Identified and classified hospital-acquired cases
In CO-CIN, using a symptom onset-based definition, we found 7% (n = 65,028) of COVID-19 cases in acute English Trusts were identified and classified as a hospital-acquired infection (having a symptom onset more than 7 days after admission and before discharge) before 31st July 2020. By adjusting for enrolment in CO-CIN (Figure 2b), we estimated that with this same cut-off there were 6,640 “hospital-onset, hospital-acquired” identified cases across acute English Trusts up to the 31st July 2020.
Proportion of infections identified
We estimated 30% (20-41%, range across weeks and sampling, Supplementary 10) of symptomatic hospital-acquired infections (using a 7 day cut-off) were identified using a symptom onset based definition for England. Across all acute English Trusts the range was 0-82% (Figure 3). The proportion identified decreased with increasing cut-off day from admission (Figure 3c). The estimates are highly sensitive to LoS distributions (Supplementary 2). These results imply that for every single identified hospital-acquired SARS-CoV-2 infection (using a 7 day cut-off) there were, on average, two unidentified symptomatic hospital-acquired infections (Figures 1&2).
Contribution of missed infections
We estimated that across England, 20,000 (mean; 95% range over 200 simulations to nearest 100: 19,200, 21,100) hospital-acquired infections were unidentified from acute Trusts if a 7 day symptom-based cut-off was used to identify hospital-acquired cases. The majority of patients with unidentified hospital-acquired infections were not identified due to the discharge of the infected patient prior to symptom onset (“missed”) (Figure 1 and 3c): 12,300 (11,400, 13,400) in total.
A proportion of the patients with unidentified hospital-acquired infections with a symptom onset after discharge returned as hospitalised cases and were misclassified: we found 1,500 (1,200, 1,900) or 2.1% (1.7%, 2.6%) of cases originally classified as “community-onset, community-acquired” should have been classified as “community-onset, hospital-acquired” for a 7 day cut-off.
We found that there could have been 47,400 (mean; 95% range over 600 simulations: 45,000, 50,000 for the time-varying R value) hospital-linked infections of individuals in the community, acquired from patients with “missed” hospital-acquired infections during the first wave. We estimated that these hospital-linked infections would result in 1,600 (1,600, 1,700) “community-onset, hospital-linked” cases with a 7 day cut-off. The values are reduced by one-third with an R constant at 0.8 (Supplementary 11). These contribute 2.3% (2.1%, 2.4%) of “community-onset, community-acquired” cases over the first wave with a 7 day cut-off and under both scenario 1 or 2 (Supplementary 11).
This contribution of community-linked infections to hospital admissions with COVID-19 varied depending on the timing of hospital admission post symptom onset (captured here by Scenarios 1-3, Table 2, Figure 4). The proportion of COVID-19 hospital admissions due to hospital-transmission was greatest when total case numbers first declined (peak in COHL in Figure 4D at ~4% in late April).
The number of unidentified hospital-acquired infections and hence reclassification levels increased or decreased under a 14 or 4 day cut-off respectively (Supplementary 11).
Contribution of hospital settings to cases, infections and onward transmission
To summarise, using a 7 day cut-off, we estimated that there have been a total of 26,600 (mean, 95% range over 200 simulations: 25,900, 27,700) symptomatic hospital-acquired SARS-CoV-2 infections in acute English Trusts (E, Figure 5) prior to August 2020. Of these, a total of 15,900 (15,200, 16,400) infections correspond to patients with COVID-19 that were identified as symptomatic cases in hospitals (B+C, Figure 5): as such only 60% of symptomatic hospital-acquired infections were identified. Over the whole first wave, we estimated that 15% (14.1%, 15.8%) of cases originally classified as community-acquired were hospital-acquired or hospital-linked ((C + F) / (A - B), Figure 5).
The estimated percentage of identified COVID-19 cases in hospitals that were hospital-acquired is then 20.1% (19.2%, 20.7%) ((B + C)/ A, Fig. 5). Accounting for onward transmission from unidentified “missed” hospital-acquired infections, we estimated that 22.1% (21.2%, 22.9%) of hospitalised COVID-19 cases were hospital-acquired or hospital-linked ((B + C + F)/A, Figure 5) using the median time-varying R value.
If 20.1% of COVID-19 cases identified in hospitals were hospital-acquired then, assuming that 3% of symptomatic cases were hospitalised, we estimated that hospital-acquired infections likely contributed to fewer than 1% of infections of the overall English epidemic of COVID-19 in wave 1.
Assuming similar levels of hospital transmission in non-acute English trusts suggests approximately 31,100 (30,300, 32,400) symptomatic infections could have been caused in total by symptomatic hospital-acquired transmission in England.
Trust-level and Sensitivity analysis
When aggregated, the results from the analysis on an individual Trust-level predicted a slightly higher proportion of cases to be hospital-acquired (25% vs 20%) (Supplementary 12). Varying the day of discharge of the unidentified “missed” infections had little impact on total case numbers, but did affect hospital-linked cases (Supplementary 11).