Ventilation during COVID-19 in a school for students with intellectual and developmental disabilities (IDD)

Abstract

Background: This study examined the correlation of classroom ventilation (air exchanges per hour (ACH)) and exposure to CO₂ ≥1,000 ppm with the incidence of SARS-CoV-2 over a 20-month period in a specialized school for students with intellectual and developmental disabilities (IDD). These students were at a higher risk of respiratory infection from SARS-CoV-2 due to challenges in tolerating mitigation measures (e.g. masking). One in-school measure proposed to help mitigate the risk of SARS-CoV-2 infection in schools is increased ventilation.

Methods: We established a community-engaged research partnership between the University of Rochester and the Mary Cariola Center school for students with IDD. Ambient CO₂ levels were measured in 100 school rooms, and air changes per hour (ACH) were calculated. The number of SARS-CoV-2 cases for each room was collected over 20 months.

Results: 97% of rooms had an estimated ACH ≤4.0, with 7% having CO₂ levels ≥2,000 ppm for up to 3 hours per school day. A statistically significant correlation was found between the time that a room had CO₂ levels ≥1,000 ppm and SARS-CoV-2 PCR tests normalized to room occupancy, accounting for 43% of the variance. No statistically significant correlation was found for room ACH and per-room SARS-CoV-2 cases. Rooms with ventilation systems using MERV-13 filters had lower SARS-CoV-2-positive PCR counts. These findings led to ongoing efforts to upgrade the ventilation systems in this community-engaged research project.

Conclusions: There was a statistically significant correlation between the total time of room CO₂ concentrations ≥1,000 and SARS-CoV-2 cases in an IDD school. Merv-13 filters appear to decrease the incidence of SARS-CoV-2 infection. This research partnership identified areas for improving in-school ventilation.

Fig 1.. Estimation of air changes per hour using ambient CO₂ levels.

Example of CO₂ time series measurement, peak identification and fitting. A: CO₂ levels versus time in curve with Gaussian smoothing window of 10, and peak (red triangle) and valley (blue triangle) identification. Segments with a long tail were truncated at the point where subsequent readings were only 2% lower than the mean of the previous 10 readings. B. Eq 1 was fitted to each set of interval data using NonlinearModelFit function in Mathematica, and the Levenberg-Marquardt algorithm option. In those rooms with multiple (peak, valley) segments, the final room ACH value was estimated by averaging the estimates from all segments.

Fig 2.. Capacity and volume of study rooms

A. Distribution of maximum room occupancy in persons for the 100 rooms in the study. B. Room occupancy versus volume in ft³. Most rooms had ceiling heights of 9–10 feet, and the relationship between room volume and area is close to linear. Larger rooms had higher occupancy levels.

Fig 3.. CO₂ concentration measurements over time in school rooms

A. Selected patterns of CO₂ concentration over the school day. Color coding indicates CO₂ concentration stratification by concentration bands commonly used by the American Society of Heating, Refrigerating and Cooling Engineers (ASHRAE) (green, < 1,000 ppm; yellow 1,000 – 2,000 ppm; red, geq 2,000 ppm). Minimum (blue) and maximum (red) CO₂ concentration versus B. Room volume, C. Estimated ACH, and D. Maximum room occupancy.

Fig 4.. CO₂ level exposure

The distribution of CO₂ exposure during an 8 hour school day. Stacked plots indicate the amount of time spent at each CO₂ level during the school day. Horizontal grid lines indicate quartiles. Each bar represents a room, with labels coded by building. Rooms are labeled by function: classroom, therapy, treatment, activity, gym, office, kitchen, break, copy, music, and technology.

Fig 5.. Peak CO2 levels by room functional type and building

Each plot displays a density distribution, along with individual measurements, for the room types indicated. Horizontal grid lines indicate ranges of CO₂ which are acceptable (green), moderately elevated (yellow), and high (red). Distributions for room types are further categorized by building.

Fig 6.. CO₂ levels and room temperature

Plots of CO₂ levels (left axis) and temperature (right axis) showing representative room patterns. (A-C) show rooms that had CO₂ levels ≥2000 ppm, (D-F) had CO₂ peak levels 1000–2000 ppm, and (G-I) had CO₂ ≤1000 from 7AM-7PM. CO₂ are coded by ≤ 1,000 ppm (green), 1,000–2,000 ppm (yellow), and > 2,000 ppm (red). Temperature is represented by the blue line.

Fig 7.. Estimates of air changes per hour (ACH)

(A) Distribution of estimated ACH. Only four rooms had an estimated ACH ≥2.0. The recommended ACH for schools is ≥ 4.0 exchanges of fresh air per hour [5, 9]. (B) Each ACH measure was used to model the decrease in CO₂ levels over time from a starting concentration of 1200 ppm CO₂ (gray lines). Colored reference lines show expected CO₂ clearance with 0.1, 0.5, 1.0, 2.0, and 4.0 ACH. (C) Estimated room ACH plotted against recommended ACH for each room based maximum room occupancy as recommended for school rooms American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) standard 61.1 calculated from room area and occupancy [5]. Dashed line is where $AC{H}_{measured}=AC{H}_{recommended}$.

Fig 8.. Positive PCR tests per room

(A) Difference in PCR tests per room, normalized by room occupancy, between rooms with HVAC systems using MERV-11 (red) versus MERV-13 filters. Statistical comparison with the Mann-Whitney U-test ($\mathrm{p}<0.0012$). (B) The time in each room where the ambient CO₂ ≥ 1,000 ppm was plotted against positive PCR tests in that room normalized by mean room occupancy. Rooms with CO₂ ≤ 1,000 ppm for the entire day were excluded (lighter markers). MERV filter status for the building rooms is shown (red, MERV-11; blue, MERV-13). Linear regression (dashed line) had an $R^2=0.46$, with an $R^2=0.4537$ explaining 45% of the variance. (C) ACH versus positive SARS-CoV-2 PCR tests per room normalized to room occupancy. No statistically significant correlation was found $\left(R^2=0.0036\right)$.