Acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) are important causes of morbidity and mortality in pediatric intensive care units (PICUs). The incidence of pediatric ARDS varies from study to study depending on the patient population targeted [14], and mortality is usually less than 30%. In developing countries, there are very few data available regarding the incidence, clinical management and outcome of pediatric ARDS. In 2001–2002, an initial study was conducted in 15 adult ICUs in Shanghai. In this study the incidence of ARDS was approximately 2% of all ICU admissions, with a 90-day mortality >70%, almost nine times the overall ICU death rate [5]. We found a similar, high mortality in four major PICUs in Shanghai during that period [6]. In the PICU, clinical features of ARDS are different from adult patients in terms of underlying diseases, predisposing factors, management of the initial period, treatment modalities and response, and outcome. Diagnosis and treatment of ARDS based on 1994 American–European Consensus Conference (AECC) definition [7] is widely accepted. Lung protective strategies have been shown to reduce the mortality in adult ARDS, however, the benefits of such strategies have never been examined in the context of PICU in the developing world. We postulated that, by conducting a prospective epidemiologic study in a large number of PICUs, we should be able to identify the incidence and mortality of ARDS and thus provide baseline data for future studies.


Participating centers

Twenty-five PICUs with a median of 9 (range 4–32) beds participated in the study, representing municipalities and provincial cities in both southeastern and mid-western regions of China (Figure A in “Supplementary materials”). There was 24 h access to clinical radiology and laboratory investigations. Half of the centers were university based with clinical research experience and facilities, while the other half were city central hospitals with limited research commitment. All but four units admitted surgical patients, and five units admitted patients after cardiac surgery. All the centers had senior consultant pediatric intensive care specialists, and provided a clinical resident training program. Each served as the tertiary referral center in pediatric emergency and intensive care for a population of 1–5 million inhabitants. Twenty-two of these centers participated in the study from 1 January to 31 December 2004, and three from 1 June 2004 to 31 May 2005.

Study design and patient management

The study protocol was approved by the Ethics Committee of Children’s Hospital of Fudan University and adopted by each center as required by the Chinese regulations for clinical investigations. As data were collected from observational parameters and no element of specific intervention was used in the protocol, informed consent was waived. All PICU admissions for infants and children in the age group from 1 month to 14 years old were eligible for entry to the study. Only patients with a pediatric critical illness score ≤90 (see Table A in “Supplementary materials”), or who met the American guidelines for admission and discharge policies [8] were included.

Patients and respiratory care

Acute respiratory distress syndrome was defined as acute onset of respiratory failure, hypoxemia (PaO2/FiO2 ≤200 mmHg), bilateral lung infiltration, and lack of left heart failure [7]. To ensure prompt identification and accurate diagnosis of ARDS, we imposed both oxygenation criteria and chest radiographic criteria as simultaneously present within 24 h [9]. FiO2 was based on set and measured values in the ventilators for those dependent on mechanical ventilation, or on measured values obtained proximal to the patient nose and mouth using a CY100 oximeter (Lihua Scientific Instrument, Hangzhou, Zhejiang) in those who breathed spontaneously on oxygen with assisted devices. All chest X-ray films were transmitted to the coordinating center through electronic mail (within 24 h of PaO2/FiO2 ≤200 mmHg) and reviewed by two radiologists to reach agreement. In case of disagreement, X-ray evidence was re-examined in the subsequent workshop. As pulmonary catheterization was not applicable in most participating centers for determination of left atrial hypertension, echocardiographic criteria were used according to Dahlem et al. [2]. We required that echocardiographic examination should be performed within 24 h when bilateral infiltration existed on chest X-ray and when there was any doubt regarding cardiac ventricular function.

For patients with the diagnosis of ARDS who required endotracheal intubation and mechanical ventilation, tidal volume (V T), peak inspiratory pressure (PIP), positive end-expiratory pressure (PEEP), and mean airway pressure (MAP) during the first 3 days of mechanical ventilation were recorded twice daily, and the highest values were used for analysis. For those who deteriorated during non-ventilator treatment with obvious hypoxemia and respiratory distress, endotracheal intubation and mechanical ventilation were implemented. Ventilation modes were variable. The oxygenation index [OI = FiO2 × MAP (cmH2O) × 100/PaO2 (mmHg)] and ventilation index [VI = PIP (cmH2O) × ventilator frequency (cycles/min) × PaCO2 (mmHg)/1,000] were used for estimation of the severity of lung disease and the treatment response. Fluid intake or intravenous infusion was recommended initially at 70% maintenance in the first 3 days of ARDS, with additional fluids as needed for cardiac or nutritional support. For those died of respiratory failure, at least one blood gas was analyzed within 24 h before death.

Underlying diseases, complications and outcome assessment

Underlying diseases of ARDS were categorized according to routine diagnostic definitions in pediatrics. Sepsis and organ dysfunction were identified by using criteria of Levy [10] and the cut-off values of heart and respiratory rate were similar to the definitions for children from Proulx et al. [11]. Respiratory failure was defined as PaO2 ≤50 mmHg in room air at sea level or PaO2/FiO2 ≤250 mmHg under oxygen administration and in the absence of cyanotic congenital heart disease. Survivors were defined as alive at 7 days after discharge from hospitals and verified by telephone contact. Non-survivors included those who died in the hospitals or within 7 days after discharge from the hospital.

Statistical analysis

A simple tabulation was used to describe the data presented as means and SD (when normally distributed) or median and range (when not normally distributed) for continuous variables or incidence ratio in percentage and 95% confidence interval (CI) for categorical variables. Depending on the distribution of data, chi-square test and Fisher’s exact test, Wilcoxon–Mann–Whitney test, or Student’s t-test were used to make comparison of outcomes according to variables of interest and to test significance at alpha error level of 0.05.


Demographic characteristics of PICU admissions

Of the 12,018 patients admitted to the PICU during the 12-month period, 7,269 cases were severely ill and eligible for the study. Among them, 5,590 were medical cases and the others were surgical ones. Average duration of stay in the PICU was 6.6 ± 3.2 days. Four hundred and eighty five (6.7%, 95% CI 6.1–7.3%) of the severely ill patients died in hospital or within 7 days after discharge from hospital.

ARDS patients

Of 7,269 severely ill cases, 105 (1.44%, 95% CI 1.18–1.75%) were diagnosed as having ARDS. Major clinical characteristics of the ARDS patients are presented in Table 1. At the onset of ARDS 42 (40%) were less than 12 months old. In all ARDS patients, 31 (29.5%) had diffuse bilateral infiltrates or white lung on chest X-ray. None of the 47 (44.8%) patients who underwent an echocardiographic examination had elevated left atrium pressure. Thirty-four patients (33.3%) had co-existing clinical conditions (Table 2).

Table 1 Characteristics of pediatric patients at the onset of ARDS
Table 2 The co-existing clinical conditions in children with ARDS

Time and risk factors of ARDS onset

The median (interquartile range) and 95th percentile time from the onset of underlying disease to the onset of ARDS were 72 (24, 120) and 168 h, respectively. The cumulative incidences of ARDS for patients with pneumonia, non-pulmonary sepsis and other disorders are shown in Fig. 1.

Fig. 1
figure 1

Time to onset of ARDS from point of clinical risk identification for patients with sepsis, pneumonia, and other risk factors, expressed as cumulative percentage of patients ultimately developing ARDS

Predisposing illnesses associated with the development of ARDS are presented in Table 3. One of the major clinical risk factors for the development of ARDS was identified as predominant for each of the 105 ARDS patients. The major etiologies of ARDS were pneumonia and non-pulmonary sepsis (Table 4), and 62 (59%) patients had ARDS of pulmonary origin.

Table 3 Distribution of predisposing factors in incident cases and deaths of children with ARDS
Table 4 Relative risk of death of children with ARDS caused by pneumonia and sepsis

Respiratory support of ARDS

At the time of ARDS diagnosis, 79 (75.2%) patients were receiving supplemental oxygen through mechanical ventilation, 5 (4.8%) via nasal continuous positive airway pressure, 7 (6.7%) with face mask, 8 (7.6%) with head mask and 6 (5.7%) by nasal cannula. Forty (38.1%) received 100% (FiO2 = 1.0) supplemental oxygen at the time of ARDS diagnosis. PaO2/FiO2 was 95.7 ± 46.3 mmHg (range 14.4–197.5 mmHg) at the onset of ARDS (Table 1). Of 26 non-ventilated patients at inclusion, 18 eventually required intratracheal intubation and mechanical ventilation. In those patients not intubated and ventilated, four were because the attending physicians had not adequately judged the severity of illness, and in the other four palliative therapy was requested by the parents in view of underlying malignant diseases.

Ninety-seven patients received mechanical ventilation, including seven patients with noninvasive mechanical ventilation. The most commonly used ventilatory modes were pressure control (n = 60), synchronized intermittent mandatory ventilation (n = 20), pressure regulated volume control (n = 5), volume control (n = 4), bi-level positive airway pressure (n = 4), nasal continuous positive airway pressure (n = 3) and high frequency oscillation (n = 1). The measurements of lung mechanics at onset of invasive mechanical ventilation are listed in Table 1. The mean (SD) values of the highest PIP, PEEP, MAP and V T during the first 3 days of ventilation were 32.6 ± 10.1 cmH2O, 10.8 ± 6.8 cmH2O, 18.2 ± 6.5 cmH2O and 10.4 ± 5.4 ml/kg, respectively. During this time, approximately 23–28% patients received V T of 6–8 ml/kg, 20–30% below 6 ml/kg, and 40–50% >8 ml/kg. Approximately 30–45% patients were using PEEP levels of ≤5 cmH2O, while 60–80% had PIP ≤30 cmH2O. No statistically significant difference in ventilation parameters was found between survivors and non-survivors.


Overall, 485 (6.7%) of PICU patients admitted to the study, died in hospital or within 7 days after discharge. The non-survivors (n = 64) of ARDS represented 13.2% of all non-survivors in these PICUs, i.e., ARDS patients had ten times higher relative risk of death than non-ARDS patients. No difference in mortality rate was found between the university teaching hospitals (37/64, 57.8%) and the non-teaching hospitals (27/41, 65.9%, χ 2 = 0.679, p = 0.410), or between the municipal/provincial hospitals (53/89, 59.6%) and prefectural/regional hospitals (11/16, 68.8%) (χ 2 = 0.482, p = 0.487). Of 64 ARDS non-survivors, 23 (35.9%) died in the first 24 h after the onset of ARDS. Seven (30.4%) of these died within 24 h without mechanical ventilation. No difference in age specific mortality was found (χ 2 = 4.206, p = 0.389). Mortality varied from 79.2% among those with non-pulmonary sepsis, to 60.3% with pneumonia. As shown in Table 4, the relative risks of death in patients with pneumonia or non-pulmonary sepsis were 14.4 and 5.1 times higher, respectively, in those with ARDS.


This study is the first to report on the incidence of ARDS in PICUs in China. There are approximately 50 PICUs in China. In the coastal regions, PICUs have been established in major cities for populations of 5–15 million, while in the mid-west regions each province has at least one PICU in the provincial children’s hospital. The participating centers in this study are representative of PICUs in the whole country, with eight of them being national university hospitals, and six of them provincial children’s hospitals of prosperous regions.

We have documented the incidence of, predisposing factors for, mortality and its relative risks from pediatric ARDS in China according to the 1994 AECC definition of ARDS. Although this definition was developed for both adult and pediatric patients, few multicenter, prospective, epidemiologic studies using this definition have been reported in PICU-based patient populations.

Approximately 40% of all patients admitted to PICU during the study periods were found to have relatively mild levels of illness. Among those who were regarded as severely ill, and comparable to those admitted to American PICUs, we found an average incidence of ARDS of 1.4%. This figure is compatible with many of the published reports of ARDS in PICU with incidences varying from 0.7 to 4.2% [1215]; but most of these studies represent single site and retrospective descriptions of PICU admissions. In a 3-month survey in 92/94 ICUs in Germany, only 12 children with ARDS were identified [4], while Flori et al. [3] recruited 221 ARDS patients in 2 American PICUs over a 4-year period. While 1.9% of PICU admissions in New Zealand and Australia during a 1-year period had ALI [16].

The 61% mortality reported in this study is higher than that reported in 2004 in a 2-month prospective study in 36 PICUs from 7 countries [17]. In that study, there was a mortality of 50% in pediatric ARDS patients receiving mechanical ventilation, which contrasts with the mortality of 26% in the study of Flori et al. [3], but this study included non-ventilated patients. Recently Erickson et al. [16] reported a mortality of 38% in patients who received ventilatory support for ALI in New Zealand and Australia. The wide variation in mortality could be related to many factors including: infrastructure for referral, admission and discharge policy in PICU, severity of illness, patient mix, quality of patient care, resource availability, etc.

The underlying disease pattern may differ in pediatric and adult patients who develop ARDS. In adult ARDS patients, sepsis, pneumonia, trauma, and aspiration of gastric contents are the most common underlying conditions [5, 1820]. In children with ARDS, pneumonia, sepsis, intoxication and asphyxia are the major causes [14, 21]. In general, respiratory and infectious diseases are the two major causes for hospital and ICU admission in infants and children [3]. Tables 3 and 4 illustrate significantly higher risks of mortality for those with ARDS due to pneumonia or sepsis than those without ARDS. Subsequently, these patients had 14 and 5 times higher relative risk of mortality, respectively, when compared to those without ARDS. In this study, only 17 patients had malignant diseases. A limitation of the study is that we did not look at age-stratified death risks compared to respective admissions and disease pattern. Our data suggest that age, underlying diseases, early and adequate management, co-existing clinical conditions such as immuno-compromised subgroup, may be related to different risk of mortality in ARDS.

The diagnostic criteria for pediatric ARDS based on the 1994 AECC definition have only recently been applied in prospective, multicenter clinical trials using surfactant, inhaled nitric oxide, and high frequency oscillation [2226]. The AECC definitions were also used in clinical epidemiologic studies based on PICU admission [13], or regional population profile [4]. Compared to these reports, a limitation of this study is the lack of systematic evaluation of mechanical ventilation. Respiratory care, mechanical ventilation and supportive management were not strictly controlled.

The high mortality within the first 24 h of onset of ARDS may reflect inadequate intervention in the PICU, however, the details of therapy are not fully available. Some deaths occurred because of a conscious decision that it was not appropriate in the context to escalate therapy. A significant proportion of patients received ventilatory patterns (low PEEP, high V T and high PIP) that have been associated with poor outcomes.

In clinical trials of ALI and ARDS in PICU patients [2226], mortality has ranged from 19–60% in both intervention and control arms. In clinical epidemiologic studies [1, 3], the mortality of ARDS or severe hypoxemic respiratory failure was <30% although in general mortality was higher than in clinical trials [17]. Using a multivariate model, Flori et al. [3] identified three major risk factors that were independent predictors of death and prolonged ventilation: non-pulmonary organ system dysfunction, central nerves system dysfunction and severe hypoxemia at early onset of ARDS, i.e., ALI phase, irrespective of whether mechanical ventilation is required. It appears that differences between our study and that of Flori’s are at least partly due to differences in study design and patient population (e.g., single or multiple centers, observational time span, ALI and/or ARDS), and social, cultural and economic background [14, 15, 21, 27].

The effective lung protective ventilation strategies in ARDS in adults include small tidal volume mechanical ventilation [1820], non-invasive positive pressure ventilation, prone position [28, 29], and possibly therapies such as exogenous surfactant [22, 23], inhaled nitric oxide [29], high frequency oscillation ventilation [30], and extracorporeal membrane oxygenation [31]. Some of these therapies have been effective in neonates and infants but still require more studies in pediatrics.

During this study, the participating centers had various levels of modern respiratory care facilities, and attending staff steadily improved skills as part of the collaboration. It is possible that uniform implementation of evidence-based protocols for lung protective ventilation could improve patient outcomes with ARDS, and this may warrant an interventional study.

Factors in developing countries affecting the etiology, prognosis and outcome for pediatric ARDS vary from medical resource to social and economic conditions. Organization of collaborative studies in developing countries may have the potential to improve mortality from respiratory diseases in infants and children; the low mortality of ARDS in the developed world is associated with high costs and possibly efficient care. Implementation of advanced respiratory therapies and standardization of care for pediatric ARDS is a challenging and a long-term task for PICU professionals in developing countries. This collaborative study will enable the development of both infrastructure in pediatric respiratory and critical care, and structures to allow for in depth investigation of ARDS and other challenging respiratory diseases in this context.