The variant‐specific burden of SARS‐CoV‐2 in Michigan: March 2020 through November 2021

Abstract Accurate estimates of the total burden of severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) are needed to inform policy, planning, and response. We sought to quantify SARS‐CoV‐2 cases, hospitalizations, and deaths by age in Michigan. Coronavirus disease 2019 cases reported to the Michigan Disease Surveillance System were multiplied by age and time‐specific adjustment factors to correct for under‐detection. Adjustment factors were estimated in a model fit to incidence data and seroprevalence estimates. Age‐specific incidence of SARS‐CoV‐2 hospitalization, death, vaccination, and variant proportions were estimated from publicly available data. We estimated substantial under‐detection of infection that varied by age and time. Accounting for under‐detection, we estimate the cumulative incidence of infection in Michigan reached 75% by mid‐November 2021, and over 87% of Michigan residents were estimated to have had ≥1 vaccination dose and/or previous infection. Comparing pandemic waves, the relative burden among children increased over time. In general, the proportion of cases who were hospitalized or who died decreased over time. Our results highlight the ongoing risk of periods of high SARS‐CoV‐2 incidence despite widespread prior infection and vaccination. This underscores the need for long‐term planning for surveillance, vaccination, and other mitigation measures amidst continued response to the acute pandemic.

require medical attention. Further, approximately one-third of all infections may be completely asymptomatic. 1,2 Under-detection is also expected to vary by age and over time related to testing availability and testing behaviors. 3 Seroprevalence studies can be helpful for surveillance and estimating the total burden of infection. 4,5 However, seroprevalence estimates provide a snapshot of past and recent infections that can be difficult to disentangle and can underestimate the cumulative incidence of infection due to waning antibodies. 6 The state of Michigan experienced four waves of SARS-CoV-2 transmission during the COVID-19 pandemic through December 2021.
Each pandemic wave has affected the general population and healthcare systems in different ways suggesting changing patterns of infection and severity by age. Michigan is also one of few states that experienced substantial transmission of both the SARS-CoV-2 Alpha (B.1.1.7 lineage) and Delta (B.1.617.2 lineage) variants. 7 We sought to quantify the burden of SARS-CoV-2 cases, hospitalizations, and deaths by age and geography over time in Michigan by integrating public health surveillance data, serial seroprevalence estimates, and genomic surveillance data. Burden estimates were used to examine how the risk of hospitalization and death varied over time by age and SARS-CoV-2 variant. We used a model, adapted from Shioda et al., 6 to estimate the cumulative incidence of infection from MDSS case incidence data and Michigan seroprevalence data from the CDC's Nationwide Commercial Lab Seroprevalence study while accounting for waning (Supporting Information Methods: Figure S1). 8  were taken as the 2.5th and 97.5th percentiles. As in Shioda et al., time from illness onset to seroconversion was assumed to follow a Weibull distribution with a mean of 11.5 days and SD of 5.7 days. 9 We estimated the average time from seroconversion to seroreversion to have a mean of 229.7 days (7.6 months) and SD 105.3 days by fitting a Weibull distribution, using a weighted least squares method, to published data on the duration of seropositivity measured by the Abbott ARCHITECT SARS-CoV-2 anti-nucleocapsid immunoglobulin G immunoassay (Supporting Information: Figure S2). 10,11 This assay was used in the Nationwide Commercial Lab Seroprevalence study in Michigan. 4 Daily MDSS cases were multiplied by the age-and time-specific case adjustment factors and their 95% CrI to estimate a range of total infections. Adjusted total infections were aggregated by age group and week to state and public health preparedness region ( Figure 1).

| Hospitalizations
Weekly, facility-level adult and pediatric inpatient admissions for confirmed or suspected COVID-19 were identified from HHS Protect. 12 Facilities were mapped to the public health preparedness region, and weekly admission counts were aggregated to state and region levels. Hospitalized cases were also identified from MDSS; because admission status is ascertained at the time of case investigation and individuals may not have been hospitalized yet, these data represent an underestimate of total hospitalizations, but age and geographic data are available. The total number of weekly admissions from HHS Protect was multiplied by week-and regionspecific age distributions of hospitalized MDSS cases to estimate total hospitalizations in each age group, region, and week.

| Deaths
Weekly estimates of deaths were made using National Center for Health Statistics (NCHS) excess death estimates 13 and confirmed and probable COVID-19 deaths reported to MDSS with those reported in the Michigan Department of Corrections system excluded. Weekly Michigan deaths were estimated as the total COVID-19 deaths reported to MDSS or the upper NCHS estimate of excess deaths, whichever was higher. The total combined MDSS/NCHS weekly deaths were multiplied by the week-and region-specific age distributions of deaths from MDSS to estimate the total deaths in each age group, region, and week.

| Variant prevalence
Proportions of Alpha and Delta variant and ancestral lineage SARS-CoV-2 viruses among all characterized viruses in Michigan were obtained from covariants.org for 2-week periods. 14 Counts of cases, hospitalizations, and deaths were multiplied by the week-specific variant proportions to estimate cases, hospitalizations, and deaths attributable to Alpha, Delta, and ancestral viruses.

| Vaccination
The proportion of the population receiving ≥1 dose of vaccine by age and region was calculated by week from publicly available data reported to the Michigan Care Improvement Registry. 15 Vaccination data were available by the following age groups: 5-11, 12-15, 16-19, 20-29, 30-39, 40-49, 50-64, 65-74, ≥75 years. Vaccination counts were reassigned to the analysis age groups described under Cases as follows. If a vaccination age group spanned two analysis age groups, vaccination counts were assigned to each of the analysis age groups according to the proportion of age years contained in each age group. For example, 2/3 of the vaccinations in the 50-64 year vaccine age group were attributed to the 50-59 year analysis age group and 1/3 to the 60-69 year analysis age group. Vaccination counts from the ≥75 years vaccination age group were assigned to the 70-79 and ≥80 years analysis age groups according to the actual proportion of ≥75-year-olds in Michigan who are also ≥80 years old (57%). had not peaked at the time of analysis. Age-specific estimated cases, hospitalizations, and deaths were plotted and compared across the waves, and by predicted variant status. Age-and region-specific proportions of cases who were hospitalized or died were also compared across waves. The effective reproduction number (R e ) was estimated for Alpha, Delta, and ancestral lineage viruses over rolling 2-week intervals using the R EpiEstim package with a serial interval of mean 5.68 and SD 4.77 days. 16,17 All analyses were completed using R (R Foundation for Statistical Computing; version 4.1.1). to be Alpha variant-specific; the likelihood of hospitalization did not vary by lineage in any age group in the spring ( Figure 4B) Figure 4A,C). This also did not appear to be driven by Alpha variant circulation in Spring 2021 as the proportion of cases that died was similar comparing Alpha and ancestral infections ( Figure 3D). Across all age groups, those with Delta variant infection in Fall 2021 were less likely to be hospitalized or die than those with ancestral or Alpha variant infections in previous waves.

| RESULTS
Cumulative incidence by age group was tightly clustered between 3% and 8% just before the Fall 2020 wave, with the 0-17 year age group most likely (8%) and the 70-79 year age group least likely to have been previously infected (3%) ( Figure 5A). Following the Fall 2020 wave, there was a wider range of age-specific cumulative incidence (13%-28%) that generally decreased with wave, the overall cumulative incidence was 40% and ranged from 22% to 52% by age group. By November 12, 2021, the overall cumulative incidence was 75% and ranged from 99% to 42% by age group. Cumulative incidence of infection did not vary as much by Michigan Public Health Preparedness Region as it did with age.
However, the more rural Regions 7 and 8 consistently had the lowest cumulative incidence throughout the pandemic ( Figure 5B).  . The total height of the bars is the total number of cases in each age group that is, hospitalizations and deaths were subtracted from the case counts. However, we do not know the proportion of hospitalized cases that died or the proportion of deaths that were not hospitalized so the combined height of those segments overestimates the total number of severe outcomes. SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.
Delta in summer 2021, and we estimated that the Delta R e was almost twice that of Alpha in periods of cocirculation. These estimates are somewhat lower than previous reports that found Alpha to be 40%-100% more transmissible than ancestral SARS-CoV-2 viruses, 18,19 but similar to reports of increased transmissibility of Delta relative to Alpha. 20,21 In the time since this analysis was carried out, Delta infections continued to rise in Michigan before being rapidly overtaken by Omicron resulting in record high daily infections. Michigan's experience in the winter of 2022 makes it clear that combined levels of prior infection and vaccination that exceed 80% are not sufficient to reach herd immunity. Suboptimal vaccine coverage, waning of natural and vaccine-induced immunity, and the emergence of more transmissible variants have facilitated ongoing transmission. [22][23][24] Indeed, it is unlikely that true herd immunity will be reached entirely ending SARS-CoV-2 transmission just as descendants of the 1968 and 2009 influenza pandemics continue to circulate today. This underscores the need for long-term planning in policy, public health capacity, and research priorities. These results also suggest that nonpharmaceutical mitigation measures may be needed during times of high transmission going forward.
We observed that the proportion of cases who were hospitalized or died decreased with each subsequent wave. Because of the ecologic nature of this analysis, we are unable to attribute reduced severity to a specific cause. However, our results do not suggest differences in the severity of Alpha and ancestral viruses in Spring 2021. Although differential severity has not been conclusively demonstrated, some studies have estimated Alpha and Delta are more severe than ancestral SARS-CoV-2, at least among unvaccinated individuals. 25,26 COVID-19 vaccines have been effective in preventing severe outcomes of infection. [27][28][29] This suggests that at least through the emergence of Delta, age-specific reductions  A notable exception to the general trend of decreasing severity is that adults <50 years were at increased risk of hospitalization in Spring 2021 relative to Fall 2020. It is possible that patterns in vaccine uptake could confound differences in severity over time.
Among those prioritized for early vaccination, younger adults with chronic conditions had a similar low intention to vaccinate as "essential workers" generally. 30 Racial disparities in both vaccine uptake and severe outcomes of SARS-CoV-2 have also been demonstrated. 31,32 If healthier adults were more likely to be vaccinated earlier, the remaining susceptible population may have been more likely to be hospitalized given infection. Factors associated with vaccine uptake warrant further investigation and consideration in analyses of differential severity.
Our results should be interpreted in the context of multiple limitations. (1) We were unable to account for reinfection, and vaccination was assumed to be independent of prior infection. (2) We relied on Michigan-specific data from the Nationwide Commercial Lab Seroprevalence study to calculate case adjustment factors to correct for under-detection. That study has its own limitations, 4