Ambulatory and hospitalized childhood pneumonia: a longitudinal study in a peri-urban low-income community with high vaccination coverage in Sub-Saharan Africa

Abstract

Background Child pneumonia is a substantial cause of childhood mortality and morbidity; it is the largest single cause of under-5 mortality outside the neonatal period. Incidence of child pneumonia, and pneumonia mortality, have decreased substantially due to improved socio-economic attainment, improved HIV programs, coverage of new conjugate vaccines against Streptococcus pneumoniae (PCV) and Haemophilus influenzae type B (Hib), access to early antibiotic therapy, and changing prevalence of pneumonia risk factors. Measurement of community-based pneumonia incidence is difficult; risk factors for pneumonia incidence and factors associated with pneumonia mortality are poorly described in low- and middle-income countries. Careful measurement of pneumonia incidence, and prospective analysis of risk factors is necessary to appreciate the evolving complexities of pneumonia causality and mortality. The aim of this work was to describe the incidence and severity of pneumonia in a birth cohort of children in the first 2 years life; and identify risk factors for pneumonia and for severe outcomes. Methods A prospective birth cohort, the Drakenstein Child Health Study, enrolled mother-infant pairs in two communities outside Cape Town, South Africa. Pregnant women were recruited and followed through pregnancy, labour and delivery, and the first 2 years of the child's life. Comprehensive data collection of risk factors was done through the first 2 years of life. A community pneumonia surveillance system was established; active case finding was used for birth cohort participants over 4 respiratory seasons. Children were examined at scheduled visits and at the time of pneumonia events. Pneumonia or severe pneumonia was diagnosed according to revised World Health Organisation (WHO) guidelines. Chest x-rays were classified according to WHO guidelines. Predictors of ambulatory and hospitalized pneumonia were explored with Poisson regression using generalized estimating equations clustered on mother-infant pairs. Factors associated with death or admission to intensive care unit were analysed with prevalence ratios from modified Poisson regression with robust variance estimation. Findings From March 2012 to March 2015, 1137 pregnant women were enrolled, delivering 1143 live-born infants. Household environmental tobacco smoke exposure was common: 82% of children were exposed in the first 6 months of life. Maternal HIV infection was common: 249 (22%) of 1143 children were HIV-exposed, but only 2 children became HIV-infected. Coverage of primary series of hexavalent vaccine, PCV and Hib was excellent (92%). During the study period (2012 to 2017), there were 795 pneumonia episodes (621 (78%) ambulatory, 274 (22%) hospitalised) in the first 2 years of life. Pneumonia incidence was higher in the first year of life (0.51 episodes per child year (e/cy)) and decreased to 0.25 e/cy in the second year. Active case finding in the birth cohort was more accurate than passive surveillance performed at the community clinics; pneumonia incidence measured by passive surveillance was significantly lower (incidence rate ratio 0.72, 95% CI 0.58 – 0.89) compared to active surveillance. Pneumonia mortality was low: 1.7% of hospitalised cases, and 0.35% of all clinical cases. There was marked variability in pneumonia incidence from year to year during the study. Many risk factors for pneumonia did not have fixed effects, but had different impacts at different ages, and variable effect on ambulatory and hospitalised pneumonia. In multivariable regression, adjusted incidence rate ratios were calculated for 5 risk factors (age< 6 months, male sex, low birth weight (<2500g), maternal smoking, delayed vaccines), which were associated with consistent effects on ambulatory and hospitalised pneumonia. Risk factors for serious outcomes of pneumonia (death or admission to intensive care unit) were identified: age under 2 months, low birth weight and hypoxia. Conclusion In this birth cohort, with low socio-economic status but high vaccination coverage, we demonstrated higher-than expected incidence of pneumonia, but very low mortality, with specific risk factors identified. Active surveillance was important for accurate detection of pneumonia. Children born at low birth weight are at increased risk for pneumonia and for serious outcomes. Pulse oximetry to detect hypoxia, and access to oxygen for children with hypoxic pneumonia, should be included in guidelines. These data will have global applicability for estimation of child pneumonia incidence in regions where direct measurement is impossible. These data can be applied to epidemiology and disease-modelling for child health; they will contribute to long-term morbidity follow-up studies; and they will contribute to understanding the constantly-evolving epidemiology of child pneumonia.