Interplay between C-reactive protein responses and antibiotic prescribing in people with suspected infection

Abstract Background Serial measurements of C-reactive protein (CRP) are often taken in hospitals to assess recovery from infection, but their utility remains debated. Previous studies, including our development of CRP centile reference charts for suspected bloodstream infections (BSI), suggest variability in CRP responses across infection types. Here we investigated the association between serial CRP percentile changes, antibiotic prescribing patterns, and patient outcomes in a large cohort with suspected infection, acknowledging that CRP is one of multiple factors in clinical decision-making. Methods We analysed 51,544 suspected infection episodes (defined by blood culture collection) from 36,578 patients in Oxfordshire, UK (2016–2021). Episodes were categorised by blood culture results: Gram-positive, Gram-negative, polymicrobial, contaminants, or culture-negative (having previously shown that 51% culture-negatives have CRP responses indistinguishable from culture-positives). The spectrum of antibiotic prescriptions and their changes over time were tracked. Multinomial logistic regression, adjusted for clinical covariates, assessed the association between CRP percentile changes and subsequent prescribing decisions. Linear mixed models evaluated CRP trajectories post-prescribing, and logistic regression associations between early CRP changes (days 1–4) and 5–30-day mortality. Results Broad-spectrum antibiotics were predominantly used within the first three days after blood culture collection, followed by a notable shift to narrow-spectrum antibiotics for Gram-positive infections, but with slower de-escalation for Gram-negative and polymicrobial infections. CRP percentile changes were modestly associated with subsequent antibiotic adjustments; in particular, suboptimal recovery, indicated by an increase in CRP centiles, was associated with a higher rate of antibiotic escalation (16.5% vs. 10.7% in expected recovery) and, conversely, faster than expected recovery in CRP was associated with de-escalation (23.6% vs. 17.2%). However, 61.8% of decisions were unchanged despite CRP trends. The relationship between various prescribing decisions and subsequent CRP percentile changes was complex and challenging to estimate, likely due to testing bias. CRP percentile changes during the 4 days post blood culture collection were strongly associated with 5–30-day mortality, highlighting their potential utility as a prognostic indicator. Conclusions While CRP monitoring can inform antibiotic stewardship, its association with prescribing decisions is probably only modest, underscoring the need to integrate a range of clinical factors to optimise infection management.