Background
Malaria transmission in low endemic areas tends to cluster geographically and temporally. Also, in low transmission settings, a higher proportion of infected individuals is asymptomatic, and therefore will not be detected by passive surveillance which occurs through health facilities. Reactive case detection (RACD), or active case finding among households and neighbours of passively identified index cases, is a strategy to address this challenge. With the goal of RACD being the identification and treatment of asymptomatic or other infections that would not otherwise present through the passive surveillance system, it is considered a key malaria strategy for gathering surveillance information [
1] and may reduce transmission and facilitate the achievement of malaria elimination [
2,
3]. RACD is widely implemented [
4], yet it is operationally challenging requiring significant human resources, commodities, and time for an “on-call” team to conduct screenings in villages, often traveling long distances to reach remote locations. There are also limitations with the standard diagnostics used, microscopy or rapid diagnostic test (RDT), to detect low-density infections. Highly sensitive diagnostics are available but the costs and cost-effectiveness of using them is unclear. Further, the costs of conducting RACD in general have not been systematically documented [
5,
6].
The detection limit of microscopy or RDTs is typically 100 parasites/µL, and in low endemic settings, a high proportion of asymptomatic infections fall below this threshold [
7,
8]. Outside of use in research settings, more sensitive detection methods such as polymerase chain reaction (PCR) are impractical due to high cost, sophisticated training, resources required, and long turnaround time (several hours). Another molecular method called loop-mediated isothermal amplification (LAMP) provides the sensitivity of PCR with fewer requirements. Testing is not point-of-care and requires use of a laboratory, but the assay is simple, does not require sophisticated equipment, and can be performed in half a day to one full day [
9,
10]. However, the costs and cost effectiveness of using LAMP in RACD are not clear.
To enable national malaria programmes to assess the feasibility of implementing RACD in their low burden setting, this study aimed to: (1) estimate the cost of implementing RACD using data from the malaria-eliminating district of Aceh Besar, Indonesia; and (2) compare the cost and cost-effectiveness of using a standard diagnostic (microscopy) to more sensitive LAMP for identifying infections during RACD. This information can help guide strategic and budgetary planning for local malaria control programmes, and also help to inform the research, development, and implementation of highly sensitive diagnostics for malaria.
Methods
Study design
The study was a prospective economic analysis of costs and cost-effectiveness. The main cost measures were cost per RACD event, per person screened, per population at risk (PAR; defined as total population in each sub-district), and per infection found. The main cost-effectiveness measure was incremental cost-effectiveness ratio (ICER), expressed as cost per malaria infection detected by LAMP versus microscopy.
Study location
Aceh Besar District is located in the western part of Aceh Province on Sumatra Island, Indonesia. Malaria transmission occurs year-round with the higher transmission occurring from January to July. In the past,
Plasmodium falciparum and
Plasmodium vivax were reported as the main malaria species, but
Plasmodium knowlesi has also been reported from the area in 2016 [
11]. Due to intensified efforts in case management, vector control, and surveillance, Aceh Besar District has successfully reduced annual malaria incidence from 2.6 cases per 1000 population in 2006 to 0.2 per 1000 in 2014 (population in 2014 was 371,412 [
12]). It is one of 114 low endemic districts categorized by the Indonesian government as ‘eliminating’, with a goal to interrupt transmission by 2020 [
13,
14]. As part of the malaria elimination strategy, the District Health Office in 2010 initiated RACD, which is locally referred to as ‘contact survey’. After a microscopy-confirmed malaria case is diagnosed at the health facility, designated surveillance staff visit the village of the index case to perform malaria testing among household members and neighbours. Based on WHO and national guidelines, RACD is performed using microscopy and all subjects residing within 500 m of the index case are targeted.
Study population
Index cases were enrolled from four sub-district level health facility sites that reported 78% of all reported cases in Aceh Besar in 2013: Indrapuri, Kuta Cot Glie, Lhoong, and Saree. An additional health facility Lhoknga was also included purposively due to high case burden prior to 2013. RACD was conducted in the villages where index cases resided, and the people reached during RACD were enrolled for this study. The study was conducted over a period of 20 months from May 2014 to December 2015.
Health facility and field procedures
All subjects presenting to health facilities with suspected malaria were assessed for malaria infection by microscopy of blood smears. Microscopy confirmed cases, also referred to as an index case, had a subsequent venous blood draw to generate a dried blood spot (DBS) for LAMP and PCR. Within one to 7 days of the index case report, RACD was conducted among individuals residing in households located within 500 m of the index cases. During an RACD event, blood was collected from individuals by finger prick to prepare slide blood smears and DBS. The RACD team consisted of five people: a microscopist and surveillance officer from the health facility, two community health workers from the village (a midwife plus a village malaria worker or village leader), and the study field coordinator. Up to two return visits were conducted to include any missing residents. The coverage goal of RACD in each of the target area was to recruit a minimum of 40 subjects or at least 80% of the residents within each of the closest five households. Additional details of index case enrolment and RACD have been previously described [
11].
Laboratory testing
The microscopists collected blood smears in the field and transported them in closed slide boxes for subsequent examination at the health facility. Slides were fixed and stained with 3% Giemsa. A slide was determined malaria positive if at least one clear form of any malaria parasite species was found after examination of the whole spread of the thick smear. Parasite densities were measured by counting the number of asexual parasites per 200 or 500 white blood cells (WBC) and calculating parasites/µL assuming a WBC count of 8000 parasites/µL in the thick smear. If positive, an additional 100 high-powered fields in the thin smear were examined to determine species [
15]. Cross-check of the malaria positive slides for species identifications were done by the district level microscopist. A provincial expert level-certified microscopist did quality assurance (QA) for all positive slides, 10% of randomly selected negatives and to resolve any discrepancies in results between the clinic and district-level microscopists [
13].
All DBS samples collected were dried overnight then stored in sealed plastic bags with desiccant. Extraction of DNA and LAMP testing on all samples were performed at the Aceh Provincial Health Laboratory. DNA was extracted from DBS using the Saponin/Chelex-100 method [
16]. Using 15 µL of chelex-extracted DNA solution, Pan-LAMP testing followed by Pf-LAMP specific testing for Pan-LAMP positive samples was performed using a commercial Loopamp detection kit [
17,
18] in accordance to manufacturer’s instructions (Eiken Chemical, Japan) (Limit of detection (LOD) between 1 and 5 parasites/µL). As previously detailed, using DNA chelex-extracted from a second DBS, Pan-LAMP positive and 10% of Pan-LAMP negative samples underwent further molecular testing for QA at the Eijkman Institute for Molecular Biology using nested PCR methods (LOD between 0.1 and 1 parasites/µL) targeting the
cytochrome b gene followed by
AluI enzyme digestion for species identification and 18S rDNA nested PCR, with positive samples undergoing further
P. knowlesi-specific PCR testing [
11].
Costing data collection
Costs were collected for RACD programme set-up, training, coordination, and all activities beginning from when an index case was reported from a health facility (e.g. preparing supplies and contacting the index case household) through to sample processing, analysis, and quality assurance conducted after the RACD event. Time spent by personnel on microscopy or LAMP, including the quality assurance by subsequent microscopy and PCR, respectively, was allocated to diagnostic-specific costs as appropriate. Follow-up costs including a return visit to provide treatment to subjects that were initially microscopy-negative but subsequently found to be LAMP-positive were not included.
Data on costs were collected by a local field study coordinator through review of expenditure records. All costs were converted from the local currency to US dollars (USD) using the average mid-year exchange rate for 2015 (1 USD = 13,432 Indonesian Rupiah, from Oanda.com). These data were inputted to a costing tool developed in Microsoft Excel 2010. Costs were organized by location at which the expenses occur (health facility, provincial laboratory, and national laboratory); the type of diagnostic test used (microscopy or LAMP); and the month in which the cost was incurred. Each cost input was grouped into one of five major categories.
Personnel
Personnel costs covered project staff salaries for trainings, coordination, field activities and laboratory work. Project staff included one full time field coordinator, a part time field team in each of the five sub-districts including one surveillance officer and a microscopist, three part-time laboratory technicians at the provincial and national levels, and one laboratory technician at the national level. Most of the personnel had duties outside the scope of this study, and only supported RACD as index cases presented to health facilities. Thus, specific time contributions of all personnel involved in the RACD activities were logged and later used for approximating the cost of human resources. Time required for initial diagnosis of index cases at health facilities was not included.
Trainings
Costs associated with any trainings conducted for this study were captured and include costs for training supplies, room rental, and participant per diems.
Services
The cost of services included utilities such as internet, communication, courier services, and vehicle rental/fuel for sample transportation used in conducting RACD. Services utilized prior to the study period but for project setup were included.
Capital
Capital costs were initial investments to set up RACD and included the cost of owned motor vehicles used for travel and transportation; electronic devices such as computers, computer accessories, printers, tablets, refrigerators; and laboratory equipment such as microscopes, centrifuges, heat blocks, and PCR machines. If the capital was used for activities other than RACD, a time use percentage was assigned. The cost of capital was valued for the study duration (20 months) after accounting for depreciation based on the useful life years of each capital, and discounted using a rate of 3%. The remaining values of capital at the end of the study period were subtracted. The cost of office space and furniture were not included due to limitations in data availability.
Consumables
Consumables included field and laboratory supplies such as slides, lancets, alcohol swabs, plastic bags, filter paper, LAMP kits, laboratory reagents, centrifuge tubes, etc. Per unit costs of these consumables were multiplied by the amounts utilized during the study period to estimate the total cost of consumables.
Analysis
The main outcome measure was number of infections detected by LAMP versus microscopy in RACD. Other measures included number of RACD events and number of individuals screened per RACD event.
To estimate the costs associated with microscopy versus LAMP, RACD costs were separated into ‘general RACD costs’ and ‘diagnostic-specific costs’. General RACD costs covered routine activities related to RACD such as the field visit, including time for preparation, travel, and screening households. Diagnostic specific costs for microscopy and LAMP were calculated by isolating any personnel, training, consumables, processing time, services from general RACD costs. As a key activity for an elimination surveillance programme, particularly for programmes considering the use of highly sensitive diagnostics, QA was included for both microscopy- and LAMP-specific costs [
1]. Trainings for microscopy and microscopy QA were conducted in conjunction with the surveillance trainings and included in the general RACD training costs. Cost proportions for each input category were compared for microscopy-specific, LAMP-specific and general RACD costs.
Cost per RACD event and cost per individual screened were calculated by dividing the total cost of conducting RACD using microscopy and/or LAMP during the study period by the total number of RACD events and screened individuals. Cost per population at risk (PAR; defined as total population in each sub-district) was calculated at the sub-district level. Cost per PAR per year was calculated by dividing the total cost of RACD using both diagnostics by the population and by the number of months the study was conducted (20 months). The resulting montly cost per PAR, was then multiplied by 12 (months in a year) to determine the cost per PAR per year. Costs per infection identified were calculated as cost of RACD using microscopy and/or LAMP, divided by the number of infections detected by microscopy and/or LAMP, respectively. The total cost for the study period and annual recurrent programmatic cost of the RACD programme were also calculated. The annual recurrent programmatic cost was approximated by excluding any capital non-recurrent costs from the total cost.
To compare the cost-effectiveness of the different diagnostic methods, the incremental cost-effectiveness ratio (ICER), defined by the difference in cost between two interventions divided by the difference in their effect, was estimated by comparing the cost and outcomes associated with LAMP and microscopy. As general RACD costs were applied to both methods, the differences in costs were due to diagnostic-specific costs and respective yield.
To explore how costs might vary across sites with different index case burdens or prevalence of infection in RACD [
19], cost measures were compared across the sub-districts. The cost drivers of RACD across the study sites were also assessed. To consider how changes or differences in diagnostic sensitivity or prevalence of infection in RACD might affect findings, the costs and ICER for different scenarios was estimated based on the costs and outcomes observed in this study.
Discussion
RACD is a strategy widely implemented to gather surveillance information in low transmission settings and it may contribute to reductions or interruption of transmission. However, the costs of doing RACD have not been well evaluated. This study investigated the cost of conducting RACD and the cost-effectiveness of different diagnostic methods in identifying additional malaria infections. During the study period of 20 months, 1495 individuals were screened in 36 RACD events yielding six additional cases by microscopy or LAMP, at an estimated total cost of $42,418. The average cost of conducting RACD using microscopy and LAMP was $1178 per event, with the main cost driver being personnel. The average cost of RACD for each individual was $11 for general costs, with an additional $0.62 and $16 per person for microscopy and LAMP, respectively. As a public health intervention, RACD using both diagnostics cost an average of $0.42 per PAR per year. LAMP was more costly but more cost-effective for the detection of infections mainly due to higher diagnostic sensitivity. For RACD using microscopy only, the cost per infection found was $8930 compared to $6915 for RACD using LAMP only. The ICER for RACD using LAMP versus microscopy was $5907 per additional malaria infection detected. As malaria programmes consider active case detection and the integration of highly sensitive diagnostics, these findings may inform strategic and budgetary planning.
Due to the intensified interventions required, it is well-established that malaria elimination compared to control is more costly but that longer term benefits make the pursuit worthwhile [
21,
22]. A systematic review on the costs and cost-effectiveness of control interventions by White et al. [
23] found the median cost of preventative interventions (insecticide-treated bed nets, indoor residual spraying, intermittent preventative treatment) was $0.60–$6.70 per person and the median costs for case management services (diagnosis, treatment) ranged from $4.32–$30.26 per patient, with the most expensive service being treatment of severe malaria. Our finding that RACD with microscopy and/or LAMP cost $12–$28 per individual screened is consistent with previous reports on the higher cost for elimination versus control interventions. The high cost of $7070 per infection identified also raises the issue of the value of RACD as a low yield activity where case detection rates are generally less than 2% even with highly sensitive diagnostics [
24]. However, the value of an intervention should be considered in a broader context. As a public health intervention, the annual cost of RACD using both diagnostics was low, at an average of $0.42 per PAR at the sub-district level.
Despite the high costs, RACD using LAMP versus microscopy led to a lower cost per infection identified (23%) and the ICER of $5907 per additional infection detected favors the use of LAMP or a LAMP-equivalent assay in RACD. Based on our analyses of cost inputs, there would opportunities to mitigate costs. Personnel and LAMP-specific consumables, and not capital, training, or services, accounted for the highest cost shares [
25]. Personnel costs could be minimized by better integrating staff in the health system. LAMP-specific consumable costs could be decreased if prices were negotiated or subsidized, or a less expensive highly sensitive diagnostic could be used. In addition, the higher sensitivity of LAMP made it possible to detect
P. knowlesi in the study area for the first time. This important benefit may outweigh the high costs of RACD using LAMP.
To explore how costs might vary across sites with different index case burdens or prevalence of infection in RACD, cost measures were compared across the sub-districts. In sub-districts with more RACD events (such as Saree and Lhoong) overall costs were higher than in sub-districts with fewer cases, and consequently the cost per individual screened was lower. Despite having zero cases during the study period, Lhoknga still needed to maintain costs for training “on-call” local health facility personnel as well as capital in the event an index case were to present and RACD would need to be conducted.
To further inform how costs and cost-effectiveness might vary with differences in prevalence of infection or the sensitivity of the diagnostic used, estimated projections show how costs per infection identified along with ICER decrease with increases in prevalence of infection and yield of a diagnostic to detect infections, relative to microscopy (Figs.
3 and
4). As seen in the Fig.
3, the curve began to plateau when the test positive rate or prevalence of infection in RACD increased above 1.6% suggesting that the highest relative cost-savings will be realized in low transmission settings. The curve also began to plateau when the yield of the diagnostic test exceed fivefold–tenfold relative to microscopy, suggesting diminishing returns with use of more sensitive diagnostics [
26] when the prevalence is already low. As more sensitive diagnostics are being developed or come to market, these considerations can help to inform decision-making on investments and strategic planning.
There were some limitations of this study. Estimation of the personnel costs relied on self-reported time allocations. Further, due to the few cases in this low transmission setting, limitations exist for the generalizability of the study and precision of the ICER estimate. Extrapolations to consider the influence of prevalence of infection and diagnostic yield were performed, but assumed other fixed factors (epidemiological, or cost-related). In real world implementation, several relevant factors could change and affect costs. For example, cost-effectiveness of LAMP could be improved with discounted prices for LAMP or other less costly diagnostics (e.g. a new highly sensitive histidine-rich protein 2-based rapid diagnostic test [
27] could be useful in
P. falciparum predominant settings and has lower costs due to a discounted price and does not require laboratory work nor travel back to the households to inform on results). In a more remote and underserved setting, there may be higher costs associated with establishing laboratory infrastructure to conduct LAMP. A new diagnostic may have higher costs than those required for LAMP. These scenarios would increase costs per infection identified and reduce cost-effectiveness. Also, the estimations assume very low parasite densities and do not take into account that in other settings, due to endemicity or species (e.g.
P. falciparum presents with higher parasite densities than
P. vivax), parasite densities may be different and thus impact the sensitivities and yield of microscopy and LAMP. More empiric data from other settings along with use of more sophisticated modeling techniques could improve generalizability. Additionally, while microscopy and LAMP both require time to acquire results, processing time for LAMP may be longer resulting in a delay in treatment and therefore transmission-blocking, which were not taken into consideration in this analysis. Finally, this study only assessed costs from service provider’s perspective, including cost sharing and in-kind donations, and costs are specific to this situation. In other contexts, absolute costs may be different. Future studies can build upon these findings to quantify costs and benefits of conducting RACD by including the broader costs to society when additional infections are not detected and may result in further transmission.
This study had several strengths. Firstly, a detailed and prospective collection of RACD costs was carried out. Retrospective data collection can introduce bias and limit the granularity of the costing data. Importantly, this study fills a critical gap on the economics of focal screen and treat. Previous studies have documented a framework for evaluating the costs of RACD [
25], conducted cost analyses of mass screen and treat [
28], or identified potentials for operational efficiencies [
29], but only one other study has measured the actual costs and that study was from a higher transmission setting where the RACD test positivity rate was 19% [
25]. To the best of the authors’ knowledge, this study is the first on RACD or focal screening and treatment to report costs from a low transmission setting, costs using a highly sensitive diagnostic, and cost-effectiveness of using a highly sensitive versus standard diagnostic in active case detection.
In summary, in the low transmission setting of Aceh Besar, RACD costs per individual screened were found to be high, though costs per PAR were low. Compared to microscopy, the use of LAMP in RACD was more costly but more cost-effective for the detection of infections, with diminishing returns observed when findings were extrapolated to scenarios with higher prevalence, or using more sensitive diagnostics when prevalence is very low. These findings can inform strategic and budgetary decisions faced by the many countries that are pursuing malaria elimination and considering active case detection with new diagnostics [
1,
30,
31].
Authors’ contributions
MSH, RB, and BZ conceived the paper and prepared the manuscript. BZ coordinated inputs to the draft. RB and BZ conceived the design and conducted the analysis. FC and CC contributed to study design. FC managed sample analysis. FC and IZ managed the local data collection. All authors read and approved the final manuscript.