Bulletin of the World Health Organization

Epidemiology and etiology of childhood pneumonia

Igor Rudan, Cynthia Boschi-Pinto, Zrinka Biloglav, Kim Mulholland, Harry Campbell

Volume 86, Number 5, May 2008, 408-416

.

Step Methods and models
Step 1 The Child Health Epidemiology Reference Group (CHERG) working group on pneumonia did an extensive review of the research on childhood pneumonia that was subsequently synthesized in a database including more than 2200 sources of information. Further details on search strategies, inclusion criteria and methods are published elsewhere.7–9
Step 2 A review of the database with 2200 CHERG studies identified risk factors for pneumonia at the community level. Only studies that investigated the role of several risk factors at the same time at the community level using a multivariate design and that included more than 500 children were initially used to establish definite, likely and possible risk factors. This step was needed to avoid confounding issues and publication bias typical of studies that use univariate design, study a single risk factor or are simply based on a too-small sample and lack power.
Step 3 Four studies were identified10–13 and several other methodologically sound studies were used as supporting evidence.14–21On the basis of those studies, we established three categories of risk factors for development of childhood clinical pneumonia in the community. The risk factors were then defined as:(i) definite (the large majority of evidence consistently pointing to the role of the risk factor):malnutrition (weight-for-age z-score < –2)low birth weight (≤ 2500 g)non-exclusive breastfeeding (during the first 4 months of life)lack of measles immunization within the first 12 months of lifeindoor air pollutioncrowding(ii) likely (most evidence consistently pointing to the role, but with some opposing findings; or scarce but consistent evidence of the role):parental smokingzinc deficiencymother’s experience as a carerconcomitant diseases (e.g. diarrhoea, heart disease, asthma)(iii) possible (with sporadic and inconsistent reports of the role in some contexts):mother’s educationday-care attendancerainfall (humidity)high altitude (cold air)vitamin-A deficiencybirth orderoutdoor air pollution.
Step 4 We decided to use only the prevalence of exposure to definite risk factors for distributing global number of pneumonia cases by individual countries. However, we decided to exclude measles immunization coverage on two grounds: because the coverage has approached high levels in recent years (while the studies that identified it as an important risk factor were done mainly in the 1980s and 1990s), so this factor is less discriminative than in was several years ago; and because there is no theoretical justification for including it apart from historically serving as a proxy for health system functioning (but this cannot justify its inclusion).
Step 5 After defining five risk factors that will be used to distribute all cases of childhood pneumonia that occur in 1 year globally, this total number was computed. It was derived by:(i) multiplying the number of all children aged less than 5 years living in developing countries (this includes WHO regions AFR D, AFR E, AMR B, AMR D, EMR B, EMR D, SEAR B, SEAR D and WPR B)a in the year 2003 with incidence of 0.28 episodes per child-year, as estimated by Rudan et al.;8(ii) multiplying the number of all children aged less than 5 years living in the most developed regions of the world (this includes WHO A regions AMR A, EUR A and WPR A) in the year 2003 with incidence of 0.03 episodes per child-year, as estimated by Rudan et al.8; the children in EUR B by 0.07 and in EUR C by 0.09;8 and(iii) adding up all the cases predicted from the first two calculations.National under-5 population information was obtained from the United Nations Population Division.22
Step 6 Three parameters were then used to distribute the global number of episodes into regional and national estimates:(i) national under-5 population;22(ii) prevalence of five of the definite pneumonia risk factors (underweight, low birth weight, non-exclusive breastfeeding during the first 4 months of life, indoor air pollution, and crowding); and(iii) estimates of relative risks for each of these five risk factors.
Step 7 Data on the prevalence of children underweight (weight-for-age z-score < –2), low birth weight (≤ 2500 g), and non-exclusive breastfeeding (during the first 4 months of life) were obtained from the Demographic and Health Surveys (DHS) or from the Multiple Indicators Cluster Surveys (MICS).23–25 Both DHS and MICS are nationally representative household surveys with large sample sizes, generally carried out every 3–5 years. Together they cover most developing countries and provide data on demographic and health indicators.Data for the prevalence of indoor air pollution were collected from WHO’s document: Indoor air pollution (national burden of disease estimates)26 as ”percentage of population using solid fuels”.Data on crowding prevalence (defined as ≥ 5 people per household) were obtained from national official governmental information retrieved country-by-country from the internet. For the countries where information could not be obtained, national data on the prevalence of exposure to specific risk factors were replaced with the mean value calculated for that particular region.We aimed to collect the information on the prevalence of exposure to these five risk factors for the year 2001–2003.The table below summarizes the availability of data on the prevalence of risk factors used in this analysis for the year 2003 and for developing countries, where more than 95% of pneumonia episodes occur.
Risk factors No. countries with available population-based data on prevalence of exposure Prevalence ofexposure(%) Estimated number of children aged 0–4 years exposed(million)
Malnutrition (weight-for-age z-score < –2) 98 27 141
Low birth weight (≤ 2500 g) 107 16 83
Non-exclusive breastfeeding (for first 4 months) 98 64 337
Indoor air pollution 117 66 342
Crowding (≥ 5 people per household) 52 53 278
Step 8 Estimates of relative risks for each of the five definite risk factors for malnutrition, low birth weight, non-exclusive breastfeeding and crowding were obtained from available studies10–21 and from Dherani et al. for indoor air pollution (personal communication; review is a part of this theme issue). Relative risk was set to the median of relative risks (or odds ratios) reported in these studies. We ensured that the definitions of risk factors were the same in the studies estimating relative risk as well as in the surveys that measured prevalence of exposure to these risk factors. Relative risks (RR) for the five definite risk factors were applied as follows:- malnutrition (weight-for-age z-score < –2), RR = 1.8- low birth weight (≤ 2500 g), RR = 1.4- non-exclusive breastfeeding (during the first 4 months of life), RR = 1.9- indoor air pollution, RR = 1.8- crowding, RR = 1.4
Step 9 The global number of new episodes of clinical pneumonia was calculated for each developing country with a model based on the epidemiological concept of potential impact fraction27 as follows:Ne/cy = (Pop < 5yrs) x (Inc.DevW) x {1 + Σ (RF = 1→n) [(PrevRFn - PrevRFnDevW) x (RRRFn - 1)]},where Ne/cy is the number of new clinical pneumonia episodes per year in each developing country, Pop < 5yrs is the population of children less than 5 years in each developing country, Inc.DevW is the estimated incidence of clinical pneumonia in the developing world, PrevRFn is the prevalence of exposure to n-th risk factor among under-fives in the developing country of interest, PrevRFnDevW is the prevalence of exposure to n-th risk factor among under-fives in all developing countries and RRRFn is the relative risk for developing clinical pneumonia associated with the n-th risk factor.
Step 10 Cautionary notes on limitations of this approach:(i) In our calculations, we used the child population estimates for the year 2000 and the prevalence of exposures to risk factors relevant to the years 2001–2003; however, the global childhood pneumonia incidence estimate is based mostly on studies conducted in the 1980s and 1990s, and so are relative risks associated with different risk factors.(ii) Prevalence of malnutrition, low birth weight and lack of exclusive breastfeeding mostly comes for MICS and DHS data that were made available in 2003–2004, but relevant to the years 2000–2001; indoor air pollution information comes from the World Bank’s source and refers to 2002–2003, while the search of the information for crowding was also done during 2002; we decided that it is most appropriate to present national-level estimates for the year 2000, as these then ensure consistency and complement the papers on global incidence of childhood pneumonia8 and global mortality from childhood pneumonia.7(iii) Our model, described in step nine, does not necessarily assume that the five risk factors are independent, because we applied relative risks derived primarily from the studies of multivariate design; however, it does assume that the magnitude of the five chosen risk factors is constant over the whole range of countries, which may not be the case in different environments with different combinations of risk factor exposures.

AFR, African Region; AMR, Americas Region; EMR, Eastern Mediterranean Region; EUR, European Region; SEAR, South-East Asia Region; WPR, Western Pacific Region.a WHO regions are subdivided based on child and adult mortality strata: A, very low child and very low adult mortality; B, low child and low adult mortality; C, low child and high adult mortality; D, high child and high adult mortality; E, high child and very high adult mortality.