Ventilation strategies based on an aerodynamic analysis during a large-scale SARS-CoV-2 outbreak in an acute-care hospital

Background This study aimed to investigate ventilation strategies to prevent nosocomial transmission of coronavirus disease 2019 (COVID-19). Methods We conducted a retrospective epidemiological investigation of a severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) outbreak in a teaching hospital (February–March 2021). The largest outbreak ward was studied, and measurements were taken to determine the pressure difference and air change per hour (ACH) of the rooms. Airflow dynamics were assessed using an oil droplet generator, indoor air quality sensor, and particle image velocimetry in the index patient's room, corridor, and opposite rooms, by varying the opening and closing of windows and doors. Results During the outbreak, 283 COVID-19 cases were identified. The SARS-CoV-2 spread occurred sequentially from the index room to the nearest room, especially the opposite. The aerodynamic study demonstrated that droplet-like particles in the index room diffused through the corridor and the opposite room through the opening door. The mean ACH of the rooms was 1.44; the air supply volume was 15.9% larger than the exhaust volume, forming a positive pressure. Closing the door prevented diffusion between adjacent rooms facing each other, and natural ventilation reduced the concentration of particles within the ward and minimised their spread to adjacent rooms. Conclusions Spread of droplet-like particles between rooms could be attributed to the pressure difference between the rooms and corridor. To prevent spread of SARS-CoV-2 between rooms, increasing the ACH in the room by maximising ventilation and minimising the positive pressure through supply/exhaust control and closing the room door are essential.


Introduction
In a hospital environment, healthcare-associated infection outbreaks likely occur owing to frequent patient and patient-healthcare worker interactions [1]. Due to risk factors, such as age and comorbid conditions, patients who become ill during a respiratory outbreak at a medical institution are more likely to develop a severe infection, which could lead to social and economic losses, such as prolonged hospitalisation, reduced long-term performance, and increased mortality rates [1][2][3].
An in-hospital epidemic occurred in the Republic of Korea (ROK) when respiratory infections were prevalent in the community owing to the H1N1 influenza pandemic in 2009, Middle East Respiratory Syndrome outbreak in 2015, and coronavirus disease 2019 (COVID-19) outbreak in 2020 [1,4,5]. In the hospital wards in the ROK, where most rooms are multi-bed rooms, a structural problem of the dense environment exists, which is vulnerable to the spread of infectious diseases [6]. Additionally, nosocomial severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2 outbreaks are a critical issue worldwide [7,8]. Therefore, establishment of a strategy by medical institutions to control respiratory infection outbreaks especially during the outbreak in the local community, is crucial [9]. Although a number of cluster infections have been reported in the hospital environment, as described above, limited studies exist predicting how infectious diseases spread through aerosol based on the aerodynamic information to guide specific preparedness strategies [10].
Our study aimed to identify the cause of large-scale outbreaks between February and March 2021 using epidemiological and aerodynamic investigations of a teaching hospital.

Study institution
The study institution was a 725-bed referral-teaching hospital. A total of 13 wards were present in three buildings. The main building consisted of 10 floors. Outpatient and emergency rooms were located on the 1st and 2nd floors, operating rooms on the 3rd floor, inpatient rooms on the 5th, 6th, 7th, 8th, and 9th floors, and a physical therapy room on the 10th floor. Each floor had 58-70 beds. Among them, 96.2% (304/ 316) were multi-bed rooms and 90.2% (285/316) were five-or sixpatient rooms. Two heating, ventilation, and air conditioning (HVAC) systems were installed in the main building of the study hospital. In the annex and mother-child health centres, one HVAC system was installed (Supplementary Figure 1).

COVID-19 screening policy
In the study hospital, patients were only admitted to the study hospital if they tested negative for SARS-CoV-2 on an RT-PCR test taken within 3 days prior to admission starting on 16 Apr, 2020. From 5 Jan, 2021, an RT-PCR test was conducted on the 5th day after admission, regardless of the presence of symptoms. Family guardians and caregivers could stay if they tested negative for SARS-CoV-2 within 3 days before admission. The 6th floor was a Comprehensive Nursing Service Unit where nurses provided care without individual caregivers. Other wards allowed one family guardian or caregiver per patient to prevent SARS-CoV-2 transmission.

Detection and response to the COVID-19 outbreak
Since the policy of routine screening of patients on the 5th day of admission implemented on 5 Jan, 2021, SARS-CoV-2 cases have not been detected. Two cases were first detected on 12 Feb 2023. Patient A and Patient B were admitted to the hospital on 6 Feb, 2021 and 7 Feb, 2021, respectively. They were scheduled to undergo medical treatment in the Orthopedic and Neurosurgery Wards, which are located on the 8th and 9th floors, respectively. Both of them did not have any respiratory symptoms. Patient A stayed in Room J of the 8th floor. They also tested negative for SARS-CoV-2 within 3 days prior to admission. On 12 Feb 2021, they tested positive for SARS-CoV-2 on the 5th day of admission. The corresponding Ct values were E gene 28.42 and RdRp gene 28.54 and E gene 23.08 and RdRp gene 24.24, respectively. On the same day, another SARS-CoV-2 case was detected among our patients, which could be related to the outbreak. Patient C sought medical advice by visiting the Emergency Medical Centre located on the 1st floor. He was tested positive for SARS-CoV-2, and Ct values demonstrated E gene 17.65 and RdRp gene 17.86. Patient C was previously admitted to the hospital in Room B of the 8th floor on 2 Feb, 2021 and was discharged on 8 Feb, 2021. After the epidemiological survey, we identified that the cases with the earliest date of COVID-19 related symptom onset were from Rooms A and B (8 Feb, 2021). However, the exact patient for the index case is not known.
All the patients in the wards on the 7th, 8th, and 9th floors of the main building, all the family guardians/caregivers, and healthcare workers were tested for SARS-CoV-2 on the same day. A total of three screening tests were performed for all healthcare workers (1st: 13-15 Feb, 2nd: 17-19 Feb, and 3rd: 24-26 Feb). If a patient was confirmed to be positive for COVID-19 in a shared room, the patient was quarantined in a single room. The 8th floor became a dedicated cohort ward for COVID-19 patients, and the outbreak was contained by 4 Mar, 2021.
As of 11 Feb, 2021 the day before the outbreak, 403 new patients with COVID-19 were recorded in the ROK, and the cumulative number of confirmed cases was 82,837. In Seoul Metropolitan City, where the study institution is located, 158 cases were recorded. Vaccination against COVID-19 was initiated on 26 Feb, 2021 in the ROK. Therefore, no individuals had received SARS-CoV-2 vaccinations at the time of the outbreak.

Data collection
We retrospectively collected the investigation results of the COVID-19 transmission outbreak in the hospital from 12 Feb to 5 Mar, 2021. The estimated exposure locations were classified according to the inhospital and community transmission. In-hospital transmission was defined and classified as transmission from patients, family guardians, caregivers, or healthcare workers in hospital based on the epidemiological investigations, and only includes those who tested negative for SARS-CoV-2 at the time of admission. Cases that spread to family members or acquaintances outside of the hospital after diagnosis with COVID-19 were classified as community transmissions. Exposures in the dialysis rooms, outpatient clinics, and operating rooms were classified as others.

Aerodynamic research
The main building had a ventilation system for supply and exhaust with a capacity of approximately six times per hour compared to that in the ward; the toilet in the ward hallway consisted of a separate exhaust system. To determine the vertical airflow through the elevator shaft of the main building, a differential pressure gauge and an anemometer were used to measure the pressure difference and flow velocity for the elevator doors of each floor. The airflow around the ward on the 8th floor of the main building was visualised, and the diffusion characteristics of the particulate matter were analysed. The diffusion characteristics of the particle matter were measured under the following conditions: windows closed, patient room door opened, windows and patient room door closed, and windows open and patient room door open. The measurement locations were as follows: the centre of Room A (Room A) where the first patient with COVID-19 related symptoms was located (S1), the middle of a nearby hallway (S2), and the centre (S3) of an adjacent room (Room B) (Fig. 1). An oil droplet generator (TSI 9307) with a size of 0.3-10 μm, similar to the size of a droplet or droplet nuclei generated in the patient's respiratory tract, detection sensor of particle concentration by particle size (TSI 3330; measurement height, 1.2 m), and particle image velocimetry (PIV) were used to visualise the airflow and measure the concentration of diffusion of particulate matter in each space to identify the possibility of infection by air diffusion in the 8thfloor ward (Supplemental Figures 2, 3, and 4).

Ethical statement
The Institutional Review Board and Ethics Committee of the study hospital approved this study (No. 2021-10-014) and waived the requirement for informed consent.

Demographics and distribution of the COVID-19 outbreaks
During the outbreak, 283 COVID-19 cases were identified. The median patient age was 53 years (interquartile range, 34-67 years). A total of 175 women (61.8%) were included. The largest number of cases in the wards was on the 8th floor of the main building with 85 cases, followed by 34 and 23 cases on the 9th and 7th floor of the main building, respectively. The highest number of COVID-19 infections were observed in the family guardians (N = 122), followed by patients (N = 83), healthcare workers (N = 43), and caregivers (N = 21) ( Table 1). A total of 112 (39.6%) and 171 (60.4%) cases were in-hospital and community transmissions, respectively (Fig. 2). Of the 283 cases, 112 (39.6%) were asymptomatic at the time of initial testing, except for 10 patients whose symptoms were unknown on the day of the diagnosis. It was observed that the transmission of the SARS-CoV-2 virus initially occurred to the neighboring rooms and then spread in a sequential order based on the dates of symptom onset or confirmation of infection in asymptomatic cases in the 8th floor of the main building (Fig. 3). A similar pattern was observed in the other wards in the main building (Supplemental Figure  5).

Characteristics of vertical and inter-floor airflow diffusion in the main building
The outdoor temperature at the time of on-site measurement was approximately 4.4 • C (39.9 • F), and the vertical pressure difference and airflow could result in the elevator hall owing to the stack effect attributed to the temperature difference indoors and outdoors. The neutral zone of the main building to be measured had six to seven floors. As a result of measuring the differential pressure and airflow velocity in the elevator door, airflow entered into the vertical elevator shaft through the elevator door from the elevator hall to the 1st-5th floors, and the airflow left the elevator hall through the elevator door from the elevator door to the 8th-10th floors. The airflow velocity of the elevator door gap was 0.28-1.1 m/s, and the differential pressure across the elevator door was 0. 8-5.7 Pa. This indicates that a small amount of the  indoor airflow on the 1st-5th floors can ascent through the elevator shaft and flow into the 8th-10th floors. However, the possibility of the spread of COVID-19 infection between the floors owing to vertical airflow was not high since the entrance door that divides the elevator hall and the hospital room (corridor) was installed separately on most floors.

Measurement result of the ventilation volume (frequency of ventilation) in the ward of the main building
The building had an HVAC system for air supply and exhaust with a capacity of air change rate of 6 times per hour in the ward; however, the actual supply and exhaust air volume of the 8th-floor ward were measured to be approximately 1.44 times per hour on average based on a 6-person room. The supply air volume was found to be larger than the exhaust air volume in all the measured rooms except room G. The supply air volume was 15.9% larger than the exhaust air volume on an average (Table 2). Accordingly, positive pressure was created in a ward without a toilet; thus, the contaminants in the ward could spread to other spaces.

Diffusion characteristics of a droplet-like particle surrounding the 8th-floor ward according to the opening and closing of the windows and doors
The conditions for all the hospital room windows to be closed and the door to the hospital room to be opened (usual conditions) shown in Fig. 4A and Table 3 (Case A) indicate the conditions of the study hospital (8th-floor ward). Especially in the winter season, all the windows of the hospital room are closed and ventilated using the HVAC system without natural ventilation; at this time, the door of the hospital room is usually open. In this case, particles generated from Room A flowed into the hallway and opposite Room B through the entrance door of the hospital room. The maximum value of the droplet particle, particulate matter (PM 2.5 ), concentration measured in the corridor was 15% of the maximum concentration in Room A, which rose to 4% in Room B.
The conditions for opening the window of the hospital room and closing the entrance to the hospital room (conditions for improvement) shown in Fig. 4B and Table 3 (Case B) are the proposed improvement operation plan: all doors of the hospital room should be closed, and all windows of the hospital room should be opened for natural ventilation in each hospital room. In this case, the particle concentrations of the corridor and Room B does not increase despite increase in the particle concentration of Room A.
The condition wherein the hospital room windows and doors are both open (improved ventilation conditions), was another proposed improvement operation plan shown in Fig. 4C and Table 3 (Case C): it is necessary to open the windows of the hospital room, entrance to the hospital room, and emergency stairs at both ends of the corridor to ensure cross-ventilation. In this case, the particles generated in Room A partially flowed into the corridor and Room B; however, the amount indicated 50% of the condition shown in Fig. 3A. In addition, as most of the particle matter exit through the window, they tend to gradually decrease over time, and the concentration in Room A could be reduced to the lowest level among the three conditions in which the experiment was conducted.

Discussion
This meticulous epidemiological and aerodynamic study identified nosocomial SARS-CoV-2 transmission according to the airflow. Despite performing routine universal screening tests before admission and 5-day follow-up tests, the occurrence of a large-scale outbreak could be attributed to the spread of the virus from room to room, especially in crowded spaces with poor ventilation. In the present study, we found that airflow attributed to the opening and closing of the hospital windows and entrance doors could be a possible route for SARS-CoV-2 transmission. As shown in the experimental study by visualising the spread of droplet-like particles, the droplet-like particles spread to the corridors and adjacent wards owing to the opening and closing of the external windows and external environment. We identified that in the study hospital, closing the entrance room and opening windows had the benefit of not transmitting particles to the corridor and other rooms. Occasionally, both windows and entrance rooms can be opened to maximise the natural ventilation effect.
According to an aerodynamic study of a hospital building conducted in France, airflow occurred outside the patient rooms owing to the heating effect of solar radiation in summer rather than winter [11]. The outbreak occurred during winter in the present study. The aerodynamic experiment demonstrated that positive pressure was formed in most patient rooms in the hospital. This could be an effect of a hot air fan positioned in a hospital room, and a change could be attributed to a difference in the amount of air intake and exhaust. Moreover, a hospital room located in the centre of the building could be less affected by the temperature of the outer hospital building wall; therefore, the airflow may vary. In medical institutions with vulnerable patients, regularly assessing the ventilation status and airflow in hospitals and preparing a ventilation strategy to prevent the outbreak of respiratory infectious diseases are necessary.
As observed in our study, even if positive pressure was produced in the ward, the droplet-like particles did not diffuse in the hallway or in the opposite direction if the ward door was closed. However, the concentration of droplet-like particles remained high when the outside window was closed in the hospital room. Therefore, keeping the door on the corridor side closed and frequently ventilating the outside window could be a possible strategy to minimise infection in the ward. According to previous studies, the probability of transmission of SARS-CoV-2 in the same room is very high (approximately 30-40%) [12,13]. Therefore, if a patient or caregiver has respiratory symptoms, wearing a mask and testing for COVID-19 promptly during the COVID-19 epidemic are necessary. In addition, natural or mechanical ventilation alone may not be sufficient to prevent transmission; therefore, the application of air purifiers should be actively considered [14].
In this study, the average air change rate of the ventilation in the ward was 1.44 times per hour. According to a study on the architectural design guidelines for medical institutions in the ROK, to satisfy 36 m3/h per person (based on the amount of external air intake), and based on the amount of air supplied, six times and greater air change rate per hour is recommended [15]. However, in the study hospital, since caregivers resided in most of the patient rooms, maintaining a greater air change rate should be considered. For this purpose, mechanical, natural, or mixed ventilation strategies are warranted. In the study hospital, we maximised natural ventilation by opening windows and closing doors at least three times a day with additional ventilation during mealtimes. Complete ventilation with all doors and windows open was recommended once a day. Interestingly, the number of COVID-19 cases was relatively small in the integrated nursing care ward, which was not densely populated. Therefore, to reduce the crowdedness in a situation where methods to maximise ventilation are limited, it is necessary to actively consider the introduction of changes in the medical environment that could minimise the number of caregivers or family guardians, such as an integrated caregiver service or restriction on the occupancy of caregivers.
The present study had a limitation. The epidemiological link was identified using contact investigations without phylogenetic analysis. However, on the floors in the hospital where the epidemiological investigation was conducted, the pattern of transmission was consistent with an epidemiological link, based on the distance from room of the index patient.
In conclusion, SARS-Cov-2 could be transmitted between different rooms. Despite conducting a screening test, a large-scale epidemic could occur in medical institutions with no appropriate ventilation strategy. To prevent the spread of various infectious diseases, establishing ventilation strategies that minimise the density suitable for each hospital is necessary.

Declaration of Competing Interest
All authors declare no competing interests.