2.1 Patient selection
This multi-center retrospective study was approved by our institutional review board. Written informed consent was waived.
Using an institutionally compiled COVID-19 dataset, we identified 367 patients over the age of 18 with reverse transcription polymerase chain reaction (RT-PCR) confirmed COVID-19 who were intubated for at least two days from March 1 to April 8, 2020 in three hospital centers within the Mount Sinai Health System in New York City. Using the mPower ™ (Nuance, Burlington, Massachusetts, United States) search and analytics platform, radiology information system (RIS) data were extracted for each patient and the initial chest radiograph on the day of intubation was identified for review. Patients with findings consistent with air leak on initial imaging were excluded (N = 4). After this exclusion, 363 patients were included for analysis. Based on the provided intubation date, subsequent chest radiograph reports were extracted for each patient and were reviewed to identify the primary outcomes of interest. Each patient was followed to discharge. The earliest intubation period was 3/3/2020 and the last discharge date was 7/9/2020 covering the study period. A total of 2178 chest radiographs were reviewed.
The primary outcomes of interest were pneumomediastinum or subcutaneous emphysema with or without pneumothorax, pneumoperitoneum, or pneumoretroperitoneum. The secondary outcomes of interest were length of intubation and death.
The radiology reports were manually reviewed for any mention of pneumomediastinum or subcutaneous emphysema and 43/363 (12%) patients with at least one of these findings were identified. An image review of all radiographs (n = 419) belonging to these patients was performed.
Clinically relevant variables obtained from the institutionally compiled COVID-19 dataset included age, sex, ethnicity, race, body mass index (BMI), asthma, chronic obstructive pulmonary disease (COPD), hypertension (HTN), diabetes, cancer, chronic kidney disease (CKD), heart failure, ARDS, and smoking history. Additional variables including length of intubation and ventilator settings (tidal volume, fraction of inspired oxygen (FiO2), peak end expiratory pressure (PEEP), respiratory rate, plateau pressure) were obtained through chart review.
2.2 Image review
In the subset of 43 patients with pneumomediastinum and/or subcutaneous emphysema without preceding pneumothorax, we conducted an image review of all 419 radiographs to determine the timeline of pneumomediastinum development and resolution in relation to mechanical ventilation.
All chest radiographs were reviewed by a cardiothoracic radiologist with >10 years of experience and a cardiothoracic imaging fellow using a picture archiving and communication system (PACS) workstation. Radiographs were reviewed independently, and final decisions were reached by consensus. The readers were not blinded to the diagnosis of COVID-19, however, they were blinded to the clinical report including the presence of pneumomediastinum, subcutaneous emphysema, and pneumothorax as well as the clinical characteristics of each case. For patients with identified barotrauma, radiographs both before and after the identified barotrauma were identified. The severity of disease on radiograph was rated utilizing an established rating system on a scale from 1 to 6.
- Toussie Danielle
- Voutsinas Nicholas
- Finkelstein Mark
- Cedillo Mario A.
- Manna Sayan
- Maron Samuel Z.
- Jacobi Adam
Clinical and chest radiography features determine patient outcomes in young and middle-aged adults with COVID-19.
All chest radiographs were performed portably with the patient in the anterior posterior (AP) projection in the emergency department or intensive care unit setting.
2.4 Statistical analysis
Bivariate analysis of continuous variables, such as BMI, was performed using the Kruskal-Wallis H Test. Bivariate analysis of categorical variables such as patient race, patient sex, smoking history, and comorbidities was performed utilizing chi-squared test. A multivariable logistic regression model adjusting for sociodemographic variables and comorbidities and including any barotrauma as an independent variable was performed for the outcome of death. In order to ensure usability of as many records as possible in multivariable analysis, missing BMI (N
= 37, 10.2%) were imputed using predictive mean matching using models that included the outcomes of interest, demographic information, and clinical variables.
Statistical matching using file concatenation with adjusted weights and multiple imputations.
These values were then utilized in the multivariable model through multiple imputation according to Rubin's rules.
In analyzing the severity scores of chest radiographs performed prior to and after barotrauma, weighted Cohen's kappa coefficient was used to assess agreement in scoring between the two cardiothoracic radiologists. A paired t
-test was performed to identify differences in severity prior to and after barotrauma. A p
-value of <0.05 (two-tailed) was considered statistically significant. All analysis was completed using R version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).