Introducing malaria rapid diagnostic tests in private medicine retail outlets: A systematic literature review

Background Many patients with malaria-like symptoms seek treatment in private medicine retail outlets (PMR) that distribute malaria medicines but do not traditionally provide diagnostic services, potentially leading to overtreatment with antimalarial drugs. To achieve universal access to prompt parasite-based diagnosis, many malaria-endemic countries are considering scaling up malaria rapid diagnostic tests (RDTs) in these outlets, an intervention that may require legislative changes and major investments in supporting programs and infrastructures. This review identifies studies that introduced malaria RDTs in PMRs and examines study outcomes and success factors to inform scale up decisions. Methods Published and unpublished studies that introduced malaria RDTs in PMRs were systematically identified and reviewed. Literature published before November 2016 was searched in six electronic databases, and unpublished studies were identified through personal contacts and stakeholder meetings. Outcomes were extracted from publications or provided by principal investigators. Results Six published and six unpublished studies were found. Most studies took place in sub-Saharan Africa and were small-scale pilots of RDT introduction in drug shops or pharmacies. None of the studies assessed large-scale implementation in PMRs. RDT uptake varied widely from 8%-100%. Provision of artemisinin-based combination therapy (ACT) for patients testing positive ranged from 30%-99%, and was more than 85% in five studies. Of those testing negative, provision of antimalarials varied from 2%-83% and was less than 20% in eight studies. Longer provider training, lower RDT retail prices and frequent supervision appeared to have a positive effect on RDT uptake and provider adherence to test results. Performance of RDTs by PMR vendors was generally good, but disposal of medical waste and referral of patients to public facilities were common challenges. Conclusions Expanding services of PMRs to include malaria diagnostic services may hold great promise to improve malaria case management and curb overtreatment with antimalarials. However, doing so will require careful planning, investment and additional research to develop and sustain effective training, supervision, waste-management, referral and surveillance programs beyond the public sector.

Background Provision of artemisinin-based combination therapies (ACTs) and other antimalarials to patients without confirmed malaria frequently results in overtreatment, potentially delaying diagnosis and treatment of other causes of illness and reducing availability of ACTs for true malaria cases [1,2]. Overuse of antimalarials by patients without malaria has been estimated to be half of global demand [3].
Prompted by recommendations from the World Health Organization in 2010 [4], national malaria programs in most endemic countries revised their diagnosis and treatment guidelines to emphasize the use of parasite-based diagnosis of malaria before treatment for all suspected malaria cases [5,6]. Since then, procurement of malaria rapid diagnostic tests (RDTs) has increased significantly in the public health care sector across much of sub-Saharan Africa [5,7]. In contrast, availability and use of diagnostic testing in the private medicine retail sector has remained low. Efforts to improve or expand malaria case management in the private sector, as demonstrated in the Affordable Medicines Facility-malaria (AMFm) pilot, focused on treatment delivery, but did not promote the use of diagnostic testing [8]. Evidence shows that RDTs or microscopy are available in less than 20% of pharmacies and drug shops selling antimalarials in six out of eight sub-Saharan African countries surveyed in 2013 or 2014 [9]. Though treatment-seeking practices vary greatly between countries, overall approximately one-third of febrile children obtaining malaria drugs are treated by private providers with limited access to malaria diagnostic services [3].
The private health care sector consists of private not-for-profit and private for-profit health providers, with the latter including private health facilities, diagnostic centers, private medicine retailers and informal practitioners [10]. Private medicine retail outlets (PMRs), a large category of for-profit private health providers in many countries [11], include outlets that specialize in medicines such as pharmacies and drug stores, as well as general stores or itinerant vendors that sell medicines along with other household merchandise [12]. In many countries, PMRs play a dominant role in the distribution and sale of antimalarials [9]. Typically, the outlets that specialize in selling medicines have storefronts, product displays, and a counter. Some may have a small room in the back, separated by a curtain or door, for examinations and treatment. Skills and qualifications vary among staff working in these outlets and include physicians, pharmacists, nurses and drug sellers with little to no formal health training [13]. PMRs are allowed to only carry over the counter drugs and in some cases a limited number of prescription drugs such as antimalarials and certain antibiotics. They are typically not allowed to perform diagnostic services, but government regulations vary amongst countries and are often poorly enforced [14][15][16].
Given the importance of PMRs as a first source of care and antimalarial treatment, several endemic countries in sub-Saharan Africa and Southeast Asia are considering introducing and scaling up RDTs in these outlets to achieve universal access to prompt parasite based diagnosis prior to treatment [17]. Introducing blood-testing in these outlets is not without controversy, and evidence to guide decisions on how and where to scale up RDTs amongst PMRs is currently lacking [18]. PMRs are often poorly supervised, rarely report into health information systems and are not equipped to manage severe illnesses [12]. Although the procedure to perform RDTs does not require specialized training, operators are required to draw and transfer an exact quantity of blood, apply a specific number of buffer drops, wait the required time before a result can be read (i.e. 15 or 20 minutes) and appropriately dispose of the hazardous infectious waste. Without adequate oversight, public health officials fear that PMRs may misdiagnose patients or not treat patients according to malaria guidelines, providing antimalarials or antibiotics to patients that test negative for malaria [19]. PMRs may also use substandard RDTs, affecting the trust in the result of the test and hence adherence to its results [20]. There is also a concern that improper handling of hazardous waste may lead to the spread of other infectious illnesses [21].
This review identifies and synthesizes available evidence and explores how it can help inform decisions about scaling up RDTs in PMRs.

Objectives
We undertook a systematic review of published and unpublished intervention studies to evaluate available evidence of the implementation and impact of RDT introduction in PMRs (pharmacies, drug stores, general stores, and/or itinerant vendors that sell medicines along with other household merchandise). The review aimed to: 1. Examine outcomes pertaining to RDT uptake, provider adherence to test results, referral, cost and safety.
2. Review characteristics of each intervention to introduce RDT use (e.g. the length and content of trainings, supervision frequency, referral guidelines, demand generation activities and retail price of RDTs) to explore factors that are associated with RDT uptake and provider adherence to test results.

Registration and eligibility criteria
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (http://www.prisma-statement.org/) and registered with PROSPERO (2013:CRD42013006146). We used the following inclusion criteria: • Participants: Any PMR providers and their patients We excluded studies that took place outside of PMRs among other private for-profit, private not-for-profit, and public health care providers (e.g., private health facilities, mission or non-governmental facilities, community health workers, and public health facilities); that reported only on the accuracy of RDTs (such as laboratory-based performance comparisons); where RDTs were not introduced into routine practice (if not performed by outlet staff or used only for reference by a research team); that evaluated the use, presence or proportion of outlets stocking RDTs without implementing any interventions to introduce RDTs; and studies based on hypothetical scenarios or modeling. To increase the evidence base, recent studies yet to be published at the time of the search were also included in the review. Principal investigators from unpublished studies were asked to extract specific testing and treatment outcome data to enable analysis across studies. Principal investigators of published studies were also asked to provide clarifications and data on additional outcomes not reported in the publication.

Search methods
We performed a systematic literature search of electronic databases on November 16, 2016, including PubMed/Medline, Cochrane Library Online, WHOLIS internet databases, IBSS, Web of Science and Ovid (EMBASE, Global Health, and Journals at Ovid). Studies which were yet to be published were identified at a Roll Back Malaria (RBM) Case Management Working Group, Informal Private Sector Task Force meeting in April 2013 [22] and a consultative working meeting on fever case management in the private health care sector in Africa, organized by ACT Consortium in October 2015 [17].
Search terms. Literature searches used synonyms and MESH terms for three concepts (i) 'malaria' (ii) 'rapid diagnostic test' and (iii) 'private sector'. No search terms or filters for methods were included. Table 1 provides an overview of the search terms.
Study selection. For published studies the resulting titles and abstracts were reviewed independently by two authors (TV and KB) to select papers or reports to read in full text. Discrepancies were resolved by a third author (KM). Papers that were clearly irrelevant were excluded after reading title and abstract. The remaining papers were read in full and excluded if they did not match the inclusion criteria after agreement between TV, KB and KM. Remaining papers were included in the systematic review.
For inclusion of unpublished studies, investigators were contacted initially to ascertain whether studies met the eligibility criteria, whether data would be available and/or computed within a given time frame and to reach agreement with investigators to include their unpublished findings in the review. Studies that met each of these criteria were subsequently included in the review and investigators asked to contribute results from their studies.

Data outcomes and extraction
Data extraction tables were used to collate information from both published and unpublished studies. The following diagnosis and treatment outcomes were compared across studies: 1. Uptake: the proportion of patients seeking treatment for fever or suspected malaria who were tested with an RDT 2. RDT positivity: the proportion of patients receiving a positive RDT result 3. ACT provision: the proportion of patients seeking treatment for fever or suspected malaria who were sold ACTs, regardless of whether or not they were tested 4. Adherence to negative or positive test results: the proportion of patients that were sold ACTs in the presence of a positive RDT result or the proportion of patients that that were not sold ACTs or other antimalarials in the presence of a negative RDT result 5. Antibiotic provision: the proportion of patients who were sold antibiotics in the presence of a positive RDT result; or the proportion of patients who were sold antibiotics in the presence of a negative RDT result 6. Referrals: the proportion of patients referred to a public facility by the provider for further care 7. Accuracy and safety: the proportion of PMR providers who accurately performed the RDT, read the result, and adequately disposed of the infectious hazardous waste

Median retail price of a RDT
We reported outcomes as proportions with comparable denominators where possible. In studies that provided cluster and individual level outcomes, we chose to use individual outcomes to enable comparison across studies. Where the same outcome was reported by more than one method of data collection, we chose the most complete data set, or presented neither if results for an outcome substantially differed between methods. To explore factors that appear to have supported RDT uptake and provider adherence to test results, outcomes across study arms were reviewed in terms of the characteristics of each intervention (length and content of trainings, supervision frequency, demand generation activities, recommended RDT retail price and referral policy). We did not make statistical comparisons between studies because of the different methodologies and outcomes used.

Study selection
A total of 1645 titles from published studies were identified through the search strategy (Fig 1). After removing duplicates, 904 titles and abstracts were screened and 136 publications were reviewed in detail. Of these, two studies focused on Cambodia [23,24], where RDTs had been scaled-up for over a decade. However, these studies did not directly evaluate the impact of implementation of RDTs on any outcomes comparable with other studies. Two other studies Each of these pairs are presented together and counted as one study. In addition, eight unpublished studies were identified. Two of these, a study in Madagascar and a study in Angola, were excluded, as data were not available at the time of this review. In all, six published [25][26][27][28][29][30][31][32] and six unpublished studies (please refer to the supporting information file S2 File) were included in the review, for a total of 12 studies. . The studies took place in areas of medium to high malaria transmission [33] and in rural, peri-urban, and urban settings. Outcomes were assessed using various data collection methods: provider records, exit interviews, mystery shoppers, direct observation, supervision visits, and household surveys (Table 3). Regulations in all study countries except Myanmar did not permit RDTs to be performed by providers in PMRs; studies were granted waivers or special permission from governments. , where RDTs were not subsidized. Gloves and infectious hazardous waste disposal units (i.e. a sharp box) were provided free of cost in most studies. RDTs were distributed either directly to a participating provider from a research warehouse or through a pre-selected wholesaler, importer or government entity. The length of the training varied from half a day to five days and often combined lectures and practice in performing the RDTs. Training content typically covered the symptoms of malaria and the recommended policies on antimalarial treatment and safety. In most intervention arms, training emphasized adherence to test results, including guidance on referral to nearby public health facilities for patients with signs of severe disease. Most studies also recommended referral when patients tested negative. Exceptions included a study in Uganda, where providers were trained on ACTs as first-line malaria treatment and how to perform RDTs but were not given specific algorithms on when to use RDTs or how to manage positive and negative results (Cohen et al. , 2015. In another study in Uganda, the RDT introduction was part of a five day integrated community case management (iCCM) training that included treatment of malaria, pneumonia, and diarrhoea (Awor et al. 2014(Awor et al. , 2015. In one arm of a study in Nigeria, training focused only on how to perform RDTs (Onwujekwe et al. 2015). The frequency of supportive supervision also differed, but in most studies research staff visited participating facilities monthly or quarterly to evaluate stock management, waste management practices and how RDTs were performed, stored,  Introducing malaria RDTs in private medicine retail outlets Testing and treatment outcomes Table 4 provides the diagnosis and treatment outcomes included in the review. Table 5 provides a summary of diagnosis and treatment outcomes by study arm alongside a summary of the supporting interventions implemented, ordered by RDT uptake (high to low).

Study design and characteristics
Uptake of RDTs. All studies reported on uptake (the proportion of eligible patients for whom an RDT was undertaken), which ranged from 8% (96/1279, exit interviews) in the provider and school-based intervention arm of a study in Nigeria    ) Ordered for 'Uptake' from high to low. Color coding of the different outcomes and interventions is based on the relative magnitude of the outcome (i.e., higher uptake is green, lower uptake is red) and the relative intensity of the intervention (i.e., shorter trainings are red, longer trainings are green). doi:10.1371/journal.pone.0173093.t005 The lowest adherence to positive results (30%, 128/421, household surveys) was found in a study in Uganda that also found relatively poor uptake and adherence to negative test results (Cohen et al. , 2015. Eight studies reported on antibiotic use. The proportion of RDT-negative patients receiving antibiotics ranged from 0.3% (6/1854, exit interviews) in a study in Ghana (Ansah et al. 2015) to 45% (51/113, household survey) in a study in Uganda (Mbonye et al. 2015). Three studies reported antibiotic use exceeding 20% (Cohen et al. , 2015Mbonye et al. 2015;Poyer et al. unpublished data), with the remaining studies reporting below 20%. Similarly, the proportion of patients with a positive RDT result receiving antibiotics varied from 0% (0/1351, provider records) in Ghana (Ansah et al. 2015) to 31% (129/441, household survey) in a study in Uganda (Cohen et al. , 2015. Studies that reported relatively high provision of antibiotics to RDT-negative patients also reported high provision to RDT positive patients (Cohen et al. , 2015Mbonye et al. 2015;Poyer et al. unpublished data).
Referrals. Only five studies reported on the proportion of patients who were referred to public hospitals or clinics. In all of these studies, providers were instructed to refer all RDT negative cases. In a study in Ghana (Ansah et al. 2015), 62% of the 1088 referred patients interviewed by phone reported attendance at the referral facility. The public health facilities had been made well aware of testing going on in the drug shops and accepted the referred patients.
In the remaining four studies, referrals were 10% or less of the cases. Reasons given for low rates of referral were explored explicitly in a qualitative component to the study in Uganda (Mbonye et al. 2015), although investigators in other studies reported similar challenges. Providers in the Ugandan study were reluctant to refer except when it was considered medically imperative because of fears that public health workers were unwilling to take patients referred from drug shops or would question the competence of outlet providers, thereby damaging their reputations [34]. Vendors were also concerned that clients might go to another shop rather than to a public facility, and they would lose their clientele. In almost all settings, formal referral systems from private medicine retail outlets to public facilities had not been established. In those studies where providers were encouraged to refer patients with severe illness or if a clear diagnosis could not be made, there was anecdotal evidence of poor or contradictory treatment at the receiving facility.
Safety & accuracy in performing RDTs. Nine studies provided data on how RDTs were administered using a check list. In general, most providers were able to accurately perform the RDT, read its results and dispose of the hazardous infectious waste appropriately. In six studies where this outcome was recorded, approximately 85% or more of the providers performed the test safely and correctly (Aung et

Discussion
The introduction of RDTs in PMRs, a primary source of care in many settings, aims to improve case management of febrile illnesses through prompt and appropriate treatment of malaria and a reduction in delays to diagnosis and treatment of other illnesses. This review demonstrates that while RDT introduction can achieve this goal, such outcomes are not guaranteed. Although studies with more intensive interventions generally produced better outcomes, it is unclear whether such efforts could be maintained or scaled up to national level.
The three studies that showed the highest uptake and the highest adherence However, there are notable exceptions to these trends. A study in Nigeria (Onwujekwe et al. 2015) compared the uptake of RDTs and adherence to national malaria guidelines under different training scenarios and found that longer and more comprehensive training (two days covering diagnosis and treatment versus one day with only a demonstration on how to use RDTs) did not appear to affect uptake or adherence. In contrast, classroom-based trainings on malaria case management in a study in Myanmar (Aung et al. 2015), were relatively short (only 0.5 days), but uptake and adherence were better than in some studies with longer trainings. Similarly, in a study in Tanzania (Maloney et al. under review), subsidizing the retail price of RDTs by over 50% did not increase uptake compared with an unsubsidized price. Factors that may limit the comparability of outcomes to the intensity of the related interventions include study setting and context (e.g., prior exposure of provider to malaria case management training), the timing of the evaluation (e.g., 3 months vs. 12 months after implementation), the method of data collection (e.g., mystery shopper vs. provider records), the number of outlets included in the studies (e.g., 18 vs. 1502 outlets) or unique events that affected study outcomes (e.g., in a study in Nigeria (Streat et al. Nigeria unpublished data), leakage from public sector into the private sector flooded the market, negatively impacting the uptake'of 'project' RDTs).
None of the studies deployed interventions that could be scaled-up easily at the national level. For example, a highly effective intervention in Uganda (Mbonye et al. 2015) included four day trainings, weekly supervision visits for the first two months and a free, continued supply of RDTs, gloves and sharp boxes. Studies that implemented less intense but perhaps more scalable interventions often produced poorer outcomes. For example, in a study in Uganda (Cohen et al. 2015 that showed low RDT uptake and adherence, PMRs were free to choose the price at which the RDTs were sold and free to make treatment recommendations as they wished. Another study, where RDT stock outs were recorded in a study in Tanzania (Maloney et al. under review), chose not to control the supply of RDTs; PMRs were simply informed where they could procure RDTs. Schools in an intervention arm of a study in Nigeria (Onwujekwe et al 2015) were supported to organize malaria events to promote uptake and adherence, but only half of the participating schools did so.
Heterogeneity in outcomes following RDT introduction is not unique to the private sector [35][36][37]. While many public health facilities that increased diagnostic testing for malaria through the use of RDTs also reported reductions in ACT provision, the availability of RDTs alone does not seem sufficient to ensure the appropriate use of ACTs [38,39]. Public and private providers alike have rationales for providing antimalarials to patients with a negative RDT result. Anxiety over the potential for patients to worsen without being given antimalarials seems paramount [40,41], just as with antibiotics in other settings [42]. This is accentuated in contexts where antimalarials are expected, or even demanded, by patients or customers [43], and where clients can take their business elsewhere [44]. Overstretched providers may find the time it takes to perform the RDT prohibitive and choose to assist other customers instead of performing the RDT or waiting for its result [45]. In all sectors, behavior change is likely to require sustained efforts.
Experience from these studies showed that requirements for introducing RDTs at scale in PMRs should be viewed as the introduction of a comprehensive service, not just another commodity. However, evidence on how to do so remains limited in many operational aspects. First, evidence is needed on how to integrate malaria testing into case management beyond malaria. Where negative cases are expected to be referred, this may be challenging: clients have chosen the retail sector, providers are keen to make a sale, and public sector workers may be unwelcoming to patients referred from PMRs [34]. Guidelines for managing RDT negative adults and children require specific development, based on levels of expertise, resource availability, and local regulations. Second, evidence is needed on how to train and supervise PMRs, given the size and heterogeneity of the sector as well as rapid staff turnover [13]. It may not be feasible or even desirable to train and supervise all PMRs. Some studies experimented with innovative supervision approaches to prioritize certain shops over others based on sales volume or performance [Poyer et al. unpublished data], but little is understood how to find, select or retain these 'successful' providers. Some studies in the review [Poyer et al. unpublished  aimed to provide more sustainable mechanisms (i.e. having professional associations instead of research team members conduct supervision visits, not subsidizing the RDT or controlling the supply chain) that could be scaled and exist beyond the length of the study, but scale up was not tested. New innovative approaches that build on existing structures and programs in the private sector, rather than building parallel infrastructures, require exploration. Third, evidence is needed on how to deal with hazardous waste from testing at scale in these non-clinical settings. In most studies in this review, research teams were responsible for this. One study in Tanzania (Maloney et al. under review) that instructed providers to visit public health facilities to drop off waste had mixed success; many providers instead chose to bury or burn their waste. A study in Uganda (Streat et al. Uganda unpublished data) contracted a private firm to collect the waste at each of the participating outlets, but poor uptake combined with frequent collection visits caused cost overruns. New innovative approaches to waste disposal require development and evaluation under real world conditions. Finally, additional consideration must also be given to issues beyond malaria control, such as role of PMRs in the wider health system and the legal and regulatory frameworks for in vitro diagnostics. A sustained scale up of RDTs in the private retail sector would require recognition from stakeholders, including regulatory bodies, that PMRs are a viable alternative to public sector provision of quality care for uncomplicated malaria.

Limitations
This review employed a broad search strategy to identify all eligible studies where RDTs were introduced in the private medicine retail sector. We did not include studies that included formal private health facilities such as clinics or hospitals. Since countries are making decisions now about if or how RDTs can be introduced in PMRs, we decided that waiting for more studies to complete the publication process was deemed too much of a delay. While it is possible an eligible but unpublished study could have been missed, this is unlikely given the involvement of extensive contacts identified through the two convened stakeholder meetings in 2013 and 2015 and the large group of authors involved. Some studies did not have data on all the outcomes assessed in the review. The mix of study designs (i.e. differences in intervention and control arm design) and evaluation methods (i.e. mystery shopper or provider records) made formal comparison of point estimates inappropriate. Differences in expectations of RDT positivity and patient demand for diagnosis across studies further limited comparability. The studies included in the review were all small scale trials or pilots with short durations. Most studies evaluated outcomes at a single point in time, which may not be representative of embedded and sustained effects. Finally, studies included in the review did not address the potential regulatory and policy barriers of introducing RDTs to PMRs. All of the studies, except in Myanmar, were provided a waiver to perform RDTs.

Conclusions
Supporting the introduction of malaria rapid diagnostic testing in private medicine retail outlets has the potential to target antimalarial drugs more effectively. This review shows that a range of private providers in different countries can incorporate RDTs into their practice, although with varying degrees of uptake and influence on case management. This review suggests investment in training and supervision may be beneficial to supporting RDT implementation. However, substantial gaps remain in the evidence for systems that will allow for RDT implementation at scale.