Introduction

Globally, pneumonia is the leading cause of death among under-five children with more than 90% of these occurring in resource-limited settings [1]. Implementation of feasible and effective interventions has reduced under-five pneumonia death substantially from 13·6 per 1000 livebirths in 2000 to 6·6 per 1000 livebirths in 2015 [2, 3], yet, pneumonia ranked top in mortality. Childhood death due to pneumonia occurs disproportionately in low-and middle-income countries (LMICs) with the greatest number observed in South Asia and sub-Saharan African countries [3, 4]. Compared to deaths due to other childhood diseases, pneumonia related deaths are declining at a slower rate [5]. Rudan et al. reported the incidence of clinical pneumonia in under-five children was approximately 0.29 episodes per child-year in LMICs of the world [6]. This means, every year 151.8 million incidences with 13.1 million (8.7%) episodes progressing from pneumonia to severe pneumonia that require hospitalization [6, 7].

According to World Health Organization (WHO), United Nations Children’s Fund (UNICEF), and others, risk factors for severe childhood pneumonia in LMICs include bacterial etiology, young age, low birth weight, malnutrition, household crowding, exposure to indoor air pollution, and low-level schooling of mothers [8–10]. Factors associated with severe pneumonia in different contexts and their changes over time in LMICs are incompletely defined [6, 11, 12]. In Bangladesh, severe pneumonia is the main cause of hospitalization among under-five children. Among 156,847 admitted children identified in a surveillance performed in seven hospitals in Bangladesh from May 2004 through April 2007, the most frequent diagnosis was pneumonia (32%) [10].

We analysed a large dataset of children with severe pneumonia/pneumonia from an urban facility-based study in Bangladesh to explore factors associated with severe pneumonia compared to pneumonia. We included clinical, socio-demographic characteristics, and care-seeking behaviour of the study population in analyses to identify the associated factors of severe pneumonia among under-five children.

Materials and methods

Results

The study comprised of 2597 children aged 2–59 months for the analysis. Of these, 1693 were cases and 904 were controls (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Enrolment profile of the study participants.

https://doi.org/10.1371/journal.pone.0265871.g001

The median age of study children was 9.2 months (IQR, 5.1–17.1), 1576 (60%) were male, and 1534 (59%) were in the 2-11-month age group. Two thirds of study children had more than two siblings. Cases often had higher fever (41% vs 20%), and higher pulse rate (17% vs 16%) compared to controls. Hypoxemia was only observed in cases (11% of cases). Regarding socio-demographic, care seeking and environmental variables; we also observed differences in child’s sex, household income, received treatment prior to enrolment, child’s mode of delivery during birth, treatment method of drinking water, toilet facility, housing (wall, floor), cooking source (gas, wood). Care seeking behaviours were observed nearly two times more frequent in cases than in their control counterparts (73% vs. 43%). Similarly, monthly household income was higher in cases compared to controls. Further, we found minimal differences between the groups in immunization status, nutritional status, household size, father’s education, or presence of diarrhoea (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Comparison of characteristics of the study population with severe pneumonia (cases) and pneumonia (controls) according to the WHO classification¹.

https://doi.org/10.1371/journal.pone.0265871.t001

In unadjusted analysis, factors significantly associated with severe pneumonia included longer duration of illness, temperature ≥ 38°C, increased pulse rate/min, male sex, 2–11 months old age group, wealth index, received treatment prior to enrolment, child’s mode of delivery during birth, housing (wall, floor), cooking source (gas, wood), drinking water treatment method, and toilet facility (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Factors associated with severe pneumonia compared to pneumonia among under-five children.

https://doi.org/10.1371/journal.pone.0265871.t002

Factors significantly associated with cases after adjusting for duration of illness, temperature, pulse rate, child’s age, child’s sex, stunting (nutritional status), received prior medical care, wealth index, mother’s education, cooking fuel source (wood, gas) in logistic regression analysis included: duration of illness ≥3days, temperature ≥38°C, severe stunting, male sex, wealth index, received prior medical care. Most children from the wealthiest quintile (60%) received care from private or public facilities whereas those from the poorest quintile (70%) sought care from pharmacies/drug stores and other facilities (data not presented). Cases sought care more often from appropriate facilities compared to the control group. In the final multivariable model, Hosmer-Lemeshow goodness-of-fit test showed it to be non-significant (p- value = 0.26) indicating the model fitted well. The VIF values of all independent variables were 2.81 or less and the mean VIF was 1.79. The value under the ROC curve was 0.7483. We have also tested for interaction. No effect modification was observed.

Discussion

Currently, in-depth understanding of childhood pneumonia based on research is greatest in context of high-income countries compared to LMICs. Definition of epidemiology of the illness, causal pathogens, and key prognostic factors is more robust in such settings, many of which are different in LMICs. According to available literature, knowledge of factors and especially clinical risk factors associated with severe paediatric pneumonia at the country as well community level in LMICs is limited [19].

Male children were predominant in the severe pneumonia group (63%) in our study. Yet in the pneumonia group their proportion reached 56% as well; overall male children thus were more affected significantly by severe pneumonia corroborating findings of other studies [7, 8, 20]. The cause behind the high susceptibility of male children could be either genetic, or higher reporting for male children by the mothers due to gender bias, which potentially causes mothers to notice symptoms due to a higher attention to male children particularly for seeking health care much earlier than female children [21]. Nevertheless, boys have greater likelihood of being affected or of care seeking in general for common acute respiratory illness than girls, as reported in several studies from Bangladesh [22–24]. Another possibility of male children to be in the high risk of infection could be the testosterone suppressing the immune response [23]. However, clarification of this trend is multifaceted, as the role of social determinants of health, such as sex, socio-economic status (SES), and water, sanitation, and hygiene (WASH) practices, with disease are often not included in studies.

For safety reasons children with severe acute malnutrition were excluded from the main effectiveness trial although children with moderate malnutrition were included as mentioned earlier. One systematic review suggested that like the severely malnourished children, children with moderate degrees of malnutrition may also be at increased the risk of death due to pneumonia [25, 26]. Studies that evaluated the impact of moderate malnutrition are comparatively few as moderate degree of malnutrition in health facilities in developing countries is not recorded as an admission diagnosis. Notably, we found severe stunting to be associated with severe pneumonia. Stunting is established to have long-term sequels on lung development and growth and to be associated with prolonged acute course of pneumonia treatment and delay in recovery [27]. Malnutrition in children results in an immunocompromised state and subsequent increase in infectious morbidity and mortality due to impairments in multiple aspects of the immune system including cell mediated and complement responses, inefficient chemotaxis, reduced mature T cells, compromised phagocytic activity, among others. The results of our study highlight the importance of stunting in children being treated for severe pneumonia [28, 29].

Hypoxaemia, a major indicator of disease severity, was observed in almost 11% children with severe pneumonia by pulse oximeter, consistent with findings from a systematic review of childhood pneumonia in LMICs [7]. The need for application of pulse oximeter and related staff training to accurately identify and monitor children with hypoxaemia is therefore self-evident [7, 30].

On the World Pneumonia Day on Nov 12, 2015, action was sought to improve the early identification and treatment of childhood pneumonia at community and outpatient level to reduce disease severity and deadly outcomes [31]. It was apparent that the case–mortality rate in untreated children with pneumonia is high, sometimes reaching as high as 20%, and deaths can occur as early as 3 days after illness onset [31]. We also found that duration of illness at home for 3 days or more was significantly associated with the likelihood of disease progression to severe pneumonia. The same observation was reported from Kenya [32]. In our study fever was significantly associated with severe pneumonia. Studies in diverse LMICs like South Africa, Papua New Guinea, and Indonesia reported no association between fever and pneumonia severity [19, 33–35]. However, one study in USA indicated temperature to be associated with severe pneumonia (not defined by WHO classification) [36] and another study reported duration of fever (at day 6) was associated with severity [37]. Although WHO did not consider fever in their pneumonia severity criteria, the British Thoracic Society (BTS) includes fever in their guideline [13, 38].

This study also depicted a noteworthy association of care-seeking behaviour with socio-economic status variables which may be new information. Although, few reports are available on severe pneumonia and those two variables according to most literature on children with pneumonia. We found care seeking behaviour proportion to be higher in much wealthier cases compared to controls. Overall 62.3% households looked for care from different health care facilities prior to enrolment in our study. Parents from the highest wealth index quintile were observed to seek care from qualified providers (in private or public facilities). These observations have also been confirmed by earlier studies [39, 40]. Not surprisingly, increased awareness and ability to meet the expense of care play a remarkable role in decision making, as people from higher wealth quintiles are often more aware and generally able to afford health care cost [40]. In contrast to the wealthiest group, the poorest groups sought health care from the pharmacies and other nearby facilities including the traditional healers. This pattern of behaviour has been reported in other recent studies [40, 41]. However, severe outcome of pneumonia and increased mortality rates have consistently been associated with low compared with high SES [42]. The Integrated Management of Childhood Illness (IMCI) guidelines [43] suggested awareness raising interventions to improve family and community practices through the education of mothers, fathers, other child caretakers, and members of the community, with a focus on timely care seeking from appropriate facility, compliance, initial care at home, and overall health promotion. This finding is difficult to explain may be more resources and information are available in the richest quintile to access nearby facilities. They could be much aware and often presented to the appropriate facility in urban area.

Exclusive breastfeeding, parent’s education, immunization, household size, number of siblings, housing type, and cooking fuel (gas) were not significantly associated with severe pneumonia in our study. Nevertheless, several studies showed some of these factors being significantly associated with pneumonia/severe pneumonia [8, 19, 44]. A possible explanation could be our participant’s baseline admission characteristics were identical in both the groups and thus validate the study results. The majority of the urban households used gas stove for cooking purposes and have same type of housing. A comparison of urban with rural data might possibly could reveal variation in housing type, cooking fuel use, etc. Pentavalent vaccine and PCV coverage were 90% in our patient population, known interventions that reduce severe bacterial pneumonia [45]. In Bangladesh we have higher vaccine coverage (>90%), thus we are in the right track in terms of our “Expanded Programme of Immunization” coverage [45, 46]. No significant association was found between diarrhoea as comorbidity and the risk of severe pneumonia. However, in our study population, only a small proportion of patients presented with diarrhoea as co-morbidity.

Conclusion

This study analysed a large number of clinical pneumonia episodes among under-five children in Bangladesh. We found male children, longer duration of illness, fever, received prior medical care, and severe stunting as significantly associated factors for severe pneumonia compared to pneumonia. The results of this study have the potential to help refine decisions about case management in resource-limited countries by facilitating decisions about the most appropriate site of treatment (i.e. home vs. hospital) or the need for additional supportive care. There is a pressing need for further research in larger populations to further define significant risk factors of severe pneumonia from the community and address key knowledge gaps in order to enable optimal management strategies with the potential for substantial reductions in morbidity and mortality.

Acknowledgments

We would like to thank study participants, data collectors, and supervisors for their unreserved contribution during data collection. We gratefully acknowledge our core donors for their unrestricted support and commitment to icddr,b’s research efforts include the Governments of Bangladesh, Canada, Sweden and the UK.