The burden of nosocomial covid-19: results from the Wales multi- 1 centre retrospective observational study of 2518 hospitalised adults. 2

Objectives – To define the burden of nosocomial (hospital-acquired) novel pandemic coronavirus (covid-19) infection among adults hospitalised across Wales. 38 Design - Retrospective observational study of adult patients with polymerase chain reaction (PCR)- 39 confirmed SARS-CoV-2 infection between 1 st March – 1 st July 2020 with a recorded hospital admission 40 within the subsequent 31 days. Outcomes were collected up to 20 th November using a standardised 41 online data collection tool. Setting – Service evaluation performed across 18 secondary or tertiary care hospitals. Participants - 4112 admissions with a positive SARS-CoV-2 PCR result between 1 st March to 1 st July 44 2020 were screened. Anonymised data from 2518 participants were returned, representing over 60% of 45 adults hospitalised across the nation of Wales.


Key Messages 61
What is already known on this topic 62 We searched PubMed and ISI Web of Science up until 31-December-2020 for studies reporting on 63 patient outcomes following hospital-acquired infection due to the severe acute respiratory syndrome 64 coronavirus 2 (SARS-CoV-2). We identified a range of case-definitions for hospital-acquired infection, 65 based on timing of diagnostic testing 5 to 15 days following admission. The largest and only multi-66 centre study concluded individuals with nosocomial infection are at a lower risk of death from SARS-67 CoV-2 than those infected in the community, however, was performed early in the pandemic and 68 utilised a conservative definition of nosocomial infection. 69

What this study adds 70
Our multi-centre observational study represents the largest assessment of clinical outcomes for patients 71 with nosocomial covid-19 in the UK to date, and suggests the burden of nosocomial SARS-CoV-2 72 infection has been underestimated. Nosocomial-infection occurred in older, frailer, and multi-morbid 73 individuals, and was consistently associated with greater inpatient mortality than amongst those who 74 were infected in the community across a spectrum of case-definitions. Our findings support 75 implementation of enhanced infection control measures to reduce this burden during future waves, 76 especially given the recent emergence of novel viral variants with enhanced transmissibility. 77 Furthermore, roughly half of the patients meeting the Public Health Wales definition of definite 78 nosocomial SARS-CoV-2 infection had been admitted for 30 days prior to diagnosis, highlighting a 79 potential window of opportunity for inpatient pre-exposure and/or post-exposure prophylaxis. 80 81 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Introduction 82 Health-care-associated infections represent an enduring and serious threat to patient safety in hospital 83 and healthcare facilities 1,2 and have been estimated to cost the National Health Service £1 billion every 84 year. 3 The transmission of respiratory viruses such as influenza in the healthcare environment are a well 85 recognised cause significant morbidity and mortality at the individual patient level, 4 however little is 86 known regarding the significance of in-hospital (nosocomial) transmission of the novel pandemic 87 coronavirus, SARS-CoV-2. Such information is essential to assess the effectiveness of recent infection 88 control measure guidance. 5 A national report published by the Healthcare Safety Investigation Branch 89 on 29th October 2020 stated: "At the time of writing, there was no publicly available data to provide 90 evidence of the scale of nosocomial of covid-19 in the NHS". 6 In addition, acknowledgement of risk is 91 timely to inform prioritisation of public health measures such as vaccination of healthcare workers and 92 vulnerable populations, and decisions regarding continuation of routine NHS service provision. 93 We identified limited studies reporting on the prevalence and outcomes of medical patients '  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. Participants 129 Positive SARS-CoV-2 polymerase chain reaction (PCR) results recorded between 1 st March 2020 and 130 1st July 2020 with a record of hospital admission were obtained from Public Health Wales and used to 131 identify patients for retrospective notes review information by local clinical teams. 132

133
Demographic variables with prognostic significance were collected as mandatory data fields for the 134 index admission using an online tool including: date of positive PCR swab collection, dates of 135 admission and discharge (when complete), patient age, sex, comorbidity count, and outcome (death or 136 discharge). Supplementary fields were also collected, as described in the text. 137 Outcomes 138 The primary outcome for analysis was descriptive in nature, with comparison of all-cause mortality rate 139 and cohort demographics based upon probable origin of SARS-CoV-2 infection. Secondary analyses 140 based on time to discharge or death were also investigated. 141

Missing data 142
Missing data were included in tables and descriptive analyses, allowing denominators to remain 143 consistent in calculations. Imputation of missing data was not performed. 144

145
Notes were selected randomly by site audit/record departments, to minimise potential for systematic 146 bias prior to data entry. 147

148
Frailty was assessed using the pre-admission Clinical Frailty Scale (CFS), as recommended by the 149 National Institute for Clinical Excellence. 18 For the purposes of the analyses, CFS scores were grouped 150 into clinically meaningful groups: 1-4, 5-6, and 7-9. Welsh Index of Multiple of Deprivation (WIMD) 151 was derived from the patients' post-code and evaluated by quartiles. 152

153
Baseline demographic and clinical characteristics assessed at admission were grouped by outcome and 154 probable infection origin, based on a) clinician-recorded case definition, b) standardised case definitions 155 (Supplementary Table 2). Continuous variables are presented as median [inter-quartile range, IQR] and 156 categorical variables as n (%), unless otherwise stated. Time-to-event data were analysed using 157 cumulative incidence curves, acknowledging the competing risk of death or discharge. We used time in 158 hospital following a covid-19 diagnosis as a standard measure to avoid introducing survivorship-bias, 159 defining day 0 of the at-risk period as the most recent of either admission date or SARS-CoV-2 160 diagnostic testing. 161 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. Vienna, Austria) with packages including tidyverse, and survival and GraphPad Prism (version 6.07). 170 Sensitivity analyses were performed to assess the effect of varying the clinical case definition for 171 nosocomial-acquired infection, and to minimise potential survivorship bias associated with previously-172 used case definitions. 173 This project constitutes the evaluation/audit arm of the National Pathway for Managing covid -19 174 Infections in Secondary Care in Wales and was approved by Welsh Government. Information 175 governance policies on data protection were followed to record data securely at each site. Anonymised 176 data handling and storage was supported by The Institute of Clinical Science and Technology, Cardiff 177 UK. 178 Patient and public involvement: These data were generated as part of a rapid appraisal process and as 179 such patients were not involved in the setting of the research question or interpretation of the study. 180 Findings will be disseminated via the www.covid-19hospitalguideline.wales.nhs.uk tool. 181 182 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021.  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021.  As 95% of patients will display symptoms between 2 to 11.5 days of exposure, 19 the COPE-study 237 definition represents a conservative estimate of the burden of nosocomial-cases infection. We therefore 238 considered the effect on mortality and case numbers by varying the case definition across the incubation 239 period, taking admission as the earliest potential nosocomial exposure (Supplementary Table 4). As 240 shown in Figure 2A, inpatient mortality rates for nosocomial covid-19 ranged from 38% to 42% and 241 remained consistently greater than in patients with community-acquired infection (31% to 34%). By 242 contrast, varying the case definition resulted in significant changes in case numbers and deaths ( Figure  243 2B/C  Competing Risk analysis provides differing conclusions to Kaplan Meier

253
Kaplan Meier survival analysis has been utilised by several recent groups to compare outcomes for 254 hospital and community-acquired covid-19 in UK hospitals to date. 8,9 However, the differing discharge 255 kinetics observed in patients with nosocomial-and community-acquired covid-19 led us to question the 256 suitability of this statistical methodology without consideration of the competing risks of discharge to 257 death. As shown in Figure 3, nosocomial-infection is associated with a greater cumulative incidence of 258 inpatient mortality than community-acquired, in line with the crude case inpatient mortality rates 259 described. This contrast with Kaplan Meier analysis (Supplementary Figure 2), confirming the potential 260 for conflicting interpretations of the same data set when employing different statistical methodologies. 261 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21249433 doi: medRxiv preprint Patients with probable nosocomial COVID-19 infection are a more vulnerable population than 265 community acquired cases 266 We hypothesized that patients with nosocomial-covid-19 might represent a more vulnerable population, 267 a finding suggested by previous groups, but potentially confounded by introduction of selection bias 268 due to the requirement for a prolonged inpatient admission period within the case definition of 269 nosocomial-SARS-CoV-2 infection. To address this, we explored demographic and clinical 270 characteristics in patients based on a diagnostic cut-off of 2 days post-admission, a definition commonly 271 used for hospital-acquired bacterial pneumonia. Individuals with a positive SARS-CoV-2 test taken >2 272 days following admission were typically older and frailer, with a greater median CFS of 6 (IQR 4.5-273 7.5), compared to 3 (IQR 1-5) in those with a positive SARS-CoV-2 test taken up to 2 days following 274 admission. Marked differences in multi-morbidity were evident, with 35.4% of "nosocomial" patients 275 having at least 4 comorbidities compared to 26.6% of "community-acquired" patients (Supplementary 276 3.38). Taken together, these observations confirmed our expectation that patient with nosocomial-293 acquired SARS-CoV-2 represented a more vulnerable cohort, and as such were less likely to be 294 managed using more invasive care such as intubation or CPAP. However, we found no evidence that a 295 ward-level ceiling of care was an independent predictor of mortality. 296 Monthly prevalence of nosocomial covid-19 297 We determined monthly trends for the frequency of total and new nosocomial diagnoses 298 (Supplementary is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. where definitely or probably hospital-acquired cases had a fatality rate of 36%. 13 Similarly, the monthly 314 prevalence estimates found in our study are also in line with the 20% modelled for a typical UK 315 hospital. 21 316 Our study is also the first to investigate the impact of applying a range of case definitions when 317 quantifying the burden of hospital-acquired covid-19. By adopting a clinical case definition based on 318 the median incubation period for SARS-CoV-2 of 5 days, 20 we identified 14.2% additional cases, and 319 13.6% more deaths attributable to nosocomial covid-19 than with the >7-day threshold currently used 320 by Public Health England. This increased burden is likely an overestimate but may be relevant given 321 that even low-risk clinical assessment areas may contain patients with SARS-CoV-2. 22 Thus, in the 322 absence of sufficient isolation cubicles for patients undergoing assessment and adequately sensitive 323 rapid turn-around diagnostic testing, the risk of patient-to-patient transmission is likely to persist despite 324 infection control practices of cohorting and universal screening. We also confirm that patients with 325 hospital-acquired covid-19 appear a disproportionately vulnerable group, seemingly independent of 326 diagnostic delay from admission. Advanced age, frailty, and multi-morbidity are well-recognised risk 327 factors for adverse outcome in community-acquired covid- 19. 23,24 We suggest increased frailty is also 328 likely to be causally associated with an increased personal care requirement and thus predispose to 329 nosocomial exposure of SARS-CoV-2 from healthcare workers, a conclusion supported by recent 330 whole-genome sequencing of nosocomial outbreaks. 14 As highlighted in a recent evaluation of serial 331 PCR screening in healthcare workers, 3% of those attending work tested positive for SARS-CoV-2. 25 332 Serological studies suggest the rate of staff infection is far higher on wards with known nosocomial 333 outbreaks, in studies conducted in hospitals not undertaking staff screening. 26 Finally, we also 334 considered the impact of ceiling of care on outcome, as few other UK multi-centre studies have explored 335 this. Importantly, we found no evidence that ceiling of care determined outcome once clinical risk 336 factors were accounted for, a finding that may be reassuring for patients and families when discussing 337 this issue. 338 This study is primarily limited by its retrospective nature, which might predispose to ascertainment bias 339 (for instance, notes being more readily available for patients who have died than awaiting discharge). 340 This risk is common to other multi-centre studies, 8,23,24 and sites adopted random notes retrieval to 341 minimise this. We further sought to mitigate this by extending the data collection period well beyond 342 the end of the first wave, enabling capture of late discharges and including nosocomial outbreaks. 343 Record entry by a body of motivated clinicians followed a standardised online tool with a limited core 344 requirement, resulting in minimal missing core data. As a result, coverage exceeded 60% of hospital 345 admissions with covid-19 across the nation of Wales (and ≥80% in 10 of 18 participating sites). This 346 compares favourably to recent descriptive studies. 8,23,24 We recognise that our findings represent crude 347 inpatient mortality rate estimates, based on all-cause mortality. As the total number of patients at risk 348 of infection was unknown, they should not be used to infer the risk of acquiring SARS-CoV-2 within 349 hospital. Future studies making use of linked datasets are suggested to estimate excess mortality 350 associated with nosocomial infection and identify routes of transmission are required. However, with 351 the NHS facing winter-pressures and further surges in SARS-CoV-2 community circulation rates, our 352 findings serve as a timely reminder as to the dangers faced by patients within hospital as well as those 353 within care homes or shielding within the community. This concern is heightened by the spread of a 354 novel antigenic variant with evidence for enhanced transmissibility 27 355 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. ; https://doi.org/10. 1101/2021 Together, our results highlight the burden associated with extremely infectious agents within the 356 healthcare environment. Although the incidence of nosocomial covid-19 cases decreased significantly 357 following the peak of community-acquired cases, the importance of enhanced infection control 358 measures cannot be overstated. 28 Indeed, current Public Health Wales data indicates an active rise in 359 probable hospital-acquired case numbers, mirroring the first wave. 29 We highlight the opportunity for 360 pre-exposure and post-exposure prophylactic measures, including vaccination and/or enrolment in 361 clinical trials employing pharmaceutical and biological antiviral agents such as antibody therapy. 30, 31 362 This follows the observation that many of those dying with hospital-acquired covid-19 had typically 363 been in hospital for over a month prior to diagnosis of infection. Inpatient vaccination has long been 364 considered a missed opportunity for improving influenza vaccine uptake. 32 Finally, we caution that 365 research is urgently required to assess the efficacy of vaccination within vulnerable groups of 366 individuals within hospital and care homes, who are likely to differ from those enrolled in initial trials 367 in age, frailty, and immunocompetence. 33 368 369 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21249433 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.18.21249433 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021.   is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 20, 2021. ; https://doi.org/10. 1101/2021