Importance: Assessing relative disease severity of SARS-CoV-2 variants in populations with varied vaccination and infection histories can help characterize emerging variants and support healthcare system preparedness.
Objective: To assess COVID-19 hospitalization risk for patients infected with Omicron (BA.1 and sublineages) compared with Delta SARS-CoV-2 variants.
Design: Observational cohort study.
Setting: New York City Department of Health and Mental Hygiene population-based COVID-19 disease registry, linked with laboratory results, immunization registry, and supplemental hospitalization data sources.
Participants: New York City residents with positive laboratory-based SARS-CoV-2 tests during August 2021-January 2022. A secondary analysis restricted to patients with whole-genome sequencing results, comprising 1%-18% of weekly confirmed cases.
Exposures: Diagnosis during periods when ≥98% of sequencing results were Delta (August-November 2021) or Omicron (January 2022). A secondary analysis defined variant exposure using patient-level sequencing results.
Main outcomes and measures: COVID-19 hospitalization, defined as a positive SARS-CoV-2 test 14 days before or 3 days after hospital admission.
Results: Among 646,852 persons with a positive laboratory-based SARS-CoV-2 test, hospitalization risk was lower for patients diagnosed when Omicron predominated (16,025/488,053, 3.3%) than when Delta predominated (8,268/158,799, 5.2%). In multivariable analysis adjusting for demographic characteristics and prior diagnosis and vaccination status, patients diagnosed when Omicron relative to Delta predominated had 0.72 (95% confidence interval [CI]: 0.63, 0.82) times the hospitalization risk. In a secondary analysis of 55,138 patients with sequencing results, hospitalization risk was similar for patients infected with Omicron (2,042/29,866, 6.8%) relative to Delta (1,780/25,272, 7.0%) and higher among those who received two mRNA vaccine doses (adjusted relative risk 1.64, 95% CI: 1.44, 1.87).
Conclusions and relevance: Illness severity was lower for patients diagnosed when Omicron (BA.1 and sublineages) relative to Delta predominated. This finding was consistent after adjusting for prior diagnosis and vaccination status, suggesting intrinsic virologic properties, not population-based immunity, accounted for the lower severity. A secondary analysis demonstrated collider bias from the sequencing sampling frame changing over time in ways associated with disease severity. Investing in representative data collection is necessary to avoid bias in assessing relative disease severity as new variants emerge, immunity wanes, and additional COVID-19 vaccines are administered.