Fig. 3: Factors associated with healthcare-associated SARS-CoV-2 in patients, HCWs and predictive distributions.

a,b, Additive effects associated with categories of host infections and multiplicative effects of vaccine coverage in HCWs, Alpha prevalence and trust characteristics (posterior means and 50% and 90% CrIs are shown and results are on the basis of 1,661 hospital trust weeks of data from 96 different hospital trusts). For multiplicative effects, values below 1 indicate an association with reduced infection rates. Note that in the model for infections in patients (a), HCW vaccine coverage acts by modulating transmission associated with infected HCWs, whereas in the model for infections in HCWs (b) it has a global effect, modulating the overall rate of infection. c,d, Associated posterior predictive distributions for the number of detected infections by week in the 20 largest trusts are shown for infections in patients (c) and in infections in HCWs (d). Bold solid lines correspond to observed values, shaded regions correspond to 50% and 90% CrIs and the central lines within the shaded regions are median values from the posterior sample. e–g, For all trusts, classifications of detected infections by week (e) and contributions to predicted hospital-acquired infections in patients (f) and HCWs (g) from the three categories of infected hosts predicted by the full negative binomial regression models accounting for HCW vaccination and Alpha variant effects are shown. When the dependent variable is healthcare-associated SARS-CoV-2 infection in patients, these results use the ECDC definitions of definite and probable healthcare-associated infection (see Supplementary Information section 2 and the Supplementary Results for models using other definitions).