Study setting
The data used for this study were collected between October and November 2015 during the 2015 NMIS that was jointly implemented by the National Malaria Elimination Programme (NMEP), the National Population Commission (NPopC), the National Bureau of Statistics (NBS), and the Malaria Partnership in Nigeria. Nigeria is a tropical country (2.66°E–14.68°E longitudes and 4.28°N–13.89°N latitudes) located at the Gulf of Guinea on the west coast Africa occupying almost a million square metres. Nigeria climates range from arid to humid equatorial. There are wide climatic variations in different parts of Nigeria. Towards the coast, temperature ranges 26–32 °C with high humidity but a much hotter temperature is prevalent in the North. Nigeria experiences a wet season from April to October with lower monthly temperatures and a dry season from November to March, with average midday temperatures of about 38 °C. Rainfall varies in Nigeria with 70 inches in the western coasts, 170 inches in the eastern coasts and 20 inches in the extreme north [15]. The diversities notwithstanding, Nigeria climate is generally very conducive for mosquitoes which are the malaria carriers. Plasmodium falciparum is the primary cause of malaria in Nigeria [4]. Nigeria has an annual population growth rate of 2.6% and estimated 180 million inhabitants who are greatly diversified culturally, socially and otherwise with only one-third residing in urban areas. Politically, Nigeria is divided into 36 administrative states and the Federal Capital Territory (FCT). These states were grouped into 6 regions on the basis of their location on Nigeria geographical landscape.
Sampling
The sample design was cross-sectional and nationally representative. The most current 2006 Nigeria National Population and Housing Census were used as the sampling frame. The states were subdivided into local government areas (LGAs), with subsequent sub-divisions into localities and convenient areas, also called census enumeration areas (EAs). The EAs, which are referred to as clusters, are used as the primary sampling unit in the 2015 NMIS. There were 138 clusters in urban areas and 195 in rural areas totalling 333 clusters. A two-stage sampling strategy was used for the survey. At the first stage, 9 representative clusters were randomly selected from each of the 36 states and the FCT. In the second stage, 25 households each were randomly selected in each cluster. Thereafter, all women aged 15–49 residents in the selected households were interviewed together. Also, all children aged 6–59 months in the selected households were tested for malaria and anaemia.
Anaemia testing
Anaemia was included in the 2015 NMIS for children aged 6–59 months as a result of the documented strong relationship between malaria infection and anaemia. With the use of a single-use retractable, spring-loaded, sterile lancet, finger/heel-prick, blood samples were drawn from every participating children and put in a microcuvette. Also, the Haemoglobin analysis was carried out on site using a battery-operated portable HemoCue® analyser. The results were given to each child’s parent or guardian in both verbal and written forms within a minute of testing. Referrals for follow-up care were offered according to guidelines. The anaemia test results were recorded on the Biomarker Questionnaire, and the households counselled appropriately.
Malaria testing using RDT
Using the same blood sample collected for anaemia testing, a drop of blood was tested immediately with the SD BIOLINE Malaria Ag P.f (HRP-II)™ (Standard Diagnostics, Inc.) RDT, being a qualitative test “to detect histidine-rich protein II antigen of Plasmodium falciparum in human whole blood” [4]. The test procedures were handled by well-trained field laboratory scientists in accordance with RDT manufacturer’s instructions. The RDT results were provided to each child’s parent or guardian in oral and written forms within 15 min and were recorded on the Biomarker Questionnaire. Children that tested positive to malaria and not currently on treatment with artemisinin-based combination therapy (ACT) or who had not completed a full course of the ACT during the preceding 2 weeks were given full treatment according to the Nigeria national malaria treatment guidelines [4].
Malaria testing using blood smears
In addition to the RDT, thick and thin blood smears were prepared in the field. Each blood smear slide was labelled according to guidelines and transmitted to the laboratory. The thick and thin smear slides were stained at zonal staining and taken to the ANDI Centre of Excellence for Malaria Diagnosis, University of Lagos, Nigeria for logging and microscopic reading. Other details of the testing procedures have been reported earlier [4].
Description of variables
The outcome variable in this study is the result of the RDT and microscopy malaria tests while the independent factors considered are child’s household wealth quintiles, child age, and sex of children, mother’s educational attainment, place of residence, region, sleeping under a long-lasting insecticide-treated net or any ever treated nets recently, experience of fever within 2 weeks preceding the survey, and the level of anaemia as used in earlier studies [3, 16]. The ages of the children were categorized into 0–6, 7–23, and 24–59 months as used in earlier studies on under-five children [17, 18].
Data analysis
There were a total of 7011 children aged 6–59 months across all the households visited during the survey. Basic descriptive statistics were used to describe the under-five children with respect to the characteristics of their mothers. The McNemar paired test and the Kappa statistics were used to test the hypothesis of non-agreement and to determine the level of agreement between the outcomes of the diagnostic tests respectively. Diagnostic test parameters including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) was used to determine the accuracy of the malaria RDT results in comparison with the “Gold standard” results from microscopy tests. The choice of microscopy as the gold standard is because it is the best available method for establishing the presence or absence of malaria. Also, the microscopy test results have been used as the gold standards in similar previous studies [7, 8, 10, 13, 16]. The Receiver Operating Curves (ROC) were used to estimate Area Under Curves (AUC).
The discriminatory accuracy of a diagnostic test is measured by its ability to correctly classify known cases (normal) and non-cases subjects (abnormal) [14]. The sensitivity, known as the True Positive Rate (TPR), is computed as proportion of True Positives (TP) among those that truly have malaria [TP + False Negative (FN)] according to the gold standard while specificity, known as the True Negative Rate (TNR), is the proportion of True Negatives (TN) among those that do not have malaria [False Positives (FP) + TN]. In contrast, the False Negative Rate (FNR) and the False Positive Rate (FPR) are (1-Sensitivity) and (1-Specificity) respectively. Generally, the higher the sensitivity and specificity, the better the RDT test.
Let X denote the true state of a person (microscopy test results) with microscopy positive = D+ and negative = D−. Also, let Y be the outcome of the RDT test, with RDT positive = T+ and negative = T−. Then,
$${\text{Sensitivity }} = P (Y = {\text{T}}^{ + } | {\text{X}} = {\text{D}}^{ + } ); \quad S{\text{pecificity}} = P(Y = {\text{T}}^{ - } |X = {\text{D}}^{ - } ) .$$
The positive likelihood ratio (LR+) and the negative likelihood ratio (LR−) are calculated as
$$LR + = \frac{TPR}{FPR} = \frac{\text{Sensitivity}}{{1 - {\text{Specificity}}}}\quad {\text{and}}\quad LR - = \frac{FNR}{TNR} = \frac{{1 - {\text{Sensitivity}}}}{\text{Specificity}}$$
as proposed by Simel et al. [19]. The unique statistics produced by the likelihood ratios have made it the optimal choice for reporting diagnostic accuracy for clinically meaningful thresholds [14].
However, there is a need to determine the predictive values since sensitivities and specificities are not measures of prediction [14]. Predictive values depend on disease prevalence, and their conclusions are transposable to other settings. The predictive values help to determine how likely the disease is, given the test result. The PPV is the probability that the disease is present, given that the diagnostic test is positive. It is computed as TP/(TP + FP) while the NPV is the probability that the disease is not present given that the test is negative, computed as TN/(TN + FN). A diagnostic test could be said to be perfect if it can predict perfectly, i.e., if PPV = NPV = 1. The PPV decreases with decreasing prevalence.
$${\text{PPV}} = P ( {\text{D}}^{ + } | {\text{T}}^{ + } );\quad {\text{NPV}} = P ( {\text{D}}^{ - } | {\text{T}}^{ - } ).$$
The accuracy of a test = (TP + TN)/(TP + TN + FP + FN) and its confidence intervals are the standard logits as given by Mercaldo et al. [20]. The pretest probability is the same as the prevalence as determined by the gold standard, the pretest odds is prevalence/(1 − prevalence), posttest odds = pretest odds * likelihood ratio and the posttest probability = posttest odds/(1 + positive odds).
The ROC analysis is used in diagnostic screening evaluation to quantify the accuracy of diagnostic tests [21]. The “roccomp” and “rocgold” implemented as ado-files Stata version 12 were used to plot the ROCs and compare the ROCs among different categories of children’s characteristics. The Lorenz curve, a measure of inequality and can be inscribed in the area between the curve and the diagonal line, was computed and quantified by the Gini index and the Pietra index. In all, data were weighted, the significance level was set at 5% with confidence intervals were estimated using earlier proposed methods [19, 22].