Background
Adequate diagnosis and prompt treatment of malaria remain major difficulties in rural settings in sub-Saharan Africa. In these areas, more than 70% of individuals with symptoms suggestive of malaria treat fevers with anti-malarial drugs without visiting the formal health sector for diagnosis [
1,
2]. If people do visit a health facility, an accurate diagnosis is not guaranteed. Microscopic examination of a blood smear is the gold standard method for the diagnosis of malaria, but is often unavailable at sub-district facilities [
3]. In the absence of microscopy, malaria diagnosis is based on clinical symptoms that are known to lack specificity [
4], or treatment is administered presumptively. Even if microscopy is available, there is substantial overdiagnosis of malaria [
5]. There is an urgent need to review these diagnostic and treatment practices with the wide-scale implementation of the relatively expensive artemisinin-combination therapy (ACT) and in the light of changing patterns of malaria transmission.
Presumptive treatment of malaria remains a useful tool for malaria control in areas where malaria is common, the risk of progression to severe malaria is significant, and diagnostic facilities are lacking. In these areas, presumptive treatment is an effective strategy to increase the coverage of anti-malarials and may act to reduce transmission of malaria [
6]. However, the strategy may not be justifiable for low endemic areas where the majority of febrile episodes are not due to malaria [
7] and over-diagnosis and over-treatment of malaria are very common [
4,
8,
9]. Knowledge of malaria transmission intensity can guide clinicians in defining algorithms for dealing with febrile patients. However, reliable estimates of malaria transmission intensity are frequently unavailable and transmission intensity can change over time as a result of interventions [
10,
11] or natural fluctuations [
12]. Across the African continent, there are reports of recent reductions in malaria transmission intensity (reviewed in [
6]).
Investigating how rural health systems function in the current climate of changing transmission and efforts to control and eliminate malaria seems pertinent. Here, diagnostic and treatment practices in rural health centres in Mto wa Mbu are described, an area that was historically hyperendemic for malaria [
13], but where no recent assessments of transmission intensity have been conducted.
Discussion
In the study area of Mto wa Mbu that is characterised by low transmission intensity, an unrealistically high perceived burden of malaria was observed. Forty percent of outpatients were treated for malaria while this diagnosis was not supported by blood slide or rapid test and was improbable in the light of the actual level of transmission intensity.
The level of malaria transmission intensity, as determined by rapid entomological assessment and MSP-1
19 age-seroprevalence data [
20], was low in the area of Mto wa Mbu. The EIR estimate by entomological assessment was 0.7 infectious bites per person per month. This estimate was obtained in the period after the long rains and probably represents the peak exposure to infected mosquitoes. It is, therefore, not surprising that it exceeds the EIR estimate based on the MSP1
19 age-seroprevalence curve (0.7 infectious bites per person per year) although the confidence intervals for both methods overlap. In general, EIR by entomological assessments is susceptible to seasonal fluctuations [
12], fluctuations that are smoothened out by the immunologic assessment [
20,
22] that may, therefore, give a more robust estimate. Either method will lead to the conclusion that transmission intensity in Mto wa Mbu is low. This is in contrast with cross-sectional surveys from 1981 [
13], when parasitological surveys indicated that the area was having a level of transmission intensity 'certainly equal to and probably higher than that found in Muheza-Ubembe' [
13], an area that is still known to be hyperendemic for malaria [
23]. Malaria transmission intensity clearly decreased in last 25 years but the period in which the reduction took place is unclear. (Recent) reductions in transmission intensity can be detected in age seroprevalence curves when the force of infection (λ) is allowed to change over time [
22]. However, a variable λ did not improve the fit of our curve (data not shown), and we found no indications for a recent reduction in EIR. Similar to its timing, the reasons for the reduction in EIR are unclear. The use of chloroquinized salt can not explain this since it was used in the 1960s and 70s and parasite rates of 50–60% in children < 10 years of age were still reported in 1981 [
13]. Bed net coverage was high in the study area, but the coverage with ITNs has only increased in recent years [
24] and also provides no plausible explanation. Similar to other areas in Africa [
6], the drop in transmission intensity is not easily explainable. The high use of anti-malarials for any fever treatment, as demonstrated in this study, could have acted as mass prophylaxis and reduced transmission over time [
6], but this cannot be proven.
Despite the low transmission intensity in the area of Mto wa Mbu, more than forty percent of all outpatients who attended two major clinics over a period of almost two years were diagnosed with malaria. There was no seasonal pattern in malaria diagnoses, as is commonly observed [
12,
14]. Prospective data from more than 400 suspected malaria cases demonstrate that there was a massive overdiagnosis of malaria in the two clinics. Despite the availability of microscopes, experienced microscopists and clinicians, who frequently requested slides, the targeting of anti-malarials appeared to be unimproved [
5,
25]. One major explanation for this was that slide readings were unreliable in the clinics included in this study: less than 1% of slides that were regarded as malaria parasite-positive by clinic microscopists were confirmed by trained research microscopists. In addition, clinicians appeared to use blood slide results more as a tool to confirm their clinical suspicion rather than to rule out malaria [
25]. All individuals with 'positive slides' received anti-malarial treatment although a 'negative slide' did by no means rule out treatment [
9]. This could be a result of the deeply entrenched belief in slide negative malaria. It is true that clinical malaria can result from low density parasitaemia in low endemic areas and malaria can therefore not always be ruled out in slide-negative cases [
4]. However, these are exceptional cases and restricting anti-malarials to true microscopy-positives is a safe approach, even in areas of low endemicity [
8]. The prescription of anti-malarials in Mto wa Mbu was clearly out of proportion. An astounding 99.6% (248/249) of the individuals who were treated with artemether-lumefantrine (AL) was free of malaria parasites.
Although a recently published hypothesis suggests that overdiagnosis and overtreatment of malaria can have a beneficial prophylactic effect in malaria control [
6], the current findings are worrying for several reasons. Firstly, our area is of low endemicity and a beneficial prophylactic effect is unlikely [
6]. Secondly, diagnosis of malaria and according treatment may simply be a 'convenient' clinical strategy avoiding the more complicated search for other causes of the presenting illness [
25]. Treatment of all febrile episodes as malaria is likely to result in underdiagnosis of other fever-causing disorders such as childhood pneumonia [
2]. Thirdly, over-treatment occurred with the expensive AL. AL and other ACTs are typically 10-times more expensive then previously used drugs as sulphadoxine-pyrimethamine [
26,
27] making reliable diagnosis crucial for cost-effective use [
28]. The artemisinin component in ACT also do not have the prophylactic effect that was suggested to be beneficial in 'opportunistic presumptive treatment' [
6]. Artemisinin is eliminated from the circulation in a matter of hours [
29] leaving the partner drug, in this case lumefantrine, unprotected. That leads us to the fourth reason for unease. There is concern for a reduced susceptibility of
P. falciparum parasites for ACT [
29] and the spread of parasites with reduced susceptibility to ACT may be enhanced by irrational drug use [
30]. Reports on allelic selection after artemether-lumefantrine [
31] provide additional warnings against over-use of ACT.
For a way forward, it is important to understand why there is so much overdiagnosis and overtreatment in the study area. A recent study elsewhere in Tanzania has demonstrated that patient pressure, traditionally mentioned as a major contributor [
32], may not be important in overtreatment [
33]. Patients often prefer to be slide-tested and treated in line with results [
33]. In the clinics in Mto wa Mbu, slide reading was clearly inadequate. Microscopists will need additional training and in addition, more objective diagnostic tools such as RDTs can play a role in the improvement of diagnosis. Clinicians in this study felt uncomfortable to rule out malaria based on a negative RDT (and one patient in our study was in fact a false RDT negative). Time will be needed to make them consider RDTs as useful diagnostic tools [
34]. Perhaps delivering a message to health workers and the public explaining that malaria control has been successful in some areas and malaria has truly been reduced may improve the situation. If both health staff and public understand that this fever may not be due to malaria quality of care may improve and "routine overdiagnosis" may be a story of the past.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
CM, AN, BM and SM were responsible for data collection and were involved in data analysis and manuscript preparation; SS and HM were involved in statistical data analysis and manuscript preparation; FM and RS in study design; CD, RG and TB were responsible for study design, data interpretation and manuscript preparation.