COVID-19 Mortality Differences: Patient-related Data and Intensive Care Unit Load Are Prerequisites

To the Editor:

We read with great enthusiasm the study by Nishikimi and colleagues, evaluating the mortality differences in mechanically ventilated patients with coronavirus disease (COVID-19) hospitalized in traditional versus expanded intensive care units (ICUs) (1). In comparing the initially higher mortality rate found in the expanded ICUs (departments converted to ICUs during the pandemic) than in original ICUs, this difference disappeared when the analysis considered certain confounders. The authors performed unadjusted and adjusted Cox regression analyses to evaluate in-hospital 28-day mortality. The adjusted analysis included patient-related data and, most important, information on disease severity upon ICU admission and data on hospital and ICU patient load. Adjusted analyses smoothed the initially observed mortality differences (1).

The authors should be congratulated on their work. They conducted manual chart reviews, carefully collected demographic data and comorbidities, and assessed illness severity (1). We believe that studies reporting mortality, especially mortality variations across disparate settings, should always consider factors that may be possible confounders. Among these, disease severity is of no doubt. This is true not only when comparing expanded with original ICUs but also when reporting regional or even between-ICU mortality differences. Patient-related data are of great importance in mortality analysis. Bravata and colleagues examined variations in between-center mortality, considering past medical history, vital signs, and laboratory examination results (2). They found that despite ICU mortality variation across facilities (from 0 to 100%), the heterogeneity resulted only from differences in patient characteristics (i.e., comorbidities, clinical disease severity) (2).

Especially for critically ill patients, mortality analyses should always consider multiple aspects; otherwise, the reports may be misleading. Disease severity, multiorgan failure upon ICU admission, different criteria and strategies concerning the optimal intubation time in COVID-19 acute respiratory distress syndrome (partly depending on ICU bed availability), and ICU admission policies (admitting nonintubated patients as well) have a substantial impact on ICU mortality (3). We have shown that when patients experienced respiratory distress (hypoxemia, arterial oxygen tension/fraction of inspired oxygen <100 mm Hg; and tachypnea, respiratory rate >25 breaths/min) for more than 7 hours before intubation, they presented impaired respiratory system mechanics, greater multiorgan involvement (depicted through the severity scores on the Sequential Organ Failure Assessment and Acute Physiology and Chronic Health Evaluation II), and higher mortality than patients intubated earlier (3). Intrahospital mortality differences may even be present, arising from different treatment protocols (i.e., excessive immunosuppression). The wide use of excess immunosuppressive treatments in regions with an increased incidence of multidrug- and pan–drug-resistant microorganisms may further affect mortality, increasing the risk for secondary bacterial infections in COVID-19, as was lately shown in a multicenter study (4). A recent report pointed to significant regional variations in mortality on a national level, but without considering patient-centered data (5). These superficial analyses may lead to conflicts between healthcare workers in the same country or may even affect decisions on a national level.

A second major factor that should always be considered in mortality reports is the patient load. Recently, it was shown that the hazard ratio for all-cause mortality was increased to 1.67 when the ICU load was greater than 75–100%, reaching 2.35 when the load was 100% and more (6). Nishikimi and colleagues calculated the hospital occupancy at each time point and adjusted mortality for this covariant. Mortality differences disappeared after adjusting for these two major factors: disease severity and hospital occupancy. Thus, the newly expanded ICU presented mortality rates similar to the classical units even when the ICU load was impressive; maintenance of a nurse-to-patient ratio of 1:2 and being cared by certified specialists in critical care, factors not considered in many studies (5), may have contributed (1).

In conclusion, when dealing with mortality, reports should always consider patient-related outcomes such as disease severity and hospital strain; otherwise, the results should always be considered with skepticism.