Article Text
PDF
Abstract
Introduction The burden of acute lower respiratory infections (ALRI), and common viral ALRI aetiologies among 5–19 years are less well understood. We conducted a systematic review to estimate global burden of all-cause and virus-specific ALRI in 5–19 years.
Methods We searched eight databases and Google for studies published between 1995 and 2019 and reporting data on burden of all-cause ALRI or ALRI associated with influenza virus, respiratory syncytial virus, human metapneumovirus and human parainfluenza virus. We assessed risk of bias using a modified Newcastle-Ottawa Scale. We developed an analytical framework to report burden by age, country and region when there were sufficient data (all-cause and influenza-associated ALRI hospital admissions). We estimated all-cause ALRI in-hospital deaths and hospital admissions for ALRI associated with respiratory syncytial virus, human metapneumovirus and human parainfluenza virus by region.
Results Globally, an estimated 5.5 million (UR 4.0–7.8) all-cause ALRI hospital admissions occurred annually between 1995 and 2019 in 5–19 year olds, causing 87 900 (UR 40 300–180 600) in-hospital deaths annually. Influenza virus and respiratory syncytial virus were associated with 1 078 600 (UR 4 56 500–2 650 200) and 231 800 (UR 142 700–3 73 200) ALRI hospital admissions in 5–19 years. Human metapneumovirus and human parainfluenza virus were associated with 105 500 (UR 57 200–181 700) and 124 800 (UR 67 300–228 500) ALRI hospital admissions in 5–14 years. About 55% of all-cause ALRI hospital admissions and 63% of influenza-associated ALRI hospital admissions occurred in those 5–9 years globally. All-cause and influenza-associated ALRI hospital admission rates were highest in upper-middle income countries, Asia-Pacific region and the Latin America and Caribbean region.
Conclusion Incidence and mortality data for all-cause and virus-specific ALRI in 5–19 year olds are scarce. The lack of data in low-income countries and Eastern Europe and Central Asia, South Asia, and West and Central Africa warrants efforts to improve the development and access to healthcare services, diagnostic capacity, and data reporting.
- pneumonia
- epidemiology
- paediatrics
- systematic review
- respiratory infections
Data availability statement
Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as supplementary information. Details of studies included in this review and references of these studies are uploaded as supplementary information. More details will be uploaded to Edinburgh DataShare (https://datashare.ed.ac.uk/) after publication.
This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.
Statistics from Altmetric.com
Key questions
What is already known?
Acute lower respiratory infections (ALRI) caused an estimated 55 800 deaths, accounting for 5% of all-cause deaths in children aged 5–19 years globally in 2019.
No global estimates are available to quantify hospital admissions due to ALRI in this age group, and only the Institute for Health Metrics and Evaluation estimated global influenza-attributable hospital admissions (about 2.0 million).
What are the new findings?
Based on data from 202 studies, we estimated that globally about 5.5 million all-cause ALRI hospital admissions occurred annually between 1995 and 2019 in the 5–19 years age group and 87 900 in-hospital deaths; 19% (1.1/5.5 million) of ALRI hospital admissions were associated with influenza virus and 4% (231 800) for respiratory syncytial virus in the 5–19 years age group; human metapneumovirus (105,500) and human parainfluenza virus (124,800) was associated with 2% of ALRI hospital admissions in the 5–14 years age group separately.
55% of ALRI hospital admissions and 63% of the influenza-associated ALRI hospital admissions occurred in children aged 5–9 years.
Estimates of all-cause ALRI hospital admissions were developed in 39 countries and estimates of influenza-associated ALRI hospital admissions were developed in 26 countries, accounting for 57% and 52% of the global population 5–19 years of age.
Key questions
What do the new findings imply?
Incidence and mortality data for all-cause and virus associated ALRI in 5–19 year olds are scarce; the lack of data in low-income countries and Eastern Europe and Central Asia, South Asia, and West and Central Africa warrants efforts to improve the development, access to healthcare services, diagnostic capacity and data reporting.
The estimates provide evidence to support introduction of ALRI intervention programmes in the 5–19 years age group.
Introduction
Although the global acute lower respiratory infection (ALRI) mortality burden for 5–19 years reduced by about 43% between 1990 and 2019, it was estimated that ALRI caused around 56 000 deaths in children and adolescents aged 5–19 years globally in 2019, accounting for 5% of global all-cause deaths in this age group.1 Several efforts have been made to improve estimation of all-cause and virus-specific ALRI burden for children under 5 years; however, the burden in those aged 5–19 years are less well understood. In the absence of a baseline disease burden estimate, it is difficult to understand the impact of change in virulence of an existing pathogen or emergence of a novel pathogen (eg, SARS-CoV-2) and their impact on all-cause/aetiology-specific ALRI disease burden. No global estimates of all-cause ALRI hospital admissions have been made for 5–19 years, and very few global estimates quantify the burden of common ALRI viral aetiologies for this age group. Estimates have only been made for influenza attributable ALRI burden in one study by Institute for Health Metrics and Evaluation (IHME).2 This study estimated that globally in 2017, approximately 2.0 million lower respiratory infection hospital admissions attributable to influenza for 5–19 years.2 No regional or country-specific estimates were reported in the paper. These estimates were developed through a complex modelling process. Despite the importance of models in global burden estimation, data allowing for direct estimation of influenza-related hospital admissions, for example, hospital admission rate for 5–19 years, offer an opportunity to reduce model-related uncertainties and to improve the robustness of estimates.
Given the differences in access to health services between countries, we included country-level layer in the analytical framework when data were sufficient. Hospital admission and in-hospital mortality estimates at country and region levels would inform the burden on health services and provide information for intervention impact analyses. In resource-limited settings, hospital care data would inform allocation of healthcare recourses, and warrant efforts to improve access to healthcare and development of healthcare registration systems.
Methods
Systematic review and assessment of risk of bias
We conducted a systematic review of burden of all-cause ALRI and of cause-specific ALRI associated with four viruses (ie, ALRI with laboratory-confirmed influenza virus, respiratory syncytial virus, human metapneumovirus, or human parainfluenza virus) in children and adolescents aged 5–19 years (online supplemental appendix pp1-4). The four viruses are commonly associated with ALRI in children under 5 years.3 We searched Medline (Ovid), Embase (Ovid), Global Health (Ovid: 1973 onwards), Cumulative Index to Nursing and Allied Health Literature, Web of Science, Global Health Library, three Chinese language databases (CNKI, Wanfang and Chongqing VIP), and Google search (for grey literature). No language or publication restrictions were applied, and three reviewers (XW, XM and EB) screened the titles and abstracts for eligibility, and extracted data independently. Disagreements were resolved by discussion between reviewers.
Supplemental material
We included studies that were published between 1 January 1995, and 31 December 2019, and reported any of the following data for 5–19 years: hospital admission rates of all-cause and cause-specific ALRI; in-hospital case-fatality ratios (hCFRs) of all-cause ALRI; and proportions of hospitalised ALRI cases positive for the four respiratory viruses. Community-based studies with data on incidence rates of all-cause ALRI and hCFRs of virus-associated ALRI were also identified (online supplemental appendix p23; 42–44), but were not included in the quantitative analysis due to the small number of studies.
Studies using hospitalised ALRI (pneumonia and bronchiolitis) or similar definitions (eg, acute respiratory infections or chest X-ray confirmed pneumonia requiring hospital admission) were eligible (online supplemental appendix pp20-22). Studies that reported data on hospital admission rates and hCFRs of all-cause ALRI, and proportion of virus positives were required to report data for at least a 12 consecutive month period. Studies that reported data on hospital admission rates of virus-associated ALRI were required to show data for at least one full virus season if in a temperate region with defined seasons, or at least a 12 consecutive month period. There were no restrictions on the duration of studies reporting hCFRs of virus-associated ALRI. We excluded studies: where the denominator was unavailable or could not be calculated; reporting modelled burden estimates; in which viral infections were diagnosed based on serology alone; which reported only single infections or only coinfections of a virus; or which only including population subgroups with high-risk conditions. For influenza, we additionally excluded studies which reported data during the 2009 pandemic (the 2009–2010 season in temperate regions and the year 2009 in tropical regions) to estimate the burden of influenza during average seasons; and those reporting data on a non-dominant influenza subtype. This study was registered with PROSPERO, CRD42018093336.
We used a modified Newcastle-Ottawa Scale to assess the risk of bias in six domains, including study design, adjustment for healthcare utilisation, patient groups excluded, case definition, and sampling strategy and test methods for studies with virus-specific data (online supplemental appendix p45).4 5
Statistical analysis
ALRI hospital admissions
Figure 1 summarises our approach to estimation of ALRI hospital admissions and influenza virus-associated ALRI hospital admissions. We stratified analyses by three 5-year age groups (5–9 years, 10–14 years and 15–19 years) for which data were most frequently reported. Not all studies reported ALRI hospital admissions rates for the three age groups. We imputed missing data for any of the three age groups for each study using a multiple imputation approach; the input data are rate ratios between the three age groups estimated using a network meta-analysis and United Nations Population Division country population structures by single-year of age (online supplemental appendix pp5-7).5–7 We estimated average rate ratios based on studies where age-stratified data were available (nine studies for ALRI and eight studies for influenza-associated ALRI). The rate ratios were then extrapolated to other studies to estimate hospital admission rates by age groups. We assumed the hospital admission rate was similar within each 5 year age bands. Estimates were comparable after the imputation was done (online supplemental appendix p7).
Estimation of global, regional, and national all-cause acute lower respiratory infections (ALRI) hospital admissions in children and adolescents aged 5–19 years. ‘Orange’ boxes show input data, and ‘green’ boxes show the outputs. All analyses were done by three age bands (5–9, 10–14 and 15–19 years). Analyses were done at national level and regional level (World Bank income region and UNICEF region). Global estimates were calculated as sum of estimates by age and World Bank income region. Estimation of influenza-associated ALRI hospital admissions were generally similar to estimation of ALRI hospital admissions, with the hospital admission rates of influenza-associated ALRI as the input. We made one adaptation: prior to meta-analyses, influenza hospital admission rates were adjusted for testing practice in each study where available. y: year.
Meta-analyses were conducted by country and by World Bank income regions and UNICEF regions.8 9 Due to the insufficient data for South Asia, we combined this region with the region of East Asia and the Pacific (as Asia-Pacific); we combined the region of Eastern and Southern Africa and the region of West and Central Africa (as Sub-Saharan Africa) due to insufficient data in West and Central Africa. First, we pooled hospital admission rates for each country using a generalised linear mixed model (binomial-normal model: within-study binomial distribution and between-study normal distribution) when there were two or more studies.10 When there was only one study for a country, the reported hospital admission rates in the study were used to yield country-specific hospital admissions. The weighted average of available country-specific rates and the sum of weighted uncertainties in each region was extrapolated to yield the number of ALRI hospital admissions in the region.11 Global results were calculated as the sum of estimates by age group and World Bank income regions. We also reported estimates by UNICEF regions where data were available. The numbers of cases and deaths were rounded to the nearest hundred.
After meta-analyses, the number of all-cause ALRI hospital admissions was estimated using the Monte Carlo Simulation to combine the meta-estimates of ALRI hospital admission rates and UN Population Division census estimates for 2019 as done previously.5 6 The median value of 1000 samples simulated from a log–normal distribution was used as the point estimate and the 2∙5th and 97∙5th percentiles as the 95% uncertainty ranges (UR) for disease burden estimates.
All-cause ALRI in-hospital mortality
The available hCFR data for all-cause ALRI were insufficient to allow for disaggregation by 5-year age group or country. We pooled hCFR for the overall age group (5–19 years) by World Bank income (as well as UNICEF) regions, and applied the hCFR meta-estimate to the number of ALRI hospital admissions to yield all-cause ALRI in-hospital mortality in each region as previously.5
Influenza-associated ALRI hospital admissions
Estimation of influenza-associated ALRI hospital admissions is generally similar to that for all-cause ALRI hospital admissions except for one adaptation: hospital admission rates of influenza were additionally adjusted for levels of testing per study (where such data were available) before meta-analyses. We did not make this adjustment if this was already accounted for and reported by the study.
ALRI hospital admissions associated with respiratory syncytial virus, human metapneumovirus and human parainfluenza virus
For the three above-mentioned viruses, data on hospital admission rates were limited; thus, we estimated virus-associated ALRI hospital admissions using a proportion-based approach, by combining the proportion positives of a virus and the number of all-cause ALRI hospital admissions. URs were estimated using the Monte Carlo Simulation. We conducted analyses by World Bank income (and UNICEF) regions where data were available. For respiratory syncytial virus, we estimated hospital admissions for the overall age group 5–19 years; for human metapneumovirus and parainfluenza virus, we estimated the hospital admissions for 5–14 years as data were most commonly reported for this age group.
Sensitivity analyses
For all-cause ALRI hospital admissions, we excluded potential outlier studies in the sensitivity analysis. For influenza-associated ALRI hospital admissions, we excluded potential outlier studies and studies using the definition of ‘International Classification of Diseases (ICD)-coded laboratory confirmed respiratory hospital admissions’ in sensitivity analyses. Additionally, we estimated influenza-associated ALRI hospital admissions by combining the pooled estimates of proportion positives of influenza and estimates of all-cause ALRI hospital admissions in a sensitivity analysis. This study was conducted and reported in accordance with the Guidelines for Accurate and Transparent Health Estimates Reporting recommendations (online supplemental appendix p59).12
The role of funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. XW and HN had full access to all the data in the study and HN had final responsibility for the decision to submit for publication.
Patient and public involvement
Patients and the public were not involved in this study.
Results
We identified 202 studies with relevant data between 1995 and 2019 (figure 2). There were 90 studies with data for all-cause ALRI (ie, incidence rates, hospital admission rates, and hCFRs), 114 studies for influenza virus, 66 studies for respiratory syncytial virus, 37 studies for human metapneumovirus, and 42 studies for human parainfluenza virus. By World Bank income region, six (3%), 38 (19%), 62 (31%) and 96 studies (47%) were from low-income, lower middle-income, from upper middle-income, and from high-income countries, respectively. By UNICEF region, most of the studies were from high-income countries (41%) and East Asia and the Pacific (27%). Figure 2 shows the study selection for the systematic review. Quantitative analyses were not conducted for incidence rates of all-cause ALRI or hCFRs of virus-specific ALRI given the paucity of data. Details of all the included studies are in the online supplemental appendix pp19-43.
Flow diagram for study selection. ALRI, acute lower respiratory infection. hCFR, in-hospital case-fatality ratio.
There were 69 studies with data on all-cause ALRI hospital admission rates, including 7 studies reporting data for pre-2000 period, 34 studies for 2001–2010, and 28 studies for 2011–2018. Data were from 39 countries, and nationally representative data (as reported in individual studies) were found in 15 countries. Of these studies, 33 studies had a low risk of bias in study design, 64 studies had a low risk in adjustment for healthcare utilisation, and 68 studies had a low risk of bias in patient groups excluded. Thirty-nine studies used definitions similar to that we specified, 26 studies used less specific definitions, and 4 studies used more specific definitions. Across 39 countries, all-cause ALRI hospital admission rate meta-estimates ranged from 0.1 (95% CI 0.1 to 0.1) per 1000 children per year to 42.7 (95% CI 35.1 to 51.8) for 5–9 years, from 0.1 (95% CI 0.1 to 0.1) to 17.4 (95% CI 12.9 to 23.6) for 10–14 years, and from 0.1 (95% CI 0.1 to 0.1) to 18.1 (95% CI 13.1 to 25) for 15–19 years (table 1). By World Bank income regions, the estimated all-cause ALRI hospital admission rate for 5–19 years was highest in upper middle-income countries (3.9 per 1000 persons per year (95% CI 2.6 to 5.6)), followed by lower middle-income countries (2.9 (95% CI 2.1 to 4.1)), low-income countries (1.9 (95% CI 1.4 to 2.6)), and high-income countries (1.4 (95% CI 1.2 to 1.7)). By UNICEF regions, we estimated higher all-cause ALRI hospital admission rates for 5–19 years in Asia-Pacific, and Latin America and Caribbean (3.0–3.6 per 1000 persons per year), followed by Sub-Saharan Africa (2.2 (95% CI 1.9 to 2.6)), high-income countries (1.3 (95% CI 1.2 to 1.6)), and Middle East and North Africa (0.7 (95% CI 0.6 to 1.1)). Estimates were not calculated for Eastern Europe and Central Asia due to the non-availability of data. Globally, we estimated 5.5 million (UR 4.0–7.8) all-cause ALRI hospital admissions in children and adolescents 5–19 years of age annually, and about 55% (3.0/5.5 million), 23% (1.3 million) and 23% (1.3 million) of the hospital admissions occurred among 5–9 years, 10–14 years and 15–19 years, respectively.
Estimates of acute lower respiratory infections (ALRI) hospital admission rates (per 1000 persons/year) and hospital admissions in children and adolescents aged 5–19 years, by three age groups, country, UNICEF region, and World Bank income region
We identified multiyear (4 years or more) all-cause ALRI hospital admission rates in 21 countries between 1981 and 2017 (online supplemental appendix pp15-16). The trend of all-cause ALRI hospital admission rates varied by geographical locations: we found no significant change in all-cause ALRI hospital admission rates in 70% of these countries, a decrease in Brazil, Canada, France and Kenya, and an increase in Denmark, the Netherlands, and Taiwan, China.
There were 27 studies with data on hCFRs of all-cause ALRI, including 16 studies reporting data for 5–14 years and 12 studies for 5–19 years. hCFRs ranged from 0.05% (95%CI 0.05 to 0.06) to 11.76% (95%CI 7.11 to 18.85) across studies (online supplemental appendix pp24-25). Of the 27 studies, 9 studies had a low risk of bias in study design, and all studies had a low risk of bias in patient groups excluded. Seventeen studies used definitions similar to that we specified, eight studies used less specific definitions, and two studies used more specific definitions. Using all data from a mixture of age groups, we estimated an hCFR of 1.6% (95% CI 0.7 to 4.1), 1.6% (95% CI 0.6 to 4.2) and 0.4% (95% CI 0.2 to 0.6) for low-income and lower middle-income, upper middle-income, and high-income countries, respectively (table 2). These estimates translated to 87 900 (UR 40 300–180 600) all-cause ALRI in-hospital deaths for children and adolescents aged 5–19 years per year globally. By UNICEF regions, the hCFR estimates had wide CIs in the region of Sub-Saharan Africa, Latin America and Caribbean, and Asia-Pacific.
Estimates of acute lower respiratory infections (ALRI) in-hospital case-fatality ratios (hCFRs, %) and in-hospital mortality in children and adolescents aged 5–19 years by UNICEF region and World Bank income region
We identified 48 studies (26 countries) with data on influenza-associated ALRI hospital admission rates (table 3). Thirty-three of the 48 studies had a low risk of bias in study design, 47 studies had a low risk in adjustment for healthcare utilisation, and 45 studies had a low risk of bias in patient groups excluded, 14 studies had a low risk of bias in sampling strategy, and 30 studies had a low risk in test methods. Thirteen studies used definitions similar to that we specified, 34 studies used less specific definitions (including 4 studies using hospitalised ICD-coded confirmed influenza respiratory cases), and 1 studies used more specific definitions.
Estimates of influenza-associated acute lower respiratory infections (ALRI) hospital admission rates (per 1000 persons per year) and hospital admissions in children and adolescents aged 5–19 years, by three age groups, country, UNICEF region and World Bank income region
Influenza-associated ALRI hospital admission rate meta-estimates ranged from 0 to 8.1 (95%CI 4.5 to 14.5) per 1000 persons per year for 5–9 years across 26 countries, from 0 to 3.5 (95% CI 1.6 to 7.5) for 10–14 years, and from 0 to 1.7 (95% CI 0.4 to 8.2) for 15–19 years. Similar to all-cause ALRI hospital admission rates, by World Bank income regions we estimated highest influenza-associated ALRI hospital admission rate in upper middle-income countries (0.9 per 1000 persons per year (95% CI 0.4 to 2.1)), followed by lower middle-income countries (0.5 (95% CI 0.2 to 1.5)), low-income countries (0.2 (95% CI 0.1 to 0.3)), and high-income countries (0.1 (95% CI 0.1 to 0.2)). By UNICEF regions, influenza-associated ALRI hospital admission rates for 5–19 years were higher in Latin America and Caribbean, and Asia-Pacific (0.7–1.5 per 1000 persons per year) compared with other regions (0.1–0.2) except for Eastern Europe and Central Asia where estimates were not calculated due to the unavailability of data. Globally, we estimated 1 078 600 (UR 456 500–2 650 200) influenza-associated ALRI hospital admissions among 5–19 years annually, accounting for 19% (about 1.1 million of 5.5 million) of all-cause ALRI hospital admissions in this age group. About 63% (675 000/1 078 600), 25% (270 500), and 13% (144 400) of the hospital admissions occurred among 5–9 years, 10–14 years and 15–19 years, respectively.
We identified 54 studies with data on proportion positives of respiratory syncytial virus-associated ALRI, including 3 studies for 5–9 years, 34 studies for 5–14 years, and 17 studies for 5–19 years. The proportion positives of respiratory syncytial virus-associated ALRI ranged from 0% to 47.3% (95% CI 45.2 to 49.4) across studies (online supplemental appendix pp34-36). The meta-estimates of proportion positives of respiratory syncytial virus-associated ALRI were similar by World Bank income regions (3.9%–4.5%) (table 4). Based on these estimates, the annual global number of respiratory syncytial virus-associated ALRI hospital admissions was 231 800 (UR 1 42 700–3 73 200) among 5–19 years. By UNICEF regions, the estimated proportion positives of respiratory syncytial virus was higher in Sub-Saharan Africa, Eastern Europe and Central Asia, and Latin America and Caribbean than other regions. The estimates in these regions were based on limited data, therefore need to be verified with additional data.
Estimates of proportion positives of acute lower respiratory infections (ALRI) associated with respiratory syncytial virus, human metapneumovirus and human parainfluenza virus, and the hospital admissions in children and adolescents by World Bank income region and UNICEF region
We identified 32 studies with data on proportion positives of human metapneumovirus-associated ALRI, and 38 studies for proportion positives of human parainfluenza-virus-associated ALRI (online supplemental appendix pp37-40). Using the proportion-based approach, an estimated 105 500 (UR 57 200–181 700) human metapneumovirus-associated ALRI hospital admissions and 124 800 (UR 67 300–228 500) human parainfluenza-virus-associated ALRI hospital admissions occurred in children and adolescents aged 5–14 years annually (table 4).
In a sensitivity analysis, we excluded two potential outlier studies with the highest rates, and the estimate of global all-cause ALRI hospital admissions (5.1 million (UR 3.9–6.8) for 5–19 years) was similar to the main analysis (online supplemental appendix p10). The estimated global influenza-associated ALRI hospital admissions was 811 800 (UR 4 00 300–1 867 400) when excluding two potential outlier studies with the highest rates, and was 1 077 100 (UR 4 57 200–2 643 100) when excluding studies using the definition of “ICD-coded influenza-confirmed respiratory hospital admissions” (online supplemental appendix pp12-13). Based on proportion positives of influenza-associated ALRI in 70 studies, the estimate of global influenza-associated ALRI hospital admissions was 796 600 (UR 525 900–1 177 500) using the proportion-based approach (online supplemental appendix p14).
Discussion
Globally, an estimated 5.5 million (UR 4.0–7.8) all-cause ALRI hospital admissions occurred annually between 1995 and 2019 in children and adolescents aged 5–19 years, leading to 87 900 (UR 40 300–180 600) in-hospital deaths. Influenza virus and respiratory syncytial virus were estimated to account for 19% (1 078 600/5 546 100) and 4% (231 800) of ALRI hospital admissions in the 5–19 years age group. Human metapneumovirus (105 500) and human parainfluenza virus (124 800) accounted for about 2% of ALRI hospital admissions in the 5–14 years age group globally. Age-stratified results suggest that an estimated 55% (3 050 800) of ALRI hospital admissions and 63% (675 000) of influenza-associated ALRI hospital admissions occurred in children aged 5–9 years. By adding a country-level layer to our analytical model, we also report national estimates of all-cause ALRI hospital admissions for 39 countries and influenza-associated ALRI hospital admissions for 26 countries, accounting for 57% and 52% of the global population 5–19 years of age. We found that all-cause ALRI hospital admission rates for 5–19 years were similar over years in 15 of 21 countries where four or more consecutive years’ data were identified, and most of the countries are from World Bank upper middle-income and high-income regions. Additional multiyear data from low-income and lower middle-income countries are needed to understand the trend of all-cause ALRI hospital admission rates in these regions.13
Several factors could have affected the national, regional and global burden estimates, including the lack of data and the presence of methodological differences between studies. For many countries, data on all-cause ALRI hospital admissions and deaths and influenza-specific hospital admissions for 5–19 years are not available in published reports. In particular, the large uncertainty in our global all-cause ALRI in-hospital mortality estimate (87 900 (UR 40 300–180 600)) reflects the difference in hCFRs of all-cause ALRI between studies and the paucity of hCFR data, especially the lack of data by narrow age groups. By region, the paucity of data is most extreme in World Bank low-income region and in the UNICEF regions of Eastern Europe and Central Asia, South Asia, and West and Central Africa. Estimates for these regions were based on extrapolation of estimates in neighbouring regions, and could have been biased. For example, estimates for the region of West and Central Africa were based on estimates for the region of Eastern and Southern Africa; the estimates in the two regions could have been different given differences in the disease epidemiology (eg, seasonal variations, HIV prevalence). About 60%–70% of the country-specific estimates were based on subnational research studies, which might not be generalisable to the entire country. Additionally, all-cause and influenza-specific ALRI hospital admissions for some countries were estimated using historic data (eg, Germany, Israel and New Zealand for all-cause ALRI; Finland, Greece and Switzerland for influenza-associated ALRI), therefore need to be interpreted with caution.
Differences in case definitions between studies are noted, including less specific definitions (eg, hospitalised acute respiratory infections with/without fever), and more specific definitions (eg, hospitalised chest X-ray confirmed pneumonia). The heterogeneity in case definitions is more pronounced between studies reporting influenza-specific hospital admission rates. For example, 60% of the included studies used ‘hospitalised acute respiratory infections’ and severe acute respiratory infections as defined by WHO (or hospitalised acute respiratory infection with fever) and a further 10% used hospitalised ICD-coded influenza-confirmed respiratory hospital admissions.14 In a sensitivity analysis, exclusion of the studies using the least specific case definition (eg, hospitalised ICD-coded influenza-confirmed respiratory hospital admissions) yielded an estimate of global influenza-associated ALRI hospital admissions similar to that in the main analysis. Further restriction by case definitions in analyses was impossible, and the WHO definition of severe acute respiratory infections, developed for use in influenza surveillance in resource-limited settings, can be used more frequently in future influenza studies.14 Generally, the use of the less specific definitions could have caused an overestimation of influenza-associated ALRI hospital admissions. On the other hand, influenza-associated ALRI hospital admissions could have been underestimated due to the underdetection of influenza virus and the use of non-polymerase chain reaction (non-PCR) tests. Although we have adjusted for underdetection, levels of underdetection are unavailable in 33% of the included studies, likely causing an underestimation of the influenza-associated ALRI hospital admissions. Additionally, the adjustment is based on the assumption that the proportion positives of influenza virus in patients who were tested was the same as that in those untested. The estimates could have been biased as the assumption may not hold in 38% of the included studies. Thirty-eight per cent of the included studies used culture, immunofluorescence, rapid influenza test and mixed methods, which could have caused an underestimation to influenza hospital admissions as they have lower sensitivity and similar specificity compared with PCR.15 Most of the virus-specific data are from 2000 onwards, and the lack of data for the period before 2000 could have biased our estimates.
Countries in the region of Sub-Saharan Africa have the highest HIV prevalence in young people.16 The analyses by UNICEF regions show that about 5.1 million all-cause ALRI hospital admissions and 996 000 influenza-associated ALRI hospital admissions occurred globally excluding the region of Eastern Europe and Central Asia (96% of the global population 5–19 years of age), suggesting that our estimates of global all-cause and influenza-associated ALRI hospital admissions are generally unaffected by HIV.
Our estimate of global influenza-associated ALRI hospital admissions for 5–19 years is lower than the IHME estimate for 2017 (which was about 1.9 million influenza-attributable hospital admissions).2 Our estimate indicates that 862 900 ALRI hospital admissions could be attributed to influenza after accounting for the same influenza-specific causal fraction as used by IHME (80%). In our analysis, influenza-specific hospital admissions were modelled independently from all-cause ALRI hospital admission estimates, while the IHME estimate for influenza was modelled on ALRI hospital admission estimates. Despite the higher number of influenza-specific hospital admissions by IHME, their estimated proportion of influenza-positive hospital admissions in all-cause ALRI hospital admissions (about 11%) was lower than our estimate (19%). Therefore, the higher estimate by IHME is speculated to be largely driven by the underlying estimate of all-cause ALRI hospital admissions that they used. Further exploration of the difference is impossible as it requires additional information on the estimation of all-cause ALRI hospital admissions by IHME, specifically for 5–19 years.
Viruses are usually isolated in upper respiratory tract samples. Isolation of viruses in upper respiratory tract does not indicate a causal relationship as viruses can also be isolated in upper respiratory tract of healthy children. According to a systematic review in children under 5 years, the attributable fraction in influenza-positive ALRI is 80%; the fraction is 90% for respiratory syntactical virus, 73% for human metapneumovirus-positive ALRI, and 70% for human parainfluenza virus-positive ALRI.17 After accounting for the attributable fraction in our estimates for laboratory confirmed cases of ALRI positive for these 4 viruses, about 16% of all-cause ALRI hospital admissions can be attributed to influenza virus (862 900) and 4% attributable to respiratory syntactical virus (208 600) in 5–19 years. Using the same approach, 1%–2% of all-cause ALRI hospital admissions are attributable to human metapneumovirus (77 000) and human parainfluenza virus (87 400) in 5–14 years. A comparison between global all-cause and virus-specific ALRI estimates for 5–19 years and those for children under 5 years suggests that the estimates for 5–19 years (by 5-year age intervals where available) are at least 65% lower except for influenza-associated hospital admissions in 5–9 year olds (675 000) where it is only 20% lower than for children under 5 years (870 000).5 13 18
Conclusions
Our estimates provide evidence to support evaluation of all-cause ALRI intervention programmes for 5–19 years. In particular, the estimated high proportion of identification of influenza (19%) in all-cause ALRI hospital admissions for 5–19 years, especially in 5–14 years (22%) indicates the potential benefit of influenza vaccination in this age group.19 20 In addition to vaccines, a decrease in the prevalence of risk factors (eg, HIV and malnutrition) may also contribute to the reduction of all-cause ALRI burden in this age group.21–24 To improve the precision of the burden estimates and to track trends over time, more high-quality data are needed. Our regional and global estimates were based on extrapolation of available data from limited countries. The paucity of data as well as the low hospital admission rates in low-income countries reflects the limited access to healthcare, lack of high-quality healthcare systems, and the limited diagnostic capacity, which present a challenge to both the reduction and tracking of ALRI burden in this setting. Given these constraints, it is likely that the number of influenza-associated ALRI cases requiring hospital admission is underestimated in low-income countries. The data gaps for all-cause ALRI burden can be reduced as high-quality electronic healthcare registration systems are developed, and data are increasingly reported in places where such systems have been present.25 Increasing the diagnostic capacity in low-income and lower middle-income countries would improve the availability of viral burden data.26 The imputation by 5-year age groups was based on data from a small proportion of studies (nine studies for ALRI and eight studies for influenza-associated ALRI; online supplemental appendix p7). Increased reporting of data by standardised 5 year age groups will help improve estimates of all-cause and virus-specific ALRI hospital admissions, and allow for estimation of ALRI in-hospital deaths by age groups. Assessment of the trend of all-cause ALRI burden requires more than multiyear hospital admission rates; collection and reporting of concurrent hCFR data will help quantify the progress in reducing ALRI mortality in 5–19 years of age and improve ALRI mortality estimates. In terms of methodology, standardisation of case definitions, and assessment and reporting of underdetection of viruses will help improve all-cause and virus-specific ALRI burden estimates. Using existing influenza surveillance systems to incorporate testing for other viruses may help fill the larger data gaps, we observed for non-influenza viruses (ie, respiratory syncytial virus, human metapneumovirus and human parainfluenza virus). Our study includes the four viruses as among others, they are shown to be the viruses most strongly associated with ALRI in children under 5 years and therefore likely to result in substantial disease burden in older children as well.3 There are other viruses that are associated with ALRI (eg, adenovirus, rhinovirus, enterovirus and seasonal coronavirus), and the burden of these viruses needs to be investigated in future studies.
Data availability statement
Data are available in a public, open access repository. All data relevant to the study are included in the article or uploaded as supplementary information. Details of studies included in this review and references of these studies are uploaded as supplementary information. More details will be uploaded to Edinburgh DataShare (https://datashare.ed.ac.uk/) after publication.
Ethics statements
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Handling editor Seye Abimbola
Twitter @XinWang1991
Contributors HN and HC conceptualised the study. XW led the literature review with inputs from XM and EB. XW led the data analysis with contributions from YL. XW wrote the first draft of the report with input from HN, HC and YL. All authors reviewed the report and approved the last draft for finalisation.
Funding This study was supported, in whole or in part, by the Bill & Melinda Gates Foundation (OPP 1172551) through a prime award to John Hopkins University. Under the grant conditions of the Foundation, a Creative Commons Attribution 4.0 Generic License has already been assigned to the Author Accepted Manuscript version that might arise from this submission. HN and HC are members of the Respiratory Syncytial Virus Consortium in Europe, which has received funding from the Innovative Medicines Initiative 2 Joint Undertaking (116019) that receives support from the EU Horizon 2020 research and innovation programme and the European Federation of Pharmaceutical Industries and Associations. YL is funded by RESCEU.
Competing interests HC reports grants and personal fees from WHO, Bill & Melinda Gates Foundation, Johns Hopkins University, Sanofi during the conduct of the study. HN reports grants and personal fees from Bill & Melinda Gates Foundation during the conduct of the study; grants from Innovative Medicines Initiative, WHO, UK National Institute for Health Research, grants and personal fees from Foundation for Influenza Epidemiology, Sanofi, personal fees from Janssen, AbbVie, and Reviral outside the submitted work. YL reports grants from WHO outside the submitted work.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.