In this study of over 7000 patients who received outpatient or inpatient care for COVID-19 between August 2020 and May 2021, adjusted mortality risk increased significantly from the Fall pre-surge period to the Winter surge period—corresponding with the very rapid rise in patient volume. Adjusted mortality risk then returned to pre-surge levels during the Spring post-surge period—corresponding to a subsequent similarly rapid decline in patient volume. The excess mortality risk during the winter COVID-19 surge was not adequately explained by conventional sociodemographic or pre-existing risk traits or easily measurable variables. However, we did observe that the excess risk recovered rapidly in the setting of targeted facility resources. Notably, however, first-encounter patient-level measures of illness severity appeared higher during surge compared to non-surge periods—suggesting that timing of patient presentation, as related to timing illness onset, may have contributed along with external socioeconomic or other epidemiological factors to augmenting risk for adverse outcomes during the surge period.
Our findings extend from numerous earlier scientific and lay reports that have chronicled the overwhelming nature of the initial COVID-19 surge that began the United States in March of 2020 [13,14,15,16]. This first wave was compounded by multiple factors including lack of knowledge around appropriate treatment of SARS-CoV-2 infection, unprepared resource supply chains, and a lack of adequately trained personnel in highly impacted communities. Advances in standards of care including the use of monoclonal antibodies, steroids and Remdesivir [17], among others, as well as more robust supply chains [18, 19] were present during the Winter surge period evaluated in the current study—allowing for a more focused evaluation of the excess patient volume effect on COVID-19 case outcomes at a health system level. Further, the presence of ‘valleys’ in patient volume during non-surge time periods allowed for comparison of surge case fatality rates to those when excess patient volume was not a predominant factor.
Expanding longitudinally from the earlier reports, our analysis from the Winter 2020 surge found that patients treated for COVID-19 during the surge period had higher odds of death, both overall and when stratified by maximum level of care required (outpatient, inpatient, and ICU). Importantly, we observed that patients presenting for care during the surge were more likely to be older, male, Hispanic/Latinx, and with a greater burden of comorbidities than those presenting during the non-surge periods; all these factors have been linked to greater severity of COVID-19 illness [4, 20,21,22]. Nonetheless, odds of death remained elevated during the surge period even when adjusting for these risk factors. We also observed apparently modest but statistically significant differences in laboratory characteristics among patients hospitalized during the surge. It is well described that during periods of high COVID-19 activity in the community, patients delay seeking care due to fear of becoming ill or spreading the virus themselves [23,24,25,26,27]. As such, delayed presentations, particularly among vulnerable patient populations, with subsequent late initiation of COVID-19 specific therapies, may well have contributed to at least a portion of the observed excess mortality risk.
Previous studies have also found that discernible non-patient factors contribute to measurable variation in COVID-19 outcomes. Increased COVID-19 case rates [3, 28, 29], ICU strain [5, 12], and limited hospital resource availability, including number of hospital beds and staff [7], have been linked to increased case fatality, though these studies examine outcomes solely during the initial stages of the pandemic. Nonetheless, these phenomena are known to continue to impact COVID-19 outcomes across in at-risk regions and communities. Fortunately, although surges in COVID-19 patient volume required the transformation of previously non-critical care environments into advanced care locations within our health system, we were ultimately able to house and medically accommodate all patients requiring advanced care including intubation, mechanical ventilation, and mechanical circulatory support. These advanced care needs were met through redirecting staffing support from non-critical care to critical care settings. Beyond facility-level factors, lack of statistically significant differences in case fatality between pre- and post-surge periods suggest that the increase in case fatality was not related to secular trends in COVID-19 outcomes during the study period, such as improvements in the standard of care and the circulation of regional COVID-19 variants linked to increased mortality [9, 10]. In fact, while not statistically significant, we observed a trend towards slightly higher mortality during the post-surge compared to pre-surge period. This finding could have been related to a bias towards more severely ill patients presenting for medical encounters over time, or a residual excess in hospitalized patients; further studies using more detailed data are needed to clarify the factors contributing to variations in post-surge recovery periods.
Several limitations of this study merit consideration. Our data were derived from a single healthcare system, and thus our findings may not be generalizable to other populations, especially those outside the United States. However, our patient cohort was found to be diverse, both demographically as well as clinically, and our institution is a high-volume center serving a large and diverse urban population. Reliance on EHR data to identify deaths may result in misclassification, particularly by undercounting deaths occurring outside of the hospital, though we would expect this to attenuate rather than confound our results. We were unable to systematically capture data on timing of illness onset, which precluding assessment of symptom duration prior to presentation to medical care. We recognize that all the factors driving as well as correlated with delayed patient presentations (i.e. delays in patients seeking or receiving medical attention), especially during COVID-19 surge periods, are critically important to identify and yet not easily measured in the real-world community setting. Detailed data on temporal trends in hospital occupancy, medical care staffing (e.g. nurse-to-patient ratios), and medical care supplies and other resources were not available for the current analysis and will be important for future investigations of excess mortality during surge periods. Finally, we were unable to control for COVID-19 vaccination status as vaccines were not available for the majority of the cohort until the post-surge period, and vaccine uptake in the post-surge period may have lowered risk for severe outcomes. However, given that all patients in our cohort were COVID-positive and that reported breakthrough infection rates are relatively low [30], it is unlikely that enough patients in the post-surge period were vaccinated to have biased our results.