Global hospital admissions and in-hospital mortality associated with all-cause and virus-specific acute lower respiratory infections in children and adolescents aged 5–19 years between 1995 and 2019: a systematic review and modelling study

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.


Appendix 2 Data preparation and imputation
For hospital admission rates of ALRI and influenza-associated ALRI, we extracted the number of cases and population-at-risk per study. Rates were adjusted for healthcare utilisation where available. Influenza-specific rates were additionally adjusted for levels of testing where available.
We imputed data for any of the three age bands (5-9 years, 10-14 years, or 15-19 years) where missing. Since data on rate ratios between the three age bands were only available for hospital admission rates of ALRI and influenzaassociated ALRI, rates and the number of cases were imputed for the two outcomes. We used a multiple imputation approach was as used previously for influenza virus burden estimation. 1,2 The imputation was done at the study level following three steps: (1) imputing the denominator; (2) imputing the rate; (3) calculating the case number by combining the denominator and rate. Details of each step of imputation are presented below.
(1) We imputed denominator for any of the three age bands using denominators reported in each study and the country-level population structure by single year of age. 3 (2) We imputed rates using a multiple imputation approach by assuming the rates were missing at random. 1,2 Figure S2.1 shows the imputation of rates. First, we pooled rate ratios between three age bands using the network meta-analysis, with the age band 5-9 years as the reference. 4 Second, the pooled rate ratios were assumed to follow log-normal distributions, and 10 samples of each rate ratio were simulated. Third, we estimated rate ratios for any other age bands (e.g., 5-14 years versus 5-9 years) using the rate ratios between the three age bands and the UN Population Division country population structures by single year of age (Table S3.1). 3 We assumed the hospital admission rate was the similar within each five-year age band. Fourth, with samples of rate ratios and the observed rates in each study, we calculated 10 samples of rates for each of the three age bands for each study. Fifth, numbers of cases for each study were calculated using the denominator and imputed rates. Following this strategy, 10 datasets were imputed for each of the three age bands. Sixth, meta-analysis was done for each dataset, and the meta-estimates were combined together using the Rubin's rules. 5,6 BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) BMJ Global Health doi: 10.1136/bmjgh-2021-006014 :e006014.    Hospital admissions 6584000 (4727700-9222300) ‡ 5546100 (3967500-7782700) ‡ The slightly higher estimate when imputation was not done could be due to the more frequent report of data for 5-9y, which were higher than older age groups (10-14y).
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance

Appendix 3 Sensitivity analyses 1. ALRI hospital admissions
We excluded the two potential outlier studies with the highest rates (one in Bolivia and the other one in China) in a sensitivity analysis (Table S3.1). After excluding the two studies, we were unable to estimate burden in Bolivia as there were no other studies; the estimate in China after the exclusion was similar to that in the main analysis, with uncertainty ranges overlapping. The estimate of global ALRI hospital admissions remained similar after the exclusion.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

Appendix 4 Trend of hospital admission rates of ALRI
Data on ALRI hospital admission rates over four or more consecutive years are plotted in Figure S4.1. The trend of ALRI hospital admission rates varied by geographic locations: a decrease was found in Brazil (annual percent change, -3%), Canada (-3%), France (-3%), and Kenya (-8%); an increase in Denmark (+4%), Netherlands (+4%), and Taiwan, China (+14%). We did not found significant change of ALRI hospital admission rates over time in other locations.
We plotted multi-year studies with influenza-associated ALRI hospital admission rates for four or more consecutive years in Figure S4.2.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance  For each region/country, the points represent the rates for each year, and the lines connecting these points represent the rates over years for one site or one age band. Different colours represent data from different sites or age bands. For example, in Canada the purple line represent the national rates for 5-19 years, and the blue line represent the rates among children and adolescents aged 5-17 years in Ontario, Canada. For each region/country, the points represent the rates for each year, and the lines connecting these points represent the rates over years for one site or one age band. Different colours represent data from different sites or age bands.  * ARI: any respiratory symptoms (e.g., cough, runny nose, sore throat, shortness of breath, or difficulty in breathing) requiring hospital admission. ARI AND fever: fever or history of fever AND any respiratory symptoms requiring hospital admission. ALRI: any of physician-diagnosed pneumonia (and bronchiolitis), ICD-coded pneumonia (and bronchiolitis), or fever or history of fever AND cough or sour throat AND shortness of breath or difficulty breathing. CXR-pneumonia: chest radiograph confirmed pneumonia. y: years.   NA NA * ARI: any respiratory symptoms (e.g., cough, runny nose, sore throat, shortness of breath, or difficulty in breathing) requiring hospital admission. ARI AND fever: fever or history of fever AND any respiratory symptoms requiring hospital admission. ALRI: any of physician-diagnosed pneumonia (and bronchiolitis), ICD-coded pneumonia (and bronchiolitis), or fever or history of fever AND cough or sour throat AND shortness of breath or difficulty breathing. CXRpneumonia: chest radiograph confirmed pneumonia. y: years. † Data were reported for other age bands, so are not presented in this  124 ALRI NPW or NPA; PCR NA 0 NA * ARI: any respiratory symptoms (e.g., cough, runny nose, sore throat, shortness of breath, or difficulty in breathing) requiring hospital admission. ARI AND fever: fever or history of fever AND any respiratory symptoms requiring hospital admission. ALRI: any of physician-diagnosed pneumonia (and bronchiolitis), ICD-coded pneumonia (and bronchiolitis), or fever or history of fever AND cough or sour throat AND shortness of breath or difficulty breathing. CXRpneumonia: chest radiograph confirmed pneumonia. y: years.  NA NA * ARI: any respiratory symptoms (e.g., cough, runny nose, sore throat, shortness of breath, or difficulty in breathing) requiring hospital admission. ARI AND fever: fever or history of fever AND any respiratory symptoms requiring hospital admission. ALRI: any of physician-diagnosed pneumonia (and bronchiolitis), ICD-coded pneumonia (and bronchiolitis), or fever or history of fever AND cough or sour throat AND shortness of breath or difficulty breathing. CXR-pneumonia: chest radiograph confirmed pneumonia. y: years.   Given the paucity of influenza-specific hCFR data, we did a primary analysis for influenza-associated ALRI inhospital deaths for 5-14 years. Table S6.11 shows the preliminary estimates. An estimated 10,500 (UR 2,200-49,800) influenza-associated ALRI in-hospital deaths occurred among children and adolescents aged 5-14 years globally when including all hCFR data (for 5-14 years or 5-19 years). When restricted the analysis to including only hCFR data for 5-14 years, we estimated 18,900 (UR 4,700-75,600) influenza-associated ALRI in-hospital deaths globally for 5-14 years. We were unable to estimate influenza-associated ALRI in-hospital deaths for 5-19 years due to the lack of data. We were unable to estimate in-hospital deaths for the other three viruses given the paucity of data on hCFRs of respiratory syncytial virus (four studies), human metapneumovirus (one study), and human parainfluenza virus (one study). § Estimates from meta-analyses. ** Estimated by combining hCFRs and hospital admissions of IFV-associated ALRI by regions. † † hCFR estimates were based on all data for 5-14 years or 5-19 years. ‡ ‡ hCFR estimates were based on the data for 5-14 years.

Low
Other definitions that are more specific or less specific, e.g., ARI or fever requiring hospital admission; hospitalised with infection infections (ICD-code)

High
Sampling strategy (only for studies with data on viruses) The proportion of cases tested is available AND either of the following:  For all data inputs from multiple sources that are synthesized as part of the study: 3 Describe how the data were identified and how the data were accessed. Page4 4 Specify the inclusion and exclusion criteria. Identify all ad-hoc exclusions. Page5 5 Provide information on all included data sources and their main characteristics. For each data source used, report reference information or contact name/institution, population represented, data collection method, year(s) of data collection, sex and age range, diagnostic criteria or measurement method, and sample size, as relevant.
Appendix pp18-41 6 Identify and describe any categories of input data that have potentially important biases (e.g., based on characteristics listed in item 5).

Appendix pp42-55
For data inputs that contribute to the analysis but were not synthesized as part of the study: 7 Describe and give sources for any other data inputs. Appendix pp18-41 For all data inputs: 8 Provide all data inputs in a file format from which data can be efficiently extracted (e.g., a spreadsheet rather than a PDF), including all relevant meta-data listed in item 5. For any data inputs that cannot be shared because of ethical or legal reasons, such as third-party ownership, provide a contact name or the name of the institution that retains the right to the data.
Data have been presented in the supplementary material. Data will be made available on Edinburgh Datashare (https://datashare.is.ed.ac.uk/) later.

Data analysis 9
Provide a conceptual overview of the data analysis method. A diagram may be helpful. Figure 1  10 Provide a detailed description of all steps of the analysis, including mathematical formulae. This description should cover, as relevant, data cleaning, data pre-processing, data adjustments and weighting of data sources, and mathematical or statistical model(s).

11
Describe how candidate models were evaluated and how the final model(s) were selected. Page 7 12 Provide the results of an evaluation of model performance, if done, as well as the results of any relevant sensitivity analysis.

13
Describe methods for calculating uncertainty of the estimates. State which sources of uncertainty were, and were not, accounted for in the uncertainty analysis.
14 State how analytic or statistical source code used to generate estimates can be accessed.
Major codes used in this study will be made available upon request.

Results and Discussion 15
Provide published estimates in a file format from which data can be efficiently extracted.
Estimates can be easily extracted in main table and supplementary  table. Main tables will also be provided on Edinburgh Datashare (https://datashare.is.ed.ac.uk/) later. 16 Report a quantitative measure of the uncertainty of the estimates (e.g. uncertainty intervals).
Uncertainty is reported for burden estimates in main tables, supplementary tables and Results.

17
Interpret results in light of existing evidence. If updating a previous set of estimates, describe the reasons for changes in estimates.

18
Discuss limitations of the estimates. Include a discussion of any modelling assumptions or data limitations that affect interpretation of the estimates.