Laura is the editorial director of The American Journal of Managed Care® (AJMC®) and all its brands, including The American Journal of Accountable Care®, Evidence-Based Oncology™, and The Center for Biosimilars®. She has been working on AJMC® since 2014 and has been with AJMC®'s parent company, MJH Life Sciences, since 2011. She has an MA in business and economic reporting from New York University.
Abstracts presented at CHEST 2020 looked at improving diagnosis of acute respiratory distress syndrome (ARDS) with machine learning and the development of intensive care unit delirium in hospitalized patients with ARDS.
A machine learning algorithm could help clinicians diagnose acute respiratory distress syndrome (ARDS) faster, according to an abstract presented at CHEST 2020.1 In the investigation, the researchers found that the algorithm recognized ARDS more frequently than clinicians at the bedside; however, when clinicians did recognize ARDS, they were faster than the algorithm.
They noted that delayed or missed diagnoses of ARDS can have significant adverse clinical consequences. The clinical data they used came from electronic health records of intubated patients with ARDS at Montefiore Hospital, and they compared the rate and time of recognition of ARDS by the machine learning algorithm and by bedside clinicians.
The study included 1305 patients: 511 (39.2%) had mild ARDS, 458 (35.1%) had moderate ARDS, and 229 (17.5%) had severe ARDS. The remaining 107 patients (8.2%) had indeterminate appearance of ARDS on chest radiographic imaging.
The algorithm was able to identify 959 cases (73.5%) of ARDS compared with 433 cases (33.2%) recognized by the bedside clinician. The more severe the disease, the better the algorithm and the clinician were at identifying ARDS. The rates were:
“The idea of applying machine learning to recognizing and improving response times to ARDS treatment is important to explore, as ARDS is deadly in an estimated 40% of patients who develop it, and mortality increases with disease severity,” Jee Young You, MD, lead researcher and CHEST 2020 presenter, said in a statement. “Early recognition may very well save lives.”
In a second abstract, researchers found that patients with ARDS with psychiatric illnesses are more likely to develop intensive care unit (ICU) delirium, particularly when receiving ventilatory support.2
Delirium is not uncommon among patients in the ICU, and patients with ARDS are often subjected to prolonged duration of mechanical ventilation and frequently require deep sedation.
“Previous studies have associated a higher incidence of psychiatric illnesses arising among ARDS survivors,” the authors wrote. “However, very little is known about the impact of pre-existing psychiatric illness on the hospital course of ARDS patients.”
The researchers performed a retrospective study of 150 patients with ARDS who were admitted to the ICU between January 2016 and December 2016, with the purpose of determining the association between preexisting psychiatric illness and ICU delirium among these patients. The patients included in the study were older than 18 years and were on mechanical ventilation for more than 48 hours. Preexisting psychiatric illnesses were defined as major depressive disorder, generalized anxiety disorder (GAD), bipolar disorder, schizophrenia, or posttraumatic stress disorder prior to admission.
Of the patients included in the study, 58 (38.7%) received a diagnosis of ICU delirium. Nearly half (48.3%) of patients with ARDS who had a preexisting psychiatric illness received this diagnosis compared with 32.6% of patients without a preexisting psychiatric illness (odds ratio [OR], 1.93 [0.98-3.79]; P = .055).
GAD was associated with the development of ICU delirium (OR, 3.00 [1.32-6.83]; P = .007), and 22.1% of patients with ICU delirium were on anxiolytics prior to their admission to the ICU.
“A possible explanation for the higher rate of delirium in patients with preexisting psychiatric illnesses could be the need to use higher doses of benzodiazepines,” said Saminder Kalra, MD, researcher and CHEST presenter. “The increase is required to achieve adequate sedation putting them at a higher risk of withdrawal symptoms. In addition, home psychotropic medications are often withheld in the ventilated patients. They are often receiving other sedatives that potentially predisposes them to withdrawal symptoms, which often manifests as delirium.”
1. You JY, Gong M, Chen J-T. Machine learning algorithm increases recognition of ARDS. Presented at: CHEST 2020; October 18-21, 2020.
2. Jaber J, Reddy R, Alzghoul B, et al. Preexisting psychiatric illness is associated with higher incidence of delirium in ARDS. Presented at: CHEST 2020; October 18-21, 2020.