Modeling the COVID-19 pandemic in King County Washington

This page illustrates modeling projections for SARS-CoV-2 infections in King County Washington during 2021. Model projections are based on documented assumptions about circulating variant prevalence and infectivity, potency of vaccines, vaccination rates, and approximations of exposure contact rate based on state social distancing policies in reaction to local incidence and mask use.

The model (black line below) was calibrated to diagnosed cases, hospital admissions, and deaths from King County in the period between January 2020 through March 2021 (points), stratified by age. Model projections extend beyond this date, with a vaccination program with protection starting on January 15, 2021 (colored lines below).

The darkest model lines represent our best estimate of future events but should not be interpreted as a precise forecast. Other lines convey different possible outcomes given different assumptions about degree of social distancing and vaccination rate. Projections are based on best available data but may change. New variants with increased infectivity and/or mortality may emerge. Vaccine efficacy may shift and prior immunity may wane. In particular, human behavior is difficult to predict and may alter future epidemic waves.

Model projections were last updated on April 28, 2021.

Details of model projections and parameters can be found on the Model description page. Detailed effects of specific management decisions and vaccine effects can be found on the Adjustable model projection page.


Model version 1.5


User adjustable version of the COVID-19 King County Model

This page allows the user to specify values for the model, including assumptions about emerging new variants, societal reopening, and vaccination rate.

The model was calibrated to diagnosed cases, hospital admissions, and deaths from King County in the period between January 2020 through March 2021 (black line below), stratified by age. Model projections extend beyond this date, including a vaccination program with protection beginning January 15, 2021 (colored lines below).

Management options are defined by the user.

Assumptions about viral variants are estimates and can be found in the table below and will be updated as new data accrues.

Details of model projections and parameters can be found on the Model description page.

Note that all simulations with the B.1.1.7 variant remain below a two-week lockdown threshold of 500 cases per 100,000 people. Therefore this threshold level approximates a policy with no further lockdowns.


Model version 1.5


Results from COVID-19 model (calibration and projection)

Management option A
Management option B
Management option C

Selected outcome

The selected outcome over time for a given variant (described in the table below), under three different management selections.

Individuals fully vaccinated

Our model contains several assumptions about vaccination. 1) A fixed number of doses are given each day, which corresponds to fully vaccinated individuals. 2) Coverage is not complete. Children (<20 years old) are not vaccinated, and only 80% of all individuals accept the vaccine. 3) Vaccines are prioritized to the elderly (>70 years old).

Social distancing: contacts relative to pre-pandemic

Social distancing follows the logic of the model description. We use a social distancing parameter that varies from 0 (pre-pandemic interactivity) to 1 (complete lockdown, no contacts). This parameter captures all non-pharmaceutical interventions including closure of businesses/schools, mask wearing, sanitization, outdoor meetings, and general reduction in contacts. When case levels exceed certain thresholds (see panel immediately right), we impose reductions in contacts through partial lockdowns and increase the value of the parameter.

Visualization of cases and case thresholds

Illustration of social distancing implementation. When cases rise above the lockdown threshold or drop below the reopening threshold, social distancing is increased or decreased, respectively.

Model parameter values for general management scenarios

King County COVID-19 Modeling Group

We are an academic group led by Fred Hutchinson Cancer Research center researchers. We are funded by the CTSE and the NIH.

Our work is dedicated to modeling the ongoing SARS-CoV-2 pandemic, with a particular focus on matching and projecting local data from King County Washington, USA.

The model considers the epidemiology of COVID-19 including social distancing, vaccination, and emerging variants.

Modeling team

Chloe Bracis (Grenoble)

David Swan (Fred Hutch)

Mia Moore (Fred Hutch)

Daniel Reeves (Fred Hutch)

Eileen S. Burns (Indepedent)

Dobromir Dimitrov (Fred Hutch/UW)

Joshua T. Schiffer (Fred Hutch/UW)

This website was created by Eileen S. Burns


Model version 1.5


COVID-19 Epidemiological model

We have developed a mechanistic mathematical model to describe the epidemiological dynamics of COVID-19 since March 2020.

Model fitting is detailed here (link to be added).

Mathematical details are presented below.

Fig 1. Detailed model schematic. The model is a modified SIR model that includes age groups, different types of vaccination, and different variants.

Fig 2. Modeling vaccination. We allow for a constant number of vaccination doses to be delivered daily. For example, President Biden has said 1.5 million doses per day. King County population 2.2mil and USA population 330mil means 0.7%β‰ˆ10,500 vaccines will go to King County daily, and most must be given as 2 doses such that a coarse estimate of our daily rate would be ~5000 doses per day. We assume that 80% of early doses will be given to elderly, and include the possibility that some individuals will not get vaccines using the maximum coverage. Once elderly are vaccinated to this coverage, 100% of vaccines go to adults. Children are not vaccinated.

Fig 3. Vaccine mechanisms. Three variables measure vaccine efficacy, including protection against infection (VESUSC), against symptoms given infection despite vaccination (VESYMP), and against secondary transmission despite vaccination (VEINF). Each of these vaccine efficacies theoretically ranges from 0-100% (0-1 in our model). Vaccine efficacy against symptomatic disease (VEDIS), which represents a combination of VESUSC and VESYMP, was measured in clinical trials and is 0.95 for the two Pfizer and Moderna licensed vaccines, though it is not known if these results were driven primarily by VESUSC or VESYMP. VEINF could lower secondary transmission considerably if VESUSC is low, but its values has also not been measured. Moreover, the efficacy of the Pfizer and Moderna vaccines against new variants remain unknown. The efficacy of the Novavax and Johnson and Johnson products were 90% and 69% against the consensus variant but decreased against the South Africa B.1351 lineage. We therefore consider low (0.1), medium (0.5) and high (0.9) values for each these variables.

Fig 4. Partial lockdowns. We expect social distancing levels 𝜎t will continue to vary according to government policy and public behavior. We include several parameters to reflect this uncertainty. The first is the case threshold to trigger partial lockdown (Cmax). If two-week number of diagnosed cases per 100,000 people rises above Cmax, a β€œpartial-lockdown” (𝜎tβ†’πœŽPL) is mandated. Partial lockdown is defined by an enforced social distancing of 𝜎PL=0.6 in the 3 younger age cohorts and 𝜎PL=0.8 in seniors(>= 70 years). These values reflect prior model estimates of 𝜎t during 2020 partial lockdowns. Values of Cmax were selected based on previous partial lockdowns implemented in Washington State. We vary Cmax in our analysis between 200 and 650, both to reflect heterogeneity in severity of the ongoing third wave across states which occurred due to implementing partial lockdowns at varying thresholds, and to represent future uncertainty. 𝜎min is the level of social distancing maintained after a period of societal reopening. Unless otherwise noted, this value is maintained at 0.2 to capture persistent features such as masking, work from home and avoidance of large social gatherings, which inherently limit the number of interpersonal contacts relative to pre-pandemic levels. 𝜎t gradually decreases (at 10% every two weeks) back to 𝜎min once the two-week average of daily cases falls below Cmin. Cmin is the case threshold (two-week number of diagnosed cases per 100,000 people) below which 𝜎t is lowered gradually from 𝜎PL to 𝜎min in the 3 younger cohorts (𝜎min+0.2 in seniors). We test 20, 60 and 100 as possible values for Cmin.

Mathematical formulation


Model version 1.5



Rapid vaccination and early reactive partial lockdown will minimize deaths from emerging highly contagious SARS-CoV-2 variants.


The goals of SARS-CoV-2 vaccination programs are to maximally reduce cases and deaths, and to limit the amount of time required under lockdown. Using a mathematical model calibrated to data from King County Washington but generalizable across states, we simulated multiple scenarios with different vaccine efficacy profiles, vaccination rates, and case thresholds for triggering and relaxing partial lockdowns. We assumed that a contagious variant is currently present at low levels. In all scenarios, it rapidly becomes dominant by early summer. Low case thresholds for triggering partial lockdowns during current and future waves of infection strongly predict lower total numbers of COVID-19 infections, hospitalizations and deaths in 2021. However, in regions with relatively higher current seroprevalence, there is a predicted delay in onset of a subsequent surge in new variant infections. For all vaccine efficacy profiles, increasing vaccination rate lowers the total number of infections and deaths, as well as the total number of days under partial lockdown. Due to variable current estimates of emerging variant infectiousness, vaccine efficacy against these variants, vaccine refusal, and future adherence to masking and physical distancing, we project considerable uncertainty regarding the timing and intensity of subsequent waves of infection. Nevertheless, under all plausible scenarios, rapid vaccination and early implementation of partial lockdown are the two most critical variables to save the greatest number of lives.

Key results figure