Airborne infection from aerosolized SARS-CoV-2 poses an economic challenge for businesses without existing heating, ventilation, and air conditioning (HVAC) systems. The Environmental Protection Agency notes that standalone units may be used in areas without existing HVAC systems, but the cost and effectiveness of standalone units has not been evaluated.

Cost-effectiveness analysis with Monte Carlo simulation and aerosol transmission modeling.

We built a probabilistic decision-analytic model in a Monte Carlo simulation that examines aerosol transmission of SARS-CoV-2 in an indoor space. As a base case study, we built a model that simulated a poorly ventilated indoor 1000 square foot restaurant and the range of Covid-19 prevalence of actively infectious cases (best-case: 0.1%, base-case: 2%, and worst-case: 3%) and vaccination rates (best-case: 90%, base-case: 70%, and worst-case: 0%) in New York City. We evaluated the cost-effectiveness of improving ventilation rate to 12 air changes per hour (ACH), the equivalent of hospital-grade filtration systems used in emergency departments. We also provide a customizable online tool that allows the user to change model parameters.

All 3 scenarios resulted in a net cost-savings and infections averted. For the base-case scenario, improving ventilation to 12 ACH was associated with 54 [95% Credible Interval (CrI): 29–86] aerosol infections averted over 1 year, producing an estimated cost savings of $152,701 (95% CrI: $80,663, $249,501) and 1.35 (95% CrI: 0.72, 2.24) quality-adjusted life years (QALYs) gained.

It is cost-effective to improve indoor ventilation in small businesses in older buildings that lack HVAC systems during the pandemic.

SARS-CoV-2 may be transmitted person-to-person via exhaled respiratory aerosols that accumulate within poorly ventilated spaces [

However, older buildings tend not to have HVAC systems installed. When HVAC systems are not present, the Environmental Protection Agency (EPA) recommends alternative means of disinfecting the air [

Given that standalone HEPA filtration units may be only marginally effective and are relatively expensive, we evaluated their cost-effectiveness. In this paper, we provide data for an example setting (a poorly ventilated restaurant), but our online model can be modified for any scenario.

We built a decision-analytic model that is designed to assist local and federal regulators in setting standards for improving the indoor air ventilation in poorly ventilated indoor commercial spaces for the prevention of SARS-CoV-2 infections via aerosolized particles. The model is designed to compute the incremental cost-effectiveness ratio (ICER), which is the net cost of an intervention divided by the number of quality-adjusted life years (QALYs) gained [

Each restaurant, café, or bar is unique with respect to the size, number of customers, hours of operation, and the time that customers spend in the establishment. This variation presents challenges for understanding the airflow and filtration needs for any given business. In this paper, we used a small, poorly ventilated restaurant space as an example so that the reader can get a general idea of the cost-effectiveness of standalone ventilation. In addition, our customizable interface can allow both regulators and restaurant owners to obtain estimates for a range of settings.

The standardized restaurant was open for a total of 3 h for lunch service and 6 h for dinner service. We assumed that the restaurant has a seating capacity of 30 occupants in a 1000 square foot space and a ceiling height of 9 feet, and that each occupant is seated for one-hour at lunch and 1.5-h during dinner. The model assumptions are listed in Table

Model assumptions for evaluating the cost-effectiveness of improving ventilation in commercial spaces for the prevention of SARS-CoV-2

Assumptions |
---|

The standardized room of 1000 square-foot with a ceiling height of 9 feet has 0.8 air changes per hour, primarily from the door opening and closing and the food vent running |

For lunch, the restaurant is open for 3 h. Each of 30 occupants is seated for one hour. We modeled 3 consecutive lunch events each for a duration of 1 h. In each event, the restaurant is at the full seating capacity |

For dinner, the restaurant is open for 6 h. Each of 30 occupants is seated for 1.5 h. We modeled 4 consecutive dinner events each for a duration of 1.5 h. In each event, the restaurant is at the full seating capacity |

Between lunch and dinner hours, the restaurant is closed for enough time so that the virus concentration in the indoor air dropped to zero as workers opened doors and moved throughout the space |

The restaurant is operating 7 days a week with similar lunch and dinner hours |

The model is built under well-mixed conditions for an infected individual present in an indoor space and there is dynamic airflow in unpredictable patterns associated with the movement of people and an overhead fan [ |

We assumed that transmission through the close-range mode—that is, when infectious aerosols were inhaled directly from the exhaled breath of an infected individual by a susceptible person in its vicinity—is on par between the comparison arms. Thus, only infection through the inhalation of accumulated aerosols, often referred to as the long-range mode of airborne transmission, is modeled and close-range transmission is not modeled [ |

We assumed that infected symptomatic Covid-19 cases would quarantine for 14 days. We also assumed those infected cases who required hospitalizations would quarantine for 21 days |

All wages were valued at the median hourly wage in the US [ |

In a previous published work, two co-authors developed a model [^{10} virus RNA copies/ml. While the mean virus RNA copies/ml of the infected sputum for the original strains of Covid-19 was estimated as 7 × 10^{6} [^{10} virus RNA copies/ml for the Delta variant. We then applied a conversion factor of 0.01 to estimate PFUs (infectious units) from RNA copies [^{2} restaurant area size with a ceiling height of 9 ft;

For an average susceptible individual sitting in the restaurant, we calculated the risk of SARS-CoV-2 infection based on the number of viral PFUs that the individual is exposed to for the duration of a lunch event (1 h) or a dinner event (1.5 h). A susceptible individual is defined as a person who is disease-free at the start of lunch or dinner service and is at risk of contracting the disease while sitting in the restaurant. We assumed that if

We calculated the risk of infection (denoted by

The above modeling approach considers only infection through the inhalation of accumulated aerosols, often referred to as the “long-range” mode of airborne transmission. Thus, transmission through the close-range mode—that is, when infectious aerosols were inhaled directly from the exhaled breath of an infected individual by a susceptible person in its vicinity—was assumed to be on par between the comparison arms and was disregarded.

We modeled the cost of installing standalone air filtration units with HEPA filters, which trap ultrafine particles down to the sub-micrometer size [

We modeled direct and indirect costs of hospitalizations due to Covid-19 [

Model input parameters for evaluating the cost-effectiveness of improving ventilation in commercial spaces for the prevention of SARS-CoV-2

Parameter | Base case value | Probability distribution |
---|---|---|

Number of people sitting in the restaurant at once | 30 | - (Changed in the sensitivity analysis from 20 to 40) |

Average age of the people sitting in the restaurant | 45 | - (Changed in the sensitivity analysis from 35 to 55) |

Probabilities and rates | ||

Probability of infection for one PFU unit exposed [based on ID50 of 280 (95% CI 130–530) PFU units] [ | 0.0024 | Beta (15.9592, 6633.707) |

Proportion of asymptomatic cases among all exposed people (excluding the ones initially asymptomatic but became symptomatic eventually) [ | 0.25 | Beta (18.5, 55.5) |

Probability of long Covid-19 among symptomatic cases [ | 0.133 | Beta (86.567, 564.3127) |

Infection hospitalization rate [ | Age-dependent: 0.019 for the average age of 45 years old | Beta (98.081, 5064.077) |

Infection mortality rate [ | Age-dependent: 0.001 for the average age of 45 years old | Beta (99.899, 99,799.1) |

Relative rate of symptomatic infection with Delta among the fully vaccinated (relative rate of 0.22 is equivalent of 78% reduction in symptomatic infection; the value represents the average effectiveness of the BNT162b2 and ChAdOx1 nCoV-19 vaccines against Delta variant) [ | 0.22 | Beta (14.8808, 52.7592) |

Direct costs (US dollars in 2020 USD) | ||

Improving room ventilation rate to 12 ACH (by installing 5 standalone air filtration units with HEPA filters trapping ultrafine particles down to the sub-micrometer size that are uniformly installed in the room and produce an equivalence of 12 ACH for a 1000 ft | $3750 | Gamma (100, 0.02667) |

Covid-19 hospitalization [ | $23,489 | Gamma (100, 0.00426) |

Indirect costs (U.S. dollars in 2020 USD) | ||

Covid-19 infection without hospitalization for symptomatic cases (losses of productivity over 2 weeks of self-isolation) | $2800 | Gamma (100, 0.036) |

Covid-19 hospitalization (losses of productivity over 3 weeks) | $4200 | Gamma (100, 0.024) |

Premature mortality due to Covid-19 (calculating losses of annual average wage of $50,000/year beyond the age at death of 45 years old in the base case model until the age of 65 years; future values were discounted at 3%) | $793,874 | Gamma (100, 0.000126) |

Health-related quality of life | ||

Losses of QALYs associated with a Covid-19 symptomatic case [ | 0.008 | Beta (99.192, 12,299.81) |

Losses of QALYs associated with a long Covid-19 infection [ | 0.034 | Beta (96.566, 2743.61) |

Losses of QALYs associated with a Covid-19 hospitalization [ | 0.020 | Beta (97.970, 4776.154) |

Losses of QALYs associated with a Covid-19 death (calculated based on an average age of 45 years at death, life expectancy of 80 years, age-dependent QALYs of the US general population, and discounting future values at 3%) [ | 18.33 | Normal (18.33, 1.83) |

We modeled losses of QALYs associated with a Covid-19 symptomatic infection and Covid-19 hospitalization [

We compared two interventions: (1) no improvement in the baseline ventilation rate of 0.8 ACH (‘status quo’), and (2) improving the room ventilation rate to 12 ACH. Our mathematical model was probabilistic and was developed in a Monte Carlo simulation of 5000 iterations, with each iteration randomly drawing from probability distributions of the input parameters. Table

We performed our analyses for different conditions defined by the mean year-round prevalence of actively infectious cases in the surrounding communities where the restaurant is located and the proportion of patrons that are vaccinated. For the base case model, we assumed a 2% mean year-round prevalence of actively infectious cases in the surrounding community and a 70% full-vaccination rate among customers sitting in the restaurant, defined as 2 doses of an FDA approved vaccine in the US. We modeled the random daily incidence rate from a normal distribution and summed the daily incidence rates over the past 12 days to obtain the daily prevalence of actively infectious cases. This assumes an average of 12 days of infectiousness for an exposed individual beginning 2 days prior to symptom onset (for symptomatic cases) plus 10 days following the initial symptom onset [

For the best-case scenario (minimum number of infections), we assumed a year-round prevalence of actively infectious cases of 0.1% in the surrounding community and a 90% full-vaccination rate among customers sitting in the restaurant.

For the worst-case scenario (maximum number of infections), we assumed a year-round prevalence of actively infectious cases of 3% in the surrounding community and a 0% full-vaccination rate among customers sitting in the restaurant.

The time horizon of the model was one year and the analytic horizon was lifetime. The outcomes of the model were incremental direct and indirect costs, infections averted, QALYs gained, and ICER for improving the ventilation rate. We also conducted one-way sensitivity analyses over all core input parameters of the model to measure the robustness of model outcomes against changes in these parameters.

(2% prevalence of disease in the surrounding community where the restaurant is located and 70% full-vaccination rate among of restaurant customers). Improving the room ventilation rate to 12 ACH was associated with 54 [95% credible interval (CrI): 29, 86] infections averted in the standardized restaurant over one year. This produced cost savings of $152,701 (95% CrI: $80,663, $249,501), and 1.35 (95% CrI: 0.72, 2.24) incremental QALYs gained. Table

Model outcomes including infections averted, incremental costs, incremental QALYs, and incremental cost-effectiveness ratio for upgrading the room ventilation rate from 0.8 to 12 ACH

Airborne infections averted Mean (95% credible interval) | Net cost ($) Mean (95% credible interval) | Cost saving ($) Mean (95% credible interval) | Losses of QALYs Mean (95% credible interval) | Incremental QALYs Mean (95% credible interval) | ICER | |
---|---|---|---|---|---|---|

Base-case scenario (2% prevalence of disease in the surrounding community where the restaurant is located and when 70% of the customers are vaccinated) | ||||||

Room ventilation rate of 0.8 ACH | $185,579 ($100,099, $300,430) | 1.6 (0.85, 2.66) | ||||

Improve room ventilation rate to 12 ACH | 54 (29, 86) | $32,877 ($19,394, $50,877) | $152,701 ($80,663, $249,501) | 0.25 (0.13, 0.42) | 1.35 (0.72, 2.24) | − $113,126/QALY (dominant, dominant) |

Best-case scenario (0.1% prevalence of disease in the surrounding community where the restaurant is located and when 90% of the restaurant customers are vaccinated) | ||||||

Room ventilation rate of 0.8 ACH | $6824 ($3524, $11,356) | 0.06 (0.03, 0.11) | ||||

Improve room ventilation rate to 12 ACH | 2 (1, 4) | $4821 ($3930, $5865) | $2003 (− $881, $5968) | 0.01 (0.01, 0.02) | 0.05 (0.03, 0.09) | − $38,104/QALY (dominant, $30,503/QALY) |

Worst-case scenario (3% prevalence of disease in the surrounding community where the restaurant is located and when no customer is vaccinated) | ||||||

Room ventilation rate of 0.8 ACH | $544,521 ($298,694, $875,492) | 4.35 (2.34, 7.14) | ||||

Improve room ventilation rate to 12 ACH | 135 (76, 213) | $89,243 ($50,540, $141,203) | $455,277 ($247,879, $734,424) | 0.68 (0.37, 1.12) | 3.66 (1.98, 6.02) | − $124,294/QALY (dominant, dominant) |

The model outcomes are calculated for the base-case scenario (mean year-round prevalence of 2% in the surrounding community where the restaurant is located and when 70% of the customers are vaccinated), best-case scenario (mean year-round prevalence of 0.1% in the surrounding community where the restaurant is located and when 90% of the restaurant customers are vaccinated), and worst-case scenario (mean year-round prevalence of 3% in the surrounding community where the restaurant is located and when no customer is vaccinated)

Negative ICERs in this table represent a cost-saving scenario, meaning the comparator intervention saves money and improves health

^{a}Quality-adjusted life years, which is equal to the product of the number of years of life gained and the health-related quality of life score

^{b}The incremental cost-effectiveness ratio (ICER) is equal to the incremental cost divided by the incremental QALYs gained

^{c}Air exchanges per hour. In this iteration of the model 0.8 is used as the baseline

(0.1% prevalence of disease in the surrounding community where the restaurant is located and 90% full-vaccination rate among restaurant customers). This scenario reflects conditions far better than were observed for the United States for the pandemic through October of 2021. It was associated with cost savings of $2,003 (95% CrI: − $881, $5968) and 0.05 (95% CrI: 0.03, 0.09) QALYs gained.

(3% prevalence of disease in the surrounding community where the restaurant is located and 0% full-vaccination rate among restaurant customers). In this scenario, improving the room ventilation rate to 12 ACH was associated with 135 (95% CrI: 76, 213) infections averted, $455,277 (95% CrI: $247,879, $734,424) savings in costs, and 3.66 (95% CrI: 1.98, 6.02) increases in QALYs gained.

Figure

One-way sensitivity analysis (tornado diagram) for each of the core input parameters of the model. The range of each value represents the incremental cost-effectiveness ratio associated with varying model input parameters over a range of plausible values for the base-case model scenario (1000 ft^{2} restaurant space, 2% prevalence of actively infectious cases, a 70% vaccination rate, and an upgrade from 0.8 ACH to 12 ACH).

Figure

The cost-effectiveness plane representing the incremental costs versus incremental QALYs for improving the ventilation rate of an exemplary 1000 ft^{2} restaurant space to 12 ACH for:

Because the variants of SARS CoV-2 are constantly changing (e.g., transition from Delta to Omicron BA.2), we conducted a scenario analysis to test the cost-effectiveness for different values of the airborne transmissibility of SARS CoV-2 (by changing the probability of infection per exposure to one viral PFU). When the airborne transmissibility was reduced by 50% (for variants with 50% lower rates of airborne spread), the standalone units were associated with $74,327 savings in costs and 0.67 increases in QALYs gained. When the airborne transmissibility was increased by 50% (for variants with 50% higher airborne spread), the standalone units were associated with $230,557 savings in costs and 2.02 increases in QALYs gained. Our online model allows the user to change the airborne transmissibility of SARS Cov-2 as new evidence emerges.

In this study, we evaluated the cost-effectiveness of improving ventilation in commercial indoor spaces using standalone HEPA filtration units as a method of preventing the transmission of airborne SARS-CoV-2. We built our probabilistic model using a Monte Carlo simulation so that the average model outcomes account for uncertainties and represent different ranges of variability in model input parameters and assumptions. Our probabilistic analyses showed that under all scenarios—even when the mean year-round prevalence of actively infectious cases was as low as 0.1% and 90% of the restaurant’s patrons were fully vaccinated—improving the ventilation rate of the indoor spaces by standalone air-filtration units would result in cost savings and QALYs gained. Our model was robust to changes across a range of inputs and assumptions, suggesting that policy mandates for HEPA filtration system use would be prudent in most situations.

There is a growing body of research modeling airborne transmission of SARS CoV-2 via aerosolized particles in indoor spaces [

The underpinning risk transmission model has been previously vetted [

Our study was limited in several ways. First, we modeled the hospitalization rate and mortality rate only as a function of age in line with the previous studies [

Even in the absence of SARS-CoV-2, poor ventilation systems in commercial spaces pose significant public health risks [

N/A.

ZZ developed the study idea, designed the study, performed the statistical analysis and programming, and wrote the first draft of the manuscript; PMDO and SG helped with the methodology, implementation, and interpretation of findings. PAM helped with the study design, implementation, and interpretation of findings. CE helped with data collection. All authors contributed to the writing. All authors read and approved the final manuscript.

This study did not have funding.

We build an accompanying online dashboard for the model. Other data will be available upon request from the authors.

This study was based upon modeling of previously published data and was exempt from ethics approval.

N/A.

Authors declared no competing interests. The study was requested by officials advising the governor of New York and the mayor of New York City and the last author is advising Columbia University on Covid-19 transmission models. No authors have any financial or other stake in the outcomes of the analysis.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.