American Journal of Respiratory and Critical Care Medicine

Rationale: Estimates of the incidence of the acute respiratory distress syndrome (ARDS) in high- and middle-income countries vary from 10.1 to 86.2 per 100,000 person-years in the general population. The epidemiology of ARDS has not been reported for a low-income country at the level of the population, hospital, or intensive care unit (ICU). The Berlin definition may not allow identification of ARDS in resource-constrained settings.

Objectives: To estimate the incidence and outcomes of ARDS at a Rwandan referral hospital using the Kigali modification of the Berlin definition: without requirement for positive end-expiratory pressure, hypoxia cutoff of SpO2/FiO2 less than or equal to 315, and bilateral opacities on lung ultrasound or chest radiograph.

Methods: We screened every adult patient for hypoxia at a public referral hospital in Rwanda for 6 weeks. For every patient with hypoxia, we collected data on demographics and ARDS risk factors, performed lung ultrasonography, and evaluated chest radiography when available.

Measurements and Main Results: Forty-two (4.0%) of 1,046 hospital admissions met criteria for ARDS. Using various prespecified cutoffs for the SpO2/FiO2 ratio resulted in almost identical hospital incidence values. Median age for patients with ARDS was 37 years, and infection was the most common risk factor (44.1%). Only 30.9% of patients with ARDS were admitted to an ICU, and hospital mortality was 50.0%. Using traditional Berlin criteria, no patients would have met criteria for ARDS.

Conclusions: ARDS seems to be a common and fatal syndrome in a hospital in Rwanda, with few patients admitted to an ICU. The Berlin definition is likely to underestimate the impact of ARDS in low-income countries, where resources to meet the definition requirements are lacking. Although the Kigali modification requires validation before widespread use, we hope this study stimulates further work in refining an ARDS definition that can be consistently used in all settings.

Scientific Knowledge on the Subject

The acute respiratory distress syndrome (ARDS) is a common and lethal condition. Although the Berlin definition is a robust and reproducible tool for identifying ARDS, it cannot be applied in low-income countries and may result in estimates of only “treated incidence” in high-income countries.

What This Study Adds to the Field

We propose an alternative definition of ARDS using criteria that have been individually validated. We then apply the alternative definition to estimate the incidence and outcomes of ARDS in a Rwandan referral hospital. We find an estimated ARDS incidence of 4.0% of hospital admissions and a mortality rate of 50.0%. The capacity to identify ARDS is critical to ensuring that treatment approaches are developed and applied in both resource-rich and resource-constrained areas of the world.

The acute respiratory distress syndrome (ARDS) is a syndrome of acute hypoxia and bilateral pulmonary infiltrates not attributed to cardiac failure and associated with pulmonary and nonpulmonary insults (1). Estimates of hospital-based incidence of moderate to severe ARDS vary from 1.6 to 7.7% of all intensive care unit (ICU) admissions and 8 to 19.7% of all ventilated patients (26). Estimates of the population-based incidence of ARDS vary from 10.1 to 86.2 cases per 100,000 person-years (3, 5, 710), with the most rigorous study suggesting the highest incidence (10). Incidence and outcomes of ARDS have not been reported for a low-income country (11) at the level of the population, hospital, or ICU.

Although the Berlin definition has been developed with resource-rich settings in mind, it may not allow identification of patients with ARDS pathophysiology in resource-constrained settings because of inaccessibility of mechanical ventilators, arterial blood gas diagnostics, and chest radiography (12, 13). This limitation could have the unintended consequence of underestimating and undertreating the burden of ARDS in many countries (14). It also means that even high-income country estimates are limited to “treated incidence,” excluding patients outside the ICU who may otherwise meet the criteria (1416). The necessity for positive end-expiratory pressure (PEEP) in the definition of ARDS remains controversial (14, 17), a validated estimate of PaO2/FiO2 can be obtained from SpO2/FiO2 using pulse oximetry (18, 19), and ultrasound has been validated as a substitute for chest radiograph in ARDS (20).

To determine the hospital incidence of ARDS in a low-income country, we applied modified criteria, aiming to approximate the Berlin criteria, without a requirement for PEEP, hypoxia cutoff defined by SpO2/FiO2 less than or equal to 315, and use of either lung ultrasound or chest radiograph for the determination of bilateral opacities (Table 1). We applied this definition to determine the incidence and outcomes of ARDS in a referral hospital in Kigali, Rwanda.

Table 1. Berlin Criteria for ARDS, Challenges in Resource-Poor Settings, and Kigali Modification of the Berlin Criteria to Address These Challenges

 Berlin CriteriaChallenges in Resource Poor SettingsKigali Modification of the Berlin Criteria
TimingWithin 1 wk of a known clinical insult or new or worsening respiratory symptomsNoneWithin 1 wk of a known clinical insult or new or worsening respiratory symptoms
OxygenationPaO2/FiO2 ≤300Scarcity of arterial blood gas diagnosticsSpO2/FiO2 ≤315
PEEP requirementMinimum 5 cm H2O PEEP required by invasive mechanical ventilation (noninvasive acceptable for mild ARDS)Scarcity of mechanical ventilatorsNo PEEP requirement, consistent with AECC definition
Chest imagingBilateral opacities not fully explained by effusions, lobar/lung collapse, or nodules by chest radiograph or CTScarcity of chest radiography resourcesBilateral opacities not fully explained by effusions, lobar/lung collapse, or nodules by chest radiograph or ultrasound
Origin of edemaRespiratory failure not fully explained by cardiac failure or fluid overload (need objective assessment, such as echocardiography, to exclude hydrostatic edema if no risk factor present)NoneRespiratory failure not fully explained by cardiac failure or fluid overload (need objective assessment, such as echocardiography, to exclude hydrostatic edema if no risk factor present)

Definition of abbreviations: AECC = American-European Consensus Conference; ARDS = acute respiratory distress syndrome; CT = computed tomography; PEEP = positive end-expiratory pressure; SpO2 = oxygen saturation as measured by pulse oximetry.

Study Oversight

The study was conducted at the University Teaching Hospital of Kigali, which is affiliated with the University of Rwanda College of Medicine and Health Sciences. The scientific committee of the University of Rwanda and the ethical committee of the University Teaching Hospital of Kigali approved the study, as did the Committee on Clinical Investigations at Beth Israel Deaconess Medical Center in Boston, Massachusetts. Requirement for individual patient-level consent was waived given the determination of minimal risk to patients.

Study Design and Setting

We performed a prospective observational cohort study at the University Teaching Hospital of Kigali. The primary outcome was incidence of ARDS among all hospitalized patients. We also described the incidence of hypoxia among all hospitalized patients, and sex distribution, median age, predisposing clinical insult frequency, hospital ward frequency, time to development of ARDS, and mortality rate for patients with ARDS. Finally, we performed sensitivity analyses to test the impact of using different definition requirements and cutoffs on the hospital incidence of ARDS.

The study hospital is one of two public referral hospitals in a country of 11 million people, with 560 total beds, approximately 12,000 admissions per year, and an average of 260 adult patients present on any given day. The study hospital receives about 60% of all referrals from the public district and provincial hospitals in the country, with the remaining 40% seen at the other public referral hospital.

The study hospital has a six-bed ICU with six mechanical ventilators available. About 85% of all ICU patients are ventilated at some time during their ICU stay, and 73% are admitted to the ICU specifically to receive mechanical ventilation for respiratory failure. Noninvasive ventilation is available only via masks attached to the existing six ventilators, and it is rarely used. The ICU has intermittent access to central venous catheters but not arterial catheters. Vasopressors are available, and were used in about 40% of ICU patients at the time of this study. Hemodialysis was not available in the hospital at the time of the study. Sick patients in this hospital are often managed in the emergency room for extended periods, and one mechanical ventilator is available in the emergency room. In-hospital mortality for ICU patients in this hospital is about 50%.

Data Collection and Quality Control

We screened every adult patient (age >15 yr old) in the hospital for 6 weeks (February 25 to April 4, 2014) for hypoxia every day using a Lifebox AH-MX pulse oximeter (Acare Technology Company, Ltd., New Taipei City, Taiwan) (21, 22). We used this cutoff for adulthood in keeping with prior ARDS studies (3, 7, 8, 10); it is also the age cutoff used in the hospital to distinguish between the adult and pediatric wards. We checked the oxygen saturation of each adult in the hospital once daily; the research assistants moved through the wards in a similar pattern each day, but the time an individual patient’s saturation was checked could vary from day to day. Clinical nursing staff were informed when a patient was found to have a saturation less than 90%. For every patient with either an oxygen saturation less than 90% or receiving supplemental oxygen, we collected data on demographics, ARDS risk factors, and hospital survival.

We also recorded the oxygen saturation, oxygen delivery device if any, and oxygen flow rate or fraction of inspired oxygen and PEEP if mechanically ventilated. We then performed ultrasound on every patient with hypoxia once daily; again, the time of ultrasound for an individual patient could vary from day to day. Each day, the chart and clinical team of each patient with hypoxia were queried for whether a chest radiograph had been performed in the previous 24 hours. If so, it was evaluated by the research team for bilateral opacities. Patients received chest radiographs only when the radiograph machine was functional, the patient could afford it, and the clinical team deemed it necessary. We did not perform research-specific chest radiographs, because this would have overwhelmed the capacity of the radiology technologist staff of the hospital.

Demographic data, ARDS risk factors (pulmonary or nonpulmonary infection, trauma, surgery, and stroke), and survival data were obtained from patient charts. Exclusion of cardiac failure was based on chart review including echocardiogram results when available. Chest radiograph interpretation and chest ultrasound were performed by three research assistants, all with previous experience in performing these tasks. These assistants were given additional training using previously developed training material (20, 2325; LIPS-A [Lung Injury Prevention Study with Aspirin], personal communication), supervised practice at the bedside, and evaluation at the bedside before study initiation. Ultrasound examination and radiograph results were independently reviewed by a study investigator (E.D.R.) in a random subset of cases within 6 hours of initial evaluation; a subset of ultrasound and radiograph images were also reviewed by an off-site investigator (D.S.T.).

Ultrasound was performed over six areas on each side of the chest (two anterior, two lateral, and two posterolateral) with the patient in a semirecumbent position, using the technique outlined by Lichtenstein and Mezière (25). Ultrasound was performed using two identical SonoSite M-Turbo machines, with curvilinear probes (FUJIFILM SonoSite, Inc., Bothell, WA). Although ultrasound procedures included systematic review of 12 areas of the chest and chest radiographs were interpreted based on examination of four quadrants, agreement for quality assurance was based only on the final interpretation of presence or absence of bilateral opacities.

The total number of admissions to the hospital in the initial 4 weeks of the study was determined from hospital admission logs in the emergency room, obstetrics ward, and direct-admit outpatient logs. Data were entered first on paper forms, then into a secure, web-based application, the Research Electronic Data Capture electronic data capture tool (26).


We proposed the Kigali modification to the Berlin definition, which contains the following changes to the Berlin definition: lack of a PEEP requirement, oxygenation cutoff of SpO2/FiO2 less than or equal to 315, and use of lung ultrasound or chest radiograph for the determination of bilateral opacities (Table 1). Each of these modifications addresses a criterion that cannot be consistently assessed in resource-constrained settings. We did not require a minimum PEEP of 5 cm H2O, in keeping with the prior American-European Consensus Conference (AECC) criteria for ARDS (17). We used SpO2/FiO2 less than or equal to 315 (with a requirement of SpO2 ≤ 97%) for the hypoxia cutoff, using results from Rice and coworkers (19) that established this SpO2/FiO2 ratio to correspond to a PaO2/FiO2 ratio of 300. We estimated FiO2 from all oxygen delivery devices using the conversion table from the Extended Prevalence of Infection in Intensive Care (EPIC II) study (27).

We allowed bilateral opacities to be defined by either chest radiograph or ultrasound, based on a study that found ultrasound to be more accurate than chest radiograph when compared with computed tomography among patients with ARDS and control subjects (20). Ultrasound findings of “B lines” and/or consolidation without associated effusion, found in at least one area on each side of the chest, were defined as consistent with bilateral opacities. B lines are an artifact consistent with alveolar-interstitial filling, showing more than two vertical lines arising from the pleural line and extending to the screen edge, whereas alveolar consolidation is seen as hypoechoic areas resembling tissue (e.g., the liver) and sometimes with hyperechoic punctiform lesions corresponding to air bronchograms (see Figure E1 in the online supplement) (20). We did not include consolidations seen in association with pleural effusions because these could represent atelectasis caused by the effusion. Chest radiographs were labeled as bilateral opacities when opacities were found in a minimum of two quadrants, at least one on each side of the chest.

We performed sensitivity analyses to test the impact of varying definitions and cutoffs on the hospital incidence of ARDS. We compared outcomes using the Kigali modification of the Berlin definition with those using the Berlin definition itself, and partial modifications of the Berlin definition in which one of our three modifications was excluded (i.e., true PaO2/FiO2 ratio required, chest radiograph opacities required, or PEEP required). We also used alternative cutoffs for the PaO2/FiO2 ratio: estimated PaO2/FiO2 ratio less than or equal to 300 (unadjusted Berlin definition using estimated SpO2 from conversion tables from the EPIC II study [27]); estimated PaO2/FiO2 ratio less than or equal to 251 with the same estimates but adjusted for the altitude of Kigali (Kigali barometric pressure of 636 mm Hg was divided by sea level barometric pressure of 760 mm Hg and multiplied by the 300 cutoff); and SpO2/FiO2 less than or equal to 370 based on another study that estimated the correlation between PaO2/FiO2 ratio and SpO2/FiO2 (18). We also used an alternate method of estimation of FiO2 from oxygen flow from a recent study (28) (FiO2 = 0.21 + 0.03 × [oxygen flow in liters/minute]), which resulted in lower estimated FiO2 values than the EPIC II study conversion table. In each of these analyses, we excluded cases with an SpO2 greater than 97% because the hemoglobin oxygenation curve flattens above 97% (19).

Statistical Analysis

Characteristics and outcomes of patients with ARDS were described for all patients meeting criteria for the Kigali modification of the Berlin definition at any time during the 6-week screening period. Results for demographic, clinical, and outcome variables are presented as percentages for categorical variables and as medians with interquartile ranges (IQRs) for continuous variables.

Incidence calculations were performed by taking the number of patients who met criteria for ARDS and who were admitted between February 25 and March 24, 2014, divided by the total adult hospital admissions during that month. We did not include patients admitted before or after this 4-week period to calculate an incidence among admissions rather than a prevalence (see Figure E2). The rationale for doing 6 weeks of screening to capture a 4-week incidence rate was that the additional 2 weeks of screening provided an opportunity to capture patients who were admitted during the incidence window but developed ARDS after the initial 4-week window.

Quality control was assessed by κ statistic for interobserver agreement for chest radiograph and ultrasound assessments for those patients for whom two independent assessments were made. κ agreement greater than or equal to 0.70 roughly indicates adequate interrater agreement (29). Sensitivity analyses using the above-noted cutoffs for the PaO2/FiO2 ratio and different combinations of Berlin and modifications of the Berlin definition were performed. All analyses were performed using SAS version 9.3 (SAS Institute, Inc., Cary, NC).

A total of 1,046 adults were admitted to the hospital in the first 4-week period of the study. Of these, 126 (12.0%) were hypoxic on at least 1 day, with mortality of 49.2% (Figure 1; see Table E1). The criteria for the Kigali modification of the Berlin definition for ARDS was met by 42 (4.0%), with mortality rate for these patients of 50.0% (Table 2). Using the full Berlin definition, no patients qualified as having ARDS. Using two of the Kigali modifications, but requiring chest radiograph as in the original Berlin definition, resulted in a hospital incidence of 1.6%, because less than half of all patients with hypoxia in the study ever had a chest radiograph. Using two of the Kigali modifications but requiring a minimum PEEP of 5 cm H2O as in the original Berlin definition, resulted in a hospital incidence of 1.0% because less than a third of patients with ARDS ever received mechanical ventilation. Using two of the Kigali modifications but requiring a true PaO2/FiO2 ratio less than or equal to 300 resulted in a hospital incidence of zero because arterial blood gases were not available during the study.

Table 2. Hospital Incidence and Mortality of ARDS by the Berlin and Kigali Modification of the Berlin Criteria, with Sensitivity Analyses

DiagnosisIncident Cases [n (%)]Mortality [n (%)]
Incident ARDS by definition
 ARDS by Berlin criteria0 (0)
 ARDS by Kigali modification of the Berlin criteria42 (4.0)21 (50.0)
  Modified definition but require chest radiograph17 (1.6)10 (58.8)
  Modified definition but require minimum PEEP10 (1.0)6 (60.0)
  Modified definition but require PaO2 from arterial blood gas*0 (0)
 ARDS by Kigali modification of the Berlin criteria with alternate hypoxia cutoffs  
  Estimated PaO2/FiO2 ≤30043 (4.1)21 (48.8)
  Estimated PaO2/FiO2 ≤25140 (3.8)21 (52.5)
  SpO2/FiO2 ≤37043 (4.1)21 (48.8)
 ARDS by Kigali modification of the Berlin criteria with alternate FiO2 estimation (FiO2 = 0.21 + 0.03 × [oxygen flow in liters/minute])39 (3.7)20 (51.3)

Definition of abbreviations: ARDS = acute respiratory distress syndrome; PEEP = positive end expiratory pressure; SpO2 = oxygen saturation as measured by pulse oximetry.

ARDS defined by the Kigali modification of the Berlin criteria is as noted in the fourth column of Table 1. Modifications are: SpO2/FiO2 ≤315, no minimum PEEP requirement, and use of chest radiograph or ultrasound to determine bilateral opacities.

*Arterial blood gas measurements were not available for the duration of the study.

Using different cutoffs for the estimated PaO2/FiO2 ratio produced very similar results. Estimating PaO2 from SpO2 and using the Berlin definition of PaO2/FiO2 less than or equal to 300 resulted in an incidence of 4.1%. Modifying this estimated ratio for the altitude of Kigali, Rwanda (estimated PaO2/FiO2 ≤251) resulted in an incidence of 3.8%. A cutoff of SpO2/FiO2 less than or equal to 370 resulted in an incidence of 4.1%. Using a standard formula to estimate FiO2 rather than the EPIC II conversion table yielded an incidence of 3.7%.

In 6 weeks of screening, 219 patients were hypoxic on at least one day of screening, and 68 patients met the Kigali modification of the Berlin criteria (Table 3). Thirty-seven (54.4%) were male. Median age was 37 (IQR, 26–49) years. The most common underlying diagnosis was infection (44.1%), with over half of these being pulmonary. Trauma (29.4%) and surgery (25.0%) were the next most common diagnoses. Half of patients were diagnosed with ARDS in the emergency room, and only 30.9% were ever admitted to the ICU. Median time to the development of ARDS from admission was 0 (IQR, 0–2) days. For the 21 patients with ARDS who went to the ICU, median ICU length of stay was 5 (IQR, 2–10) days. Median hospital length of stay was 19 (IQR, 13–37) days for survivors and 6 (IQR, 5–15) for nonsurvivors.

Table 3. Demographic and Hospital Stay Characteristics for Patients with ARDS

 SurvivorsNonsurvivorsAll Patients with ARDS
Number of patients, n303868
 Male, n (%)15 (50.0)22 (57.9)37 (54.4)
 Age, yr, median (IQR)28 (21–40)44 (32–60)37 (26–49)
 Home outside the referral hospital province, n (%)16 (53.3)26 (68.4)42 (61.8)
Oxygenation ratios
 SpO2/FiO2, median (IQR)245 (163–278)184 (101–258)233 (137–272)
 Estimated PaO2/FiO2, median (IQR)203 (143–288)158 (81–223)187 (118–267)
Clinical insults predisposing to ARDS, nonexclusive categories, n (%)
 Infection12 (40.0)18 (47.4)30 (44.1)
  Pulmonary8 (66.7)9 (50.0)17 (56.7)
  Nonpulmonary4 (33.3)8 (44.4)12 (40.0)
  Unknown0 (0)1 (5.6)1 (3.3)
 Trauma9 (30.0)11 (29.0)20 (29.4)
 Surgery5 (16.7)12 (31.6)17 (25.0)
 Stroke0 (0)3 (7.9)3 (4.4)
 Other5 (16.7)4 (10.5)9 (13.2)
Location at time of diagnosis, n (%), n = 56*
 Emergency room13 (50.0)15 (50.0)28 (50.0)
 ICU4 (15.4)9 (30.0)13 (23.2)
 Obstetric ward or recovery4 (15.4)1 (3.3)5 (8.9)
 Medical ward2 (7.7)1 (3.3)3 (5.4)
 Isolation (TB ward)2 (7.7)1 (3.3)3 (5.4)
 Surgical ward or postoperative recovery1 (3.9)1 (3.3)2 (3.6)
 Neurology ward0 (0)1 (3.3)1 (1.8)
 Private ward0 (0)1 (3.3)1 (1.8)
Patient days spent in each location, n (%)
 Emergency room96 (20.9)43 (19.6)139 (20.5)
 ICU70 (15.2)59 (26.9)129 (19.0)
 Surgical ward or postoperative recovery81 (17.6)38 (17.3)119 (17.6)
 Medical ward78 (17.0)31 (14.2)109 (16.1)
 Neurology/neurosurgery ward33 (7.2)35 (16.0)68 (10.0)
 Isolation (TB ward)58 (12.6)2 (1.0)60 (8.8)
 Obstetric ward or recovery43 (9.4)5 (2.3)48 (7.1)
 Private ward0 (0)6 (2.7)6 (0.9)
In ICU at any time during hospitalization6 (20.0)15 (39.5)21 (30.9)
 ICU LOS, d, median (IQR)12 (5–13)4 (2–6)5 (2–10)
Received mechanical ventilation7 (23.3)14 (36.8)21 (30.9)
 Time on mechanical ventilation, d, median (IQR)6 (2–9)4 (2–5)4 (2–6)
Time from admission to development of ARDS, d,  median (IQR), n = 56*1 (0–2)0 (0–1)0 (0–2)
Hospital LOS, d, median (IQR)19 (13–37)6 (5–15)11 (6–20)
Primary diagnoses/causes of death
 Trauma7 (23.3)12 (31.6)19 (27.9)
 Infectious diseases4 (13.3)7 (18.4)11 (16.2)
 Gastrointestinal system3 (10.0)6 (15.8)9 (13.2)
 Respiratory3 (10.0)4 (10.5)7 (10.3)
 Neoplasms2 (6.7)3 (7.9)5 (7.4)
 Cardiovascular2 (6.7)2 (5.3)4 (5.9)
 Pregnancy/peripartum2 (6.7)2 (5.3)4 (5.9)
 Other or unknown7 (23.3)2 (5.3)9 (13.2)

Definition of abbreviations: ARDS = acute respiratory distress syndrome; ICU = intensive care unit; IQR = interquartile range; LOS = length of stay; SpO2 = oxygen saturation as measured by pulse oximetry; TB = tuberculosis.

Median PaO2/FiO2 and SpO2/FiO2 ratios use the worst ratio for each patient.

ARDS defined by the Kigali modification of the Berlin criteria is as noted in the fourth column of Table 1.

*Location at time of diagnosis and time from admission to development of ARDS include only patients with hospital admission dates during the study period. Patients admitted to the hospital before the study period and found to have ARDS during the study were excluded because they could have developed ARDS before the study’s beginning.

Patient days spent in each location include all hospital days for patients with ARDS, including days when they did not meet criteria for ARDS.

Interobserver agreement for the random sample of 54 of 1,069 (5.1%) ultrasounds independently performed by two operators was 96.3% (bilateral B lines or consolidation without effusion vs. unilateral or none) with κ statistic 0.92 (95% confidence interval, 0.82–1.00). For the 58 of 165 chest radiographs (35.2%) reviewed by two independent readers, agreement for bilateral opacities was found in 87.9% of cases, κ statistic 0.73 (95% confidence interval, 0.55–0.92).

Our study is the first to estimate the incidence and outcomes of ARDS in a low-income country (11). Our data suggest ARDS is a common and fatal condition in this setting, affecting 4% of all admissions to a Rwandan referral hospital, with mortality of 50%. The current Berlin definition would overlook all of these cases of ARDS because of a requirement for PEEP, arterial blood gas measurements, and chest radiographs.

Compared with prior studies in high-income countries, Rwandan patients were younger, with median age 37 (IQR, 26–49) years versus 62 (IQR, 48–75) years in one study in King County, Washington (10). The identified etiologies underlying ARDS also may be different in low-income settings in comparison with North American populations; the same King County study demonstrated a higher proportion of infection and a lower proportion of trauma than our study (10). ARDS-related mortality in this Rwandan referral hospital also seems to be higher than among studies predominantly from high-income countries (30). Despite being a much younger population, patients with ARDS in Rwanda are much less frequently admitted to fewer ICU beds, and have poorer outcomes.

Somewhat surprisingly, we found mortality to be similar among patients with ARDS and all patients with hypoxia (see Table E1). This is, however, similar to preliminary findings from the global LUNG SAFE study of 3,595 patients across hundreds of ICUs, in which mortality was almost identical between the patients with non-ARDS severe acute respiratory failure and patients with mild or moderate ARDS (31). We also found it initially surprising that almost 50% of the 88 patients with SpO2/FiO2 less than or equal to 315 and oxygen saturation less than or equal to 97% met criteria for ARDS (Figure 1); however, this finding is consistent with a recent study of high-flow oxygen that found that 79% of all patients with PaO2/FiO2 ratio less than or equal to 300 had bilateral infiltrates on chest radiograph (28) and the LUNG SAFE study, which found that 61% of patients with severe acute respiratory failure met full criteria for ARDS (31).

Our study has several limitations. First, we created and used the Kigali modification of the Berlin definition. Each of the three pieces of the modification has been validated previously, but the modification as a whole has not been validated. The first adaptation is the removal of the requirement for a PEEP level of at least 5 cm H2O, which could overestimate ARDS in patients whose oxygenation would improve to a PaO2/FiO2 greater than 300 with PEEP. Both the AECC panel in 1994 and the Berlin panel in 2012 recognized that PEEP has a significant impact on the PaO2/FiO2 ratio but also that the size of its impact in a given patient is unpredictable and that requiring PEEP excludes some patients in places with fewer mechanical ventilators (13, 14, 17, 32). Given these considerations, AECC chose to exclude a minimum PEEP requirement and Berlin chose to include it. Although many intensivists would intuitively consider ARDS to exist only in patients requiring mechanical ventilation with PEEP, it is nonetheless controversial whether a PEEP of 5 cm H2O should be required to meet ARDS criteria. The second adaptation is the use of an estimation of the PaO2/FiO2 ratio based on a prior study that compared PaO2/FiO2 with SpO2/FiO2 ratios (19). That study excluded patients from sites at high altitude, and our study site is at 1,500 m of altitude. It also included only patients with PEEP greater than or equal to 5 cm H2O. However, our hospital incidence estimates were consistent across several plausible cutoffs of the PaO2/FiO2 ratio. Similarly, we estimated FiO2 from oxygen flow rate for patients not receiving mechanical ventilation, but were reassured that results were similar using both more liberal and conservative estimates (27, 28). The final adaptation is to allow either ultrasound or chest radiograph to determine the presence of bilateral opacities. One study of patients with ARDS found accuracy in diagnosing alveolar-interstitial syndrome to be 95% by ultrasound and only 72% by chest radiograph, and alveolar consolidation 97% by ultrasound versus 75% by chest radiograph when compared with computed tomography scan in 384 examined lung regions (20). Other studies also suggest that ultrasound is likely more reliable than chest radiograph for assessing alveolar filling and consolidation (25, 33). The use of ultrasound in our study confirms prior studies suggesting the feasibility, affordability, and effectiveness of ultrasound in resource-constrained settings (34). Although we believe each modification to the Berlin definition is justified by prior studies, it would nonetheless be helpful to compare results using the Berlin definition and modified definition in a setting in which both definitions could be fully assessed.

A second limitation is that this was a small single-center study in one country in sub-Saharan Africa, during one rainy season. Rwanda has a relatively low HIV rate for its region of the world (∼3% of adults), and HIV could impact ARDS incidence dramatically. Therefore, our incidence and mortality estimates may not be generalizable to other resource-constrained settings or all seasons.

A third limitation is the inability to directly compare our incidence estimates with other studies. Unlike prior studies in which a geographic area with confined referral patterns could be identified, referral patterns in Rwanda are not restricted enough to allow for population estimates based on admissions to one or a few hospitals. In addition, we screened the entire hospital for ARDS given a very small proportion of ICU beds, whereas all prior studies except one (35) screened only intubated ICU patients. It may be that hospital-screening in a high-income country would give a higher incidence of ARDS than has been reported. It is unclear whether ARDS incidence would be expected to be higher or lower in a resource-constrained setting or a resource-rich setting: a higher burden of infection and injury in low-income countries could lead to higher rates (36), but a lower rate of mechanical ventilation with its potential contribution to lung injury could lead to lower rates (37).

A fourth limitation is the possibility that we have included patients in our ARDS cohort who do not have the pathophysiology of permeability pulmonary edema. First, we could have inadvertently included some cases of bilateral opacities that were caused by cardiogenic pulmonary edema. However, in keeping with the Berlin definition itself and with prior epidemiologic studies of ARDS, we excluded patients with myocardial infarction or congestive heart failure based on review of the medical chart (10). In our setting, valvular heart disease is relatively common, “cardiopathy” an oft-cited diagnosis, and bedside echocardiography performed relatively frequently and skillfully, so we do not believe we missed many diagnoses of cardiogenic pulmonary edema. Although previous studies excluded only patients with myocardial infarction or congestive heart failure and no other predisposing clinical insult, we excluded all patients with congestive heart failure or myocardial infarction even if they had a clinical insult to produce a conservative estimate.

Second, some of our cases may represent atelectasis or pneumonia. This is a problem inherent in the Berlin definition itself. In a 2013 study that examined autopsy findings for 356 patients who met the Berlin criteria for ARDS at the time of death, only 45% had diffuse alveolar damage, the pathologic correlate to clinical ARDS (38). However, 88% of patients with ARDS were found to have either diffuse alveolar damage or pneumonia. This speaks to the fact that the Berlin definition does not correlate well with our pathophysiologic definitions of ARDS. Although we cannot change the inherent difficulty in using a clinical definition to capture a pathophysiologic entity, we tried to be as consistent as possible with the definition used in high-income countries, which is currently the best classification we have for research and treatment (37).

An important next step in understanding the impact of ARDS in different regions of the world is to replicate this study in a high-income country setting, screening all hospital patients and comparing outcomes using both the Berlin definition and the Kigali modification of the Berlin definition. Beyond more rigorously validating the method we have used, this would also serve to help address the problem of counting only “treated incidence” in all regions of the world (1416). All prior studies but one limited screening to ICU patients so that much of the observed variability in incidence estimates may be related to a large variability in ICU bed availability even among high-income countries (12).

In summary, our study is the first to estimate ARDS incidence in a low-income country setting. We found a high incidence of ARDS and high mortality associated with the syndrome in one hospital in Rwanda. Diagnosis using the Berlin criteria would have captured none of these patients. We have defined the Kigali modification of the Berlin definition for the identification of ARDS in resource-constrained settings. Although the Kigali modification requires validation before widespread use, we hope this study stimulates further work in refining an ARDS definition that can be consistently used in all settings. Real differences in ARDS incidence may exist because of the epidemiology of risk factors including mechanical ventilation practices, so finding a definition and method of screening for ARDS that is accessible to all regions of the world is critical to understanding the burden of and trends in ARDS across the world. An improved ability to recognize ARDS worldwide will provide opportunities to identify risk factors unique to various settings, apply effective treatments, and reduce morality.

The authors are grateful to Jean de Dieu Gatete, Jacqueline Ingabire, Eliane Kampimbare, Fotide Ndayishimiye, and Denis Rwabizi for excellent research assistance. They thank Georges Ntakiyiruta and Trish Henwood for generously providing ultrasound machines. They are also grateful to Meaghan Muir for her valuable contributions to a literature search. They thank Emily Fish for sharing her expertise in pulmonary ultrasound.

1. Ware LB, Matthay MA. The acute respiratory distress syndrome. N Engl J Med 2000;342:13341349.
2. Estenssoro E, Dubin A, Laffaire E, Canales H, Sáenz G, Moseinco M, Pozo M, Gómez A, Baredes N, Jannello G, et al. Incidence, clinical course, and outcome in 217 patients with acute respiratory distress syndrome. Crit Care Med 2002;30:24502456.
3. Luhr OR, Antonsen K, Karlsson M, Aardal S, Thorsteinsson A, Frostell CG, Bonde J; The ARF Study Group. Incidence and mortality after acute respiratory failure and acute respiratory distress syndrome in Sweden, Denmark, and Iceland. Am J Respir Crit Care Med 1999;159:18491861.
4. Roupie E, Lepage E, Wysocki M, Fagon JY, Chastre J, Dreyfuss D, Mentec H, Carlet J, Brun-Buisson C, Lemaire F, et al. Prevalence, etiologies and outcome of the acute respiratory distress syndrome among hypoxemic ventilated patients. SRLF Collaborative Group on Mechanical Ventilation. Société de Réanimation de Langue Française. Intensive Care Med 1999;25:920929.
5. Caser EB, Zandonade E, Pereira E, Gama AM, Barbas CS. Impact of distinct definitions of acute lung injury on its incidence and outcomes in Brazilian ICUs: prospective evaluation of 7,133 patients*. Crit Care Med 2014;42:574582.
6. Brun-Buisson C, Minelli C, Bertolini G, Brazzi L, Pimentel J, Lewandowski K, Bion J, Romand JA, Villar J, Thorsteinsson A, et al.; ALIVE Study Group. Epidemiology and outcome of acute lung injury in European intensive care units: results from the ALIVE study. Intensive Care Med 2004;30:5161.
7. Bersten AD, Edibam C, Hunt T, Moran J; Australian and New Zealand Intensive Care Society Clinical Trials Group. Incidence and mortality of acute lung injury and the acute respiratory distress syndrome in three Australian States. Am J Respir Crit Care Med 2002;165:443448.
8. Hughes M, MacKirdy FN, Ross J, Norrie J, Grant IS; Scottish Intensive Care Society. Acute respiratory distress syndrome: an audit of incidence and outcome in Scottish intensive care units. Anaesthesia 2003;58:838845.
9. Li G, Malinchoc M, Cartin-Ceba R, Venkata CV, Kor DJ, Peters SG, Hubmayr RD, Gajic O. Eight-year trend of acute respiratory distress syndrome: a population-based study in Olmsted County, Minnesota. Am J Respir Crit Care Med 2011;183:5966.
10. Rubenfeld GD, Caldwell E, Peabody E, Weaver J, Martin DP, Neff M, Stern EJ, Hudson LD. Incidence and outcomes of acute lung injury. N Engl J Med 2005;353:16851693.
11. The World Bank. Country and lending groups. 2014 [accessed 2014 Jul 8]. Available from:
12. Buregeya E, Fowler RA, Talmor DS, Twagirumugabe T, Kiviri W, Riviello ED. Acute respiratory distress syndrome in the global context. Glob Heart 2014;9:289295.
13. Ranieri VM, Rubenfeld GD, Thompson BT, Ferguson ND, Caldwell E, Fan E, Camporota L, Slutsky AS; ARDS Definition Task Force. Acute respiratory distress syndrome: the Berlin Definition. JAMA 2012;307:25262533.
14. Angus DC. The acute respiratory distress syndrome: what’s in a name? JAMA 2012;307:25422544.
15. Angus DC, van der Poll T. Severe sepsis and septic shock. N Engl J Med 2013;369:840851.
16. Linde-Zwirble WT, Angus DC. Severe sepsis epidemiology: sampling, selection, and society. Crit Care 2004;8:222226.
17. Bernard GR, Artigas A, Brigham KL, Carlet J, Falke K, Hudson L, Lamy M, Legall JR, Morris A, Spragg R. The American-European Consensus Conference on ARDS: definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994;149:818824.
18. Pandharipande PP, Shintani AK, Hagerman HE, St Jacques PJ, Rice TW, Sanders NW, Ware LB, Bernard GR, Ely EW. Derivation and validation of Spo2/Fio2 ratio to impute for Pao2/Fio2 ratio in the respiratory component of the Sequential Organ Failure Assessment score. Crit Care Med 2009;37:13171321.
19. Rice TW, Wheeler AP, Bernard GR, Hayden DL, Schoenfeld DA, Ware LB; National Institutes of Health, National Heart, Lung, and Blood Institute ARDS Network. Comparison of the SpO2/FIO2 ratio and the PaO2/FIO2 ratio in patients with acute lung injury or ARDS. Chest 2007;132:410417.
20. Lichtenstein D, Goldstein I, Mourgeon E, Cluzel P, Grenier P, Rouby JJ. Comparative diagnostic performances of auscultation, chest radiography, and lung ultrasonography in acute respiratory distress syndrome. Anesthesiology 2004;100:915.
21. Dubowitz G, Breyer K, Lipnick M, Sall JW, Feiner J, Ikeda K, MacLeod DB, Bickler PE. Accuracy of the Lifebox pulse oximeter during hypoxia in healthy volunteers. Anaesthesia 2013;68:12201223.
22. Lifebox. Specifications: handheld pulse oximeter: Model AH-MX. 2014 [accessed 2014 Aug 4]. Available from:
23. Miller A. Intensive care medicine and intensive care ultrasound: lung ultrasound. 2014 [accessed 2014 Feb 12]. Available from: ultrasound/intro/
24. Christian Medical College Hospital. ICU sonography: lung ultrasound. 2014 [accessed 2014 Feb 12]. Available from:
25. Lichtenstein DA, Mezière GA. Relevance of lung ultrasound in the diagnosis of acute respiratory failure: the BLUE protocol. Chest 2008;134:117125.
26. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap): a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 2009;42:377381.
27. Vincent JL, Rello J, Marshall J, Silva E, Anzueto A, Martin CD, Moreno R, Lipman J, Gomersall C, Sakr Y, et al.; EPIC II Group of Investigators. International study of the prevalence and outcomes of infection in intensive care units. JAMA 2009;302:23232329.
28. Frat JP, Thille AW, Mercat A, Girault C, Ragot S, Perbet S, Prat G, Boulain T, Morawiec E, Cottereau A, et al.; FLORALI Study Group; REVA Network. High-flow oxygen through nasal cannula in acute hypoxemic respiratory failure. N Engl J Med 2015;372:21852196.
29. Mucci B, Murray H, Downie A, Osborne K. Interrater variation in scoring radiological discrepancies. Br J Radiol 2013;86:20130245.
30. Phua J, Badia JR, Adhikari NK, Friedrich JO, Fowler RA, Singh JM, Scales DC, Stather DR, Li A, Jones A, et al. Has mortality from acute respiratory distress syndrome decreased over time? A systematic review. Am J Respir Crit Care Med 2009;179:220227.
31. Laffey J. Preliminary findings from the LUNG SAFE study. ESICM; 2014 [accessed 2015 Jul 1]. Available from:
32. Ferguson ND, Kacmarek RM, Chiche JD, Singh JM, Hallett DC, Mehta S, Stewart TE. Screening of ARDS patients using standardized ventilator settings: influence on enrollment in a clinical trial. Intensive Care Med 2004;30:11111116.
33. Silva S, Biendel C, Ruiz J, Olivier M, Bataille B, Geeraerts T, Mari A, Riu B, Fourcade O, Genestal M. Usefulness of cardiothoracic chest ultrasound in the management of acute respiratory failure in critical care practice. Chest 2013;144:859865.
34. Shah SP, Epino H, Bukhman G, Umulisa I, Dushimiyimana JM, Reichman A, Noble VE. Impact of the introduction of ultrasound services in a limited resource setting: rural Rwanda 2008. BMC Int Health Hum Rights 2009;9:4.
35. Ferguson ND, Frutos-Vivar F, Esteban A, Gordo F, Honrubia T, Peñuelas O, Algora A, García G, Bustos A, Rodríguez I. Clinical risk conditions for acute lung injury in the intensive care unit and hospital ward: a prospective observational study. Crit Care 2007;11:R96.
36. Adhikari NK, Fowler RA, Bhagwanjee S, Rubenfeld GD. Critical care and the global burden of critical illness in adults. Lancet 2010;376:13391346.
37. Thompson BT, Matthay MA. The Berlin definition of ARDS versus pathological evidence of diffuse alveolar damage. Am J Respir Crit Care Med 2013;187:675677.
38. Thille AW, Esteban A, Fernández-Segoviano P, Rodriguez JM, Aramburu JA, Peñuelas O, Cortés-Puch I, Cardinal-Fernández P, Lorente JA, Frutos-Vivar F. Comparison of the Berlin definition for acute respiratory distress syndrome with autopsy. Am J Respir Crit Care Med 2013;187:761767.
Correspondence and requests for reprints should be addressed to Elisabeth D. Riviello, M.D., M.P.H., Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, MA 02215. E-mail:

Author Contributions: E.D.R., A.M., and V.N. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. E.D.R., D.S.T., and R.A.F., study concept and design. E.D.R., W.K., M.M., and L.O., acquisition of data. E.D.R., A.M., and V.N., data analysis and interpretation. E.D.R. and R.A.F., drafting of manuscript. All authors, critical revision of the manuscript for important intellectual content. W.K., T.T., V.M.B.-G., L.O., D.S.T., and R.A.F., administrative, technical, or material support.

This article has an online supplement, which is accessible from this issue's table of contents at

Originally Published in Press as DOI: 10.1164/rccm.201503-0584OC on September 9, 2015

Author disclosures are available with the text of this article at

Comments Post a Comment

New User Registration

Not Yet Registered?
Benefits of Registration Include:
 •  A Unique User Profile that will allow you to manage your current subscriptions (including online access)
 •  The ability to create favorites lists down to the article level
 •  The ability to customize email alerts to receive specific notifications about the topics you care most about and special offers
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record