Diagnosis of pneumonia and malaria in Nigerian hospitals: A prospective cohort study

Abstract Background Pneumonia and malaria are the leading causes of global childhood mortality. We describe the clinical presentation of children diagnosed with pneumonia and/or malaria, and identify possible missed cases and diagnostic predictors. Methods Prospective cohort study involving children (aged 28 days to 15 years) admitted to 12 secondary‐level hospitals in south‐west Nigeria, from November 2015 to October 2017. We described children diagnosed with malaria and/or pneumonia on admission and identified potential missed cases using WHO criteria. We used logistic regression models to identify associations between clinical features and severe pneumonia and malaria diagnoses. Results Of 16 432 admitted children, 16 184 (98.5%) had adequate data for analysis. Two‐thirds (10 561, 65.4%) of children were diagnosed with malaria and/or pneumonia by the admitting doctor; 31.5% (567/1799) of those with pneumonia were also diagnosed with malaria. Of 1345 (8.3%) children who met WHO severe pneumonia criteria, 557 (41.4%) lacked a pneumonia diagnosis. Compared with “potential missed” diagnoses of severe pneumonia, children with “detected” severe pneumonia were more likely to receive antibiotics (odds ratio [OR], 4.03; 2.63‐6.16, P < .001), and less likely to die (OR, 0.72; 0.51‐1.02, P = .067). Of 2299 (14.2%) children who met WHO severe malaria criteria, 365 (15.9%) lacked a malaria diagnosis. Compared with “potential missed” diagnoses of severe malaria, children with “detected” severe malaria were less likely to die (OR, 0.59; 0.38‐0.91, P = 0.017), with no observed difference in antimalarial administration (OR, 0.29; 0.87‐1.93, P = .374). We identified predictors of severe pneumonia and malaria diagnosis. Conclusion Pneumonia should be considered in all severely unwell children with respiratory signs, regardless of treatment for malaria or other conditions.


| INTRODUCTION
Pneumonia and malaria are the leading causes of global mortality in children under 5 years. 1 In 2017, pneumonia and malaria accounted for almost one-third of deaths in children under 5 years (excluding the neonatal period). 1 The majority of pneumonia and malaria cases and deaths occur in low-and middle-income countries, with particularly high burden in sub-Saharan Africa. 1,2 In 2017, the WHO Africa region recorded 426 000 pneumonia deaths and 257 000 malaria deaths among children under 5 years of age. 1 This accounted for 53% of global pneumonia deaths, and 98% of global malaria deaths, for children under 5 years of age during that year. 1 World Health Organisation (WHO) guidelines recommend a case management approach to the diagnosis and management of pneumonia and malaria. 3,4 However, the clinical presentation of pneumonia and malaria can be similar, involving fever, fast breathing, and other signs of severe illness such as altered conscious state. 4 Previous reports have highlighted the risks of under and overdiagnosis of pneumonia and malaria, recommending that, in malaria-endemic regions, children with a positive malaria test and respiratory signs who require hospitalization should be treated for both pneumonia and malaria. 5,6 However, we have few data on the clinical overlap of pneumonia and malaria syndromes, or current diagnostic practices, among hospitalized children in high-burden countries such as Nigeria. [5][6][7] Nigeria is a populous lower-middle-income country in west Africa with high child and neonatal mortality (2017: under-five mortality 100, neonatal mortality 33, per 1000 live births). 8 Pneumonia and malaria are the major causes of child deaths, accounting for 18% and 14% of deaths, respectively. 8 Recent estimates show that Nigeria contributed one-quarter of under-five malaria deaths and one-sixth of under-five pneumonia deaths globally in 2017. 1 The aim of this study was to describe the clinical presentation of children diagnosed with pneumonia and/or malaria at secondarylevel hospitals in Nigeria and to explore clinical decision making at the point of admission. Our exploratory analysis identified potential missed cases of severe pneumonia and severe malaria, explored predictors of "correct" pneumonia and malaria admission diagnoses, and considered reasons for clinical overlap and alternative diagnoses in malaria and pneumonia syndromes.

| Design
We conducted a prospective cohort study nested within a steppedwedge trial evaluating pulse oximetry and oxygen systems in 12 Nigerian hospitals. 9 The study design has been described previously. 9 In brief, it involved (a) baseline needs assessment and retrospective clinical review, (b) introduction of pulse oximetry to all participating hospitals (November 2015), and (c) stepped introduction of a comprehensive oxygen system, with hospitals randomized to receive the intervention between March 2016 and March 2017. 9 We have previously reported results from our needs assessment, 10,11 process evaluation, 12,13 and impact evaluation. 14 This paper reports an exploratory secondary analysis using prospective data from November 2015 to October 2017 extracted from clinical records.

| Participants
We conducted this study in southwest Nigeria, recruiting 12 small-to medium-sized hospitals that were representative of secondary healthcare facilities that admit children. 9 We included government (n = 7) and mission (n = 5) hospitals of varying capacity (Table 1, more details in Appendix 1). All hospitals were low-altitude (elevation 50-500 m above sea level) and in malaria-endemic regions. Pediatric care was provided by junior doctors and generalist nurses under the supervision of family medicine physicians and/or pediatricians (Table 1). Patients requiring additional medical services were referred to larger secondary-or tertiary-level hospitals.
We included all children aged 28 days to 15 years who were admitted to participating hospitals during the study period (Table 1).

| Materials and procedures
Trained research nurses used standardized data collection forms to gather data from clinical records immediately following patient discharge. Documented data included participant demographics, signs and symptoms on admission, admission diagnoses, and clinical care practices. Admission documentation was typically performed by junior doctors (medical admission note) and generalist nurses (nursing observations and nursing admission note). Admitting doctors were typically junior doctors who did not have pediatric specialty training, who performed their pediatric clinical duties while also covering other wards, and who would complete admission documentation at the point of admission (outpatient or emergency department) or on the pediatric ward. Admitting nurses were typically generalist nurses or nursing assistants who were based on the pediatric ward and who would complete admission documentation when the child arrived on the ward.
As part of our broader pulse oximetry and oxygen intervention, 9 we trained doctors and nurses on pulse oximetry and oxygen therapy, in accordance with WHO guidelines. 15 We did not provide doctors or nurses with additional training or standardization of diagnostic, treatment, or documentation practices. We did not provide any pneumonia-or malaria-specific training or guidance.

| Data analysis
Trained data entry clerks used EpiData 3.1 16 to double-enter data from data collection forms following standard data management procedures. We performed data cleaning and analysis using

| S39
We identified children diagnosed with pneumonia or malaria on admission, permitting multiple diagnoses. We accepted all diagnostic terms relating to lower respiratory tract infection for pneumonia (including bronchiolitis) and similarly for malaria. We categorized participant diagnoses as "malaria," "malaria + pneumonia," "pneumo- we were able to explore real-life diagnostic and management practices in the critical first hours of a severely ill child presenting to hospital. We categorized those meeting WHO criteria for severe pneumonia as "detected" or "potential missed" depending on whether they had a corresponding admission diagnosis of pneumonia. We categorized those meeting WHO criteria for malaria similarly.
To understand diagnostic decision making, we compared "detected" vs "potential missed" severe pneumonia subgroups. We built univariate logistic regression models to look for associations between presenting demographic characteristics, clinical signs, investigations, treatment, and clinical outcomes, with a "detected" pneumonia and malaria diagnosis. We then used a mixed effects logistic regression model to evaluate the relative contribution of demographic characteristics, clinical signs, and investigations to pneumonia and malaria diagnostic status ("detected" vs "potential missed"). We selected variables for inclusion in the multivariate model using backward and forward stepwise regression. We used random effects to adjust for clustering at the hospital level.

| Characteristics of children diagnosed with pneumonia and malaria
Children diagnosed with pneumonia were younger than those with malaria or other diagnoses (median age 9.8 vs 26 and 18 months,

Severe pneumonia
Cough or Difficult breathing AND at least one of the following: • severe respiratory distress; • hypoxemia (SpO 2 <90%); • unable to breastfeed or drink adequately; • decreased conscious state; • seizures.

| Severe pneumonia
Of all admitted children, 1345 (8.3%) met the WHO case definition for severe pneumonia ( Figure 1, Table 2). Of these, 557 (41.4%) did not have pneumonia as an admission diagnosis, representing potential missed severe pneumonia diagnoses.
Compared with those with potential "missed" severe pneumonia diagnoses, those with "detected" severe pneumonia were younger   (Figure 1).

| DISCUSSION
Pneumonia and malaria continue to be leading causes of mortality in children in sub-Saharan Africa. Using routinely collected clinical data, this paper described children hospitalized with malaria and pneumonia syndromes in a malaria-endemic lowaltitude region of southwest Nigeria. We identified "potential missed" severe pneumonia and malaria cases, and explored associations between "detected" diagnosis and participant demographic, clinical signs, and investigations, to give insight into current diagnostic practices.
a P values derived from univariate logistic regression. b Adjusted odds ratio derived from mixed effects logistic regression model that included random effects to adjust for clustering at the hospital level, and used stepwise regression to select variables that were independently associated with 'detected' diagnosis at alpha 0.05 level (excluding treatment and outcome variables). c CXR denominator restricted to those with results available.
GRAHAM ET AL.

| S45
T A B L E 4 Associations with "detected" severe malaria diagnosis and presenting signs, treatments, and outcome while smaller studies have reported variable incidence between 14% and 60%. 31,33 In malaria, compensatory hyperventilation related to metabolic acidosis is the most common driver of respiratory signs-causing tachypnoea, chest indrawing, nasal flaring, and deep breathing-and is probably the primary mechanism for excess mortality. 30 Severe anemia may similarly result in increased respiratory effort, particularly if associated with cardiac failure. While anemia-associated cardiac failure is anecdotally common, limited data from one Kenyan study reported mild to moderate cardiomegaly in 11% (7/64) of children with malaria and severe anemia but no evidence of pulmonary edema or congestion. 5 Similar to previous studies, we found very high incidence of severe anemia among children with severe malaria (732/2261, 32%). 5, 28 We did not collect data on cardiac failure.
Adult studies suggest that acute respiratory distress syndrome (ARDS) is relatively common in severe malaria, noting that respiratory symptoms may develop following treatment when malaria parasite blood levels have dropped. 34,35 This suggests that parasitic infiltration into the lungs may not be the only cause of respiratory signs, and that posttreatment inflammatory effects (perhaps immune-mediated) may be major contributors. 34,35 However, ARDS is less common in children with severe malaria and it is unclear how much parasitic sequestration in child lungs contributes to respiratory signs. 35 4.4 | How can frontline healthcare workers disentangle severe pneumonia and malaria?
Previous studies in similar contexts have failed to find clinical or laboratory signs that reliably distinguish between severe malaria and pneumonia. 5,28 In Mozambique, younger age, dehydration, lower PCV counts, and hyperbilirubinaemia were predictive of severe malaria, while preadmission antibiotic treatment, crackles, unilateral chest signs, nasal flaring, hypoxemia, poorer nutritional status, and higher leukocyte counts were predictive of severe pneumonia. 28 In Kenya, severe indrawing, unilateral chest signs, and hypoxemia were found to be specific predictors of radiographic pulmonary consolidation, but only crackles had reasonable specificity and sensitivity. 5 Authors of the papers from Mozambique and Kenya emphasized the inability of any combination of signs to reliably distinguish between severe pneumonia and malaria, and recommended that all hospitalized children meeting criteria for severe pneumonia and malaria be treated for both-especially in the absence of chest X-ray radiography facilities. 5,28 Our findings suggest that doctors were using signs of severe respiratory distress and hypoxemia to diagnose pneumonia appropriately. However, doctors may have been overly reliant on these insensitive respiratory signs and unduly influenced away from a pneumonia diagnosis when other signs were present that could still be consistent with severe pneumonia (eg, diarrhea, altered conscious state, feeding difficulties)-or when bedside testing suggested an additional diagnosis (eg, malaria, anemia). While the proportion of "missed" severe malaria diagnoses was relatively low, our results suggest that doctors appropriately used severe anemia as a predictor of malaria diagnosis but may have been unduly swayed away from a malaria diagnosis by other signs that could still be consistent with severe malaria (eg, diarrhea, severe dehydration).

| Limitations
This study involved post hoc exploratory data analysis and is therefore primarily hypothesis generating. We relied on documented clinical information and had minimal missing data, thanks to the close attention of our research nurses at each site. A pediatrician assessed the level of detail in case note documentation before commencing data collection, but we did not independently verify the documented clinical signs or diagnostic results. Where clinical signs were not documented, we assumed that they were not present; however, this may have resulted in an underestimate of signs. We focussed comparative analysis on those meeting WHO definitions of severe pneumonia and malaria to exclude those with mild respiratory disease or incidental parasitaemia. We were limited by the lack of independent confirmatory testings, such as routine chest X-ray radiography or blood cultures, which could have more accurately defined severe pneumonia and invasive bacterial disease.
While we did not exclude children with HIV from analysis, this study was conducted in an area with low HIV-prevalence and results may not be generalizable to contexts with higher HIV prevalence.
Southwest Nigeria is relatively wealthier and better served with health services and staff than other regions of Nigeria (and sub-Saharan Africa more broadly), and participating hospitals had received particular assistance with pulse oximetry and oxygen therapy.
Diagnostic practices may be different in other areas of Nigeria and Africa that have lower socioeconomic indices, fewer healthcare workers, and less available hospital services. Routine use of pulse oximetry by nurses in our study is relatively unusual for hospitals in sub-Saharan Africa, 27 and may have influenced diagnostic treatment practices. 14