Effectiveness of rapid SARS-CoV-2 genome sequencing in supporting infection control for hospital-onset COVID-19 infection: Multicentre, prospective study

Abstract

Background: Viral sequencing of SARS-CoV-2 has been used for outbreak investigation, but there is limited evidence supporting routine use for infection prevention and control (IPC) within hospital settings.

Methods: We conducted a prospective non-randomised trial of sequencing at 14 acute UK hospital trusts. Sites each had a 4-week baseline data collection period, followed by intervention periods comprising 8 weeks of 'rapid' (<48 hr) and 4 weeks of 'longer-turnaround' (5-10 days) sequencing using a sequence reporting tool (SRT). Data were collected on all hospital-onset COVID-19 infections (HOCIs; detected ≥48 hr from admission). The impact of the sequencing intervention on IPC knowledge and actions, and on the incidence of probable/definite hospital-acquired infections (HAIs), was evaluated.

Results: A total of 2170 HOCI cases were recorded from October 2020 to April 2021, corresponding to a period of extreme strain on the health service, with sequence reports returned for 650/1320 (49.2%) during intervention phases. We did not detect a statistically significant change in weekly incidence of HAIs in longer-turnaround (incidence rate ratio 1.60, 95% CI 0.85-3.01; p=0.14) or rapid (0.85, 0.48-1.50; p=0.54) intervention phases compared to baseline phase. However, IPC practice was changed in 7.8 and 7.4% of all HOCI cases in rapid and longer-turnaround phases, respectively, and 17.2 and 11.6% of cases where the report was returned. In a 'per-protocol' sensitivity analysis, there was an impact on IPC actions in 20.7% of HOCI cases when the SRT report was returned within 5 days. Capacity to respond effectively to insights from sequencing was breached in most sites by the volume of cases and limited resources.

Conclusions: While we did not demonstrate a direct impact of sequencing on the incidence of nosocomial transmission, our results suggest that sequencing can inform IPC response to HOCIs, particularly when returned within 5 days.

Funding: COG-UK is supported by funding from the Medical Research Council (MRC) part of UK Research & Innovation (UKRI), the National Institute of Health Research (NIHR) (grant code: MC_PC_19027), and Genome Research Limited, operating as the Wellcome Sanger Institute.

Clinical trial number: NCT04405934.

Keywords: COVID-19; epidemiology; global health; healthcare-associated infection; hospital-acquired infection; human; infection control; infection prevention; infectious disease; microbiology; molecular epidemiology; viral genomics.

Conflict of interest statement

OS, JB, FM, AM, MP, AH, NM, TM, KS, TS, YT, NM, CP, AC, AT, RW, AD, DR, FF, SR, ML, KL, IM, BK, SH, RG, MB, AW, MB, MC, JH, GN, DP, MP, JP, CP, SR, LS, Td, ET, AC, JB No competing interests declared, GS has an unpaid role as Deputy Chair, British Medical Association London Regional Council. The author has no other competing interests to declare, MC received payment for anonymous interview conducted by Adkins Research Group. The author has no other competing interests to declare, EN holds grants by NIHR, EPSRC, MRC-UKRI , H2020, ViiV Healthcare, Pfizer and Amfar, and has received grants to attend meetings from H2020 and ViiV Healthcare, DS holds the following grants that are not specifically for the present work: COG-UK, PHE test and trace funded the sequencing aspect. HOCI funded a technician to support sequencing during study period. The author has no other competing interests to declare, FC received consulting fees from Next Gen Diagnostics LLC (during 2018/2019), received payment or honoria for lectures from University of Cambridge and Wellcome Genome Campus Advanced Courses, and received support for attending meeting and/or travel to meetings from European Congress of Clinical Microbiology &amp;amp; Infectious Diseases (ECCMID), The American Society for Microbiology (ASM), Microbiology Society, European Congress of Clinical Microbiology &amp;amp; Infectious Diseases (ECCMID), and the British Infection Association (BIA). The author has no other competing interests to declare, SP received consultancy fees from Pfizer (Coronavirus External Advisory Board) and Melinta Therapeutics, received payment from SVB Leerink for a round table meeting and for Mary Strauss Distinguished Public Lecture from the Fralin Biomedical Research Institute, US, and support for attending ICPIC conference, Geneva and World Health Summit, Berlin in 2021, and hold stocks or stock options in Specific Technologies (European Union Scientific Advisory Board) and Next Gen Diagnostics (Scientific Advisory Board). SP also serves as Chair, Medical Advisory Committee, Sir Jules Thorn Charitable Trust, Board member of the Wellcome SEDRIC (Surveillance and Epidemiology of Drug Resistant Consortium), and Non-Executive Director of Cambridge University Hospitals NHS Foundation Trust. The author has no other competing interests to declare, PF is a member of the SAGE hospital onset covid working group 2020-2022. The author has no other competing interests to declare

© 2022, Stirrup et al.

Figures

Figure 1.. Plots of the median turnaround time against the percentage of hospital-onset COVID-19 infection (HOCI) cases with sequence reporting tool (SRT) reports returned for the rapid (left panel) and longer-turnaround (right panel) sequencing phases across the 14 study sites.

The size of each circle plotted indicates the number of HOCI cases observed within each phase for each site, with letter labels corresponding to study site. The criteria for inclusion in our sensitivity analysis are displayed as the green rectangle in the rapid phase plot, and sites on the longer-turnaround phase plot are colour-coded by their inclusion. In the rapid phase, SRT reports were returned for 0/4 HOCI cases recorded for site H. Site N did not have a longer-turnaround phase, Site A observed 0 HOCI cases, and sites D and E returned SRT reports for 0/1 and 0/2 HOCI cases, respectively, in this phase.

Figure 2.. Plots of the proportion of returned sequence reporting tool (SRT) reports that had an impact on infection prevention and control (IPC) actions (a, b) and that were reported to be useful by IPC teams (c, d).

Data are shown for all sites in (a) and (c), and for the seven sites included in the ’per protocol’ sensitivity analysis in (b) and (d). Results are only shown up to turnaround times of ≤28 days, and grouped proportions are shown for ≥9 days because of data sparsity at higher turnaround times. Error bars show binomial 95% CIs. ‘Yes’ and ‘No’ outcomes for individual hospital-onset COVID-19 infection (HOCI) cases are displayed, colour-coded by rapid (red) and longer-turnaround (blue) intervention phases and with random jitter to avoid overplotting. ‘Unsure’ responses were coded as ‘No’ for (c) and (d).

Appendix 1—figure 1.. Flow diagram of study site enrolment and intervention implementation.

*Baseline phase extended for one site due to a complete lack of hospital-onset COVID-19 infection (HOCI) cases during the first few weeks of study period and omission of longer-turnaround sequencing phase. †Rapid sequencing phase truncated at one site due to cessation of enrolment at all sites.

Appendix 1—figure 2.. Plots of the proportion of returned sequence reporting tool (SRT) reports that had an impact on infection prevention and control (IPC) actions by study site.

Results are only shown up to turnaround times of ≤28 days, and grouped proportions are shown for ≥9 days because of data sparsity at higher turnaround times. Error bars show binomial 95% CIs. ‘Yes’ and ‘No’ outcomes for individual hospital-onset COVID-19 infection (HOCI) cases are displayed, colour-coded by rapid (red) and longer-turnaround (blue) intervention phases and with random jitter to avoid overplotting.

Appendix 1—figure 3.. Plots of the proportion of returned sequence reporting tool (SRT) reports that were reported to be useful by infection prevention and control (IPC) teams by study site.

Results are only shown up to turnaround times of ≤28 days, and grouped proportions are shown for ≥9 days because of data sparsity at higher turnaround times. Error bars show binomial 95% CIs. ‘Yes’ and ‘No’ outcomes for individual hospital-onset COVID-19 infection (HOCI) cases are displayed, colour-coded by rapid (red) and longer-turnaround (blue) intervention phases and with random jitter to avoid overplotting. ‘Unsure’ responses were coded as ‘No’.

Appendix 1—figure 4.. Weekly incidence of hospital-acquired infections (HAIs) at each site (●), with (a) proportion of all inpatients SARS-CoV-2+ve and (b) local community incidence of SARS-CoV-2+ve tests also plotted on the y-axis (purple line).

Horizontal bars show the duration of study phases (orange: baseline; blue: longer turnaround; green: rapid).

Appendix 1—figure 5.. Adustment variables for analysis of weekly incidence of infection prevention and control (IPC)-defined hospital-acquired infections (HAIs) per 100 inpatients, as described in Table 3.

Incidence rate ratios are displayed relative to the median for (a) calendar time expressed as study week from 12 October 2020, (b) proportion of inpatients with positive SARS-CoV-2 test, and (c) local community incidence of SARS-CoV-2 (government surveillance data weighted by total set of postcodes for patients at each site). The spline curves shown are estimated simultaneously within the final analysis model and show how these factors have independent contributions to the prediction of the incidence rate for HAIs. The associations for each covariable indicated by model parameter point estimates are shown as solid lines, with 95% CIs shown as dashed lines. Adjustment for (c) was not pre-specified in the statistical analysis plan (SAP), but adding this variable to the model was associated with a statistically significant improvement in fit (p=0.01). The proportion of community-sampled cases in the region that were found to be the Alpha variant on sequencing was also considered, but adding this as a linear predictor did not lead to a statistically significant improvement in model fit (p=0.78).

Appendix 1—figure 6.. Adustment variables for analysis of weekly incidence of infection prevention and control (IPC)-defined outbreak events per 1000 inpatients, as described in Table 3.

Incidence rate ratios are displayed relative to the median for (a) calendar time expressed as study week from 12 October 2020, (b) proportion of inpatients with positive SARS-CoV-2 test, and (c) local community incidence of SARS-CoV-2 (government surveillance data weighted by total set of postcodes for patients at each site). The spline curves shown are estimated simultaneously within the final analysis model and show how these factors have independent contributions to the prediction of the incidence rate for outbreaks. The associations for each covariable indicated by model parameter point estimates are shown as solid lines, with 95% CIs shown as dashed lines. Adjustment for (c) was not pre-specified in the statistical analysis plan (SAP), but adding this variable to the model was associated with a near-statistically significant improvement in fit (p=0.05) and was included for consistency with the analysis of individual hospital-acquired infections (HAIs). The proportion of community-sampled cases in the region that were found to be the Alpha variant on sequencing was also considered, but adding this as a linear predictor did not lead to a statistically significant improvement in model fit (p=0.80).

Appendix 1—figure 7.. Weekly counts of enrolled hospital-onset COVID-19 infection (HOCI) cases by date of positive test return for each site, colour-coded by intervention phases.

Horizontal bars show the duration of study phases.