SCCM 2024: Discovery, the Critical Care Research Network - Datathon Winner Presentations
The PRONE-COVID study: PRoning and Outcomes among different ethNicitiEs
The COVID-19 pandemic has highlighted healthcare and social inequities, particularly in terms of racial and ethnic disparities among SARS-CoV-2-infected patients. This study was conducted to assess the effect of non-pharmacologic COVID-19 therapies and their outcomes among different ethnicities. The study aimed to investigate outcome disparities associated with proning in COVID-19 patients. Using a COVID-19 registry with over 21,000 ICU admission entries from 28 countries; the study analyzed patients aged > 18 years under invasive ventilation, excluding pregnant subjects and those in ISARIC trials. The overall global and American cohort demographics were largely similar, except for a higher proportion of black patients in the American cohort and a smaller proportion of Asian patients. The results showed that one-third of the patients in the global cohort were proned. However, logistic regression, adjusting for severity, revealed that globally, black patients and females were prone less. When focusing specifically on America, the difference in proning based on race was not significant, with black patients being proned slightly more, though females were still prone less. There was no significant effect observed for hospital mortality when considering overall prone positioning. In conclusion, global differences in proning frequency based on race were observed, but not specifically in America. Overall, females were prone less in both American and global contexts. Race did not affect mortality in patients undergoing proning. This preliminary study highlights the significant under-documentation of proning, emphasizing the need for further research.
Does Limited English Proficiency Affect Time-to-Death for Critically ill Patients Who Expire in the Hospital?
Language barriers could lead to prolonged hospital stays and increased visits to the hospital and emergency department. The main goal of the study was to enhance the quality of time in the ICU by addressing language barriers, aiming to improve both patient comfort and care. The study hypothesized that patients with limited English proficiency who expired during admission might experience increased time to death due to delays in end-of-life care discussions. Using the Mimic-III database, which included 46,520 patients from which 5800 expired. The study found that Limited English Proficiency (LEP) patients tended to be older, and there was no significant gender difference in LEP. Interestingly, there was a higher prevalence of LEP in minority groups. The type of insurance also played a role, with more lapses in Medicaid coverage compared to other insurance types. A study on procedural utilization based on English speaking proficiency showed no significant differences. Similarly, in the probability of survival for both English and LEP, no significant differences in mortality were observed. Further analysis showed that non-English speakers had end-of-life conversations more frequently but later during their admission. Gender stratification in end-of-life discussions was significant, with the most pronounced differences noted among females. In conclusion, LEP has a less impactful effect on time-to-death than initially hypothesized. The data set is disproportionately represented by older patients and those covered by Medicaid. Notably, female patients with LEP engage in more frequent and later conversations compared to other groups.
Disparities in Glucose Measurements
This study delves into the emerging field of digital determinants of health, emphasizing the impact of digital technologies on patient outcomes, separate from social determinants. Drawing parallels with biases observed in technologies like pulse oximetry and electroencephalogram (EEG), the study focused on glucose measurement disparities. The study aimed to investigate disparities in glucose measurements, focusing on the factors influencing glucose variability in point-of-care and serum measurements. Utilizing the extensive eICU database, the study aligns glucose measurements within 15 minutes to ensure temporal proximity. The dataset, initially comprising 139,365 patients, underwent exclusions based on pairing criteria and hospital size, resulting in 46,063 patients and 190,609 pairs of glucose measurements. In investigating bias and its potential impact on specific demographic groups, an offset plot revealed substantial missingness, particularly for insulin, vasopressin, and lactate. The study focused on "occult hypoglycemia," where the point of care test shows a value > 70 mg/dl and serum glucose shows a value < 70 mg/dl. The study employed Bayesian logistic regression models for individual hospitals to assess the prediction of occult hypoglycemia, where finger stick glucose readings exceeded 70 mg/dL, but subsequent serum values were significantly lower. Posterior estimates, reflecting the effect on ethnicity, were depicted for each hospital. While some hospitals showed no ethnicity-based effect on occult hypoglycemia, a subset exhibited a high probability of causing unequal harm. The study highlights the possibility that hospitals exhibiting notable effects in the parameters could share common devices, protocols, or practices. Future work aims to delve deeper by examining newer data, specifically focusing on these distinctive hospitals.
Society of Critical Care Medicine’s Critical Care Congress (SCCM 2024), Phoenix, Arizona, USA, Jan 21-23, 2024
