Abstract Rationale Complicated parapneumonic effusion (CPPE) is a pleural infection with high morbidity and healthcare costs. Research has been limited by small single-center cohorts and lack of validated electronic health record (EHR) phenotyping methods. We developed and validated a rule-based EHR algorithm integrating structured and unstructured pleural data to identify CPPE cases and evaluate survival and hospital length of stay. Methods Adult admissions (2017-2025) across six affiliated hospitals (community through quaternary) were screened for pleural procedures and fluid studies. Procedures were identified using structured orders, note searches for “thoracentesis”, “thoracostomy”, or “chest tube”, and documented chest tube outputs from flowsheets. CPPE was defined by one or more pleural fluid findings consistent with infection (pH 7.2, glucose 60 mg/dL, positive microbiology, or turbid/purulent fluid description) and an ICD-10 code for pneumonia, empyema, sepsis, or respiratory failure. Algorithm performance was assessed by clinician chart review of randomly selected 75 admissions (50 positive, 25 negative) evenly distributed across hospitals and years. CPPE was adjudicated using predefined criteria and categorized as Definite, Likely, Possible or Not CPPE (Figure 1). Algorithm performance metrics were calculated under restrictive and inclusive definition as outlined in Figure 1, and compared to those using ICD-10 codes for empyema or pneumonia and non-malignant pleural effusion (Figure 1). Ninety-five percent confidence intervals were calculated using the exact binomial method. Results Among 35,046 admissions with pleural procedures, 9,403 had pleural fluid studies, of which 1,659 met algorithmic CPPE criteria and 843 underwent chest-tube drainage. CPPE cases were most often identified by low pleural pH (49%) or glucose (50%), followed by positive microbiology (42%), or turbid fluid (27%). Corresponding ICD-10 codes included pneumonia (62%), empyema (58%), sepsis (49%), and respiratory failure (63%). The EHR algorithm achieved substantially higher sensitivity and negative predictive value (NPV) than ICD-10 codes for either empyema or pneumonia and pleural effusion (Figure 1). Among 843 CPPE admissions, median age was 65 years (IQR 53-75), 64% were male, and 66% white. Median hospital stay was 15 days (IQR 9-31) with 12% in-hospital mortality; 46% were discharged home (two-thirds with home health), 28% to rehabilitation, and 7% to hospice. Conclusions We developed and validated a multicenter EHR algorithm that accurately identifies CPPE across six hospitals. The algorithm demonstrated markedly higher sensitivity and NPV than ICD-10-based definitions, enabling reproducible multicenter cohort creation. This approach produced one of the largest CPPE cohorts to date and confirmed persistently poor outcomes despite modern management. This abstract is funded by: T32 HL007586, HL161196, HL155159, K08CA279881
Toro et al. (Fri,) studied this question.