Study site
Malaria case data were collected for patients at two health facilities in Kilombero District, Morogoro, southern Tanzania. These facilities are a private dispensary in the village of Mwaya (S 07°51’17.4”, E 036°53’39.9”) and a government-owned public health centre in the village of Mang’ula (S 07°50’19.9”, E 036°53’32.8”). The two villages are 1.8 km apart, and adjacent to the Udzungwa Mountains National Park (Figure 1). Plasmodium falciparum is the only malaria species reported in this region[1].
According to the 2002 Tanzania Housing and Population Census, Mwaya has a population of 9,841 people and Mang’ula a population of 12,083 people[16]. The Kilombero Valley’s population is currently estimated at 321,611 people.
According to the National Park’s weather station, mean annual rainfall in this area ranges from 1,200-1,800 mm and follows a unimodal pattern falling from December to April; however, variability in this pattern has been observed in recent years (pers obs 2008-present).
Study design
A direct comparison between RDTs and microscopy was made at the private dispensary from June to mid-August 2011. The results were used to estimate retroactively the levels of over-diagnosis for much of the previous three and a half years at the nearby public health centre, which had more extensive data on monthly malaria rates, saw more patients and had a wider demographic of patients.
Study participants
At the private dispensary, eligible patients included any patient for whom the clinician judged a malaria diagnostic test necessary, based on case history and presenting symptoms. Between June and August 2011, 400 patients participated in this study and were tested with both microscopy and RDT.
The study population at the public health centre was made up of all patients who were tested for malaria during the time period for which malaria records were available, i e, from December 2007 until August 2011, excluding April 2008, September 2008 to March 2009, February 2010, and August 2010 to October 2010. Because of limited availability of malaria records at the public health centre, data during these months were never collected. All 9,175 patients tested with microscopy during this period were included in the retrospective study.
Study protocol
At the private dispensary, results of RDTs (ICT Malaria Combo Cassette, ICT Diagnostics, Cape Town, South Africa) and blood slide microscopy were directly compared. ICT Malaria Combo Cassette Test is a rapid, in vitro diagnostic test for the detection of circulating P. falciparum antigens (HRPII antigen); an antigen that is common to all five species of human malaria is aldolase antigen[17]. According to WHO RDT Product Testing (Round 1) results from 2008, the detection rate of the ICT Malaria Combo Cassette Test for identifying P. falciparum is 86% at low parasite density (<200 parasites/μ l) and 100% at a higher parasitaemia (2,000-5,000 parasites/μ l). This test is reported to have a Plasmodium vivax detection rate of 0% at low parasite density (<200 parasites/μ l) and 95% at a higher parasitaemia (2,000-5,000 parasites/μ l)[17]. Given the high sensitivity and specificity of RDTs, and because of the known inaccuracy of routine microscopy in many rural clinics, the RDT was considered the gold standard for the purposes of this study.
Each patient tested at the private dispensary completed a short questionnaire providing sociodemographic characteristics, medical history, information on malaria prevention measures, and information on the frequency and types of anti-malarial treatments used (see Additional file1). After testing with both microscopy and RDTs, patients returned to the clinician with both laboratory results and a final clinical diagnosis was made by the attending doctor and recorded by the study team.
At the public health centre, RDTs had been introduced by another research project as the sole diagnostic method for malaria in March 2011, in accordance with national policy of scaling up of RDTs at the country level. Prior to this introduction, microscopy was the sole diagnostic method. Because this clinic had ceased to use microscopy at the time of this study, the clinic did not at any point use RDTs and microscopy simultaneously. Therefore, given potential seasonal fluctuations in malaria burden, it was not possible to compare RDT results to microscopy results. Instead, a side-by-side comparison was done between the data collected at the private dispensary and the data collected previously at the public health centre. Hard copies of patient malaria records collected before and after RDT introduction were digitized to estimate the potential overuse of anti-malarials following diagnosis using microscopy. These records included patient age, sex, address, and the blood slide results (for patients tested prior to March 2011) or RDT results (for patients testing from March 2011 onwards).
Analysis of cost-effectiveness
At the public health centre, regardless of the type of diagnostic test used, the cost to the patient for malaria testing was 0.25 USD (500 Tanzanian Shillings, at 2011 conversion rates). The cost to the government of using RDT kits was 1.44 USD, adjusted for the price of the ICT Combo Cassette test (0.52 USD/test) (adapted from[18]); the cost of using microscopy was estimated at 0.59 USD[18]. These costs to the government include labour costs associated with each test and were used to calculate the average monthly costs spent by the government on malaria diagnostics. The average number of each diagnostic tests preformed was also calculated. Based on observations at the clinic, it was assumed that all patients diagnosed with malaria using either an RDT or microscopy received and used anti-malarial drugs. In conjunction with the data collected from the direct comparison at the private dispensary, it was possible to calculate: 1) an estimate of government expenditure on unnecessary anti-malarial treatments; 2) the cost to the government of providing subsidized RDTs; and, 3) the net savings resulting from accurate malaria diagnosis.
Statistical analyses
All analyses were undertaken using R version 2.13.1[19]. Only patients with complete data including age, sex, and malaria test results were included in the analysis. The breakdown of age and sex was tabulated for the data from each of the two health facilities. A binomial exact test was used to determine the rates of positive and negative microscopy results, and the rates of positive and negative RDTs at both facilities. Using the epi.tools package in R, the specificity, sensitivity, negative predictive values, positive predictive values, 95% confidence intervals, and the apparent and estimated over-treatment with anti-malarials was calculated for the private clinic. RDTs were considered the gold standard for this analysis. Using the retrospective design, the extent of over-treatment with anti-malarials and the cost-effectiveness of the introduction of RDTs were calculated for the public health centre. Because a large proportion of patients using clinics were children, a non-parametric regression was used in which subsequent age categories were double the size of the previous age category in order to allow a better representation of prevalence of malaria at a young age (one month, three months, six months, one year, two years, four years, eight years, 16 years, 32 years, 64 years, and 64+ years). Association between malaria prevalence and variables such as age and rainfall were tested with Pearson’s product–moment correlation.
Ethical considerations
Prior to participating in this study, each patient provided informed written consent. Ethical approval for this study was obtained from the Tanzania National Institute for Medical Research (NIMR/HQ/R.8a/Vol. IX/1165) and the Princeton University Institutional Review Board (IRB # 5374).