Since the onset of the coronavirus disease 2019 (COVID-19) pandemic, obesity has been linked to poor clinical outcomes from the disease.¹ Despite conflicting conclusions regarding mortality, several analyses have suggested that higher body mass index (BMI) may increase risk of hospitalization, supplemental oxygen requirement, and intensive care unit (ICU) admission.^1-3 Obesity coincides with many comorbidities independently associated with worse health outcomes in acute illness, yet the physiological mechanisms making obese patients more susceptible to COVID-19 remain uncertain. Adiposity-derived immunosuppression, increased thrombogenicity, and inflammatory hyperreactivity have all been postulated mediators of this phenomenon.^4,5 However, the most immediate explanation is that obese patients have reduced respiratory reserve from lower diaphragmatic contractility, smaller airways (making intubation more challenging), and stiffer pulmonary compliance from greater thoracic wall mass, predisposing them to increased hypoxia and mechanical ventilation requirement.^4,6

Given the extreme stress placed upon inpatient resources during COVID-19 surges, hospital length of stay (LOS) is a metric of great interest to both acute care clinicians and hospital administrators.^7,8 Furthermore, as newer variants of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) have been associated with lower virulence and mortality than those in earlier phases of the COVID-19 pandemic, alternate markers of disease severity such as LOS are becoming increasingly meaningful as clinical endpoints. Thus, to explore the contribution of non-respiratory effects of obesity on hospitalization-based outcomes, we assessed the association of BMI with LOS in patients admitted to non-ICU care with lower initial respiratory distress.

The American Heart Association (AHA) National COVID-19 Registry, a retrospective cohort of adults admitted with COVID-19 to any of over 130 participating U.S. hospitals across 30 states, was used to query index admissions between March 2020 and April 2021 for patients aged ≥18 years initially hospitalized to non-intensive care to address confounding from immediate respiratory compromise.⁹ A total of 22,915 patients met these criteria and were included in the analysis, and 5,327 patients initially admitted to the ICU were additionally analyzed as a comparison group for the regression model. We described patient and hospital characteristics by BMI group (classified as a categorical variable with levels of underweight [BMI <18.5 kg/m²], normal [BMI 18.5 to <25 kg/m²], overweight [25 to <30 kg/m²], class I obesity [30 to <35 kg/m²], class II obesity [35 to <40 kg/m²], and class III obesity [≥40 kg/m²]). To quantify the association between LOS and BMI category, we applied a multivariable Poisson regression model with identity link and reported ß coefficients for difference in LOS with 95% confidence intervals (CIs). Model covariates were chosen a priori based on clinical relevance and >80% completeness to reflect patient-, hospital-, and physician-level characteristics. Patient characteristics included age, sex, medical insurance source, medical history (pre-existing malignancy, diabetes, cerebrovascular, respiratory, and cardiac comorbidities), admission respiratory rate, oxygenation status, admission month, thromboembolic event complication, hemodialysis requirement, intubation requirement, and baseline serum laboratory results (hemoglobin, white blood cell count, absolute lymphocyte count, platelet count, alanine aminotransferase, aspartate aminotransferase, total bilirubin, creatinine, and interleukin-6). Hospital characteristics included transfer from another hospital, hospital size, and region. As physician caseloads have been shown to affect LOS, we also included a measure of hospital COVID-19 burden (if ≥20% of beds were occupied by COVID-19 patients on day of admission).⁷ Multiple imputation was performed for missing data. Analyses were performed using R v3.6.2 (R Foundation, Vienna, Austria) and the AHA Precision Medicine Platform (https://precision.heart.org). Enrollment in the registry was approved or review waived by the Institutional Review Board at each participating site, including Stanford Health Care, where it was exempted as quality improvement.

In univariable analysis, higher BMI class was associated with younger age and greater hypoxia on admission. Elevated BMI category patients had lower prevalence of cancer and cerebrovascular disease, higher prevalence of diabetes and respiratory disorders, and no difference in cardiac diseases. There were no differences in incident thrombotic events or hemodialysis. Increasing BMI category was associated with increased intubation requirement but greater disposition alive (Table 1).

In multivariable regression, overweight and obese patients showed a statistically significant dose-dependent relationship between increasing BMI category and decreasing LOS (overweight: –0.469 days [95% CI, –0.708 to –0.231; P<0.01]; class I obesity: –0.480 days [95% CI, –0.733 to –0.226, P<0.01]; class II obesity: –0.578 days [95% CI, –0.852 to –0.304; P<0.01]) (Table 2). Only class III obesity did not follow this trend (–0.320 days; 95% CI, –0.615 to –0.024; P=0.034) but still represented a shorter LOS relative to the normal weight category. The same analysis was applied to patients initially admitted to the ICU, which showed no significant association between BMI and LOS (Table 2).

Our analysis revealed a counterintuitive inverse trend between BMI and COVID-19 LOS when controlling for respiratory compromise via restriction to non-ICU-level patients and adjusting for respiratory rate, oxygenation status, and mechanical ventilation requirement. This suggests the most potent drivers of increased length of hospitalization and intensive care use associated with obesity may involve ventilatory compromise, rather than other proposed mechanisms such as the inflammatory or thrombotic sequelae of increased adiposity. Furthermore, these findings support the “obesity paradox” of non-critical COVID-19, a phenomenon positing that higher BMI may correlate with improved outcomes.^2,10,11 Though this paradox is often described for critical illness, it has been observed in non-critical illness such as pulmonary embolism and pneumonia as well as multiple chronic illnesses.^12-14 In COVID-19, it is possible that obesity provides a metabolic reserve that may benefit less ill hospitalized patients but is overwhelmed by poor respiratory mechanics in sicker patients.^2,15 Interestingly, the trend in LOS did not remain linear at the highest obesity category, suggesting that negative effects of more extreme obesity may counteract possible benefits of metabolic reserve. Notably, an analysis of the AHA registry early in the pandemic found an association between BMI category and mortality and mechanical ventilation, but not LOS.¹⁶ Our study, which was restricted to patients who did not initially require ICU care or ventilation, adds insight into a possible mechanism for these findings.

A limitation of our analysis is the use of a registry, which may select for higher disease severity. Future prospective analyses of obese individuals with COVID-19 should examine pulmonary function testing and serum inflammatory markers to further elucidate the mechanisms of this phenomenon.

The authors declare no conflict of interest.

Data for this study are from the Get with the Guidelines programs provided by the American Heart Association (AHA). The American Heart Association Precision Medicine Platform (https://precision.heart.org/) was used for data analysis. IQVIA (Parsippany, NJ, USA) serves as the data collection and coordination center for the GWTG COVID-19 CVD Registry. AHA’s suite of registries is funded by multiple industry sponsors. The overall AHA COVID-19 CVD Registry is partially supported by the Gordon and Betty Moore Foundation.

Study concept and design: WJC and AYC; acquisition of data: WJC, CGO, FR, and NR; analysis and interpretation of data: WJC, AYC, and YW; drafting of the manuscript: WJC and AYC; critical revision of the manuscript: all authors; statistical analysis: YW; obtained funding: FR; administrative, technical, or material support: WJC, AYC, CGO, NA, FR, and NR; and study supervision: WJC, AYC, and NR.