Admission respiratory status predicts mortality in COVID‐19

Abstract COVID‐19 has significant case fatality. Glucocorticoids are the only treatment shown to improve survival, but only among patients requiring supplemental oxygen. WHO advises patients to seek medical care for “trouble breathing,” but hypoxemic patients frequently have no respiratory symptoms. Our cohort study of hospitalized COVID‐19 patients shows that respiratory symptoms are uncommon and not associated with mortality. By contrast, objective signs of respiratory compromise—oxygen saturation and respiratory rate—are associated with markedly elevated mortality. Our findings support expanding guidelines to include at‐home assessment of oxygen saturation and respiratory rate in order to expedite life‐saving treatments patients to high‐risk COVID‐19 patients.

symptoms of COVID-19 admitted for non-COVID-19-related medical issues. SARS-CoV-2 infection was confirmed using qRT-PCR. Study was approved by Institutional Review Boards at both institutions.
Patient information and clinical outcomes were collected by physician chart review. Primary outcome was all-cause in-hospital mortality. Poisson regression models with robust standard errors were used to calculate relative risks (RR) and 95% confidence intervals (CI) for the associations of oxygen saturation and respiratory rate with the mortality. Models were adjusted for age, sex, race, health system, nursing home residence, smoking status, hypertension, diabetes, body mass index [BMI], pulmonary disease, and cardiovascular disease. In secondary analyses, linear regression models were used to examine associations of candidate risk factors with admission oxygen saturation and respiratory rate. Multiple imputation with chained equations was used to impute missing BMI values (n = 46).
During the study, 1,095 individuals were hospitalized with symptomatic COVID-19. Patients had mean age of 58, were mostly men (62%), and commonly had comorbidities (Table 1). While patients frequently had hypoxemia (mean oxygen saturation of 91%) and tachypnea (mean respiratory rate of 23 breaths per minute) on presentation, few reported shortness of breath (10%) or cough (25%) regardless of oxygen saturation. The most common symptom at presentation was fever (73%). Patient respiratory symptoms and symptomatic fever were not associated with mortality.
Overall, 197 patients died in hospital. After adjustment for risk factors, both hypoxemia and tachypnea were associated with mortality risk (Figure 1). Compared to normoxemic patients, those who were hypoxemic (oxygen saturation <92%) had a 1.8-to 4.0-fold increased mortality risk, depending on initial oxygen saturation. Similarly, compared to patients with normal respiratory rate (≤20 beats per minute), those with respiratory rates >22 breaths per minute were at 1.9-to 3.2-fold elevated mortality risk. Nearly, all hypoxemic (99%) and tachypneic (98%) patients required supplemental oxygen administration during hospitalization. By contrast, other clinical signs at presentation, including temperature, heart rate, and blood pressure, were not associated with mortality. Findings were similar among subgroups stratified by demographic and clinical characteristics.
Our study shows that indices of respiratory compromise at initial presentation that are readily measurable at home-oxygen saturation <92% or a respiratory rate >22 breaths per minute-were each associated with elevated mortality in hospitalized COVID-19 patients. Current CDC and WHO COVID-19 guidelines do not include While the majority of patients in our study had one or more comorbidities, the increased mortality risk associated with hypoxemia and tachypnea was present irrespective of comorbidities. 6 Obese patients have an elevated risk of silent hypoxemia and tachypnea that partly account for their elevated mortality risk and are likely to benefit most from assessing their respiratory status during acute  Study limitations include analysis restricted to hospitalized patients with comorbidities which may not generalize to nonhospitalized patients. Since patient symptoms were retrospectively abstracted from medical records, symptom prevalence may be underestimated. We cannot exclude the possibility of residual confounding. Although pulse oximetry may underestimate the severity of hypoxemia in patients with darker skin tones, 9 our study findings were robust among those of African and Hispanic ancestry. While the accuracy of at-home pulse oximeters may be compromised at oxygen saturations less than 90%, 10 ascertainment at the risk threshold identified in this study (92%) is attainable.
In summary, indices of respiratory compromise at initial presentation, that are readily measurable at home, are associated with markedly elevated risk of in-hospital mortality in COVID-19 patients.
Public health strategies emphasizing the importance of at-home surveillance of oxygen saturation and respiratory rate may identify at-risk patients earlier and enable timely institution of life-saving medical therapy.

ACK N OWLED G EM ENTS
We wish to acknowledge the frontline workers engaged in the care of patients with COVID-19 in our communities. We also wish to acknowledge the following individuals who assisted with data abstraction at the University of Washington ( Zemke MD). We would also like to thank Ayushi Gupta and the

University of Washington Allergy and Infectious Disease Research
Collaboratory for assistance with data acquisition. Finally, we are appreciative of the many friends, family, coworkers, and patients stricken with COVID-19 who have inspired this work.

D I SCLOS U R E S
The authors each report no disclosures relevant to the content of the manuscript. F I G U R E 1 Association of oxygen saturation and respiratory rate on admission with in-hospital mortality. *Multivariable adjustment is for age, sex, race, health system, hypertension, diabetes mellitus, body mass index, pulmonary disease, cardiovascular disease, smoking, and nursing home residence