The COVID-19 pandemic has been top of mind for most healthcare executives in the past three years. Unprecedented demand for inpatient and ICU beds has forced hospital operational changes, including postponing elective surgery; repurposing beds to increase surge capacity; implementing staffing changes; and managing shortages of drugs, personal protective equipment, and even mechanical ventilators.
Evidence is accumulating that pandemic COVID-19 surges in hospital case volume are associated with higher mortality in COVID-19 patients(1) as well as those hospitalized for other reasons.(2) Intensive care unit (ICU) physicians and administrators indicate that hospitals nationwide are concerned with recent unexpectedly high ICU mortality rates.
Raw mortality rates, of course, will be affected by a hospital’s or ICU’s case mix, and surge conditions would logically divert less-ill patients to step-down or other locations to preserve ICU capacity for the most critically ill. Determining the impact of the COVID-19 pandemic on reportable outcomes thus requires adjustment for age, presenting physiology, chronic health conditions, and diagnosis. With this information, expected mortality can be estimated for each patient, and a standardized mortality ratio (SMR; defined as observed/expected mortality) generated for ICU patients, thus adjusting for case mix or changes in presenting severity.
Because empiric coefficients for a novel diagnosis (COVID-19) do not yet exist, viral pneumonia would logically serve as a proxy diagnosis for COVID-19. However, it’s been previously shown that mortality in COVID-19 patients is significantly higher than expected, even after normalizing for severity of illness with APACHE IVb,(3) when using viral pneumonia with or without concurrent Acute Respiratory Distress Syndrome (ARDS) as the substitute coefficient for COVID-19.
The Acute Physiology and Chronic Health Evaluation (APACHE(4)) system uses patient age, physiologic parameters (e.g., heart rate, respiratory rate, temperature, selected laboratory values) and presence of co-morbidities to generate a summary risk score. APACHE then estimates predicted hospital mortality for individual patients based on that score and the patient’s specific clinical diagnosis, because the relationship between physiologic derangement and outcome varies by diagnosis.
By summing its observed and expected mortality rates over large groups of patients, a hospital can calculate its SMR for patients admitted to the ICU. SMRs significantly >1.00 imply that observed mortality is higher than expected, and thus provide a measure of ICU clinical performance.
Similar standardized rates can be created for ICU or hospital length of stay, ventilator days, or other outcomes. However, a novel diagnosis such as COVID-19 will not immediately have a specific diagnosis coefficient. SMR for COVID-19 patients is about 1.5 times greater than that predicted using coefficients for a “viral pneumonia” diagnosis alone or combined with ARDS).(3) With COVID-19 mortality rates improving over time,(5) the emergence of new variants, changing therapeutics, and the introduction of vaccines, it is challenging to determine a COVID-specific coefficient to accurately predict outcomes. Capacity constraints may also affect SMR.
Given those challenges, we expanded our prior analysis with additional patients and a longer timeframe to examine how COVID-19 is affecting overall standardized hospital mortality and length of stay rates, with and without the inclusion of COVID-19 patients, and during peaks of pandemic activity.
Methods
We examined APACHE outcome data from 43 hospitals consistently contributing APACHE data during 2020 and 2021. The APACHE database is maintained by Cerner Corporation (Kansas City, Missouri). To reduce potential bias introduced with new participants or organizations where operational limitations halted data collection during the pandemic, we limited the analysis to hospitals that contributed data throughout the pandemic.
Data included patients age ≥16 identified as COVID-19 positive, admitted between March 2020 and Dec. 31, 2021, plus baseline pre-pandemic data from January and February 2020. We excluded patients whose primary admission was trauma or postoperative care where COVID-19 seropositivity might have been incidental. We extracted patient demographics, mortality, occurrence and length of mechanical ventilation and length of stay. Expected mortality and length of stay were predicted for each patient using APACHE IVb national benchmarks, and standardized ratios were calculated. SMR was also examined by traditional APACHE age categories (16-44, 45-59, 60-64, 65-69, 70-74, 75-84, and ≥85) Statistical analysis was conducted using Microsoft Excel (Microsoft, Redmond, Washington).
As previous studies using this HIPAA-compliant, anonymized data have been deemed exempt under existing federal regulations, this study was not submitted for Institutional Board Review approval.
Results
From Jan. 1, 2020, through Dec. 31, 2021, there were 136,299 patients representing 143,177 ICU admissions captured by the APACHE Outcomes database in 100 ICUs at 43 hospitals. Seven were large teaching hospitals, 21 were small teaching hospitals, and 15 were non-teaching. Mean and median bed count for the 43 hospitals were 362 and 315, respectively; ICU bed capacity ranged from four to 40 beds and hospitals had between one and 10 ICUs. Of the patients, 45% were female. In all, 132,570 patient records (92.6%) contained sufficient data to generate valid APACHE predictions; 19,930 of these (15.0%) had COVID-19 documented, after excluding patients where testing positive for COVID-19 was considered incidental to an unrelated primary diagnosis such as trauma or elective surgery.
Four successive peaks of COVID-19 as a percentage of all admissions were seen, tracking temporally with U.S. national COVID-19 infection rates (Figure 1). August 2021 represented the highest peak incidence in 2020-2021, with 33.1% of ICU patients carrying a COVID-19 diagnosis.
Figure 1. COVID-19 Encounters as a Percentage of All ICU Admissions
Table 1 presents demographic characteristics and outcomes in the ICU population overall, with and without COVID, and for comparison a subset of the non-COVID-19 encounters with ARDS-viral pneumonia.
Figure 2 shows the observed versus predicted number of hospital deaths by month. Pre-pandemic, this group of hospitals was achieving standardized mortality rates below 1.0, as evidenced by actual deaths being consistently lower than predicted, a trend (not shown) that dates back several years.
Figure 2. Predicted and Actual Hospital Deaths; all ICU patients 2020–2021
Beginning in April 2020, the baseline gap between observed and predicted deaths narrows, and a large number of unexpected deaths are observed contemporaneously with each subsequent wave of the pandemic. Considering all ICU patients (not just COVID), the SMR peaks at 1.06 in July 2020,1.28 in December 2020, 1.50 in August 2021, and 1.32 in December 2021. Notably, the SMR falls below 1.00 in between surges in COVID-19 admissions.
The standardized mortality rate for COVID-19 patients was elevated in all APACHE age categories (Figure 3). While elderly (age 75+) patients were impacted most heavily in 2020, the SMR has increased over time for all age brackets and peaked in the 45–59 year-old age group during August 2021.
Figure 3. Hospital Standardized Mortality Rate (SMR) for all ICU Patients by APACHE age category.
The peaks in overall SMR, not just COVID-19 patients, intrigued us with its apparent temporal correlation to waves of the pandemic. Could capacity constraints or triage decisions during peak COVID-19 months be affecting the outcomes of ICU patients without COVID?
SMR continues to demonstrate an overall increase over the course of the pandemic. Looking at all patients in the ICU, SMR increases with each wave, although it does not exceed 1.00 in the first wave, a period when the percentage of COVID-19 patients did not exceed 11%.
Even when the COVID-19 patients with their expected high SMR are removed from the analysis, upticks in SMR are apparent with each wave. Figure 4 approaches this issue from a different perspective by plotting SMR against the percentage of COVID-19 patients each month. While the correlation coefficient is only 0.55, there appears to be a trend toward higher SMR in the non-COVID-19 patients when more than 20% of ICU beds are occupied by COVID-19 patients.
Figure 4. Standardized Mortality Ratios for COVID-19, non-COVID-19, and All Patients Admitted to 43 ICUs, 2020–2021
Hospital length of stay at 9.08 days for all ICU patients was only marginally longer (ratio 1.01) than the 8.95 days expected overall but breaks down into significantly better than expected (8.20 vs. 8.47 days) for non-COVID-19 patients; still it is more than two (14.0 vs. 11.7) days longer than expected for COVID-19 patients, generating a LOS ratio of 1.20 (p < 0.01).
ICU length of stay showed even more dramatic results, with an overall ICU excessive LOS ratio 1.11 (p < 0.01), driven by a ratio of 1.51 in the COVID-19 population, or 2.74 days longer than predicted. Ventilator days were longer than expected overall (ratio 1.12), again entirely as the result of COVID-19 patients whose 10.36 ventilation days were 3.3 days or about 1.53 times predicted.
Discussion
The COVID-19 pandemic substantially disrupted hospital and ICU operations during 2020–2021. During this period, hospital mortality rates and corresponding SMR were highly variable and tracked lower when there was lower prevalence of COVID-19 in surrounding communities.(6) A French study found that in hospital units with moderate or high levels of COVID-19 critically ill patients, non-coronavirus disease deaths were also higher.(7) A multicenter study from the United Kingdom concluded that strained ICU capacity is associated with higher acute hospital mortality even after adjusting for baseline characteristics.(8)
The current study demonstrates a similar effect on standardized outcome in non-COVID-19 patients when the percentage of COVID-19 admissions to the ICU exceeds 20%. This may be a manifestation of system overload at times of high census. Hypothetically, increased patient-to-nurse ratios; shortages of hospital beds, durable equipment, and pharmaceuticals; delays in moving patients from emergency department to inpatient beds; use of surge units and surge personnel; and provider fatigue could all be contributing to worse results when COVID-19 census is high.
ICU and hospital mortality are publicly reported outcomes, and the deterioration of outcomes during the pandemic should prompt concern among physicians, other providers, hospital management, regulators, and the public. Hospital mortality rates for ICU patients are typically about 10%, although they may approach zero in small community hospitals and exceed 20% in academic medical centers. By normalizing these raw mortality rates to expected rates based on patient presenting condition, the resulting SMR would be expected to hover around 1.00 ± 0.03 if the sample size is sufficiently large.
Our data suggest that while hospital mortality in non-COVID-19 ICU patients is relatively stable at 10.0%, the periodic influx of COVID-19 patients whose ICU mortality averages 31% and may exceed 41% is enough to drive the overall reported mortality (in this group of hospitals) to 14.7%, significantly higher than predicted using APACHE IVb as the benchmarking tool. In months where COVID-19 patients comprise >20% of the overall ICU population, COVID-19 and non-COVID-19 mortality ratios all rise. This has important implications for ICU triage decisions, disaster preparation, staffing decisions, and resource allocation, as well as for public reporting of outcomes.
Our team(3) and others,(9,10) have pointed out the limitations of SMR for benchmarking COVID-19 patients. Benchmarking tools such as APACHE, the Mortality Prediction Model (MPM), or the Simplified Acute Physiology Score (SAPS), which are based on admission physiology, may not calibrate well with COVID-19, where patients may be admitted to ICU with single organ dysfunction but slowly progress to refractory ARDS and multisystem failure over the course of their ICU stay.
Beals and colleagues demonstrated that the risk of deterioration is higher for COVID-19 patients than non-COVID-19 patients, even when stratified for admission severity of illness.(11) One possible explanation is that COVID-19 patients present with acute respiratory failure and then go on to develop other organ system failures in a pattern that differs from usual ICU patients, but this has not yet been studied.
There likely is a learning curve for clinicians treating a novel illness when initially promising therapies are discarded, and clinicians adopt different but more suitable respiratory support techniques and pharmacologic interventions. Patient characteristics, vaccination status, and disease severity are also changing over time yet do not fully explain the variability in mortality.(12) Unmeasured hospital factors appear to be associated with increased mortality observed in Black patients, even after adjustment for patient socio-demographic and clinical factors.(13)
The shifting landscape as new viral variants emerge, along with hospital-level capacity issues, also make it challenging to determine the proper coefficients to use when predicting mortality, and thus SMR, in COVID-19 patients. For the present, APACHE-IV provides only partial risk-adjustment for COVID-19 outcomes. As a general and preliminary rule of thumb, mortality in COVID-19 patients may be expected to be 50–90% higher than predicted for similar patients with viral pneumonia, and it may be associated with roughly 20% and 50% increases in hospital and ICU length of stay, respectively.
We are currently studying how the emergence of new variants may change our findings; preliminary data suggests a lower need for ICU care with the more contagious but less lethal Omicron variants.(14)
Limitations
This study was conducted using data submitted by 43 hospitals that voluntarily contribute APACHE data, and although diverse geographically, by bed size, and by teaching status, may not be fully representative.
Another limitation is that we did not have information on the specific COVID-19 variants during each surge. Nationally, Delta supplanted the original and Alpha strains as the dominant variant in June 2021, and Omicron was not designated a variant of concern until the last month of this study.(15) Further research will be needed to determine whether emerging variants will also be associated with excess SMR and length-of-stay.
Our present recommendations are to acknowledge that observed hospital mortality, and thus SMR, may be much higher than expected for COVID-19 patients during 2020-2021. Based on our estimates, SMR is roughly 1.9 times that predicted for non-COVID-19 viral pneumonia patients. It may be helpful when presenting your institution’s statistics to show baseline SMR rates before March 2020 before presenting results overall and with COVID-19 patients excluded as we have in Figure 5.
Figure 5. SMR in Non-COVID-19 Patients as a Function of % COVID-19 Patients.
Physician leaders should anticipate that SMR might be worse in non-COVID-19 ICU patients when COVID-19 patients comprise more than 1 of 5 admissions to the ICU. Similar effects on observed/expected hospital and ICU length of stay will impact bed capacity and may have follow-on effects for an institution’s ability to deliver non-COVID-19 care, particularly elective surgery.
Summary
Overall, SMR for all ICU patients has increased concurrently with each “wave” of COVID-19 as tracked by the CDC. This increase is attenuated, but not eliminated, when examining non-COVID-19 patients alone, as thus-far unmeasured capacity-related events appear to occur at times of high COVID-19 census. Hospital SMR in COVID-19 patients alone averages 1.94 times expected (range 1.21 to 2.48 based on monthly sampling across 43 ICUs) when using “viral pneumonia” coefficients for APACHE IVb risk adjustment. Hospital and ICU LOS observed to expected ratios are approximately 1.2 and 1.5 times what is predicted for viral pneumonia patients.
While we plan to recalibrate APACHE predictions to provide a proper coefficient for COVID-19 patients, such customization may presently prove unstable due to regional pandemic surges, evolving therapy, and emergence of new variants. Healthcare providers should be aware that SMR and length of stay benchmarks for all ICU patients may be significantly elevated above pre-pandemic values due to surges when more than 1 in 5 ICU admissions test positive for COVID-19.
Disclosures
Thomas Higgins is a consultant to Cerner Corporation and has been involved in developing and updating the APACHE models.
Kathy Henson, Eric Ringle, and Bridgette Collins are employees of Cerner Corporation and involved in APACHE maintenance and development.
Clinical information presented is anonymous and IRB approval was previously waived for the use of the APACHE database.
Preliminary results from this project were presented at SCCM 2021 Annual meeting (published in Crit Care Med 2021 49: DOI: 10.1097/CCM.0000000000005012) and at the American Association for Physician Leadership (AAPL) Vanguard Meeting, Spring 2021.
References
Kadri SS, Sun J, Lawandi A, et al. Association Between Caseload Surge and COVID-19 Survival in 558 U.S. Hospitals, March to August 2020. Ann Intern Med. 2021;174:1240–1251. doi:10.7326/M21-1213.
Sabbatini AK, Robicsek A, Chiu ST, Gluckman TJ. Excess Mortality Among Patients Hospitalized During the COVID-19 Pandemic. J Hosp Med. 2021;16:596–602.
Higgins TL, Stark MM, Henson KN, Freeseman-Freeman L. Coronavirus Disease 2019 ICU Patients Have Higher-Than-Expected Acute Physiology and Chronic Health Evaluation-Adjusted Mortality and Length of Stay Than Viral Pneumonia ICU Patients. Crit Care Med. 2021 Jul 1;49(7):e701–e706. doi: 10.1097/CCM.0000000000005012. PMID: 33861555.
Zimmerman JE, Kramer AA, McNair DS, Malila FM. Acute Physiology and Chronic Health Evaluation (APACHE) IV: Hospital Mortality Assessment for Today’s Critically Ill Patients. Crit Care Med. 2006;34:1297–1310.
Anesi GL, Jablonski J, Harhay MO, et al. Characteristics, Outcomes and Trends of Patients with COVID-19 Related Critical Illness at a Learning Health System in the United States. Ann Intern Med. 2021;174:613–621. doi:10.7326/M20-5327.
Asch DA, Sheils NE, Islam MN, et al. Variation in U.S. Hospital Mortality Rates for Patients Admitted with COVID-19 During the First 6 Months of the Pandemic. JAMA Intern Med. 2021;181:471–478.
Payet C, Polazzi S, Rimmele T, Duclos A. Mortality Among Noncoronavirus Disease 2019 Critically Ill Patients Attributable to the Pandemic in France. Crit Care Med. 2021. doi: 10.1097/CCM.0000 00000005215.
Wilcox ME, Rowan KM, Harrison DA, Doidge JC. Does Unprecedented ICU Capacity Strain, as Experienced During the COVID-19 Pandemic, Impact Patient Outcome? Crit Care Med. 2002; doi: 10.1097/CCM0000000000005464.
Quintairos A, Zampieri FG, Souza-Dantas VC, Salluh, JIF. The Limitations of Standardized Mortality Ratios for Coronavirus Disease 2019 ICU Patients. Crit Care Med. 2021. (online) doi: 10.1097/CCM0000000000005245.
Kurtz P, Bastos LSL, Salluh JIF, et al. SAPS-3 Performance for Hospital Mortality Prediction in 30,571 Patients with COVID-19 Admitted to ICUs in Brazil. Intensive Care Med. 2021. (online) doi:10.1007/s00134-021-06474-3.
Beals J, Barnes JJ, Durand DJ, Rimar JM, Donohue TJ, et al. Stratifying Deterioration Risk by Acuity at Admission Offers Triage Insights for Coronavirus Disease 2019 Patients. Crit Care Explor. 2021;3(4):e0400. doi: 10.1097/CCE.0000000000000400
Roth GA, Emmons-Bell S, Alger HM, et al. Trends in Patient Characteristics and COVID-19 In-hospital Mortality in the United States During the COVID-19 Pandemic. JAMA Network Open. 2021;5:e218828. doi: 10.1001/jamanetworkopen.2021.8828.
Asch DA, Islam MN, Sheils NE, et al. Patient and Hospital Factors Associated with Differences in Mortality Rates among Black and White US Medicare Beneficiaries Hospitalized with COVID-19 infection. JAMA Network Open. 2021; 4:e2112842. DOI:10.1001/jamanetworkopen.2021.12842.
Anthes E. In Omicron Hot Spots, Hospitals Fill Up, but ICUs May Not. New York Times. January 4, 2002. www.nytimes.com/2022/01/04/health/covid-omicron-hospitalizations.html . Accessed January 20, 2022.
Centers for Disease Control. COVID-19 Timeline website. www.CDC.gov/museum/timeline/covid19.html . Accessed April 29, 2022.