Re: Can we predict which COVID-19 patients will need transfer to the intensive care within 24 hours of floor admission?

To the Editor:

We were very interested to read the study by Wang et al. on predicting which ED patients with COVID‐19 will require escalation to intensive care unit within 24 hours of admission. They identified risk factors and developed triage models for the ED physician based on presenting history, signs, and symptoms. Their analysis and modeling ultimately determined that history of heart failure, initial SpO2 < 93%, and either GFR ≤ 46 or WBC ≤ 6.4 had the highest OR for risk of escalation of care. This model had a high specificity (0.98) but with low sensitivity (0.11) and an AUC of 0.54. When the model was reversed to identify low‐risk patients, the sensitivity was 0.85 with an NPV of 0.96. Thus, they determined that this could be a valuable tool for safely triaging low‐risk patients to a general medicine floor.

Interestingly, within the study cohort, being of the Hispanic race carried a more significant OR for clinical decompensation than a history of heart failure and low lymphocyte count. We know COVID‐19 disproportionately affects both the case count and the severity of disease in Hispanic patients , and warrants further exploration of the etiology. The racial disparities and social determinants of health are well described , and the preexisting health inequities in Hispanic communities have been exacerbated in the COVID‐19 pandemic. At baseline, Hispanic patients have a higher burden of diagnosed, poorly controlled, and undiagnosed chronic health conditions rooted in poor access to affordable health care, uncertain immigration status, language barriers, and reduced financial means. Due to this health inequity, these patients are more likely to have higher rates of the clinical features identified in the triage tool created by Wang and associates as well as other risk assessment tools such as the COVID‐GRAM and COVID severity index.

We strongly support the authors' initiatives in aiding the triage process for frontline workers in the ED and wish to share potential opportunities for improvement that will likely further their work of risk stratification. The triage tool created includes a history of heart failure and low SpO2; however, it is much more likely to have an undiagnosed chronic disease in Hispanic patients that may be overlooked with this tool. We propose investigating known social determinants of health care such as insurance status, immigration status, language barriers, and financial means as predictors of acute decompensation. These data may improve the risk stratification and identify modifiable factors to improve access to diagnosis, vaccination, and treatment of COVID‐19.

We commend Wang and colleagues for their work in developing this triage tool during a time of uncertainty and demand for quality patient care. Their work corroborates the racial disparity in the severity of COVID‐19 and can potentially aid in closing the gap of this health care inequity by identifying and treating modifiable risk factors. We look forward to further publications by Wang et al.