Mortality Risk Among Patients Hospitalized Primarily for COVID-19 During the Omicron and Delta Variant Pandemic Periods

United States, April 2020-June 2022

Stacey Adjei, MPH; Kai Hong, PhD; Noelle-Angelique M. Molinari, PhD; Lara Bull-Otterson, PhD; Umed A. Ajani, MBBS; Adi V. Gundlapalli, MD, PhD; Aaron M. Harris, MD; Joy Hsu, MD; Sameer S. Kadri, MD; Jon Starnes, MPH; Kristin Yeoman, MD; Tegan K. Boehmer, PhD

Disclosures

Morbidity and Mortality Weekly Report. 2022;71(37):1182-1189. 

In This Article

Abstract and Introduction

Introduction

The risk for COVID-19–associated mortality increases with age, disability, and underlying medical conditions.[1] Early in the emergence of the Omicron variant of SARS-CoV-2, the virus that causes COVID-19, mortality among hospitalized COVID-19 patients was lower than that during previous pandemic peaks,[2–5] and some health authorities reported that a substantial proportion of COVID-19 hospitalizations were not primarily for COVID-19–related illness,* which might account for the lower mortality among hospitalized patients. Using a large hospital administrative database, CDC assessed in-hospital mortality risk overall and by demographic and clinical characteristics during the Delta (July–October 2021), early Omicron (January–March 2022), and later Omicron (April–June 2022) variant periods among patients hospitalized primarily for COVID-19. Model-estimated adjusted mortality risk differences (aMRDs) (measures of absolute risk) and adjusted mortality risk ratios (aMRRs) (measures of relative risk) for in-hospital death were calculated comparing the early and later Omicron periods with the Delta period. Crude mortality risk (cMR) (deaths per 100 patients hospitalized primarily for COVID-19) was lower during the early Omicron (13.1) and later Omicron (4.9) periods than during the Delta (15.1) period (p<0.001). Adjusted mortality risk was lower during the Omicron periods than during the Delta period for patients aged ≥18 years, males and females, all racial and ethnic groups, persons with and without disabilities, and those with one or more underlying medical conditions, as indicated by significant aMRDs and aMRRs (p<0.05). During the later Omicron period, 81.9% of in-hospital deaths occurred among adults aged ≥65 years and 73.4% occurred among persons with three or more underlying medical conditions. Vaccination, early treatment, and appropriate nonpharmaceutical interventions remain important public health priorities for preventing COVID-19 deaths, especially among persons most at risk.

COVID-19 hospitalizations and in-hospital deaths during April 2020–June 2022 were identified from 678 hospitals in the Premier Healthcare Database Special COVID-19 Release (PHD-SR).§ COVID-19 hospitalizations were defined as those with the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) code U07.1 (COVID-19, virus identified [laboratory-confirmed]) listed as the primary or secondary discharge diagnosis; a COVID-19 in-hospital death was defined as a COVID-19 hospitalization with expired discharge status. COVID-19 hospitalizations were identified as being primarily for COVID-19 if they had 1) a U07.1 primary discharge diagnosis or 2) a U07.1 secondary discharge diagnosis accompanied by either treatment with remdesivir or a primary discharge diagnosis of sepsis, pulmonary embolism, acute respiratory failure, or pneumonia. Monthly cMRs (deaths per 100 hospitalizations) were calculated for COVID-19 hospitalizations (total, primarily for COVID-19, and not primarily for COVID-19) and non–COVID-19 hospitalizations.

Patient-level analyses were conducted by selecting each patient's last hospitalization primarily for COVID-19 during the Delta, early Omicron, and later Omicron periods. For each period, sociodemographic (age, sex, race and ethnicity, and insurance type), clinical (underlying medical conditions, disability status, and previous COVID-19),** disease severity (intensive care unit [ICU] admission, receipt of COVID-19 medications, noninvasive ventilation, and invasive mechanical ventilation [IMV]),†† and hospital (U.S. Census Bureau region and number of beds) characteristics were described for patients hospitalized primarily for COVID-19 and in-hospital deaths, and cMR was calculated. Descriptive analyses were also conducted for three pre-Delta periods (April–September 2020, October 2020–February 2021, and March–June 2021).

Using a generalized estimating equations model, specified as a log-linked binomial regression including all three periods, aMRDs and aMRRs for in-hospital death were estimated across periods (early Omicron versus Delta and later Omicron versus Delta).§§ aMRDs were estimated as the difference in the adjusted predicted mortality risk between periods; aMRRs were estimated as the ratio of adjusted predicted mortality risk between periods.¶¶ SEs and 95% CIs were obtained by hospital-patient clustered bootstrapping with 500 replications. Z-tests were used to compare cMR, aMRDs, and aMRRs among pandemic periods; p<0.05 was considered statistically significant. Analyses were conducted using SAS (version 9.4; SAS Institute) and Stata (version 15.1; StataCorp). This activity was reviewed by CDC and conducted consistent with applicable federal law and CDC policy.***

During April 2020–June 2022, a total of 1,072,106 COVID-19 hospitalizations and 128,517 in-hospital deaths were reported in PHD-SR. The proportion of COVID-19 hospitalizations identified as primarily for COVID-19 was relatively stable during the pre-Omicron period (83.8%, 95% CI = 83.7–83.9) and decreased during the Omicron period (62.8%, 95% CI = 62.6–63.0) (Supplementary Figure, https://stacks.cdc.gov/view/cdc/121070). cMR was 1–2 percentage points higher for hospitalizations primarily for COVID-19 than for total COVID-19 hospitalizations through December 2021; the cMR difference increased to 3–3.5 percentage points during the early Omicron period, when the proportion of hospitalizations primarily for COVID-19 and cMRs began decreasing, and returned to 1–2 percentage points in the later Omicron period (Figure).

Figure.

Crude mortality risk* for total COVID-19 hospitalizations, hospitalizations primarily for COVID-19, hospitalizations not primarily for COVID-19, and non–COVID-19 hospitalizations — Premier Healthcare Database Special COVID-19 Release,§ United States, April 2020–June 2022
*In-hospital mortality was defined by a discharge status of expired. Crude mortality risk was calculated as in-hospital deaths per 100 hospitalizations.
Total COVID-19 hospitalizations are those with a primary or secondary discharge diagnosis of COVID-19 (i.e., International Classification of Diseases, Tenth Revision, Clinical Modification code of U07.1). Non–COVID-19 hospitalizations are those without a COVID-19 discharge diagnosis. Hospitalizations primarily for COVID-19 had a primary discharge diagnosis of COVID-19 or a secondary discharge diagnosis of COVID-19 accompanied by either treatment with remdesivir or a primary discharge diagnosis of sepsis, pulmonary embolism, acute respiratory failure, or pneumonia. Hospitalizations not primarily for COVID-19 are those that did not meet criteria for a hospitalization primarily for COVID-19.
§August 2, 2022, data release. Data are from 678 hospitals that had at least one inpatient record per month during April 2020–May 2022.
Variant pandemic periods were selected based on two factors: 1) the U.S. epidemic curve for new admissions of patients with confirmed COVID-19 (https://covid.cdc.gov/covid-data-tracker/#new-hospital-admissions) and 2) the U.S. variant proportions from SARS-CoV-2 genomic surveillance (https://data.cdc.gov/Laboratory-Surveillance/SARS-CoV-2-Variant-Proportions/jr58-6ysp). Pandemic periods are defined using whole months because of date aggregation in the data source. The Delta variant (B.1.617.2) became the predominant circulating strain (representing >50% of sequenced isolates) during the week ending June 26, 2021, the Omicron B.1.1.529 subvariant became the predominant circulating strain during the week ending December 25, 2021, and the Omicron BA.2 subvariant became the predominant circulating strain during the week ending March 26, 2022. The predominant circulating strains during the early Omicron period were B.1.1.529 and BA.1 and during the later Omicron period were BA.2 and BA.2.12.1.

Among patients hospitalized primarily for COVID-19 who died in-hospital during the Delta, early Omicron, and later Omicron periods, 57.8%, 58.0%, and 51.4%, respectively, were male; 63.8%, 66.8%, and 69.1%, respectively, were non-Hispanic White (White); 53.7%, 73.5%, and 81.9%, respectively, were aged ≥65 years; 15.1%, 22.9%, and 28.9%, respectively, had a disability; and 61.7%, 70.8%, and 73.4%, respectively, had three or more underlying medical conditions (Table 1). In addition, a decreasing proportion of patients who died in-hospital had other indicators of disease severity during the Delta, early Omicron, and later Omicron periods: 76.1%, 64.0%, and 57.2%, respectively, were admitted to ICU; 93.8%, 86.8%, and 76.4%, respectively, received COVID-19 medications; 61.8%, 51.2%, and 35.0%, respectively, received noninvasive ventilation; and 71.9%, 57.6%, and 43.6%, respectively, received IMV.

The cMR among patients hospitalized primarily for COVID-19 was 15.1 during the Delta, 13.1 during the early Omicron, and 4.9 during the later Omicron periods (Table 2); cMR range was 9.9–16.1 during the pre-Delta periods (Supplementary Table, https://stacks.cdc.gov/view/cdc/121069). After adjustment, in-hospital mortality was 0.69 (95% CI = 0.68–0.70) times as likely during the early Omicron period and 0.24 (95% CI = 0.22–0.25) times as likely during the later Omicron period than during the Delta period. Adjusted mortality risk during the early and later Omicron periods was lower than it was during the Delta period for patients aged ≥18 years, males and females, all racial and ethnic groups, persons with and without disabilities, and those with one or more underlying medical conditions, as indicated by significant aMRDs and aMRRs (p<0.05); mortality risk did not differ between the Omicron and Delta periods for patients aged <18 years. Larger aMRDs were observed with increasing age and number of underlying medical conditions; aMRD and aMRR were similar in magnitude for patients with and without disabilities.

*Examples include New York (https://www.wivb.com/news/new-york/new-york-state-covid-19-update-saturday-january-22/); Massachusetts (https://www.boston.com/news/coronavirus/2022/01/21/almost-half-of-mass-covid-hospitalizations-are-now-classified-as-incidental-heres-what-that-means/); Marin County, California (https://coronavirus.marinhhs.org/surveillance#keyindicators); and various health systems in Florida, Maryland, and Texas (https://www.nytimes.com/2022/01/04/health/covid-omicron-hospitalizations.html).
Variant pandemic periods were selected based on two factors: 1) the U.S. epidemic curve for new admissions of patients with confirmed COVID-19 (https://covid.cdc.gov/covid-data-tracker/#new-hospital-admissions) and 2) the U.S. variant proportions from SARS-CoV-2 genomic surveillance (https://data.cdc.gov/Laboratory-Surveillance/SARS-CoV-2-Variant-Proportions/jr58-6ysp). Pandemic periods are defined using whole months because of date aggregation in the data source. Variants became the predominant circulating strain (representing >50% of sequenced isolates) during the following weeks: Delta (B.1.617.2) during the week ending June 26, 2021; Omicron B.1.1.529 during the week ending December 25, 2021; and Omicron BA.2 subvariant during the week ending March 26, 2022. Thus, the predominant circulating strains during the early Omicron period were B.1.1.529 and BA.1 and during the later Omicron period were BA.2 and BA.2.12.1.
§PHD-SR is a large U.S. hospital-based all-payor database (https://offers.premierinc.com/rs/381-NBB-525/images/PHD_COVID-19_White_Paper.pdf), in which patient records are linked by a unique identifier within, but not across, hospital systems. This analysis included data from 678 hospitals that had at least one inpatient record per month during April 2020–May 2022. Of these, 521 hospitals also had at least one inpatient record during June 2021. PHD-SR data are released every 2 weeks; this study used the August 2, 2022, data release. According to information provided by Premier, Inc., data completeness is estimated to be 37%, 72%, 87%, and 95% during June, May, April, and January–March 2022, respectively.
The definition of hospitalizations primarily for COVID-19 was intended to be relatively simple and replicable and used for monitoring temporal trends. Multiple iterations of the definition were evaluated, such as inclusion of additional primary discharge diagnoses and treatments. This definition was selected for its specificity in identifying patients experiencing COVID-19–related illness. The following ICD-10-CM codes were used to define sepsis (A41.89, A41.9, R65.2*), pneumonia (J12*, J18*), acute respiratory failure (J96.00, J96.01, J96.02, J96.20, J96.21, J96.22, J80, R06.03, R06.9, R09.2), and pulmonary embolism (I27.82, I26*).
**Sixteen underlying medical conditions associated with higher risk for severe COVID-19 were assessed: asthma, cerebrovascular disease, cancer, chronic kidney disease, chronic lung disease, chronic liver disease, cystic fibrosis, dementia, diabetes, heart conditions, HIV, mental health disorder, obesity, primary immunodeficiencies, transplantation, and tuberculosis (https://www.cdc.gov/coronavirus/2019-ncov/science/science-briefs/underlying-evidence-table.html). Conditions were assessed using ICD-10-CM codes listed either at or before the COVID-19 health care encounter. For each patient, the number of underlying medical conditions was summed and categorized. Presence of a disability was assessed separately from underlying medical conditions and included ICD-10-CM codes for birth defects, developmental disabilities, spinal cord injury, traumatic brain injury, and vision-, hearing-, and mobility-related disabilities. Previous COVID-19 was identified by presence of a COVID-19 diagnosis during an outpatient or inpatient encounter that occurred in the same hospital system ≥90 days before the current diagnosis. COVID-19 vaccination status was not assessed because it was underascertained in PHD-SR.
††COVID-19 medications included dexamethasone, remdesivir, baricitinib, tofacitinib, tocilizumab, and sarilumab. Noninvasive ventilation included continuous positive airway pressure and bilevel positive airway pressure.
§§The model included main effects and two-way interactions between pandemic period and the following covariates: age (0–17, 18–34, 35–49, 50–64, 65–79, and ≥80 years), sex (male and female), race and ethnicity (Hispanic or Latino [Hispanic], non-Hispanic White, non-Hispanic Black or African American [Black], non-Hispanic Asian, non-Hispanic other, and unknown), number of underlying medical conditions (0, 1, 2, 3, 4, and ≥5), and presence or absence of a disability. Additional covariates were included in the model without interaction terms: insurance type (commercial, Medicare, Medicaid, self-pay, and other or unknown), prevoius COVID-19, hospital U.S. Census Bureau region (Northeast, Midwest, South, and West), and number of hospital beds (<200, 200–499, and ≥500). Patients with unknown sex were excluded.
¶¶From the regression, the following average predicted probabilities were obtained 1) the average predicted probability of death with pandemic period set to be later Omicron and all other covariates set to their observed values (PLO), 2) the average predicted probability of death with pandemic period set to be early Omicron and all other covariates set to their observed values (PEO), and 3) the average predicted probability of death with pandemic period set to be Delta and all other covariates set to their observed values (PD). aMRD is the difference in the average predicted probabilities (PLO minus PD and PEO minus PD). aMRR is the ratio of the average predicted probabilities (PLO divided by PD and PEO divided by PD). aMRRs and aMRDs were estimated for the full sample (where all covariates other than pandemic period were set at their observed values) and for each subsample (where the corresponding covariate was set at the specific category and other covariates were set at their observed values).
***45 C.F.R. part 46; 21 C.F.R. part 56; 42 U.S.C. Sect. 241(d); 5 U.S.C. Sect. 552a; 44 U.S.C. Sect. 3501 et seq.

processing....