Estimation of COVID-19 mRNA Vaccine Effectiveness and COVID-19 Illness and Severity by Vaccination Status During Omicron BA.4 and BA.5 Sublineage Periods

Key Points Question What is the estimated vaccine effectiveness (VE) associated with first-generation COVID-19 mRNA vaccines against medically attended COVID-19 during Omicron BA.4 and BA.5 sublineage predominance? Findings This case-control study included 82 229 emergency department or urgent care encounters and 21 007 hospitalizations for COVID-19–like illness. Among hospitalized patients, estimated 3-dose VE was 68% for those with the third dose 7 to 119 days prior, but was lower by 120 days or longer after vaccination (VE, 36%). Meaning These findings suggest that first-generation COVID-19 mRNA vaccines were associated with protection against COVID-19 during the Omicron BA.4/BA.5 sublineage-predominant periods but protection declined over time.


Section 2.4: Primary Outcome Model and Covariate Adjustment
The primary outcome model to estimate the association between symptomatic medically attended laboratory-confirmed SARS-CoV-2 infection and vaccination status was a multivariable logistic regression model, with SARS-CoV-2 test result (i.e., case-control status) as the dependent variable and vaccination status as a dichotomous independent variable. Medical encounter observations were weighted by their inverse propensity to be vaccinated (if vaccinated) or unvaccinated (if not vaccinated). Four covariates were also directly included as additional independent variables in the regression model to account for possible residual confounding that remained after inverse propensity score weighting based on BRT modeling. The four variables were age (as a spline), calendar date (as spline), geographic region, and local SARS-CoV-2 circulation on the day of each medical encounter index date (as a spline). Spline functions for calendar date, local SARS-CoV-2 circulation, and age were defined as natural cubic splines with knots at quartiles. In addition, any other covariates with distributions that remained imbalanced between vaccinated and unvaccinated patients after inverse propensity score weighting, based on an absolute standardized mean or proportion difference >0.2, were also included directly in the respective regression model. The list of unbalanced variables for each model is presented in eTable 3.

Section 2.5: Relative VE
In addition to calculating ORs to estimate absolute VE (i.e., VE for receipt of vaccine compared with unvaccinated status), ORs were also calculated to estimate relative VE, for which a specific vaccinated group was compared with a different vaccinated group in order to determine the incremental benefit of receiving an additional vaccine dose when recommend. Relative VE was estimated by comparing individuals who had recently received a 3 rd or 4 th dose to those who were eligible for, but had not received, the 3 rd or 4 th dose, respectively. More specifically, the two comparisons were: (1) 3 doses with the 3 rd dose in the last 7-119 days versus 2 doses with the 2 nd dose ≥150 days earlier; and, among patients aged ≥50 years, (2) 4 doses with the 4 th dose within the last 7-119 days versus 3 doses with the 3 rd dose ≥120 days earlier.
To calculate ORs reflecting relative VE comparisons, with receipt of 2 or 3 doses serving as the referent group, a similar methodology was used: patient encounters were weighted based on their inverse propensity to be 3dose vaccinated (if 3-dose vaccinated) or 2-dose vaccinated (if 2-dose vaccinated)or 4-dose vaccinated (if 4-dose vaccinated) or 3-dose vaccinated (if 3-dose vaccinated)and a dichotomous variable for vaccination status (3versus 2-dose vaccinated or 4-versus 3-dose vaccinated) was included as the independent variable used in primary outcome models.

Section 2.6: Subgroup Analyses
The analyses described were conducted in each setting (ED or UC encounters, hospitalizations, and hospitalizations with ICU admission and/or in-hospital death) and in different subgroups within each setting. Analyses were conducted among all adults aged ≥18 years as well as separately among three age groups (18-49, 50-64, and ≥65 years). They were also conducted separately for each vaccine product(s) received ( Abbreviations: ED, emergency department; EHRs, electronic health records; N/A, not applicable; UC, urgent care. a Each site defined sub-regions that represent meaningfully distinct geographic areas within their network. Sub-region values were assigned to medical encounters based on the location of the admitting hospital, ED, or UC clinic and were used for purposes of adjustment for geographic region in multivariable regression modeling. b The vaccine record lag is the duration of time post-vaccination before vaccine records are expected to be available in sites' records contributing to data collection for this study. c For Baylor Scott & White Health, inpatient facilities were located in only 8 of the 9 sub-regions. d CARE Everywhere is an Epic electronic health record inter-hospital system for vaccination record sharing. e For Regenstrief Institute, one of the 10 sub-region values represents unknown location of facility, and inpatient facilities were located in only 8 of the 9 defined sub-regions.

eTable 2. COVID-19-Like Illness Categories and Corresponding International Classification of Diseases, Ninth Revision (ICD-9), and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), Diagnosis Codes
Description of Diagnosis ICD-10 codes ICD-9 codes  1, 2021), and local virus circulation (percentage of SARS-CoV-2-positive results from testing within the counties surrounding the facility on the date of the encounter) and weighted for inverse propensity to be 3-dose or 2-dose vaccinated or 4-dose or 3-dose vaccinated (calculated separately for each VE estimate). Generalized boosted regression trees were used to estimate the propensity to be 3-dose or 4-dose vaccinated based on the following socio-demographic, facility, and medical factors: age, sex, race, ethnicity, Medicaid status, calendar date, geographic region, local SARS-CoV-2 circulation on the day of each medical visit, urban-rural classification of facility, chronic respiratory condition, chronic non-respiratory condition, asthma, chronic obstructive pulmonary disease, other chronic lung disease, heart failure, ischemic heart disease, hypertension, other heart disease, stroke, other cerebrovascular disease, diabetes type 1, diabetes type 2, diabetes due to underlying conditions or other specified diabetes, other metabolic disease (excluding diabetes), clinical obesity, clinical underweight, renal disease, liver disease, blood disorder, dementia, other neurological/musculoskeletal disorder, Down syndrome, and the presence of at least one prior molecular or rapid antigen SARS-CoV-2 test record documented in the electronic medical record ≥15 days before the medical encounter date (pre-vaccination).