Global, regional, and national incidence and mortality burden of non-COVID-19 lower respiratory infections and aetiologies, 1990–2021: a systematic analysis from the Global Burden of Disease Study 2021

1Evidence before this study Lower respiratory infection (LRI) is a common and deadly infectious disease, particularly in children and older adults. Previous iterations of the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) and estimates from WHO and the Maternal and Child Epidemiology Estimation Group have quantified the LRI burden for select aetiologies in the pre-COVID-19 era. In addition, many studies have estimated the decrease in incidence or mortality due to LRI or select respiratory pathogens during the COVID-19 pandemic, but these studies are generally limited to one or a few surveillance networks, countries, or clinical sites. We searched PubMed with the search terms (“lower respiratory infection*”[Title/Abstract] OR “LRI”[Title/Abstract]) AND (“mortality” OR “incidence”) AND “global*” AND (“etiology” OR “pathogen”) with no language restrictions, for articles published from Jan 1, 2021 to June 16, 2023. We did not identify any studies that evaluated global levels and trends of LRI burden in all ages, attributable to a comprehensive set of aetiologies, across all countries, and inclusive of the COVID-19 pandemic's effects to the year 2021. Added value of this study This study provides two key improvements on the past GBD study: expanded aetiology estimation and evaluation of COVID-19 pandemic impact. We produced estimates of non-COVID-19 LRI burden attributable to a comprehensive set of 18 different aetiologies (Acinetobacter baumannii, Chlamydia spp, Enterobacter spp, Escherichia coli, fungi, group B streptococcus, Haemophilus influenzae, influenza, Klebsiella pneumoniae, Legionella spp, Mycoplasma spp, polymicrobial infections, Pseudomonas aeruginosa, respiratory syncytial virus [RSV], Staphylococcus aureus, Streptococcus pneumoniae, and other viruses, as well as a residual category of other bacterial pathogens). 13 of these aetiologies are newly included in the GBD study, significantly expanding our understanding of the diverse causes of LRI. In addition, this research, which models through the year 2021, estimates the reduction in non-COVID-19 LRI incidence and mortality observed during the COVID-19 pandemic period. In addition, we added many new data sources on LRI morbidity and mortality since GBD 2019, which span widely across time and geography, enabling us to revise and improve estimates from previous years. Overall, these enhancements contribute to a more comprehensive and up-to-date understanding of the global burden of LRI, incorporating previously unaccounted for aetiologies and considering the influence of the COVID-19 pandemic on respiratory infections. This information is invaluable for health-care practitioners, policy makers, and researchers in effectively developing targeted interventions to combat LRIs. Implications of all the available evidence With a comprehensive understanding of the aetiologies of LRI and their impact, health-care authorities can design targeted interventions to address specific pathogens responsible for respiratory infections. These interventions might include vaccination campaigns, improved infection control measures, and early detection and treatment strategies. This study found S pneumoniae to be the most common cause of LRI deaths in 2021, followed by S aureus and K pneumoniae. During the COVID-19 pandemic, following the implementation of non-pharmaceutical interventions such as facemask use and mobility restrictions, we observed a decline in global influenza and RSV infection incidence and mortality. Since 1990, incidence and mortality due to LRI have greatly decreased, especially in children younger than 5 years, while mortality rates in adults, especially those aged 70 years and older, have had a slower rate of decline. Our analysis particularly highlights the decrease in vaccine-preventable aetiologies, S pneumoniae and H influenzae, and the importance of maintaining and expanding vaccine coverage against these bacteria. We also found high mortality attributable to non-vaccine-preventable aetiologies, including S aureus; development of preventive therapies and vaccines for these pathogens should receive further investment and research. Furthermore, as the threat of antimicrobial resistance grows, robust pathogen surveillance, point-of-care pathogen identification, and implementation of strategies to reduce antibiotic overuse become essential. The LRI burden remains highly inequitable, with both deaths and cases highly concentrated in low-income and middle-income countries; thus, all interventions must be financially accessible and distributed to areas with a high burden of LRI.

2Evidence before this study

3Lower respiratory infection (LRI) is a common and deadly infectious disease, particularly in children and older adults. Previous iterations of the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) and estimates from WHO and the Maternal and Child Epidemiology Estimation Group have quantified the LRI burden for select aetiologies in the pre-COVID-19 era. In addition, many studies have estimated the decrease in incidence or mortality due to LRI or select respiratory pathogens during the COVID-19 pandemic, but these studies are generally limited to one or a few surveillance networks, countries, or clinical sites. We searched PubMed with the search terms (“lower respiratory infection*”[Title/Abstract] OR “LRI”[Title/Abstract]) AND (“mortality” OR “incidence”) AND “global*” AND (“etiology” OR “pathogen”) with no language restrictions, for articles published from Jan 1, 2021 to June 16, 2023. We did not identify any studies that evaluated global levels and trends of LRI burden in all ages, attributable to a comprehensive set of aetiologies, across all countries, and inclusive of the COVID-19 pandemic's effects to the year 2021.

4Added value of this study

5This study provides two key improvements on the past GBD study: expanded aetiology estimation and evaluation of COVID-19 pandemic impact. We produced estimates of non-COVID-19 LRI burden attributable to a comprehensive set of 18 different aetiologies (Acinetobacter baumannii, Chlamydia spp, Enterobacter spp, Escherichia coli, fungi, group B streptococcus, Haemophilus influenzae, influenza, Klebsiella pneumoniae, Legionella spp, Mycoplasma spp, polymicrobial infections, Pseudomonas aeruginosa, respiratory syncytial virus [RSV], Staphylococcus aureus, Streptococcus pneumoniae, and other viruses, as well as a residual category of other bacterial pathogens). 13 of these aetiologies are newly included in the GBD study, significantly expanding our understanding of the diverse causes of LRI. In addition, this research, which models through the year 2021, estimates the reduction in non-COVID-19 LRI incidence and mortality observed during the COVID-19 pandemic period. In addition, we added many new data sources on LRI morbidity and mortality since GBD 2019, which span widely across time and geography, enabling us to revise and improve estimates from previous years. Overall, these enhancements contribute to a more comprehensive and up-to-date understanding of the global burden of LRI, incorporating previously unaccounted for aetiologies and considering the influence of the COVID-19 pandemic on respiratory infections. This information is invaluable for health-care practitioners, policy makers, and researchers in effectively developing targeted interventions to combat LRIs.

6Implications of all the available evidence

7With a comprehensive understanding of the aetiologies of LRI and their impact, health-care authorities can design targeted interventions to address specific pathogens responsible for respiratory infections. These interventions might include vaccination campaigns, improved infection control measures, and early detection and treatment strategies. This study found S pneumoniae to be the most common cause of LRI deaths in 2021, followed by S aureus and K pneumoniae. During the COVID-19 pandemic, following the implementation of non-pharmaceutical interventions such as facemask use and mobility restrictions, we observed a decline in global influenza and RSV infection incidence and mortality. Since 1990, incidence and mortality due to LRI have greatly decreased, especially in children younger than 5 years, while mortality rates in adults, especially those aged 70 years and older, have had a slower rate of decline. Our analysis particularly highlights the decrease in vaccine-preventable aetiologies, S pneumoniae and H influenzae, and the importance of maintaining and expanding vaccine coverage against these bacteria. We also found high mortality attributable to non-vaccine-preventable aetiologies, including S aureus; development of preventive therapies and vaccines for these pathogens should receive further investment and research. Furthermore, as the threat of antimicrobial resistance grows, robust pathogen surveillance, point-of-care pathogen identification, and implementation of strategies to reduce antibiotic overuse become essential. The LRI burden remains highly inequitable, with both deaths and cases highly concentrated in low-income and middle-income countries; thus, all interventions must be financially accessible and distributed to areas with a high burden of LRI.

8Lower respiratory infections (LRIs) were the leading infectious cause of death globally in 2019., Gram-positive and Gram-negative bacteria, atypical bacteria, viruses, and fungi can all cause LRI. Mortality rates are highest in adults older than 70 years and in children younger than 5 years, and both incidence and mortality are generally higher in males.,, Risk factors for LRI mortality in all age groups include exposure to tobacco smoke, indoor and outdoor particulate matter, and extreme temperatures. In children younger than 5 years, wasting is estimated to be responsible for over half of LRI deaths. Among adults aged 65 years and older, host-level risk factors can include frailty and presence of comorbid conditions such as asthma., Vaccination against Streptococcus pneumoniae is protective against pneumococcal pneumonia in both infants and older adults.,

9Among community-acquired bacterial LRIs, S pneumoniae remains the most prevalent pathogen in children and adults and across different income-level settings., Historically, Haemophilus influenzae was the second-leading cause of childhood pneumonia. However, with the widespread implementation of H influenzae type b (Hib) vaccination, the incidence of H influenzae pneumonia has declined substantially over the past decade., Staphylococcus aureus, which is not vaccine-preventable, is a noteworthy cause of complicated pneumonia, with substantially higher rates of poor clinical outcomes, including sepsis and death, than S pneumoniae., S aureus also has the ability to develop resistance to multiple antibiotics, posing further barriers to care. In school-age children, the atypical bacterium Mycoplasma pneumoniae is a leading cause of pneumonia, with one review estimating that it is responsible for 4–39% of cases of paediatric community-acquired pneumonia.,

10Viruses, including influenza and respiratory syncytial virus (RSV), are highly prevalent causes of LRIs, particularly in children., A 2021 global meta-analysis estimated that influenza viruses were responsible for 14·1% of adult LRI hospitalisations, or more than 5 million hospitalisations. Another global meta-analysis estimated that RSV was responsible for 3·6 million hospitalisations in 2019 among children younger than 5 years. In addition, viral infections increase patients' risk for superimposed bacterial infections, most commonly by S pneumoniae and S aureus, causing substantial morbidity and mortality.

11Beginning in 2020, the COVID-19 pandemic promoted the adoption of non-pharmaceutical interventions, including stay-at-home orders, school and community closures, and facemask requirements. These measures effectively curbed the incidence of respiratory infections in 2020 and 2021, for both COVID-19 and other respiratory viruses.,,, RSV and influenza infection incidence declined in response to these non-pharmaceutical interventions, although some locations had outbreaks of these viruses in atypical seasons as non-pharmaceutical interventions were relaxed.,,,

12This study presents the results from the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2021, which estimates LRI incidence and mortality, combined with the findings of the Global Research on Antimicrobial Resistance (AMR) project, which estimates LRI pathogen distribution. We aimed to describe the burden and trends of LRIs and the pathogens responsible across 204 countries and territories from 1990 to 2021. Previous GBD studies included estimates of four aetiologies that were not mutually exclusive or collectively exhaustive., In the current study, we provide estimates for a comprehensive set of 18 pathogen categories across all age groups., Additionally, the estimates for 2020 and 2021 account for the reduction of LRIs seen during the COVID-19 pandemic and implementation of non-pharmaceutical interventions.

13This Article was produced as part of the GBD Collaborator Network and in accordance with the GBD protocol. GBD 2021 produced estimates of mortality and morbidity due to LRI by age and sex for 204 countries and territories between 1990 and 2021. The Global Burden of AMR study produced estimates of aetiology-specific fatal and non-fatal burdens of select infectious syndromes, including LRI. LRI is defined as acute pneumonia or bronchiolitis, not inclusive of COVID-19. ICD codes mapped to LRI in GBD are provided in appendix 1 (pp 17–18) for ICD-9 and ICD-10. The GBD case definition of LRI does not include tuberculosis, pertussis, or COVID-19; although the pathogens that cause these diseases can infect the lower respiratory tract, they are modelled separately due to their individual public health significance and are not included in the GBD category of LRI.

14GBD uses a set of modelling tools, described in the sections below, to extrapolate available data out to produce results for the entire global population, by age, sex, and year. Modelling was done at the 1000 draw level, where the point estimate was computed as the mean of 1000 draws, and the 95% uncertainty intervals (UIs) were computed as the 25th and 975th ranked values of 1000 draws. We used the GBD 2021 global population age standard to calculate age-standardised rates, which allow for comparison of rates between locations or years with different age structures. In the following sections, we summarise key methods from the GBD and Global Burden of AMR studies for the estimation of LRI and its aetiologies. More details on these methods, including a flowchart, are provided in appendix 1 (pp 4–29). Full descriptions of the GBD and Global Burden of AMR studies have been published previously.,

15All metadata for input sources described below are available on the GBD Sources Tool, found on the Global Health Data Exchange (GHDx), which readers can use to identify which sources were used for estimating an outcome in any given location. GBD 2019 complies with the GATHER statement (appendix 1 pp 30–31). Statistical code used for GBD estimation is publicly available online on the GHDx.

16As inputs to the GBD LRI mortality-estimation model, we used a total of 26 259 site-years of data: 23 062 site-years from vital registration, 825 site-years from sample vital registration, 1682 site-years from verbal autopsy, 681 site-years from surveillance sources, and 9 site-years from minimally invasive tissue sampling. Data are processed using a set of standard algorithms accounting for incompleteness, misclassification of the underlying cause of death, garbage coding, and stochastic variability.

17We estimated overall LRI mortality using the Cause of Death Ensemble model (CODEm), which evaluates a wide array of potential models using various combinations of covariates and four model classes. Each model class uses either cause fraction or death rate as the outcome variable, and either a mixed-effects linear model or a spatiotemporal Gaussian process model as the regression method. Models included fixed effects on covariates and age dummies. Random effects are applied at the levels of super-region, region, and age in the spatiotemporal model's mixed-effects structure, and at the levels of super-region, region, country, and age in the mixed-effects linear models. In mixed-effects regression, the random effects are assumed to follow a normal distribution with a mean of zero and a variance–covariance matrix that is to be estimated from the data. Models were evaluated using out-of-sample predictive validity and integrated into one ensemble model. A full list of covariates is provided in appendix 1 (pp 8–9). Final LRI mortality estimates are scaled by a procedure known as CoDCorrect to ensure consistency between the sum of cause-specific mortality and the total envelope of all-cause mortality.

18For LRI morbidity estimation, we used data from published studies identified via a systematic review (appendix 1 p 10), surveillance data, LRI mortality estimates (described above), health insurance claims data, and inpatient data. To correct for potential systematic bias among different categories of data sources, we used a standardised crosswalking technique to adjust the data to enhance comparability before modelling (appendix 1 pp 10–13). We estimated LRI incidence and prevalence using DisMod-MR 2.1, a compartmental Bayesian meta-regression model that enforces consistency among prevalence, incidence, remission, and mortality., More details on DisMod-MR, including information on priors and a full list of covariates, is provided in appendix 1 (pp 15–17).

19Data used for aetiology estimation originated from multiple cause-of-death vital registration, hospital discharges, microbial laboratory data, and published studies from the literature. Mortality and morbidity are estimated for the following causes of LRI: Acinetobacter baumannii, Chlamydia spp, Enterobacter spp, Escherichia coli, fungi, group B streptococcus, H influenzae, influenza viruses, Klebsiella pneumoniae, Legionella spp, Mycoplasma spp, polymicrobial infections, Pseudomonas aeruginosa, RSV, S aureus, S pneumoniae, and other viruses (ie, the aggregate of all viruses except for influenza and RSV), as well as a residual category of other bacterial pathogens. The ICD-9 and ICD-10 codes mapped to each cause are listed in appendix 1 (pp 18–19).

20Incidence proportions were estimated using multinomial estimation as part of a network analysis model, which allows for the inclusion of data sources that are considered to be partial observations—ie, which do not contain all pathogen groups modelled in the study. Proportions were estimated as a function of age group, infection type, Hib and pneumococcal vaccination, and Healthcare Access and Quality (HAQ) Index. These covariates vary across geography and time, creating unique predictions for each age group, location, and year. For data sources that only reported deaths, we used modelled case-fatality rates (CFRs) to retroactively estimate the number of cases. These CFRs for each pathogen were modelled using a Bayesian meta-regression tool, MR-BRT (meta-regression—Bayesian, regularised, trimmed), as a function of age group, pathogen, and HAQ Index, with random effects on data source.,, For S pneumoniae, we used a vaccine probe design as an additional input to the incidence proportion model, due to the documented challenge in the microbiological identification of this pathogen. Modelled CFRs were then used again to compute mortality proportions from case proportions. More details on aetiology estimation can be found in appendix 1 (pp 17–26). Ultimately, all estimated incident LRI cases were distributed to an estimated aetiology, even those with no aetiology detected, following the modelled aetiology distribution patterns by age, location, and year.

21We developed a multistep modelling process to estimate the reduction of incidence of influenza and RSV in 2020 and 2021. Our source data were reported cases of influenza by country, from notifications reported by countries to WHO's FluNet. First, we interpolated the number of reported cases of influenza in 2020 and 2021 by month using the RegMod framework, a Poisson model that estimates the underlying rate of infection in each month as a function of a seasonal pattern and an underlying temporal trend. Second, we calculated an under-reporting ratio in the pre-pandemic reference period, 2017–19, for each location by dividing the interpolated number of reported cases from RegMod by the GBD estimated number of cases of LRI due to influenza. Third, we estimated the pandemic disruption-free counterfactual number of reported cases, meaning the number of reported cases we would have expected during 2020 and 2021 in the hypothetical pandemic-free scenario. We did this by multiplying the under-reporting ratio by the estimated number of cases of LRI due to influenza, for 2020 and 2021, that GBD would have estimated in a pandemic-free scenario. Finally, we calculated a yearly disruption influenza scalar for each location for 2020 and 2021. This scalar was computed by dividing the interpolated number of reported cases from RegMod (result of first step) by the counterfactual disruption-free number of reported cases (result of third step).

22These influenza disruption scalars (result of final step) were multiplied by counterfactual incident cases and deaths for both influenza and RSV (result of third step), to estimate adjusted cases and deaths. More details on the adjustments are provided in appendix 1 (pp 26–29).

23The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

24Globally, in 2019, before the reductions in incidence observed during the COVID-19 pandemic, we estimated 369 million (95% UI 349–391) LRI episodes, for an all-age incidence rate of 4770 episodes (4510–5040) per 100 000 population (table 1).

25In 2021, we estimated 344 million (325–364) incident episodes of LRI globally, for an all-age incidence rate of 4350 episodes (4120–4610) per 100 000 (table 1). Across 204 modelled locations, the all-age incidence rate in 2021 ranged from 463 episodes (428–500) per 100 000 in Cyprus to 9980 episodes (9220–10 800) per 100 000 in Nepal (figure 1; appendix 2 p 5). Adults aged 70 years and older had the highest global incidence rate at 18 900 episodes (17 100–21 000) per 100 000, followed by adults aged 50–69 years at 6370 episodes (5800–6940) per 100 000 (table 1). Among children younger than 5 years, we estimated 37·8 million (33·5–43·0) incident episodes of LRI and an incidence rate of 5750 episodes (5090–6540) per 100 000 (table 1), ranging from 413 episodes (335–504) per 100 000 in the Netherlands to 12 190 episodes (10 600–13 900) per 100 000 in Pakistan (appendix 2 p 5).

26Since 1990, the all-age global LRI incidence rate decreased 19·0% (95% UI 16·0–21·9), from 5880 (5510–6250) episodes per 100 000 in 1990 to 4770 episodes (4510–5040) per 100 000 in 2019 (table 1). This decline was primarily attributable to reductions in incidence among children younger than 5 years, which decreased 59·3% (58·3–60·1), from 16 300 episodes (14 500–18 300) per 100 000 in 1990 to 6640 episodes (5900–7490) per 100 000 in 2019 (table 1). By contrast, the global incidence rate among adults aged 70 years and older declined at a lower rate from 1990 to 2019, with an overall decrease of 4·8% (0·4–9·1; table 1).

27Globally in 2019, before reductions in mortality observed during the COVID-19 pandemic, we estimated 2·55 million (95% UI 2·32–2·74) global LRI deaths and an all-age mortality rate of 32·9 deaths (29·9–35·4) per 100 000 population, representing a 41·7% decrease (35·9–46·9) in mortality rate since 1990 (table 2). Among children younger than 5 years, we estimated 693 000 (580 000–822 000) deaths, for a mortality rate of 102·2 deaths (85·5–121·3) per 100 000 in this age group in 2019 (table 2).

28In 2021, we estimated 2·18 million (1·98–2·36) deaths globally due to LRI and an all-age mortality rate of 27·7 deaths (25·1–29·9) per 100 000 (table 2). The all-age mortality rate ranged from 2·3 deaths (1·8–2·9) per 100 000 in Qatar to 104·0 deaths (81·8–129·2) per 100 000 in Chad (figure 1; appendix 2 p 86). Among children younger than 5 years, we estimated 502 000 deaths (406 000–611 000) due to LRI globally, or 76·2 deaths (61·7–92·9) per 100 000 (table 2), ranging from 0·3 deaths (0·2–0·5) per 100 000 in Andorra to 357·9 deaths (271·4–456·4) per 100 000 in Chad (appendix 2 p 86). Across the aggregated age groups, adults aged 70 years and older had the highest global mortality rate (224·6 deaths [197·8–243·7] per 100 000), followed by children younger than 5 years (table 2).

29LRI fatalities in 2021, especially among children, were concentrated in countries with a low Socio-demographic Index (SDI; appendix 2 p 5). Of 204 modelled countries and territories, 57 had an LRI mortality rate greater than 60 per 100 000 among children younger than 5 years in 2021 (appendix 2 p 86). In 2021, among children younger than 5 years, mortality rates per 100 000 population were 153·2 deaths (118·7–193·4) in low SDI countries, 94·0 (77·5–112·4) in low-middle SDI countries, 34·2 (28·8–40·3) in middle SDI countries, 8·6 (7·2–10·0) in high-middle SDI countries, and 1·9 (1·7–2·0) in high SDI countries (table 2). In total, 254 000 LRI deaths (197 000–320 000) in children younger than 5 years occurred in low SDI countries (table 2). However, although the low SDI quintile had the highest burden in 2021, these countries also showed the greatest improvement in all-age mortality rates over time (table 2).

30Globally, between 1990 and 2021, the all-age LRI mortality rate decreased by 50·9% (95% UI 45·6–55·9), from 56·5 deaths (51·3–61·9) to 27·7 deaths (25·1–29·9) per 100 000 population (figure 2). For males, it decreased by 49·4% (44·0–54·4), from 58·6 deaths (53·0–64·6) to 29·6 deaths (27·2–32·1) per 100 000. For females, it decreased by 52·7% (46·3–58·6), from 54·4 deaths (48·8–60·5) to 25·7 deaths (22·5–28·3) per 100 000 (figure 2). Analogous to incidence, the decline in mortality was largely attributable to reductions in deaths among children; LRI mortality rate decreased by 75·6% (70·7–79·8) in children younger than 5 years and 59·2% (52·7–64·2) in children aged 5–14 years (figure 2). Adults aged 70 years and older had the smallest decrease in LRI mortality rate, with a 23·8% (18·7–28·7) decline (figure 2). More detailed results on LRI incidence and mortality for additional age groups by sex, country, and year are available online via the GBD Results Tool on the GHDx.

31In 2021, the pathogen responsible for the largest proportion of LRI incident episodes globally was S pneumoniae, which caused an estimated 97·9 million (95% UI 92·1–104·0) episodes (Figure 3, Figure 4; appendix 2 p 2104). This was followed by the categories of other viruses (ie, the aggregate of all viruses studied except influenza and RSV; 46·4 million [43·6–49·3] episodes) and Mycoplasma spp (25·3 million [23·5–27·2] episodes; Figure 3, Figure 4; appendix 2 p 2104). Key pathogens varied by age and geography. S pneumoniae was responsible for the largest number of episodes in 165 of the 204 modelled countries and territories in 2021, while the category of other viruses was responsible for the largest number of episodes in 39 countries (appendix 2 p 156). For all five studied age subdivisions, S pneumoniae caused the most episodes (figure 3; appendix 2 p 2104).

32In 2019, before the COVID-19 pandemic and the decline in observed incidence of influenza and RSV, the first and second most common aetiologies were the same as for 2021, but the third most common aetiology was influenza, responsible for 36·4 million (95% UI 34·2–38·7) episodes globally (figure 4). S pneumoniae was responsible for the most LRI episodes in all five age groups in 2019, and was followed by the category of other viruses in all age groups except 70 years and older, in which influenza ranked second highest (appendix 2 p 2104). The third most common aetiology was other viruses in people aged 70 years and older, RSV in children younger than 5 years, Mycoplasma spp in children aged 5–14 years and people aged 15–49 years, and influenza in people aged 50–69 years (appendix 2 p 1971).

33In 2021, the pathogen responsible for the largest proportion of all-age LRI deaths globally was also S pneumoniae, which led to an estimated 505 000 deaths (95% UI 454 000–555 000; figure 4; appendix 2 p 2107). This was followed by S aureus (424 000 deaths [380 000–459 000]) and K pneumoniae (176 000 deaths [158 000–194 000]; figure 4; appendix 2 p 2104). In 2019, before COVID-19 impacted the transmission of influenza and RSV, the first and second most common aetiologies leading to LRI death were the same as in 2021, but the third most common aetiology was influenza, which led to 349 000 deaths (318 000–377 000) globally (figure 4). Across age groups, in both 2019 and 2021, S pneumoniae was responsible for the most LRI deaths in people younger than 70 years, whereas S aureus caused the most deaths in people aged 70 years and older (figure 3; appendix 2 p 2104). In 2019, the second largest number of deaths came from S aureus in all age groups except children younger than 5 years, for whom the second-ranked aetiology was RSV, and people aged 70 years and older, for whom the second-ranked aetiology was S pneumoniae. The third most common aetiology leading to death was influenza for all five age groups (appendix 2 p 1971). In 2021, S pneumoniae was responsible for the highest number of deaths in 103 of 204 modelled countries and territories, whereas S aureus was responsible for the most deaths in the remaining 101 countries. These differences were largely attributable to differences in age structures across countries (appendix 2 p 156).

34From 1990 to 2019, H influenzae had the largest reduction in global mortality (a 54·8% decrease [95% UI 48·8–60·6] to 59 800 deaths [53 200–66 200]), followed by S pneumoniae (48·5% decrease [42·8–53·9]; appendix 2 p 1971). Most of this improvement for both pathogens was in children younger than 5 years, with a 77·4% (72·9–81·2) decline in deaths due to H influenzae (from 102 000 [83 800–123 000] to 23 000 [18 600–27 700]) and a 76·1% (71·4–79·7) decline in deaths due to S pneumoniae (from 721 000 [621 000–843 000] to 172 000 [142 000–205 000]) during this period (appendix 2 p 1971).

35Following the onset of the COVID-19 pandemic, we estimated that from 2019 to 2021, the number of influenza episodes decreased by 60·3% (95% UI 47·1–72·9) to 14·4 million (9·81–19·4) episodes, and deaths decreased by 71·8% (63·8–78·9) to 98 200 (74 300–126 000; appendix 2 p 1951). Across 18 modelled pathogen categories, influenza fell from being the third leading cause of both LRI episodes and deaths globally in 2019, to the ninth leading cause of episodes and the eighth leading cause of deaths in 2021 (figure 4). The high-income super-region saw the largest decrease in influenza episodes from 2019 to 2021 (91·5% [86·6–94·7]), and south Asia had the smallest decrease in influenza episodes in that same time frame (44·0% [14·6–71·8]; appendix 2 p 1951).

36Similarly, since 2019, we estimated that global RSV episodes declined by 63·2% (53·1–72·7) to reach 4·59 million (3·32–6·02) in 2021, with a similar decrease in RSV deaths (66·7% [56·6–75·3]), to 31 500 (23 300–41 600) in 2021 (figure 4; appendix 2 p 1951).

37Overall, for non-COVID-19 LRIs from 2019 to 2021, we estimated an 8·6% (6·6–10·4) decline in the overall incidence rate, from 4770 episodes (4510–5040) to 4350 episodes (4120–4610) per 100 000 population (from 369 million [349–391] to 344 million [325–364] total episodes), and a 16·0% (13·1–18·6) decline in mortality rate, from 32·9 deaths (29·9–35·4) to 27·7 deaths (25·1–29·9) per 100 000 (table 2).

38This study provides comprehensive global, regional, and national estimates of LRI episodes and deaths attributable to 18 pathogen categories, by age group, from 1990 until 2021. These estimates are inclusive of the reduction in transmission of certain respiratory viruses observed during the COVID-19 pandemic and implementation of non-pharmaceutical interventions. We estimated 344 million (95% UI 325–364) incident episodes of LRIs and 2·18 million (1·98–2·36) deaths worldwide in 2021. S pneumoniae was responsible for the highest proportion of both incidence and mortality in all ages, followed by the category of other viral aetiologies and Mycoplasma spp for incidence, and S aureus and K pneumoniae for mortality. Between 2019 and 2021, during the COVID-19 pandemic, we estimated substantial declines in global influenza incidence and RSV incidence.

39Although LRIs are ubiquitous across the world, the burden disproportionately falls on people living in poverty. In 2013, WHO and UNICEF formulated the Global Action Plan for the Prevention and Control of Pneumonia and Diarrhea (GAPPD), with the ambitious goal to end preventable childhood pneumonia and diarrhoea deaths by 2025. A specific target for 2025 is to reduce mortality from pneumonia in children younger than 5 years to fewer than 3 deaths per 1000 livebirths, roughly equivalent to a mortality rate of less than 60 deaths per 100 000 people per year among children younger than 5 years. As of 2021, we estimated a global LRI mortality rate of 76·2 deaths (61·7–92·9) per 100 000 children in this age group, and that 57 countries and territories, all but one of which were low-income and middle-income countries (LMICs), had a mortality rate over the global benchmark of 60 deaths per 100 000. To reduce mortality, the action plan calls for promotion of exclusive breastfeeding in infants younger than 6 months, reduction of indoor air pollution, expanded access to health care, and ongoing pneumonia case management in LMICs—approaches that have historically driven progress towards reducing LRI child mortality.,

40The WHO and UNICEF GAPPD also calls for increased coverage of pneumococcal conjugate vaccines (PCVs) and Hib vaccines. From 1990 to 2019, H influenzae showed the largest decline in global deaths, followed by S pneumoniae, both largely attributable to vaccination. Between 2000 and 2015, use of the Hib vaccine prevented an estimated 1·2 million deaths due to H influenzae infection globally, and PCV prevented an estimated 250 000 deaths due to pneumococcal infection. However, global coverage of these vaccines shows substantial room for improvement. According to WHO–UNICEF estimates of national immunisation coverage, global final-dose coverage of PCV among 1-year-olds was 60% and Hib coverage was 76% in 2022, both of which are above 2019 levels, suggesting recovery from pandemic immunisation disruptions.,, Although these global increases are promising, they can mask substantial inequities, and many vulnerable communities remain without access to vaccination.,, Strategies described by the WHO Immunization Agenda 2030 to increase coverage—including focusing on children who have not received any routine immunisations, building trust to avert vaccine hesitancy, and increasing vaccine access across the lifespan—can help reduce pneumonia mortality in areas with the highest burden.,,,

41The age groups of children younger than 5 years and adults aged 70 years and older had the highest LRI mortality rates in 1990. Time trends showed a steep decline in mortality in children younger than 5 years between 1990 and 2021, whereas no substantial decrease was observed in adults aged 70 years and older (figure 2). This trend holds true for the more granular age groups of 70–74 years and 75–79 years (appendix 2 p 4). Decline in immune function with ageing, called immunosenescence, promotes susceptibility to LRIs, as do age-related organ system changes and the development of comorbid conditions., Influenza and pneumococcal vaccination remain effective tools to address LRIs in older adults. Pneumococcal immunisation of infant populations provides some herd protection for older adults., In addition, pneumococcal vaccine administration to adults aged 65 years and older has been shown to be cost-effective with modest efficacy,,, at least in high-income settings. Because immunosenescence limits the efficacy of some vaccines in older adults, improved vaccine efficacy has emerged as a priority. Strategies towards more effective vaccines for older adults include higher doses of vaccines, repeated vaccinations, mucosal, subcutaneous, or intradermal administration, and use of more potent adjuvants. In LMICs, vaccine access for older adults is severely limited. More research is needed to assess the potential benefits of adult vaccination, understand barriers and challenges, and establish evidence-based guidelines in these settings.,

42RSV, the second-leading cause of LRI deaths in children younger than 5 years in 2019, has historically not been vaccine preventable. The development of affordable RSV vaccines and long-acting, affordable monoclonal antibodies (mAbs) was a priority for WHO's Vaccine Product and Delivery Research Unit and an active area of research., These efforts came to fruition in 2023, when two vaccines for RSV were approved in the EU and the USA.,,, Both are for use in adults aged 60 years and older, and one is also approved for pregnant women to protect their infants. In addition, a new long-acting mAb injection, nirsevimab, was approved in 2022 in the EU and in 2023 in the USA to prevent RSV hospitalisation in both healthy and high-risk infants., Generally, mAbs, including the long-approved, short-acting, RSV-preventive palivizumab, are too costly for use in most LMICs. The affordability of long-acting mAbs for LMICs is not yet known; preliminary cost-effectiveness analyses suggest a benefit, but this cost-effectiveness will depend on multiple factors, including pricing.,,, These long-acting mAbs and RSV vaccines, available for the first time, have the potential to avert unprecedented numbers of RSV cases and deaths in the 2023–24 respiratory infection season and beyond. A 2023 modelling study forecasts that with 60% vaccine coverage, in the USA alone, up to 2·0 million symptomatic RSV respiratory infections could be averted per year in adults older than 60 years, plus another 690 000 infections in the non-vaccinated population through indirect effects. However, these benefits will only reach locations where patients can access the vaccines. For the full global benefit of these preventives to be realised, equitable distribution is essential. It will be crucial for pharmaceutical companies, non-governmental organisations, and governments to work together to reduce barriers to access in LMICs.

43We have quantified the global burden of LRI attributable to S aureus in all ages and, for the first time in a comprehensive global study, we have identified the pathogen as the second-leading cause of LRI mortality after S pneumoniae in 2021. Although S aureus is a less frequent cause of LRI cases than S pneumoniae, it has a higher incidence of complications and a higher CFR. In a multisite US study, adult patients with S aureus LRI (n=37) had worse outcomes than those with S pneumoniae (n=115), including higher rates of intensive care unit (ICU) admission (62·2% vs 34·8%), mechanical ventilation (24·3% vs 12·2%), and inpatient mortality (10·8% vs 4·4%). Because of this poor prognosis, antistaphylococcal therapy is frequently included in empirical treatment for severe pneumonia., Finding the causative pathogen in a patient with pneumonia can be challenging, and clinicians face the trade-off of balancing sufficiently broad empirical treatment with antibiotic stewardship., In a global meta-analysis of S aureus pneumonia, 51% of isolates were meticillin-resistant S aureus (MRSA). The emergence of vancomycin-intermediate and vancomycin-resistant S aureus represents an escalating concern, and multidrug-resistant S aureus is classified as high-priority on the WHO global priority antimicrobial resistance pathogen list., Antibiotic overuse, a key driver of resistance, remains an important concern across high-income countries and LMICs. Improved point-of-care diagnostics, including targeted PCR testing for MRSA, can prevent antibiotic overuse.,, In addition, although efforts to develop a vaccine against S aureus have so far been unsuccessful, ongoing research might generate a new method for prevention or treatment.,

44Our overall estimates of LRI mortality among children younger than 5 years are consistent with findings from a publication by WHO and the Maternal and Child Epidemiology Estimation Group, which estimated 740 000 (95% UI 620 000–840 000) child LRI deaths in 2019. We estimated 693 000 (580 000–822 000) child LRI deaths in 2019. Our estimates of pathogen distribution are also similar to other global reports. A 2016 study from the Global Initiative for MRSA Pneumonia (GLIMP), which included data across 54 countries, identified S aureus in 188 (6%) of 3193 adults with community-acquired pneumonia, in alignment with the current study. Likewise, a 2021 meta-analysis across eight countries estimated that 18% (95% CI 13–24) of community-acquired pneumonia cases in adults aged 50 years and older were attributable to S pneumoniae. Our estimates for the 50–69 years and 70 years and older global age groups are within the 95% CI of this meta-analysis.

45In addition, we have estimated the COVID-19 pandemic-era reduction in influenza and RSV mortality and incidence by country, applied to a comprehensive set of global LRI estimates. From 2019 to 2021, we estimated a 71·8% (95% UI 63·8–78·9) decrease in influenza deaths and a 66·7% (56·6–75·3) decrease in RSV deaths worldwide (appendix 2 p 1951). These reductions were observed following the implementation of non-pharmaceutical interventions such as facemask use and mobility restriction, which have been implicated in the reduction of transmission of respiratory viruses, including influenza and RSV, in 2020 and 2021.,,,, However, other respiratory viruses, such as rhinovirus, adenovirus, and respiratory enteroviruses, quickly rebounded within a few months and persisted despite non-pharmaceutical interventions, showing fewer fluctuations in case counts with changing policies compared with influenza and RSV. Overall, hospitals across the world have reported reductions in admissions for community-acquired pneumonia during the COVID-19 pandemic.,,

46This study has several limitations. First, we quantified the COVID-19 pandemic-attributable reduction in LRI for influenza and RSV only. New evidence from a global surveillance network including 26 countries shows a decline in incidence of invasive infections attributable to respiratory pathogens, including S pneumoniae and H influenzae, during the COVID-19 pandemic. The decline in pneumococcal disease incidence might be primarily attributable to the decline in transmission of co-infecting respiratory viruses, including influenza and RSV, rather than reduced transmissibility or serotype selection of S pneumoniae itself. For the bacterial aetiologies that are predominantly health-care acquired, evidence is mostly limited to single-site studies and mixed, with some studies showing a reduction,, others showing an increase, and others showing no change. In the post-pandemic period, studies suggest that several of the pathogens that decreased in 2020 rebounded, including RSV, influenza, and pneumococcus.,,, Other pathogens—namely, Mycoplasma spp—showed a decline that persisted for a longer duration after the COVID-19 pandemic, with continued decreased detection observed until the end of 2022, followed by a delayed re-emergence in some countries in mid-2023.,, Due to interruption of established data exchanges caused by the COVID-19 pandemic, we were unable to estimate how bacterial pathogen distributions might have changed between 2019 and 2021. As data become available for more locations, pathogens, and years, we can comprehensively quantify the indirect effects of the pandemic on the incidence of LRI and its aetiologies in future rounds of GBD.

47A second limitation is that, when estimating the effect of the COVID-19 pandemic on influenza and RSV incidence, we relied exclusively on case notification data from national and multinational surveillance networks. Our method cannot separate the effects of a true decrease in LRI incidence from the effects of a decrease in health-care-seeking behaviour; we also did not account for potential changes in reporting capability over time. Third, to calculate the reduction in RSV, we applied modelled estimates of COVID-19 pandemic-associated influenza reduction directly to RSV estimates. This decision was based on a meta-analysis of the ratio of the percentage change in influenza to the percentage change in RSV in 2020, relative to the pre-pandemic period, which showed no statistically significant difference in the reduction of the two pathogens. However, empirical studies published since the pandemic have shown that the resurgence patterns of RSV and influenza have differed.,,,, Fourth, limited data availability and quality are constraints, particularly in low-income countries, where the LRI burden is highest. Our assessment of LRI mortality in countries lacking vital registration data relies largely on verbal autopsy studies, which have modest sensitivity in accurately identifying deaths due to LRIs. Covariates and regional trends were leveraged to predict the burden of LRI and corresponding aetiologies for locations with few or no data. In selecting these covariates, some degree of model misspecification is possible due to potential omitted variables that are not captured in the dataset, which could affect the accuracy of our predictive model. Fifth, misclassification might be present in pathogen proportion data if certain pathogens are more difficult to detect than others, or if some pathogens, such as viruses in the population of older adults, are irregularly tested in a laboratory or clinical setting. Sixth, although we used a crosswalking process to adjust for systematic differences in incidence data source categories, this process might not fully account for all forms of bias. Finally, we directly applied LRI aetiology proportions from the Global Burden of AMR study to GBD estimates of LRI cases and deaths, although the two studies use slightly different definitions of LRI. In particular, the Global Burden of AMR study's definition of LRI deaths covers any event for which LRI was present in the causal chain, regardless of the underlying cause of death, whereas the GBD definition only includes instances in which LRI was the underlying cause of death.

48In summary, we have shown that, despite declines in incidence during the COVID-19 pandemic, LRIs remain a significant cause of morbidity and mortality worldwide. Increased access to existing vaccines, as well as rollout of novel vaccines and therapies, could reduce the burden of LRIs. Supporting research for low-cost interventions against S aureus could accelerate progress in reducing LRI-related mortality and incidence, especially in resource-constrained settings. In addition, the growing threat of antimicrobial resistance highlights the importance of antibiotic stewardship and investment in improved diagnostic technologies to improve the specificity and accuracy of therapy. Finally, all these interventions must come at an affordable cost, so that they can reduce inequities seen in LRI mortality, rather than exacerbate them.

49To download the data used in these analyses, please visit the GHDx GBD 2021 website.

50J A Berkley reports support for the present manuscript from research grants from the Bill & Melinda Gates Foundation, Wellcome Trust, the National Institute for Health and Care Research (NIHR), and the Medical Research Council (MRC). C Brown reports other financial support from an ad-hoc one-off market research advisory role (anonymously conducted via market research companies with no direct communication, none specifically related to lower respiratory tract infections), all outside the submitted work. K Krishan reports other non-financial support from the UGC Centre of Advanced Study, CAS II, awarded to the Department of Anthropology, Panjab University (Chandigarh, India) outside the submitted work. M-C Li reports support for the present manuscript from the National Science and Technology Council in Taiwan (112-2410-H-003-031) and other financial or non-financial support as a Technical Editor of the Journal of the American Heart Association, Review Editor of Frontiers in Public Health, and Editorial Board Member of BMC Public Health, outside the submitted work. S A Meo reports grants or contracts from King Saud University (Riyadh, Saudi Arabia; RSP-2024 R47), outside the submitted work. L Monasta reports support for the present manuscript from the Italian Ministry of Health (Ricerca Corrente 34/2017), with payments made to the Institute for Maternal and Child Health IRCCS Burlo Garofolo. C Moore reports participation with Dr Gwen Knight as a member of the advisory board for MRC grants (no payments made), with the WHO Advisory group, and with the REVIVE Advisory group as a member of the steering group; and leadership or fiduciary roles in board, society, committee, or advocacy groups (unpaid) as the co-chair of the Impact and Influence Group of the Microbiology Society, outside the submitted work. A Pollard reports grants or contracts from the Bill & Melinda Gates Foundation, Wellcome Trust, Cepi, MRC, NIHR, AstraZeneca, European Commission, and the Serum Institute of India; royalties or licenses from AstraZeneca; consulting fees from Shionogi; leadership or fiduciary roles in board, society, committee, or advocacy groups (unpaid) as the chair of the Department of Health and Social Care's Joint Committee on Vaccination and Immunisation and as a member of the WHO Strategic Advisory Group of Experts on Immunization until 2022; and receipt of equipment, materials, drugs, medical writing, gifts, or other services from Moderna, outside the submitted work. L F Reyes reports grants or contracts from MSD and Pfizer; consulting fees from GlaxoSmithKline, MSD, and Pfizer; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from GlaxoSmithKline, MSD, and Pfizer; payment for expert testimony from GlaxoSmithKline, MSD, and Pfizer; and support for attending meetings or travel from GlaxoSmithKline and Pfizer, outside the submitted work. Y L Samodra reports grants or contracts from Taipei Medical University, and leadership or fiduciary roles in board, society, committee, or advocacy groups (paid or unpaid) as the co-founder of Benang Merah Research Center, outside the submitted work. E A F Simôes reports support for the present manuscript from the Bill & Melinda Gates Foundation; grants or contracts from AstraZeneca, Merck & Co, Pfizer, and Icosavax; consulting fees from Merck & Co, Pfizer, GlaxoSmithKline, Sanofi Pasteur, Cidara Therapeutics, Adagio Therapeutics, Nuance Pharmaceuticals, Enanta, and Icosavax; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Pfizer and AstraZeneca; support for attending meetings or travel from Pfizer and AstraZeneca; and participation on a data safety monitoring board or advisory board with AbbVie, GlaxoSmithKline, the Bill & Melinda Gates Foundation, and Moderna, outside the submitted work. M Zielińska reports other financial support as an AstraZeneca employee, outside the submitted work.