Misclassification of malaria as pneumonia in children in sub-Saharan Africa

Abstract

Background: The World Health Organization (WHO) clinical case definitions for pneumonia were designed to prioritize sensitivity over specificity. In sub-Saharan Africa, the disease that is most likely to be misclassified as pneumonia is Plasmodium falciparum malaria.

Methods: By using chest X-ray positivity as an indicator for pneumonia, we estimated the extent of pneumonia misclassification due to malaria in the Pneumonia Etiology Research for Child Health (PERCH) study. Additionally, we developed a simple model to predict the proportion of pneumonia cases as defined by the WHO that could be attributed to malaria in settings with varying levels of malaria parasitaemia prevalence.

Results: In the PERCH study, the prevalence of malaria parasitaemia was low (4.7% among WHO pneumonia cases and 1.4% among controls) and we estimate that only 2.5% of WHO pneumonia cases were misclassified. However, when assuming a prevalence of malaria parasitaemia of 24%, corresponding to the average for malaria-endemic areas in Africa, we estimate that 28% of WHO pneumonia cases are misclassified. Among malaria-slide-positive WHO pneumonia cases in PERCH, lower chest wall indrawing [adjusted odds ratio (aOR) =18.1, 95% confidence interval (95% CI): 1.9, 175.8, P = 0.012], crackles on chest auscultation (aOR = 13.1, 95% CI: 1.4, 127.4, P = 0.027), and nasal flaring (aOR = 5.9, 95% CI: 1.1, 32.8, P = 0.041) were associated with chest X-ray positivity.

Conclusion: In settings that are typical of sub-Saharan Africa, we predict that one-quarter of WHO-defined pneumonia cases are malaria rather than pneumonia. Among children with WHO pneumonia who also test positive for malaria parasitaemia, clinical features that favour pneumonia include lower chest wall indrawing, nasal flaring, and crackles on chest auscultation.

Keywords: malaria; misclassification; pneumonia.

Figure 1.

Schematic of the method used to estimate the PPV of clinical pneumonia as defined by the WHO. WHO-defined cases are classified into four groups that are defined by true pneumonia and malaria-slide result, with the proportion in each group denoted $p_i (i=\mathrm{0,1},\mathrm{2,3})$. The sum of the middle two bars ($p_1+p_2$) represents the proportion of WHO-defined clinical cases that are true cases of pneumonia (PPV), whereas the last bar ($p_3$) represents the misdiagnosed proportion. Because true pneumonia status is unknown, the direct estimation of $p_i$is not possible. Instead, we use three observable quantities: (i) the proportion, $v$, of WHO cases that are malaria-slide-positive; (ii) the proportion, $q$, of WHO cases that are both positive on CXR and malaria-slide-positive; and (iii) the proportion, $r$, of WHO cases that are both CXR-positive and malaria-slide-negative. First, we estimate $p_1$ by assuming that the WHO pneumonia criteria are 100% specific among malaria-slide negatives. This implies that $p_0=0$ and $p_1=1-v$. Then, to estimate $p_2$, we divide $q$ by an estimate of the sensitivity of CXR ($\frac{r}{1-v}$) such that $p_2=\mathrm{ }q\times \frac{(1-v)}{r}$. Here, we assume that the sensitivity of CXR is independent of malaria parasitaemia. Finally, using the fact that the probabilities sum to one, $p_3$ can be estimated as $p_3=1-p_1-p_2$.

Figure 2.

Breakdown of children included in the analysis and reasons for exclusion.

Figure 3.

Misclassification of clinical pneumonia cases, as defined by the WHO, as a function of the community prevalence of malaria parasitaemia in children aged 1–59 months. Cases are categorized as either: (i) malaria-slide-negative, (ii) coinfected and attributable to malaria, (iii) coinfected and not attributable to malaria, and (iv) misdiagnosed. The PPV of the WHO pneumonia definition corresponds to the proportion of cases that are either malaria-slide-negative or coinfected.