Making Polling Places Safer From COVID-19
The pandemic has raised unprecedented fears of voting in person in the upcoming election. Despite these fears it has been predicted that a record turnout of perhaps 145 million votes will be cast, two thirds in person in largely indoor polling buildings .Using a virus transmission model developed at the University of Colorado, an analysis was made of the effects of increasing ventilation and restricting occupancy at polling places as a method of increasing the safety of such in-person voting.
More than 230,000 polling places were used in the 2018 general election, according to the 2018 Election Administration and Voting Survey (EAVS). The EAC also reported that less than 1% of those locations were election offices—the vast majority of polling places were at other sites, such as schools or churches.
The Center for Disease Control(CDC) has made a number of recommendations to improve the safety of in person voting at such locations. Among these are limiting the number of voters in the facility.The Association of Heating, Refrigerating and Air Conditioning Engineers also provided guidelines geared to the ventilation control of polling places:
Select a space with larger area for people to spread out
A good supply of outside air is a first line of defense against aerosol transmission of SARS-CoV-2.
The number of buildings involved and the fact that the major voting event is of relatively short duration limits the practical measures that can economically be employed to improve voter safety. Assuming that social distancing and mask wearing is enforced, in practice the only controls implementable on short notice are to increase the ventilation of enclosed polling spaces and to restrict the maximum number of voting occupants in the polling space at a given time.
The use of an inexpensive CO2 room monitors ($50-100 cost) in polling places may allow a rough guide to indicate how well such controls are affecting the risk of infection. Operationally if the CO2 concentration in a room rises above a certain level that would indicate that the occupancy in the room should be decreased or ventilation should be increased.
The quantitative evaluation of the effects of ventilation and occupancy controls as well as the use of CO2 concentrations as an index of virus transmissability requires the use of a model indicating the relationship of ventilation measures, occupancy levels and CO2 concentrations to virus transmissability. In order to develop such relationships, a model developed by Dr. Jimenez at the University of Colorado was employed.
The Jimenez virus estimator models the propagation of COVID-19 by aerosol transmission only.Thus transmission by droplets is assumed controlled by the wearing of masks and social distancing. The model is based on a standard model of aerosol disease transmission, the Wells-Riley model. It is calibrated to COVID-19 based on recent literature. The output of the model is expressed as the probability of infection. A more detailed discussion of the estimation model can be found here, in Dr. Jimenez’s excellent webcast.
The model contains a large number of inputs,. This investigation employs many of the default assumptions employed by Dr. Jimenez in his examples of the model but focuses only on the effects of changing the size of the polling place, and its ventilation levels and occupancy..The length of time that a voter spends in the polling room was assumed to be 20 minutes. This does not account for any exposure which may take place waiting outside in long lines in order to be allowed entrance to the polling room.
The intent of this exercise is to model two standard types of rooms used as polling places- a school gymnasiums (dimensions of 110x60x24) and a large civic meeting room (60x40x15). The model inputs will employ two levels of air exchange-one air exchange per hour and six air exchanges per hour. (Air exchange per hour refers to a measure of the air volume added to or removed from a space divided by the volume of the space). Occupancy in the polling room was varied from 25 to 150 occupants.
These assumptions resulted in estimates of expected CO2 concentrations as well as the virus transmission probability of a voter contracting the virus due to the 20 minute voting experience.. It should be stressed that given the present state of knowledge, probability estimates are at best rough approximations to risk.. The intent is to try to give at least a first order picture of the relative impact of occupancy and ventilation controls on risk. Any .reliance on absolute risk estimates from the model must be viewed very cautiously.
Results and Discussion
The major result of the modeling effort was to quantitatively reinforce the CDC and ASHRAE recommendations. Figure 1 below compares the best and worst case conditions modeled. The worst case is voting in the smaller meeting space with lower ventilation. The best case is voting in the gymnasium with high ventilation. Both cases were examined over a range of occupancies. The combined effect of employing higher ventilation measures, occupancy limits and a larger polling place can significantly reduce the probability of a voter being infected. Even given the uncertainty of the probability estimates, the fundamental lesson is that preventive actions such as seeking larger polling sites, increasing ventilation or enforcing occupancy limits to reduce the risk of voting is justified.
Even if a polling site has been selected, ventilation and occupancy controls can yield sizable benefits. If the probability of infection for the meeting room is compared across different ventilation and occupancy assumptions, a striking drop in risk is indicated. If the risk of high occupancy(150 people), low ventilation conditions is compared to high occupancy and high ventilation, the model indicates a 35% drop in virus transmission probability. If worst case conditions of high occupancy, low ventilation, are compared to high ventilation, low occupancy conditions (25 people), a 90% reduction is indicated. Simple steps can yield benefits.
The use of CO2 concentrations as a guide to occupancy limits is feasible. For the meeting room with high ventilation rate case, Figure 2 shows the probability of infection related to estimated CO2 concentration levels as measured in parts per million(ppm). A background concentration of 400 ppm was assumed. In practice the CO2 level that triggers occupancy controls should be linked to the increase in concentration above background level at the specific polling place.
The lesson from the modeling effort is not related to the level of absolute risk of voting in person but rather that simple steps can be taken to make voting places safer. The authors would be remiss if we did not recognize the tremendous efforts of Dr. Jimenez in developing the virus estimation model.
