This multicenter retrospective study used Epic Cosmos electronic-health-record data covering 2016–2025 to evaluate associations between antibacterial treatment and clinical outcomes in adults with imaging-evaluated, nonsevere, virus-positive community-acquired pneumonia. The overall cohort comprised 376,320 adults, including 275,604 hospitalized patients and 100,716 outpatients. The investigators report that underlying patient-level data accessed through Epic Cosmos cannot be shared publicly because of the data use agreement; aggregated study outputs are provided in the manuscript and online supplement, and analysis code is available from the corresponding author on reasonable request.
The study compared inpatient groups defined by antibacterial treatment during two time windows: empirical treatment within the initial 24 hours and continuation versus discontinuation during hours 24–48. Outpatients were compared by whether an antibacterial prescription was issued at emergency-department discharge. Analyses were stratified by guideline-defined comorbidity status. The primary analytic approach used propensity-score overlap weighting with source-cluster bootstrap confidence intervals to adjust for measured confounding. The authors also performed exploratory analyses evaluating associations by respiratory virus, receipt of antiviral therapy, antibacterial class, and timing of outpatient prescriptions.
Among hospitalized patients who received empiric antibacterial therapy, continuation of therapy after the initial empiric period was not associated with improved outcomes. Instead, continuation was associated with increased observed risk of 30-day adverse events and with longer post-landmark hospital stay in both strata of guideline-defined comorbidity.
Specifically, the adjusted risk difference for 30-day adverse events after continuation versus discontinuation was 1.70 percentage points (95% confidence interval, 0.80–2.39) in patients without guideline comorbidity, and 2.56 percentage points (95% confidence interval, 1.88–3.14) in patients with guideline comorbidity. Continuation was also associated with longer post-landmark length of stay, reported as adjusted mean ratios of 1.14 and 1.08 in the without-comorbidity and with-comorbidity groups, respectively.
These results indicate that continued inpatient antibacterial therapy beyond the empiric window showed no evidence of clinical benefit in this cohort and was linked to worse observed outcomes on multiple measures.
The authors report exploratory, class-specific analyses. Broad-spectrum regimens primarily targeting resistant Staphylococcus species or Pseudomonas were associated with the largest observed increases in adverse events and mortality. In contrast, regimens containing macrolides and other agents directed at atypical pathogens did not show a consistent adverse signal in these exploratory analyses.
The source article notes additional exploratory assessments by respiratory virus and by receipt of antivirals, but specific stratified numeric results for those respiratory-virus and antiviral analyses were not detailed in the text of the preprint abstract. Where detailed subgroup results are not provided in the source, those specifics were not reported.
For outpatients, issuance of an antibacterial prescription at emergency-department discharge was associated with lower observed risks compared with no outpatient antibacterial prescription. However, the authors emphasize that outpatient associations are susceptible to care-transition effects and residual confounding that may bias observed associations. The preprint explicitly cautions that care-transition processes and unmeasured differences between patients who received prescriptions and those who did not could account for some or all of the observed outpatient associations.
In this large retrospective cohort of adults with imaging-evaluated, nonsevere, virus-positive community-acquired pneumonia, continuation of inpatient antibacterial therapy after the empiric period showed no evidence of benefit and was associated with worse observed outcomes, including a higher adjusted risk of 30-day adverse events and longer post-landmark hospital stays. Exploratory analyses suggested higher adverse-event and mortality associations with broad-spectrum agents targeting resistant Staphylococcus or Pseudomonas, while macrolide-containing and atypical-coverage regimens lacked a consistent adverse signal. Outpatient prescribing appeared favorable in unadjusted associations but remains vulnerable to confounding and care-transition bias.
The authors declare no competing interests and report that institutional approvals and patient consent processes were followed. Because the underlying Epic Cosmos data are restricted by institutional permissions, raw data are not publicly available; aggregated results and the online supplement accompany the preprint, and analysis code may be requested from the corresponding author.
Clinicians and antimicrobial stewardship programs should interpret these observational associations in the context of residual confounding and the study's retrospective design. The preprint presents evidence that discontinuing antibacterial therapy after the initial empiric window warrants consideration in hospitalized adults with virus-positive, nonsevere community-acquired pneumonia, but randomized or prospectively controlled evidence would be required to establish causality and to guide definitive practice change.