when model values were set most conservatively. However, 90% of model values across iterations ranged between $55⋅58 (lower 5% threshold) to $132⋅38 (upper 95% threshold; figure 1).
Figure 1: Sensitivity simulation results for cost per life-year saved when model values were set to the most optimistic and most conservative levels.
Costs are in US$.
By far, the cost per life-year saved is most affected by changes in the average number of infants enrolled per month at the clinic, with a high value of this parameter (larger number of infants coming to the clinic) reducing the cost per life-year saved to $59⋅74; and a low value inflating the cost per life-year saved to $129⋅75. Moderate effects on the outcome were seen for variations in the cost for personnel (low $73⋅16; high $106⋅33), technical assistance support ($79⋅99; $94⋅08), and the number of exposed infants ($80⋅41; $93⋅42). Minimal effects were found from variations in the licence fee, the assumption for the number of life-years gained per infection averted, communication fees (internet connection costs), training costs, and computers and modems (figure 2).
Figure 2: Bivariate sensitivity analysis showing effect of changes in each parameter on the mean value of the cost per life-year gained, rank ordered by their magnitude of effect.
Costs are in US$. ART=antiretroviral therapy.
As seen in bivariate analysis, the multivariate analysis showed that, by far, the largest effects on the cost per life-year saved are attributed to variations in the size of the programme and the cost for clinical personnel (figure 3). Modest effects are seen for variations in cost of technical assistance support, number of infants exposed to HIV (driven by HIV prevalence among mothers), and the licence fee. Trivial effects are found for variations in the costs for training, communication (internet connection costs), and computers and modems. In sum, the most cost-effective programme is one with hospitals enrolling a large number of infants per month that have modest personnel and technical assistance costs (eg, training existing mentor mothers or data clerks for system maintenance) in settings with high HIV prevalence among birthing parents.
Figure 3: Multivariate sensitivity analysis generated from stochastic model with inputs rank ordered by each parameter’s magnitude of effect on the cost per life-year saved.
Costs are in US$. ART=antiretroviral therapy.
Discussion
Previous analyses have underscored the importance of strengthening existing early infant diagnosis services, including uptake, retention, and quality, to reduce mortality and morbidity.10,19,20 The HITSystem had improved quality (receipt of complete early infant diagnosis care) and efficiency (turnaround time of services) of early infant diagnosis in a cluster randomised trial in Kenya.12 Applying the benefits obtained in the trial to national early infant diagnosis data, we modelled the cost-effectiveness of universal adoption of the HITSystem in Kenya. The cost per life-year gained from the HITSystem intervention was US$82⋅72 (range $40⋅13–215⋅05). Thus, even at the expensive end of our estimated range, the cost is well under WHO’s threshold for very cost-effective (≤ per capita national income; $1525 in Kenya in 2016 and $1878 in 2020),21 highlighting the economic viability of wider HITSystem implementation in Kenya. Results from the multivariate sensitivity analyses indicate implementation of the HITSystem is most cost-effective in clinics enrolling a large number of infants exposed to HIV per month and those that have modest clinical personnel and technical assistance costs.
Since these data were collected, Kenyan national guidelines for early infant diagnosis testing have shifted to recommend PCR testing at birth, 6 weeks, 6 months, and 12 months, with a final serological test at 18 months.22 This shift to more frequent PCR testing will increase costs for standard of care, but the intervention costs are fairly fixed (eg, personnel, training, internet connection costs, and computers). Changes in the number of infants exposed to HIV that are tested or perinatal transmissions that occur will affect the cost-effectiveness of the intervention. The average number of infants tested (early infant diagnosis volume) and the personnel costs for HITSystem implementation had the largest effect on the cost per life-year saved, with sensitivity analyses showing that high early infant diagnosis volume (>15 infants per month) and training existing staff such as mentor mothers or data clerks to maintain the HITSystem had the lowest cost per life-year saved. These cost-effectiveness data should also be considered in balance with the findings of the HITSystem efficacy trial, which indicated the highest effect in early infant diagnosis retention was evidenced at health facilities with medium and low early infant diagnosis volume (adjusted OR 16⋅8, 95% CI⋅6·3–44⋅9 for medium volume and 9⋅4, 4⋅2–21·4 for low volume), where resources were probably minimal and the benefit of the HITSystem intervention most pronounced.12
Other ongoing developments in paediatric HIV care, including implementation of point-of-care diagnostics for infant testing are more expensive, but can greatly expedite initiation of ART among infants.23,24 Studies comparing the incremental cost-effectiveness ratio (ICER) for point of care testing with standard of care have varied substantially, ranging from $23–6188 in Zambia,23,24 and $830–1010 in Zimbabwe,25,26 with ICERs highly dependent on the lifespan of the testing instruments, use of the point-of-care platforms for other testing purposes (ie, viral load monitoring, tuberculosis testing, HPV testing), and provider time spent per test.23 Point-of-care testing is also more cost-effective in settings with lower PMTCT coverage, poorer standard-of-care early infant diagnosis outcomes (ie, long delays in test results and low ART initiation rates), and when point-of-care instruments can be integrated with other disease programmes such as those for tuberculosis.24 One model-based analysis comparing standard-of-care laboratory-based early infant diagnosis, strengthened laboratory-based early infant diagnosis, and point-of-care testing concluded that point-of-care testing was more cost-effective than strategies to strengthen existing early infant diagnosis systems.25,26 This study described a range of strategies to strengthen existing laboratory-based early infant diagnosis services, but cost data were limited to enhanced sample transport efforts and SMS printers to expedite the return of infant test results. This Article is the first publication of HITSystem costs and cost-effectiveness data, establishing the ICER of $82⋅72 per life-year saved with this early infant diagnosis strengthening strategy.
The start-up costs for national point-of-care testing are substantial and might not be practical in some low-resource settings. Given the challenges to implementing and sustaining point-of-care testing (eg, cost, cartridge stock outs, machine errors, provider training, secure machine storage, and reduced sensitivity at birth),27–29 PCR tests at birth might be more pragmatic in some settings for policy makers to add, if investing in the infrastructure and training required for birth point-of-care testing is not yet feasible. Furthermore, even with more routine point-of-care testing for early infant diagnosis, standard laboratory testing would probably remain a component of early infant diagnosis in settings where point-of-care algorithms require confirmatory testing or point-of-care testing is not feasible.23 Therefore, continued investment in cost-effective strategies to strengthen existing early infant diagnosis networks are required to gain ground in outcomes until widespread investment in point-of-care infrastructure is achieved in all countries with a high HIV burden.
This study is a rigorously conducted cost-effectiveness analysis leveraging efficacy data from a randomised controlled trial and national early infant diagnosis data from a wide range of health facilities to model the cost-effectiveness of scaling up the HITSystem intervention in Kenya. Although the number of hospitals included in the trial was small, the sites were strategically stratified to represent wide geographical variation, early infant diagnosis patient volume, and the three tiers of government hospitals that provide nearly all early infant diagnosis services in Kenya. Despite having an extended eligibility period of up to 24 weeks of infant age for early infant diagnosis testing (well beyond the recommended age of 6 weeks), the median age of testing in our sample was still 6⋅4 weeks. Although our lowest estimate for personnel costs includes a task shifting model of training and making use of existing staff to support HITSystem implementation, we recognise that this approach might not be feasible in many facilities given budget constraints and staff shortages. The sensitivity analyses allow for a range of results so implementation at various types of health facility can be better estimated. The Cost Effectiveness of Preventing AIDS Complications-Pediatric model accounts for lifelong ART costs, but the cost of lifelong ART might be underestimated given increasing life expectancy among people with HIV. Lastly, we acknowledge that circumstances such as prolonged health-worker strikes and pandemics like COVID-19 can disrupt services and might affect costs of early infant diagnosis and the cost-effectiveness of the HITSystem.
Although strengthening early infant diagnosis is a crucial component of improved paediatric HIV care, the gains in earlier diagnosis and ART initiation can be lost without strong systems to support ART adherence, viral loading monitoring, and retention in care for children, adolescents, and adults. Such challenges have been reported in South Africa.30 Furthermore, the estimates presented in this study account for infants exposed to HIV who are already engaged in early infant diagnosis at health facilities, which was estimated to be 50% at the time of the randomised controlled trial.31 To maximise the return on investment for the HITSystem—and early infant diagnosis in general—comprehensive efforts to expand access and reach of early infant diagnosis, and birthing parent–infant retention past the point of infant diagnosis are needed.
Modifications to the HITSystem were made in 2016 to support the new testing algorithm outlined in Kenya’s national guidelines, and to account for the use of point-of-care testing for early infant diagnosis at some facilities, and standard laboratory-based early infant diagnosis testing at others. Furthermore, a PMTCT module has been developed to support PMTCT retention, with automatic early infant diagnosis linkage and active tracking and follow-up of birthing parent−infant pairs for 6-week testing. We are currently evaluating the efficacy and cost-effectiveness of this system in an ongoing cluster randomised controlled trial (NCT04571684), including analysis of mediators and moderators of provider and patient behaviour to inform health policy around the HITSystem and similar eHealth interventions. With 95% of pregnant people presenting for antenatal care in Kenya,1 a system that tracks and retains pregnant people with HIV and their infants from confirmation of pregnancy through completion of early infant diagnosis has the potential to substantially increase the proportion of infants exposed to HIV who engage in early infant diagnosis. If these benefits are realised on a larger scale, the increased uptake of early infant diagnosis is likely to further justify the costs of a combined PMTCT and early infant diagnosis HITSystem intervention.
Table:
Key model input parameters for calculating cost-effectiveness of the HITSystem
| Input (sensitivity range) | Data source | |||
|---|---|---|---|---|
| Low | Most likely | High | ||
| National data | ||||
| Number of infants exposed to HIV | 74 007 | 82 230 | 90 453 | NASCOP database, July 2014 to Dec 2015 (18 months) |
| Percentage of exposed infants infected with HIV | 2% | 7% | 14% | NASCOP database, July 2014 to Dec 2015 (18 months) |
| Life-year gained per infant on ART | 26·2 | 27·6 | 28·8 | CEPAC |
| Trial data: intervention arm | ||||
| Mother-infant pairs enrolled per month | 144 | 270 | 450 | Study data |
| Received opportunistic infection prophylaxis | 90% | 100% | 100% | Study data |
| Dried blood spots taken at 6 weeks | 90% | 100% | 100% | Study data |
| Results returned to hospital | 90% | 100% | 100% | Study data |
| Mothers notified of PCR results | 89% | 99% | 100% | Study data |
| Tested for HIV again at 9 months | 87% | 97% | 100% | Study data |
| Tested for HIV again at 18 months | 76% | 85% | 93% | Study data |
| Receiving HIV ART | 80% | 100% | 100% | Study data |
| Completed early infant diagnosis | 77% | 85% | 94% | Study data |
| Trial data: control arm | ||||
| Received opportunistic infection prophylaxis | 81% | 90% | 93% | Study data |
| Dried blood spots taken at 6 weeks | 89% | 99% | 100% | Study data |
| Results returned to hospital | 87% | 97% | 98% | Study data |
| Mothers notified of PCR results | 80% | 89% | 93% | Study data |
| Tested for HIV again at 9 months | 82% | 91% | 94% | Study data |
| Tested for HIV again at 18 months | 62% | 69% | 75% | Study data |
| Receiving HIV ART | 58% | 73% | 87% | Study data |
| Completed early infant diagnosis | 55% | 61% | 66% | Study data |
| Trial data: intervention costs (HITSystem costs per 18 months per site) | ||||
| Computer and modem | $616 | $684 | $741 | Study data |
| Internet connection costs | $410 | $513 | $616 | Study data |
| Licence fee | $2052 | $2462 | $3078 | Study data |
| Training costs | $342 | $456 | $570 | Study data |
| Technical assistance support | $2052 | $3078 | $4104 | Study data |
| Clinic personnel support | $0 | $616 | $4720 | Study data |
The base model used to calculate the cost-effectiveness was deterministic and calculates progression of infants through the early-infant-diagnosis care continuum in both the intervention and standard-of-care (control) arms of the study. Differences in progression across the two study arms were used to establish the efficacy outcomes. The number of life-years gained per infant who successfully received timely ART were based on results of the CEPAC model, a computer-based, state-transition, Monte Carlo simulation model of the progression and outcomes of HIV disease that accounts for lifelong treatment.11,14 We used the CEPAC-Pediatric model application, validated and calibrated for paediatric HIV using data from east African countries, to estimate survival.11,14,15 We report 27·6 life-years gained for each infant receiving timely ART; averaging the life expectancies reported in seminal studies among African children (95% CI 26·2–28·8).11,15 Cost data for the HITSystem intervention were integrated into the model, and the incremental cost-effectiveness ratio was calculated in terms of the cost per life-year gained from the intervention. Costs are in US$. ART=antiretroviral therapy. CEPAC=Cost-Effectiveness of Preventing AIDS Complications. NASCOP=National AIDS and STI Control Program.
Research in context.
Evidence before this study
We searched PubMed and Google Scholar using the following search terms: “early infant diagnosis or EID”, “infant HIV testing” with “cost-effectiveness analyses”, “costs”, “ICER”, for publications between Jan 1, 2008, and Jan 1, 2022, excluding studies done in high-resource settings. All identified publications were in English. During that period, early infant diagnosis innovations evolved to include a range of interventions intended to strengthen the quality and efficiency of returning infant HIV test results and expediting antiretroviral therapy (ART) initiation, if indicated. In 2016, point-of-care HIV PCR diagnostics were approved by the US Food and Drug Administration and introduced in African countries. This approval revolutionised opportunities for same-day test-and-treat strategies for infants diagnosed with HIV. Multiple studies have evaluated the cost-effectiveness of point-of-care strategies in African countries. Despite higher costs per test with point-of-care versus conventional laboratory-based PCR testing, the benefits of obtaining same-day results and earlier ART initiation creates a cost-effective advantage of point-of-care testing in settings with a high HIV burden, if resources are available to support and maintain the needed infrastructure for machines, cartridges, and maintenance. Despite strong recommendations from WHO and the research community to transition from laboratory-based PCR testing to point-of-care testing for infants exposed to HIV, few countries in lower resource settings have been able to allocate the needed resources and infrastructure investment to make this transition. In such settings, strategies that optimise results from the existing laboratory-based infrastructure (eg, targeted, PCR birth testing for infants at high risk, and strategies to optimise testing efficiency and patient retention) should be prioritised. In the review of early infant diagnosis cost-effectiveness data, few studies report costs with consistent methods or report findings as an incremental cost-effectiveness ratio (ICER) for life-years saved, making comparison across studies difficult.
Added value of this study
We report the first cost-effectiveness data of implementing the HITSystem intervention, which was designed to optimise early infant diagnosis outcomes, within the existing early infant diagnosis infrastructure in Kenya. These data add to the previously reported HITSystem efficacy data that show significantly increased early infant diagnosis retention up to18 months of age, accelerated turn-around time for PCR results, and earlier ART initiation. Calculated from a health-system perspective reflecting the investment required to add the HITSystem intervention to existing laboratory-based early infant diagnosis services, the cost per infant exposed to HIV was US$31·38 and the ICER for life-years saved was $82⋅72.
Implications of all the available evidence
The HITSystem intervention was designed to be cost-effective and feasible in low-resource settings to optimise the return on investment of the existing early infant diagnosis infrastructure. When modelled to reflect national-level use in Kenya, the HITSystem is among the most cost-effective early infant diagnosis interventions reported to date. Although point-of-care HIV testing has become the optimal standard for infant HIV testing, many countries in low-resource settings do not have point-of-care infrastructure in place at the national level due to a high burden of competing health needs and limited resources. These findings could help governments in Kenya and other similar settings to balance the value, feasibility, and timing of implementing various interventions to ensure progress in early infant diagnosis outcomes remains a priority.
Acknowledgments
This study was funded by the US National Institute of Child Health and Human Development (grant number R01HD076673). The University of Kansas Medical Center, Kenya Medical Research Institute, Global Health Innovations—Kenya, and Children’s Mercy Hospital were collaborative partners in these efforts. We would like to acknowledge the members of the team who played a key role in implementation. We are grateful for the implementation support from mentor mothers and clinical staff, and all the mother–infant pairs participating in this research. We also acknowledge the crucial role of our government partners at the Kenya National AIDS and STI Control Program. We thank Prof Elijah Songok, the acting Director of KEMRI, for permission to publish this manuscript.
Funding
The US National Institute of Child Health and Human Development.
Footnotes
Declaration of interests
We declare no competing interests.
For the Consumer Price Index see https://data.bls.gov/cgi-bin/cpicalc.pl
Data sharing
Upon study completion, de-identified data can be made available upon request or deposited to public repositories.
References
- 1.UNAIDS. Kenya: country factsheet. 2020. https://www.unaids.org/en/regionscountries/countries/kenya (accessed Dec 16, 2021).
- 2.Bourne DE, Thompson M, Brody LL, et al. Emergence of a peak in early infant mortality due to HIV/AIDS in South Africa. AIDS 2009; 23: 101–06. [DOI] [PubMed] [Google Scholar]
- 3.PEPFAR. Country operational plan: (COP/ROP). Strategic direction summary, 2020. 2020. https://www.state.gov/wp-content/uploads/2020/08/Kenya-COP20-SDS-Final.pdf (accessed June 23, 2023).
- 4.Ministry of Health. Guidelines on use of antiretroviral drugs for treating and preventing HIV infection: a rapid advice. 2014. http://guidelines.health.go.ke:8000/media/Rapid_Advice_Booklet_2014_July_V5.pdf (accessed June 23, 2023).
- 5.UNAIDS. Understanding Fast-Track: accelerating action to end the AIDS epidemic by 2030. 2015. https://www.unaids.org/en/resources/documents/2014/JC2686_WAD2014report (accessed June 23, 2023).
- 6.Hassan AS, Sakwa EM, Nabwera HM, et al. Dynamics and constraints of early infant diagnosis of HIV infection in rural Kenya. AIDS Behav 2012; 16: 5–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Goggin K, Wexler C, Nazir N, et al. Predictors of infant age at enrollment in early infant diagnosis services in Kenya. AIDS Behav 2016; 20: 2141–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Goggin K, Hurley EA, Staggs VS, et al. Rates and predictors of HIV-exposed infants lost to follow-up during early infant diagnosis services in Kenya. AIDS Patient Care STDS 2019; 33: 346–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Bianchi F, Cohn J, Sacks E, et al. Evaluation of a routine point-of-care intervention for early infant diagnosis of HIV: an observational study in eight African countries. Lancet HIV 2019; 6: e373–81. [DOI] [PubMed] [Google Scholar]
- 10.Ciaranello AL, Park J-E, Ramirez-Avila L, Freedberg KA, Walensky RP, Leroy V. Early infant HIV-1 diagnosis programs in resource-limited settings: opportunities for improved outcomes and more cost-effective interventions. BMC Medicine 2011; 9: 59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ciaranello AL, Doherty K, Penazzato M, et al. Cost-effectiveness of first-line antiretroviral therapy for HIV-infected African children less than 3 years of age. AIDS 2015; 29: 1247–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Finocchario-Kessler S, Gautney B, Cheng A, et al. Evaluation of the HIV Infant Tracking System (HITSystem) to optimise quality and efficiency of early infant diagnosis: a cluster-randomised trial in Kenya. Lancet HIV 2018; 5: e696–705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Fischer M, Safaeinili N, Haverfield MC, Brown-Johnson CG, Zionts D, Zulman DM. Approach to human-centered, evidencedriven adaptive design (AHEAD) for health care interventions: a proposed framework. J Gen Intern Med 2021; 36: 1041–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Ciaranello AL, Morris BL, Walensky RP, et al. Validation and calibration of a computer simulation model of pediatric HIV infection. PLoS One 2013; 8: e83389. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Dunning L, Francke JA, Mallampati D, et al. The value of confirmatory testing in early infant HIV diagnosis programmes in South Africa: a cost-effectiveness analysis. PLoS Med 2017; 14: e1002446. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lipscomb J, Weinstein M, Torrance G. Time preference. In: Gold M, Siegel J, Russell L, Weinstein M, eds. Cost-effectiveness in health and medicine. New York: Oxford University Press; 1996: 214–246. [Google Scholar]
- 17.Basu A, Ganiats T. Discounting in cost-effectiveness analysis. In: Neumann PJ, Ganiats TG, Russell LB, Sanders GD, Siegel JE, eds. Cost-effectiveness in health and medicine. New York: Oxford Academic, 2016: 277. [Google Scholar]
- 18.Frick KD. Microcosting quantity data collection methods. Med Care 2009; 47 (suppl 1): S76–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lilian RR, Johnson LF, Moolla H, Sherman GG. A mathematical model evaluating the timing of early diagnostic testing in HIV-exposed infants in South Africa. J Acquir Immune Defic Syndr 2014; 67: 341–48. [DOI] [PubMed] [Google Scholar]
- 20.Francke JA, Penazzato M, Hou T, et al. Clinical impact and cost-effectiveness of diagnosing HIV infection during early infancy in South Africa: test timing and frequency. J Infect Dis 2016; 214: 1319–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.World Bank. GDP per capita: Kenya. 2020. https://data.worldbank.org/indicator/NY.GDP.PCAP.CD?locations=KE (accessed Dec 16, 2021).
- 22.Ministry of Health. Guidelines on use of antiretroviral drugs for treating and preventing HIV infection in Kenya. 2018. http://cquin.icap.columbia.edu/wp-content/uploads/2017/04/ICAP_CQUIN_Kenya-ARV-Guidelines-2018-Final_20thAug2018.pdf (accessed June 23, 2023).
- 23.De Broucker G, Salvatore PP, Mutembo S, et al. The cost-effectiveness of scaling-up rapid point-of-care testing for early infant diagnosis of HIV in southern Zambia. PLoS One 2021; 16: e0248217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Salvatore PP, de Broucker G, Vojnov L, Moss WJ, Dowdy DW, Sutcliffe CG. Modeling the cost-effectiveness of point-of-care platforms for infant diagnosis of HIV in sub-Saharan African countries. AIDS 2021; 35: 287–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Frank SC, Cohn J, Dunning L, et al. Clinical effect and cost-effectiveness of incorporation of point-of-care assays into early infant HIV diagnosis programmes in Zimbabwe: a modelling study. Lancet HIV 2019; 6: e182–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.McCann NC, Cohn J, Flanagan C, et al. Strengthening existing laboratory-based systems vs investing in point-of-care assays for early infant diagnosis of HIV: a model-based cost-effectiveness analysis. J Acquir Immune Defic Syndr 2020; 84 (suppl 1): S12–S21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Hsiao NY, Dunning L, Kroon M, Myer L. Laboratory evaluation of the Alere q Point-of-Care System for early infant HIV diagnosis. PLoS One 2016; 11: e0152672. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Ibrahim M, Moyo S, Mohammed T, et al. Brief report: high sensitivity and specificity of the Cepheid Xpert HIV-1 qualitative point-of-care test among newborns in Botswana. J Acquir Immune Defic Syndr 2017; 75: e128–131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wexler C, Maloba M, Brown M, et al. Factors affecting acceptance of at-birth point of care HIV testing among providers and parents in Kenya: a qualitative study. PLoS One 2019; 14: e0225642. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Kalawan V, Naidoo K, Archary M. Impact of routine birth early infant diagnosis on neonatal HIV treatment cascade in eThekwini district, South Africa. South Afr J HIV Med 2020; 21: 1084. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.UNAIDS. UNAIDS data 2021. 2021. https://www.unaids.org/sites/default/files/media_asset/JC3032_AIDS_Data_book_2021_En.pdf (accessed June 23, 2023).
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Upon study completion, de-identified data can be made available upon request or deposited to public repositories.