Agreement and comparative accuracy of instability criteria at discharge for predicting adverse events in patients with community-acquired pneumonia

Abstract

Objective Five definitions of clinical instability have been published to assess the appropriateness and safety of discharging patients hospitalised for pneumonia. This study aimed to quantify the level of agreement between these definitions and estimate their discriminatory accuracy in predicting post-discharge adverse events.Study design and setting We conducted a retrospective cohort study involving 1038 adult patients discharged alive following hospitalisation for pneumonia.Results The prevalence of unstable criteria within 24 hours before discharge was 4.5% for temperature >37.8°C, 13.8% for heart rate >100/min, 1.0% for respiratory rate >24/min, 2.6% for systolic blood pressure <90 mm Hg, 3.3% for oxygen saturation <90%, 5.4% for inability to maintain oral intake and 6.4% for altered mental status. The percentage of patients classified as unstable at discharge ranged 12.8%–41.0% across different definitions (Fleiss Kappa coefficient, 0.47; 95% CI 0.44 to 0.50). Overall, 140 (13.5 %) patients experienced adverse events within 30 days of discharge, including 108 unplanned readmissions (10.4%) and 32 deaths (3.1%). Clinical instability was associated with a 1.3-fold to 2.0-fold increase in the odds of postdischarge adverse events, depending on the definition, with c-statistics ranging 0.54–0.59 (p=0.31).Conclusion Clinical instability was associated with higher odds of 30-day postdischarge adverse events according to all but one of the published definitions. This study supports the validity of definitions that combine vital signs, mental status and the ability to maintain oral intake within 24 hours prior to discharge to identify patients at a higher risk of postdischarge adverse events.