Fig. 1: Quantifying the probability of observing hospital-acquired infections and estimating the total number of such infections.

a–c, Model inputs are shown in the top row and include the incubation period distribution³² (a), the PCR sensitivity profile¹⁶ (b) and the length-of-stay distribution (c) for patients who were not admitted with COVID-19 between June 2020 and February 2021 (solid lines in a and b show expected values and shaded regions show 95% CrIs). In c, the minimum lengths of stays needed to be classified as a probable or definite healthcare-associated infection are shown by dashed and solid vertical lines. d–f, Estimates of the probabilities that patients with hospital-acquired SARS-CoV-2 infections have a PCR positive test while in hospital under different screening policies (d), and estimates of the probabilities that they both screen positive and meet the post-14 d onset criteria to be considered a ‘definite’ healthcare-associated infection (e) or the post-7 d criteria to be classified as a probable or definite healthcare-associated infection (f) are shown in the middle row. Panels d–f are on the basis of 1,000 Monte Carlo samples, with violin plots showing median values (points), interquartile ranges (rectangles) and densities. The Public Health England screening recommendations are highlighted in green and the policy of screening all patients at 7 d intervals after admission is highlighted in blue (note that in contrast to this policy, weekly and 2× and 3× weekly policies screen on fixed days of the week). g, The estimated total number of hospital-acquired infections across adult acute NHS trusts in England linked to observed weekly number of detected post-14 d onset infections, assuming the screening policies highlighted in the middle row on the basis of recorded ‘definite healthcare-associated infections’; week numbers are counted as 1 plus the number of complete 7 d periods since 1 January 2020. Green and blue shaded regions indicate 90% CrIs and white lines are posterior means.