On-site isolation of patients with suspected COVID-19 in multiple-bed rooms, applied over a restricted timeframe, was not associated with secondary transmission events in our setting and could represent an adequate alternative isolation strategy, especially during a shortage of single-bed rooms and/or lacking availability of rapid SARS-CoV-2 QNAT results.

A similar approach was previously implemented for other viral respiratory infections and revealed similar results [9, 10]: Birrer et al. evaluated the introduction of on-site droplet precautions in a tertiary-care center during the 2018/19 influenza season and did not find an increased rate of nosocomial infections as compared with the standard single-room isolation strategy [9]. To our knowledge this is the first study reporting data on the safety and feasibility of this strategy in the setting of suspected COVID-19. In line with the World Health Organization recommendations [11], we implemented a comprehensive testing approach, thus only a small proportion of patients with symptoms compatible with COVID-19 were confirmed to have a SARS-CoV-2 infection at our center. The on-site isolation strategy allowed us to avoid single-room occupancy and prevent the subsequent in-hospital relocation of a large number of patients without confirmed infection.

The secondary attack rate among the control group was similar to previous studies performed during the same period in hospital and community settings [12, 13]. According to our results, the main determinant of SARS-CoV-2 transmission seems to be the viral load of the index patients rather than the contact time or the proximity of the contact patients to the index in the same room. In this respect, the lower viral load observed among the index-patients in the on-site isolation cohort, could have significantly contributed to the success of this strategy. This difference in viral burden, however, is in accordance with the natural course of COVID-19: patients with SARS-CoV-2 infection usually develop an exacerbation of symptoms requiring medical attention in late stages of the disease when the viral load and thus transmissibility is reduced and immunological pathogenesis is predominant [14]. On the other hand, index-patients of the control group consisted mainly of nosocomial infections and/or pre- or paucisymptomatic phase, generally related with high viral burden [15, 16]. Longer contact times, previously identified as risk factor for SARS-CoV-2 transmission [13], seemed to play a less prominent role in our cohort. However, median contact time among the on-site isolated patients was significantly shorter as compared with the control group. Nowadays, the broad availability of rapid SARS-CoV-2 QNAT results would allow for even shorter exposure times, reinforcing the safety of this approach.

Interestingly, among the pairs of index-contact patients with sequencing results available, only 57% (4/7) of the transmission events could be supported by phylogenetic data. Similar results were found in an outbreak investigation carried out at our center, where approximately half of the transmission events were confirmed by WGS analysis, suggesting that the remaining cases most likely reflect community-acquired infections randomly detected by broad screening efforts [17].

This study has several limitations including the small sample size and the retrospective design. Formally, the clinical impact of the on-site isolation would have been better evaluated through the comparison with a control group of patients with suspected COVID-19 submitted to conventional pre-emptive isolation measures. Such a control group was not available given the COVID-19 surge during the study period and the early implementation of the on-site isolation strategy. The lack of transmission events recorded in our study, points to the challenges in terms of the required sample size to adequately power a well-designed interventional study. Estimates of the secondary attack rate among the control group may have been hampered by the presence of preexisting asymptomatic or presymptomatic infection. Therefore, transmission events were first postulated based on clinical plausibility (exposure and chronological onset of symptoms and/or test positivity) and then confirmed or ruled out through sequencing, if possible. Further, this study was performed during the circulation of the wild type and pre-alpha variants of SARS-CoV-2 and in the pre-vaccine era limiting its generalizability to other variants with potentially higher transmissibility [16, 18] and populations with higher levels of immunity. Nevertheless, a continuous risk assessment of this strategy at our center did not point to an increase of healthcare acquired SARS-CoV-2 infections. SARS-CoV-2 testing was only performed in a small proportion of the contact patients in the on-site isolation group. This may be due to several reasons, such as earlier hospital discharge or more difficult follow-up in case of room or ward transfer linked to the lack of proper labelling of these patients. This aspect could introduce a detection bias and, indeed, if missing data were excluded from the analysis, there would be no significant difference in the rates of confirmed secondary transmission between the two groups. However, the hospital’s testing site was the most important testing site within the city at that time, increasing the likelihood of detecting patients with symptomatic infections also in an outpatient setting. Sequencing results were not available for all potential transmission events and therefore the secondary attack rate in the control group could be overestimated. In favor of patient safety, we choose to err on the side of assuming transmission at our institution had occurred rather than not occurred. However, the only epidemiologically suspected transmission event in the intervention group could be ruled out, supporting this approach. We further acknowledge that the laboratory developed nucleic acid test used for SARS-CoV-2 quantification was not calibrated against an international reference standard [19].