Air cleaners and respiratory infections in schools: A modeling study using epidemiological, environmental, and molecular data

Abstract

Background: Using a multiple-measurement approach, we examined the real-world effectiveness of portable HEPA-air filtration devices (air cleaners) in a school setting.

Methods: We collected environmental (CO₂, particle concentrations), epidemiological (absences related to respiratory infections), audio (coughing), and molecular data (bioaerosol and saliva samples) over seven weeks during winter 2022/2023 in two Swiss secondary school classes. Using a cross-over study design, we compared particle concentrations, coughing, and the risk of infection with vs without air cleaners.

Results: All 38 students (age 13-15 years) participated. With air cleaners, mean particle concentration decreased by 77% (95% credible interval 63%-86%). There were no differences in CO₂ levels. Absences related to respiratory infections were 22 without vs 13 with air cleaners. Bayesian modeling suggested a reduced risk of infection, with a posterior probability of 91% and a relative risk of 0.73 (95% credible interval 0.44-1.18). Coughing also tended to be less frequent (posterior probability 93%). Molecular analysis detected mainly non-SARS-CoV-2 viruses in saliva (50/448 positive), but not in bioaerosols (2/105 positive) or HEPA-filters (4/160). The detection rate was similar with vs without air cleaners. Spatiotemporal analysis of positive saliva samples identified several likely transmissions.

Conclusions: Air cleaners improved air quality, showed a potential benefit in reducing respiratory infections, and were associated with less coughing. Airborne detection of non-SARS-CoV-2 viruses was rare, suggesting that these viruses may be more difficult to detect in the air. Future studies should examine the importance of close contact and long-range transmission, and the cost-effectiveness of using air cleaners.

Keywords: air cleaner; airborne transmission; molecular detection; respiratory viruses; schools.

Fig 1.. Study setting.

Schematic study setup of the classrooms. One air cleaner was placed in the front and one in the back of the classrooms. All devices were placed at the head level of the students when they were seated. Both classrooms lacked an active HVAC (Heating, Ventilation, Air conditioning) system, but they were ventilated naturally by opening windows at the discretion of the teachers.

Fig 2.. Comparison of outcomes with vs without air cleaners.

At the top of each plot, the posterior probability for a reduction with air cleaners is shown, based on the Bayesian model. (a) Daily average aerosol number concentrations as boxplots. (b) Number of respiratory cases. (c) Daily average number of detected coughs per minute as boxplots. (d) Number of positive saliva samples.

Fig 3.. Molecular detection of respiratory viruses and transmission network based on spatiotemporal analysis of students’ saliva samples.

(a) Number of positive saliva samples by virus. (b) Daily number of positive saliva samples (colored circles) and possible transmission chains within classes (directed arrows). Positive samples are linked if they belong to the same virus and are less than 1 week apart. Positive samples from the air and filters as blank squares aligned. IFB: influenza B, HRV: human rhinovirus, AdV: adenovirus, CoV: SARS-CoV-2, MPV: human metapneumovirus, PIV: parainfluenza virus.