We have evaluated the impact of reversing any or all of eight interventions and changes occurring in NHS England hospitals over wave 1 of the COVID-19 pandemic on rates of nosocomial transmission to patients and HCWs in England. The literature search conducted to parameterise the model in this study highlighted a paucity of evidence around the efficacy of interventions, particularly increasing space between beds and of cleaning and hand-hygiene on reducing nosocomial transmission of respiratory viruses. Also, while several studies exist exploring the effect of mask-wearing, many of them are single site studies and their applicability outside of each setting is unclear (Table S1/S2). Furthermore, there is a lack of evidence around compliance with mask-wearing by healthcare workers when in communal areas. This is a key issue noting the relatively high prevalence of environmental contamination by SARS-CoV-2 in staff communal areas. There is additionally anecdotal evidence that HCW to HCW transmission may occur on travelling to/from work or between colleagues while socialising outside of work.
While the baseline results are calibrated to high-quality national datasets, the scarcity of reliable evidence on the effectiveness of individual interventions (required for simulating a scenario in which they are reversed) is a key weakness of the results presented in this study. Additionally, while the model has been parameterised to best reflect the available data, uncertainty remains around the contribution of nosocomial and community sources of SARS-CoV-2 infection. As COVID-19 becomes an endemic disease and pressures on health systems from other seasonal respiratory pathogens increase, there is a critical need for evidence on the efficacy of such interventions on reducing nosocomial spread in order to design efficient and effective infection prevention and control strategies. We note also that the capacity to isolate patients in single rooms in English hospitals is limited. Approximately 80% of hospital beds are in multi-occupancy bays, and the remaining 20% are unevenly distributed within institutions, such that the great majority of patients (COVID-19 cases, unsuspected and suspected cases) were cohort nursed.
Due to the limited data available on effectiveness of interventions, further work will include conducting a formal elicitation of expert opinion alongside the simulation study. The elicitation exercise can be used to generate parameter distributions on, for example, compliance with and efficacy and feasibility of IPC measures, which can then be used to augment model inputs using evidence from literature, and allow further understanding on the uncertainty of model inputs and hence outputs.
While highly uncertain, given the dependency on (scant available) evidence on the effectiveness of interventions, model results suggest that in a scenario with high occupancy, no testing, reduced IPC, visitors, and longer stays, approximately 5.2% (3.9-7.0) of all susceptible inpatients (140,603 (89,352- 197,977) patients in total), and 44.1% of patient-facing HCWs could have been nosocomially infected with SARS-CoV-2 over wave 1 compared with the 1.0% of patients (33,922 (24,089- 41,050) and 21.3% (16.0-28.8) of HCW observed. The most effective interventions and changes within hospitals for prevention of nosocomial infections in patients was decreased occupancy, increased spacing between beds, and testing patients on admission, resulting in prevention of 23,434 (14,544-31,341), 10,979 (2,458-16,979), and 9,505, (4588-12,823) nosocomial infections. However, every intervention evaluated had some impact in reducing transmission to inpatients, and it was the collective impact of all interventions that demonstrated greatest effect. In HCWs, the most effective single intervention was universal mask use, averting 46.0% (42.9-54.5%) of infections (39.2 vs 21.3%), resulting in 17,980 (2,772-28,450) fewer infections per 100,000 patient-facing HCWs. It is likely that the effect of this intervention was largely due to the dominance of the HCW-to-HCW transmission route [2, 16] but it is possible that there is also a cumulative effect of a reduction in HCW infections contributing to a reduction in HCW-to-patient transmission. There is evidence that the risk of patient-to-HCW and HCW-to-patient transmission events is small, presumably due to the efficacy of masking and hand-hygiene to prevent transmission [2, 3, 16], but the potential for a larger effect of a small number of patient-to-HCW transmissions seeding larger outbreaks in the HCW population cannot be ignored.
A strength of the IBM is that it captures these ‘knock-on’ benefits associated with prevention of transmission chains both within and across patient and HCW populations; such combined/bundled effects are sometimes referred to as the ‘Swiss Cheese’ infection prevention model [23, 24]. This cumulative effect of reducing transmissions is apparent when looking at the impact of removing all interventions in combination, which results in a higher number of nosocomial infections than the sum of the individual interventions. Model findings suggest that collectively the interventions introduced over wave 1 of the SARS-CoV-2 pandemic in England averted 140,603 (89,352-197,977) infections in inpatients and reduced HCW infection rates by 51.1% (43.6-55.3%).
Previous modelling studies have shown that HCW to HCW transmission was a dominant route of transmission within hospitals in wave 1 of the pandemic [16, 25, 26], and impacts transmission dynamics to the greatest degree. Universal mask wearing by HCWs is an intervention which impacts this route. The dominance of this transmission route, combined with the evidence of effectiveness of universal mask wearing being high, results in high impact of universal mask wearing by HCW. By evaluating the literature, we estimate that universal mask-wearing by HCWs with complete compliance reduces the probability of transmission by a factor of 1.26-4, with a median effect of 3.2 [7, 14]. However, there is evidence that HCWs do not perceive interactions with other HCWs as a potential risk for becoming infected with SARS-CoV-2 , and that there are certain times that preventative measures such as mask-wearing and social distancing are not possible (e.g. while eating in a shared communal area) [28, 29]. Both of these factors, and others, could reduce compliance with the guidance updated on June 15th 2020 stating that masks/face coverings should be worn universally (i.e. by HCWs, patients and visitors). We assumed that HCWs fully complied with universal masking upon the change in guidance, wearing masks for their entire shift and for every interaction with patients and other HCWs, thus the reduction in transmission resulting from wearing a mask was applied to every HCW interaction. If this was not the case the modelled impact of mask wearing around other HCWs would fall as the reduction in transmission would only be applied to those interactions where masks were worn. The impact of mask-wearing by patients was not included in the model, however implementation of masking in patients was poor during Summer 2020, i.e. throughout wave 1.
When HCWs wore masks only when treating patients and not for contacts with other HCWs (prior to June), the number of infections increased by approximately 30% (from 21.32 (15.95, 28.78) to 26.4% (18.7-41.5)). Evidence from a single site study between 01-Mar-2020 and 25-Jul-2020 suggested, that 50% of HCW infections were nosocomial, and 55.3% of non-imported cases were likely to be attributable to HCW-to-HCW transmission . The authors of this paper also examined samples from the second wave (30-Nov-2020 - 24-Jan-2021) that followed the change in guidance, and thus for this time period HCWs were advised to wear face-coverings in all areas. During this period of study the proportion of infections that were acquired in the community remained at 50%, but the proportion attributable to HCW to HCW transmission fell to 37.4%, i.e. there was a reduction by a factor of 1.48. An additional study in a Parisian hospital that was not included in the initial model parameterisation estimated that the adjusted odds ratio for HCWs that did not wear masks over the period of study (in wave 1) was 13.1; however, there was a large amount of uncertainty around this estimate . These studies are largely consistent with our estimate that over wave 1 an additional 22.7% (12.3-35.7%) of HCWs would have been infected had the guidance on mask wearing not changed.
In previous work using the same IBM as used here, we have shown that patient-to-patient transmission is the dominant source of transmission to hospital inpatients [16, 25]. This has been supported by single centre genomic studies. One of these showed that of the non-imported cases identified in patients, 63.6% of infections in wave 1 and 79.2% of infections in wave 2 were attributable to patient-to-patient transmission . Another study showed that transmission outside of a ward is unlikely to be a major contributor to nosocomial outbreaks in hospital inpatients . It is therefore expected that interventions that reduce the risk of transmission to patients sharing a bay or ward with infected patients will be the most effective in reducing transmission. We predict that the most effective interventions for reducing the risk of transmission to patients (i.e. those that had the greatest effect when they were removed), were increasing spacing between beds, increasing hand-hygiene of HCWs, decreasing the length of stay of inpatients, and testing on admission. A single trust study that compared infection rates before and after increasing bed spacing also found a significant impact of increasing space between beds in terms of reducing nosocomial SARS-CoV-2 transmission .
It should be noted that these findings are estimates for the impact of hospital interventions on nosocomial infections in patients and HCW for wave 1 of the COVID-19 pandemic only. The impacts of each of these components of control could have likely changed in subsequent waves, noting, for example, alterations in the transmissibility of virus variants, increased capacity for rapid testing upon hospital admission, and changes in host immunity, including once vaccination was introduced.
There are several limitations of this work, most importantly the aforementioned reliance on limited evidence of individual intervention effectiveness. To be transparent about the provenance and quality of the data used we have categorised effectiveness estimates according to the study type and pathogen of study. However, lack of evidence means that conclusions should be treated with caution. Several simplifying assumptions have been made in the model, including equal risk across all patients and patient-facing HCW when interacting with infected patients or HCWs, interventions being applied uniformly across the hospital and across the simulation period. It is also of note that estimates of intervention effectiveness reflect the level of compliance with the interventions in the particular studies, which was not reported. We do not adjust for compliance in our model, and use the direct estimate of efficacy reported in the literature in our simulations. If compliance in the reported studies differed from that exhibited in practice, the impact of the intervention on nosocomial transmission would change in accordance. Reversing of interventions only considered those described, additional measures such a double gloving or sessional gown use are not included in this work.
This study uses previously published parameter estimates on the efficacy of individual measures to evaluate the contribution individual interventions to the reduction of nosocomial transmission as part of the collection of interventions and hospital changes in place in wave 1. A strength of the modelling approach used here is the ability of the model to capture cumulative effects of interventions through reducing the seeding of new infection clusters. These results highlight the importance of maintaining high levels of compliance to infection prevention and control measures in hospitals and have important implications as hospitals prepare for a surge in demand due to emerging winter pressures and COVID-19.