Incidence of community-acquired pneumonia among children under 5 years in Suzhou, China: a hospital-based cohort study


  • The incidence of community-acquired pneumonia (CAP) was estimated from a longitudinal cohort, which could track the temporal trend of CAP incidence.

  • This study includes both hospitalised and outpatient cases of CAP, providing a more comprehensive estimation of the CAP disease burden.

  • The cohort was established in a single hospital, the selection bias was inevitable and the withdraw bias occurred when children visited other hospitals and the health information was unavailable.


Community-acquired pneumonia (CAP) is one of the leading causes of morbidity and mortality in children aged ≤5 years.1 It is estimated that CAP accounts for 15.5% of all deaths and 24.8% of all hospitalisation cases in children.2 3 To date, CAP is more prevalent in developing countries, with a high incidence of 280 per 1000 person years,4 leading to substantial burden for individuals and society.

Despite this massive burden caused by CAP in children, significant knowledge gaps still exist. Estimating the incidence is challenging because CAP is not a notifiable disease in most countries. Rates of hospitalisation for CAP are widely used to estimate the disease burden in hospital-based studies.3 5 However, CAP is a group of diseases that constitute a wide range of severity, with some cases being treated as outpatients and often ignored. What is more, being the capacity limitation, many CAP cases requiring hospitalisation during the epidemic seasons are treated in outpatient clinics. Thus, only using the hospitalisation rate of CAP derived from discharge data may underestimate the CAP disease burden, and may blur the seasonality variation of CAP incidence in some way.

Therefore, some researchers have employed administrative data, to estimate the incidence of CAP.6 7 For example, a study used the urban Basic Medical Insurance database to estimate the nationwide incidence of CAP in China.6 The study reported an estimated incidence of 65.64 per 1000 person years among children aged ≤5 year. And regional variations and the seasonality in the CAP incidence were identified as well. But this study only included medical insurance population and the coverage medical insurance vary across regions and populations. Other studies used the data of cross-sectional surveys to estimate the CAP incidence.8 9 For instance, a cross-sectional survey was conducted in China in 2012 reported a CAP prevalence of 2.55%, revealing urban–rural differences in the disease burden.8 Nevertheless, results from the cross-sectional survey only provide a snapshot of the CAP prevalence, making it difficult to track the temporal trend of CAP incidence and analyse its seasonality. So, a longitudinal cohort design is preferable for accurately estimating the CAP incidence and exploring the age and seasonal variation.

As the rage of COVID-19 pandemic around the world, the epidemic characteristics of many non-COVID-19 diseases were also influenced by the implementation of the non-pharmaceutical interventions (NPI), which changed the people’s lives and health-seeking behaviours.10 In America, researchers found the declines in children’s medical visits across all diagnoses among children aged ≤5 years.11 12 It has been reported that during the COVID-19 lockdown, the activity of viruses such as influenza and respiratory syncytial virus decreased sharply, and the seasonality changed13–16 while studies have shown that some respiratory virus resurged when the community reopened.15 17 Suzhou adopted stringent NPIs from 25 January, 2020, including intercity traffic controls, wearing face masks and issuing stay-at-home orders. The measures were relaxed from 13 April 2020, the primary and secondary schools reopened, but social distance and city traffic controls were still required for months. During the study period, up to June 2021, due to zero case policy in China, Suzhou had a remarkably low infection rate of SARS-CoV-2 and none of them was not identified from the community. Also, these measures might have potential influencing on the epidemiology of CAP as well. Thus, the contemporary estimate of incidence for CAP before and during the COVID-19 pandemic would be of value.

With these, we conducted a hospital-based birth cohort study in Suzhou, an economically developed city in eastern China with four distinct seasons. The study aimed to describe the seasonality and age variations of CAP incidence from 2017 to 2021 and evaluate the COVID-19 impact on the CAP incidence. The birth cohort enrolled children born in 2017 or 2018, and followed up to June 2021 to identify the CAP cases both from the hospitalised data and the outpatient clinic data in a comprehensive tertiary children hospital.

Materials and methods

Study design and participants

This study was conducted at Soochow University Affiliated Children’s Hospital (SCH), which is the only comprehensive tertiary paediatric hospital with two campuses in Suzhou. SCH serves approximately 65.2% of the Suzhou population aged ≤5 years.18 All medical records were digitised and saved for several years. The International Classification of Diseases 10th Revision (ICD-10) codes were used for discharge diagnoses (or outpatient diagnoses).

We conducted a hospital-based cohort study from 1 January 2017 to 30 June 2021, which included children born in 2017 or 2018. The children were identified from the SCH’s hospital information system (HIS), regardless of the reasons for visit. To minimise the rate of loss to follow-up, we only included the registered residents or permanent residents of Suzhou. The Suzhou immunisation platform database was used to identify resident children because of the high coverage of the Expanded Programme on Immunization in China (more than 90%), and those could not be traced in the database would be excluded. And to minimise the effects of self-paid vaccines on the CAP incidence, the children ever received pneumococcal conjugate vaccine (PCV), Haemophilus influenza type b conjugate vaccine (Hib), or seasonal influenza vaccine were excluded.

Children were finally included in the study cohort if they met the following criteria: (1) born between 1 January 2017 and 31 December 2018. (2) Received any care at the SCH in the outpatient clinic, emergency department, or hospital wards between 1 January 2017 and 30 June 2021. (3) Registered on the immunisation platform of Suzhou Centers for Disease Control and can be successfully linked by name, sex and date of birth. (4) Never received PCVs, Hib and seasonal influenza vaccine.

Children with eczema, allergic rhinitis, asthma, congenital heart disease, low birth weight and immunodeficiency are more likely to develop CAP, and defined as high-risk populations, whereas other children are low-risk populations.4 9 19–21

Case definition

The CAP episodes occurred from birth to 30 June 2021 were identified through the HIS of SCH by screening the discharge or outpatient diagnoses (ICD-10). The diagnosis of CAP was based on the symptoms (fever, coughing, expectoration, chest pain and worsening symptoms of existing respiratory diseases), pulmonary signs, laboratory examinations and chest radiographs. The CAP episodes were coded as J09–J18 (influenza and pneumonia) or J20–J22 (other acute lower respiratory infections) according to the diagnoses.3 If two consecutive hospital visits with the above diagnoses occurred more than 30 days apart, they were considered as separate CAP episodes.

Statistical analysis

All analyses were performed using R (V.4.2.1, R Foundation for Statistical Computing, Vienna, Austria). The incidence of CAP was calculated as the number of CAP episodes divided by the cumulative observational person years in the cohort. In addition, we estimated the CAP incidence stratified by age (0–6, 7–12, 13–24, 25–36, 37–48 and 49–60 months), sex, birth year, health status group (high risk or low risk), season (spring is March–May, summer is June–August, autumn is September–November and winter is December–February) and month. The corresponding 95% CIs of the incidences were estimated using a Poisson distribution assumption. We fitted a quasi-Poisson regression model to calculate the adjusted rate ratio (RR). Stratified analysis of incidence of CAP by birth month was conducted to understand the age and seasonal variation. The Mann-Whitney U test, which was two-tailed and had a significance level of 0.05, was used to compare the incidence of CAP between various groups.

Patient and public involvement

It was not appropriate or possible to involve patients or the public in the design, or conduct, or reporting, or dissemination plans of our research.


The characteristics of study population

During the study period, a total of 328 495 children born in 2017 and 2018 were identified from the SCH’s HIS. After excluding 162 046 (49.3%) children who could not be linked to the immunisation database, 166 449 (50.7%) children were considered as Suzhou residents and included. Among the resident children, 56.6% born in 2017, and 43.4% born in 2018. These proportions were similar to those of excluded children (57.6% born in 2017, 42.4% born in 2018) (online supplemental table S1). The study found that 20.2% of the resident children had been vaccinated against pneumococcus, influenza or Hib. Children born in 2018 have a higher rate of vaccination than children born in 2017 (24.5% vs 16.9%). And children at high risk for CAP had a higher uptake of the self-paid vaccination (22.7%) than those at low risk (18.9%). (online supplemental table S2).

Supplemental material


Finally, we included 132 797 children in the cohort (figure 1), of which 78 251 (58.9%) and 54 546 (41.1%) were born in 2017 and 2018, respectively. The cohort consisted of slightly more boys than girls, with boys accounting for 54.2% and girls for 45.8% of the cohort. Additionally, 44 148 (33.2%) children were identified as being high risk for CAP. The most common high-risk status was eczema (19.5%), followed by allergic rhinitis (14.4%), asthma (6.1%), congenital heart diseases (1.8%), low birth weight (1.1%) and immunodeficiency (0.1%) (table 1).

Figure 1
Figure 1

Flow chart of the inclusion process for the study cohort. Hib, Haemophilus influenza type b conjugate vaccine; HIS, hospital information system; CDC, Centers for Disease Control.

Table 1

Demographic and characteristics of the children included in the study cohort

Among the children included in the cohort, there was a difference in the distribution of birth months between the 2 years. In 2018, the children born from January to March were more than those born from October to December (30.5% vs 18.6%), while the number of children born in each month in 2017 was similar. In addition, 35.1% of children born in 2017 are at high risk, and the percentage was higher than that of children born in 2018 (30.6%) (table 1).

Incidence of CAP

From 1 January 2017 to 30 June 2021, a total of 479 734.80 person years were observed and 62 406 CAP episodes occurred. The overall incidence of CAP was 130.08 per 1000 person years (95% CI 129.06 to 131.10).

Boys had a higher incidence than girls (143.99 vs 113.59 per 1000 person years, p<0.001), with an adjusted RR was 1.22 (95% CI 1.11 to 1.35). Children born in 2018 had slightly higher incidences than children born in 2017, and the adjusted RR was 1.03 (95% CI 0.93 to 1.14), indicating that birth year may not be an independent influence factor of CAP incidence. The incidence among children at high risk was significantly higher than those at low risk (222.48 vs 83.26 per 1000 person years, p<0.001). Multivariate analysis revealed that CAP incidence in children at high risk was 2.67 times greater (adjusted RR=2.67, 95% CI 2.42 to 2.94) compared with children at low risk. Among children at high risk, those with asthma had the highest incidence (449.71 per 1000 person years, 95% CI 442.13 to 457.29), while those with eczema had the lowest (207.89 per 1000 person years, 95% CI 204.97 to 210.81) (table 2).

Table 2

Incidence of community-acquired pneumonia (CAP) among children aged 0–60 months in Suzhou, China from January 2017 to June 2021

Age variation of CAP incidence

CAP incidence exhibited significant variation across different age groups (table 2). The highest CAP incidence was observed in children aged 7–12 months (214.11 per 1000 person years, 95% CI 210.59 to 217.63), while the lowest CAP incidence was in children aged 25–36 months (68.89 per 1000 person years, 95% CI 67.45 to 70.33). Age was identified as an independent risk factor for CAP by multivariate analysis using quasi-Poisson regression models. Compared with children aged 25–36 months, the greatest adjusted RR of CAP was 3.12 among children aged 7–12 months, then followed by 2.62 in children aged ≤6 months and 2.28 in children aged 13–24 months (table 2).

Among children born from January to June, the highest CAP incidence was noted when they were 7–12 months old, while those born from July to December had the highest incidence during 0–6 months old (figure 2). Additionally, it was observed that the peaks of CAP incidence occurred from November to February regardless of birth months and age (online supplemental figure S1).

Figure 2
Figure 2

Incidence of community-acquired pneumonia (CAP) by age group stratified by birth months. (A) The CAP incidence of children born from January to March among different age groups. (B) The CAP incidence of children born from April to June among different age groups. (C) The CAP incidence of children born from July to September among different age groups. (D) The CAP incidence of children born from October to December among different age groups.

The seasonality of CAP incidence

The highest CAP incidence was observed in winter (206.7 per 1000 person years, 95% CI 204.12 to 209.28), and the lowest incidence was in summer (79.57 per 1000 person years, 95% CI 77.95 to 81.19) (online supplemental table S3). The seasonal trends in the incidence of CAP during the study period are shown in figure 3. Generally, the incidence of CAP began to rise in August, reached its peak in December and January, and then declined to a trough in July and August. Incidences from October to April of the next year were higher than average incidence. The highest incidence was in January 2019 (426.04 per 1000 person years) and the lowest incidence before the COVID-19 lockdown was in August 2019 (85.65 per 1000 person years).

Figure 3
Figure 3

The temporal change of community-acquired pneumonia (CAP) incidence of children born from 2017 to 2018. The black solid lines indicate the trend of the CAP incidence from January 2017 to June 2021. The blue-dotted line indicates the average incidence during the whole study period, and the yellow-dotted line indicates the average incidence after exclusion the pandemic period of COVID-19.

Taking the incidence in August as referent, we calculated the RR for each month by age group (online supplemental figure S2). For children aged 0–6 months, the highest RR was observed in December (RR=3.75), while for children aged 13–24 months, the highest RR was in January (RR=3.86). Children aged 7–12 months had the relatively high and similar RR from December to February (RR: 2.99–3.24). Among children aged 25–60 months, the highest RR was in December (RR=4.59) and the RR decreased deeply from January (RR=3.73) to February (RR=1.16).

The impact of COVID-19 lockdown on CAP incidence

A significant decrease in the incidence of CAP was observed from February to August 2020 because of the outbreak of COVID-19 and subsequent lockdown in Suzhou (figure 3). The average incidence was 17.19 per 1000 person years during the lockdown, which was significantly lower than that during the rest of the study period (151.83 per 1000 person years). The lowest incidence (9.35 per 1000 person years) was observed in April 2020, with only one-ten of the lowest pre-pandemic incidence in August 2019. The incidence began to rise again when the community reopened since August 2020, although it remained lower than that before the COVID-19 outbreak. Of note, the incidence in February 2021 decreased significantly to 30.24 per 1000 person years, possibly due to the Spring Festival holiday and the implementation of special non-pharmacy interventions for COVID-19.

Moreover, in order to determine whether the decrease in CAP incidence after the COVID-19 lockdown was attributed to the pandemic itself or related age, we included a post-lockdown cohort comprising 31 692 children born from August 2020 to June 2021 who visited SCH for any reasons but not excluded those getting PCV vaccination. And the 328 495 children born in 2017–2018 and had visited the SCH were defined as the pre-pandemic cohort. We found that the CAP incidences were 138.04 and 163.45 per 1000 person years in the children of the post-lockdown cohort aged 0–6 months and 7–12 months, which were slightly higher than that occurred in children of pre-pandemic cohort (0–6 months: 133.26; 7–12 months: 146.75 per 1000 person years). These suggested that it is not the interventions of COVID-19 but increasing age leading to the decrease in incidence in 2021 (online supplemental table S4).


Based on the hospital-based birth cohort, we estimated the paediatric incidence of CAP in Suzhou, and clarified which age group children are susceptible to CAP and the seasonality of CAP incidence. Our study demonstrated that disease burden of CAP in Suzhou children remained high, and children≤24 months were more susceptible to CAP. And the study indicated a strong seasonal pattern for CAP and demonstrated a significant influence from the COVID-19 lockdown.

The present study showed that the overall incidence of CAP in Suzhou (130.08 per 1000 person years) is much higher than that in developed countries (3.38 per 1000 person years in Europe), but lower than that in developing countries (260 per 1000 person years).4 The difference may partially due to the potential disparity of social–economic across geographical regions. A recent study used the Chinese Urban Basic Medical Insurance data to estimate the incidence of CAP among children aged ≤5 years, which was lower than our estimation (65.64 vs 130.08 per 1000 person years).6 These differences could be explained by medical insurance coverage, payout ratio and the different CAP episode definition employed in this national study. Another study from Fuzhou, a southeastern city, reported an even much lower incidence of CAP of 15.97 per 1000 person years among children aged <5 years in 2020.22 Not including provincial-level hospitals and recurrent CAP cases might be the main causes of underestimation in the Fuzhou study.

As expected, age is a significant factor influencing the incidence of CAP. Children aged≤24 months exhibited a more than twofold higher incidence of CAP compared with older children. Our results align with previous studies, demonstrating consistency across different regions.9 23–26 An Indian study reported that children aged ≤12 months have the highest incidence of CAP in India (273–911 per 1000 person years), which was much higher than that in our study.9 American research showed that the trend of CAP incidence with age was consistent with our result, although they reported a much lower incidence.26 After conducting a stratified analysis by birth month, we found that age trends of CAP incidence varied among children in different birth months. Children born from January to June had the highest incidence of CAP at the age of 7–12 months, whereas children born from July to December had the highest incidence at the age of 0–6 months. The seasonal trends of CAP incidence among children born in different months were consistent, peaking in January and declined to a trough in July and August, indicating that seasonality played a more important role in determining CAP incidence than age among children≤12 months.

Our study showed that the incidence of CAP began to rise in August, peaked in December and January, and then declined to a trough in July and August. From October to April of the next year were the months with higher incidence of CAP compared with the average incidence, which means the incidence is higher in winter. The seasonal pattern was consistent with the results of other studies.3 6 8 22 The highest incidence was observed in January 2019, which was also seen in other southern provinces of China,27 28 relating to the temperature, humidity, economic and social activities and personal immunity at that time. We also found that the different seasonal trend among age groups. Compared with children aged 13–24 months, the earlier outbreaks were observed in children aged 0–6 months. And the peak timing lasted longer in children aged 7–12 months. As the pathogenic heterogeneity in seasonal peaking timing had been reported previously, for example, Influenza A often precedes Influenza B, the seasonality variation may associate with the different susceptibility to pathogens among different age groups.29 30

The incidence of CAP decreased sharply during the lockdown of COVID-19. The CAP incidence slumped to 17.19 per 1000 person years between February to August 2020, which was far lower than the previous lowest incidence, and then rebounded. Also, as this study was based on a birth cohort, this declined incidence (resulted from COVID-19 lockdown) was mainly occurred among children aged 25–60 months. The trend was consistent with other researches in China, which indicated that non-pharmaceutical measures are effective measures against CAP and children’s healthy seeking behaviours had changed during the period of the COVID-19 lockdown.10 31–33 Around the time of the Chinese New Year in 2021, there was another decline in incidence, which may have been caused by the special holiday policy implemented in the post-COVID-19 lockdown period. We also found that the CAP incidence did not decrease after August 2020 in the same age group, indicating that COVID-19 plays limited impacts on CAP incidence after the COVID-19 lockdown period.

The risk of CAP among children at high-risk health status was 2.67 times that of children at low risk. Our study found that onset of CAP was closely associated with respiratory or atopic conditions, such as asthma, allergic rhinitis and eczema, which was consistent with previous researches.21 34 35 Also, children with congenital heart disease, immunological deficiencies and low birth weight are more susceptible to CAP.35

Our study had some limitations. First, due to the fact that almost half children were excluded from the cohort when identifying residents by Suzhou immunisation platform database, some residents may have been excluded erroneously. As there was not significant difference between the included and excluded residents, this exclusion may lead to a smaller cohort, while the influence on the estimation of incidence could be neglected. Second, children vaccinated against influenza, Hib and Streptococcus pneumoniae were excluded from our cohort may lead to selection bias. While because all of these are optional and commercial vaccines in China, the vaccination rates are quite low to have indirect impact on the unvaccinated children. In addition, reporting the incidence of unvaccinated children are more important to understand the disease burden and evaluate the vaccine impact in the future. Third, the cohort was established in a single hospital, and the selection bias was inevitable. The children who never seek medical care would be ignored, which might lead to the overestimate of the incidence, while the prevalence of eczema, asthma and allergic rhinitis in our cohort aligns with previous researches,36 indicating that the potential selection bias is unlikely to have a significant impact on the results. Also, the information bias occurred when children visited other hospitals and the health information was unavailable For example, the prevalence of low birth weight was lower than the previous survey,37 which may result in records related to low birth weight are predominantly found in the birth hospitals rather than SCH, and the RR of children with low birth weight might be influenced.

In conclusion, the study reveals the considerable burden of CAP among children, especially among children aged ≤24 months, and the CAP seasonal pattern, of which October–April of the next year had the higher incidence. The incidence of CAP decreased during COVID-19 lockdown and rebounded when the community reopened. These data offer reliable evidence for which population should be considered susceptible to CAP when applying prevention strategies, as well as how to adjust the strategies dynamically in accordance with the temporal variation of CAP.

Data availability statement

No data are available.

Ethics statements

Patient consent for publication

Ethics approval

This study was approved by the School of Public Health, Fudan University Human Research Ethics Committee (IRB#2017-11-0646).


We thank all the children and their guardians who have participated in this study, as well as the staff of the SCH involved in data collection.

This post was originally published on