Introduction Morbidity and mortality due to Plasmodium falciparum malaria have been increasing in sub-Saharan Africa since the early 1990s, concomitantly with spread of resistance to commonly used monotherapies, i.e., chloroquine and sulfadoxine-pyrimethamine [1], [2]. This increased resistance has necessitated that many African countries change their treatment policy to artemisinin-based combination therapy (ACT) as a first-line treatment for uncomplicated malaria. The restricted use of ACT to confirmed malaria patients is critical. Overuse of the more expensive ACTs will not only put an extra heavy financial burden on malaria control programmes in Africa, but also enhance drug resistance and prevent other causes of fever from being appropriately treated, for example, pneumonias, which require antibiotics. Symptom-based or clinical malaria diagnosis has proven to be quite unspecific [3]–[7]. Malaria diagnosis based on parasitological confirmation is therefore increasingly advocated. Integrated Management of Childhood Illness (IMCI) algorithms based on clinical symptoms could potentially be made more efficient and cost-effective if simple parasitological diagnostic methodologies were incorporated. The use of microscopy has been tried in various health care settings, but is associated with problems of logistics, sustainability, and quality control [8], [9]. The development of rapid diagnostic tests (RDTs) for P. falciparum malaria offers a potential alternative in remote and poorly resourced health facilities that are beyond the reach of high-quality microscopy services [10]–[14]. The combination of RDT and ACT provides an important strategic opportunity to reduce malaria-associated mortality in Africa, and RDT use will potentially improve treatment of other causes of fever, for example, life-threatening bacterial diseases [15], [16]. However, the evidence base is still inadequate for malaria control programmes to recommend the use of RDTs on a large scale. There are several studies on sensitivities and specificities of various malaria diagnostic methods [11]–[14], [16]–[19]. In two recently published studies on the implication of RDT use at the health facility level on drug prescription, both describe major problems with test efficiency when used in clinical practice [15], [20]. However, this may be attributed to different messages regarding the risk of withholding malaria treatment to patients with negative test results [21]. Also, these studies did not describe staff training on technique and validation of RDTs, a prerequisite for the malaria diagnostic tests to become cost effective [22]. Furthermore, and importantly, there are no randomized control trials on the health impact and cost-effectiveness of confirmatory malaria diagnosis based on RDTs [18]. Zanzibar was among the first regions in sub-Saharan Africa to introduce ACT, free of charge through public health care, as both as first- and second- line treatment for uncomplicated malaria, which are provided free of charge through public health care. In view of the fact that many patients with fever are prescribed ACT without being malaria infected, the present study was undertaken to assess, on a wide scale, the added value of RDT to clinical diagnosis (CD) alone for management of patients of all ages presenting with fever at primary health care facilities. The hypothesis was that RDT-aided diagnosis of fever patients would improve rational use of ACTs and possibly other necessary treatments, such as antibiotics to non-malaria patients, with an overall improved health impact. Material and Methods Study Area and Study Health Centres The trial was conducted in four Primary Health Care Units (PHCUs) in Zanzibar, namely, Muyuni and Uzini on Unguja Island, and Kinyasini and Mzambarauni on Pemba Island. The selection of the four study sites aimed to provide a representative picture of Zanzibar with regard to malaria epidemiology as well as previous use of RDT in Zanzibar. By the time of the trial, malaria transmission in Zanzibar was generally considered endemic [23], with recorded malaria parasite rates between 10% and 50% in different age groups (unpublished data, Zanzibar Ministry of Health). A previous clinical trial conducted in two comparable PHCUs had shown an overall malaria parasite prevalence of about 30% among febrile children aged 15 y Total CD+RDT 228/544 (42%) 93/210 (44%) 40/251 (16%) 361/1,005 (36%) CD alone 423/503 (84%) 169/196 (86%) 160/183 (87%) 752/882 (85%) Microscopy 374/1,047 (36%) 128/406 (32%) 50/434 (12%) 552/1,887 (29%) A majority of antimalarial prescriptions were for children below 5 y, 228/361 (63%) in CD+RDT group and 423/752 (56%) in CD alone group. Prescription in relation to microscopy results are presented in Table 2. 10.1371/journal.pmed.1000070.t002 Table 2 Proportions of patients receiving antimalarial drugs and antibiotics in relation to day 0 microscopy results. Drugs Received by Patients Diagnostic Testing Blood Slide Result 15 years Total Antimalarials CD+RDT BS positive 186/200 (93%) 71/72 (99%) 22/33 (67%) 279/305 (91%) BS negative 42/344 (12%) 22/138 (16%) 18/218 (8%) 82/700 (12%) CD alone BS positive 174/174 (100%) 54/56 (96%) 17/17 (100%) 245/247 (99%) BS negative 249/329 (76%) 115/140 (82%) 143/166 (86%) 507/635 (80%) Antibiotics CD+RDT BS positive 51/200 (26%) 4/72 (6%) 3/33 (9%) 58/305 (19%) BS negative 190/344 (55%) 52/138 (38%) 72/218 (33%) 314/700 (45%) CD alone BS positive 38/174 (22%) 4/56 (7%) 1/17 (6%) 43/247 (17%) BS negative 140/329 (43%) 28/140 (20%) 24/166 (14%) 192/635 (30%) A total of 607/1,887 (32%) patients were prescribed antibiotics, including mainly cotrimoxazole, but also ampicillin, amoxicillin, and erythromycin. Prescription of antibiotics was significantly higher in the CD+RDT than CD-alone group, 372/1,005 (37%) and 235/882 (27%) (OR 1.8, 95%CI 1.5–2.2, p 99% for detecting a parasite density of ≥1,000 parasites/µl, 76% and 59% for parasite densities 100–999 and 99% Specificity 88% 20% Positive predictive value 77% 33% Negative predictive value 96% 98% Antimalarial and antibiotic prescriptions in relation to age and BS results are presented in Table 2. Among a total of 552 BS-positive patients, 28 (14 children below age 5 y) were not prescribed antimalarial treatment, 26 after CD+RDT (RDT negative), and two after CD alone. Their parasite densities at enrolment were, however, relatively low (GM 174 parasites/µl blood, range 32–2029). Among patients with BS negative results a total of 82/700 (12%) were prescribed antimalarial drugs in the CD+RDT group (RDT positive) compared with 507/635 (80%) in the CD-alone group (Table 2). A total of 82/361 (23%) antimalarial treatments in the CD+RDT group and 507/635 (80%) in the CD-alone group may thus have been unnecessary according to microscopy results (Tables 1 and 2). In contrast, BS-negative patients received significantly more antibiotics in the CD+RDT group, 314/700 (45%) patients compared with 192/635 (30%) in the CD alone group (OR 2.1, 95%CI 1.6–2.6, p 15 y All ages CD+RDT group General costs 1.90 1.90 1.90 1.90 Drugs 0.39 0.63 0.69 0.51 Reattendance 0.08 0.06 0.03 0.06 Total mean costs 2.37 2.59 2.62 2.47 CD alone group General costs 1.40 1.40 1.40 1.40 Drugs 0.58 0.89 1.55 0.85 Reattendance 0.17 0.05 0.05 0.12 Total mean costs 2.15 2.34 3.00 2.37 All estimates are based on an exchange rate of USD 1 = TSh 1,100. General costs = transport (USD 0.90)+consultation (USD 0.50)+RDT (USD 0.50) = USD 1.90. Drugs = ACT (USD 0.50−1.40)+antibiotics USD (0.30−0.90)+antipyretics (USD 0.05−0.20). Reattendance costs = transport (USD 0.90)+consultation (USD 0.50)+drugs (ACT, antibiotic+antipyretics = average USD 1.10) = USD 2.50. Discussion We found an overall 2-fold reduction in prescription of antimalarial drugs and reattendance of patients due to illness during the two-week follow-up period in the CD+RDT group compared with CD-alone group. Overall costs were, however, similar in the two groups despite a significant reduction of cost among the adult patients after RDT-aided diagnosis. Almost all enrolled fever patients in the CD-alone arm were considered and treated as malaria patients, resulting in high diagnostic sensitivity (99%) but low specificity (20%). This result follows the suggestion that fever alone may be a better criterion for malaria treatment than more complicated algorithms [4]. Studies on clinical diagnostic algorithms have shown that with weighting and scoring systems for clinical signs and symptoms may result in sensitivities of 70%–88% and specificities of 63%–82% [3], [4], [17], [27]. However, these methods may be too complicated to be effective under operational conditions, and the algorithms may be site- and context-specific [4]. Health workers learnt to use RDTs correctly with relative ease, confirming that the tests are simple to perform and interpret [11]. The estimated sensitivity (>100 parasites/µl of blood) is in line with WHO recommendations [10] and is also in accordance with a recent review concluding that the accuracies of the HRP2-based test in P. falciparum–endemic areas are normally high with a mean sensitivity of 93% [13]. The specificity in our study—88%—was similar to or lower than in some previous studies [12], [13], [15], [16], [19]. Especially under field conditions, heat and time stability could be an important impediment for the optimal use of RDTs for malaria, but according to the manufacturer Parachek Pf is expected to be stable at temperatures up to 40°C for up to two years. The use of confirmatory malaria diagnosis with RDT is expected to reduce the overuse of antimalarial drugs by ensuring that treatment is targeted to patients suffering from malaria infections as opposed to treating all patients with fever. Our findings confirm this expectation, although the impact of RDT-aided diagnosis will obviously be highly dependent on the malaria incidence (prevalence of malaria in fever patients) in a given situation. Importantly, in our study, the study nurses showed great confidence in the RDT results as a guide to choice of treatment, as did the patients. This is in contrast to the assumption that care providers, although willing to perform diagnostic tests, do not always comply with the results, especially when the result is negative [15], [20]. High adherence by prescribers in relation to RDT results was, however, also reported in a recent study conducted in mainland Tanzania [28]. We believe that the high compliance and confidence in the RDT in our study may result from a successful pre-study training, although local beliefs, behaviours, and treatment traditions may also account for discrepancies between our results and those of previous publications [15], [20]. We further realise that the study situation, supervision, and incentives provided to the nurses may also affect compliance, but we do not believe it has seriously biased our results. The incentive to the nurses was consistent with common practice for project participation in Zanzibar, but whereas it represented up to approximately a 65% increment of the ordinary salary it was not influenced or affected by performance. Our results obviously need to be confirmed before RDT can be more generally recommended, but we do believe they suggest that RDT use may be efficient if local diagnostic and treatment traditions are properly addressed. Fearing false negative test results and being aware that delays in providing effective treatment can be fatal for malaria patients is reported to be the main reason to prescribe antimalarial drugs despite a negative RDT result. Importantly, in our study, the patients with malaria detected by BS but non-detectable by RDT and therefore not treated with antimalarial drugs had relatively low parasite densities and no patients developed any severe malaria manifestations during the two week follow-up. This supports a general recommendation of consistence in not treating RDT negative patients. Re-testing will, however, obviously be required if the illness remains or aggravates. Our finding of a reduction in perceived illness during a two-week follow-up in the CD+RDT group of patients is critical. This was probably attributed to improved treatment of patients with fever not associated with malaria. More antibiotics were prescribed to the RDT-negative patients. The introduction of RDT and ACT thus provides an opportunity to improve the treatment of both malaria and bacterial diseases. We did consider the potential selection bias of the four health facilities; indeed, significant heterogeneity was observed with regard to the primary effect parameter. We do, however, assume this heterogeneity was at least partly accounted for by multilevel analysis and, since the RDT effect on drug prescriptions was quite large in each PHCU, it seems unlikely to be due to selection bias. The selection of the four study sites was done to provide a relatively representative picture of both malaria epidemiology and previous use of RDT in Zanzibar. Since RDTs had already been introduced by MSF in some parts of Zanzibar we opted for including PHCUs both with previous experience (two sites) and without previous RDT use (two sites). Beside the previous RDT exposure, the selection of the four study sites was based on representing a common rural situation and representing both of the two major islands in Zanzibar, i.e. Unguja and Pemba (two PHCUs on each island). However, we do of course acknowledge that the choice of the four sites remains arbitrary and of low number and thus cannot be fully representative of an overall Zanzibar situation and even less so of an overall situation in sub-Saharan Africa, which indeed is very diverse itself with regard to epidemiology of malaria, cultural and behavioural aspects, health care structure, etc. With the understanding that four PHCUs is a very low number, we used a cross-over design of RDT versus non-RDT weeks within sites. The choice of RDT or non-RDT the first week was based on an allocation with one previous MSF/RDT site being in either arm and one non previous MSF/RDT similarly being in either arm. Still, we acknowledge that there may still be confounding effects with regard to health-seeking behaviour or even selection bias by study nurses on respective weeks by (a) patients/caretakers postponing health care attendance to a week with RDT or staff applying exclusion criteria on a CD week and instead request the patient to return on a RDT week, and/or (b) attending alternate PHCUs where RDT is performed. However, we assume that (a) is less realistic, considering that uncomplicated malaria requires urgent treatment and patients or their caretakers as well as health care workers are therefore not likely to wait and postpone treatment. We also do not believe (b) is realistic because systematic RDT use was not implemented outside study PHCUs at the time of the trial and the study sites were located far from each other. And indeed statistical analysis showed no significant difference between frequencies of fever patient attendance on RDT and non-RDT weeks. The only trend observed with regard to frequency of attendance was a tendency to a relative increase toward the later period of the study, compatible with increased malaria transmission. In summary, RDTs were well performed in peripheral health facilities with acceptable sensitivity and specificity for identifying malaria-attributable fever episodes. The RDT results were adhered to and did provide consistent and significant reduction in antimalarial treatment in parallel with an increase in prescribed antibiotics. This probably contributed to the significant reduction in reattendance due to illness during the two-week follow-up. Our results indicate that RDTs may represent an important tool for improved management of fever patients in peripheral health care settings in malaria-endemic areas, especially where ACT has been introduced for treatment of uncomplicated malaria. Supporting Information Text S1 Trial protocol. (0.15 MB DOC) Click here for additional data file. Text S2 CONSORT checklist. (0.06 MB DOC) Click here for additional data file.
