Acute Respiratory Distress Syndrome in an African Intensive Care Unit Setting. A prospective study of prevalence and outcomes

Acute respiratory distress syndrome (ARDS) is a life-threatening form of respiratory failure accounting for at least 10% of global intensive care unit (ICU) admissions (1–5). The Berlin definition for diagnosing ARDS is expensive for most resource-constrained settings to apply (6–8). There are therefore very limited data reporting on the burden of ARDS in low-income countries. We sought to determine the prevalence of ARDS among mechanically ventilated Ugandan ICU patients using the Kigali modification.

Methods
Study design, study setting, and procedure
We conducted a prospective prevalence study. Patients admitted to one of six participating Ugandan ICUs, aged ⩾18 years, who had a chest X-ray and were receiving mechanical ventilation for ⩾48 hours, were considered eligible. The setting was recently described (9). Informed consent was obtained and approval to conduct the study was obtained from the School of Medicine Research and Ethics Committee of Makerere University. Those who met these criteria were then assessed for ARDS diagnosis based on onset ⩽7 days, bilateral lung opacities on chest radiograph, and ratio of oxygen saturation as measured by pulse oximetry (SpO2) to fraction of inspired oxygen (FiO2) of <315. The study team extracted demographic and clinical data from medical records. The lowest SpO2 and other patient vitals on the day of enrollment were recorded, as was any clinician diagnosis of ARDS documented on the medical charts at any time during the follow-up period. Cardiac causes as an exclusion criteria were ascertained from medical charts or echocardiograms. Only chest X-ray interpretations as reported by radiologists were considered. Lung ultrasound was not routinely done.

Analyses
Data were analyzed using STATA (StataCorp. 2014. Stata Statistical Software: Release 14.1). The proportion of participants meeting the study definition was calculated. Data are summarized using proportions and/or percentages for categorical variables and means with standard deviations (SDs) for continuous data. In addition, continuous, nonnormally distributed variables are summarized as medians and interquartile ranges. To compare characteristics between categorical variables, the chi-square and Fisher’s exact tests were used. Logistic regression analyses were used to identify predictors of mortality among patients with and without ARDS. Bivariate analysis was performed for each of the independent predictors of interest, and all those with P values of 0.2 and less were included in a multivariable logistic model. We also included variables that are known risk factors for mortality in an ARDS population. Significance was established at a P value of ⩽0.05. A sample size of 229 was calculated.