Pleural Fluid Analytes Collected at an Academic Medical Center, February 2011 – January 2026
Description
This dataset accompanies a retrospective cohort study examining pleural fluid pH in adult thoracentesis specimens. The study had three objectives: (1) to estimate pleural fluid pH in specimens meeting prespecified non-pathological criteria, (2) to evaluate the diagnostic performance of pH alone for classifying exudative versus transudative effusions, & (3) to identify biochemical predictors of pleural fluid pH including the causes of effusion using linear mixed-effects models. Data were collected from (mostly) adult patients who underwent thoracentesis at UC Davis Medical Center in Sacramento, California, between February 2011 & January 2026. Pleural fluid pH was measured using a blood gas analyzer, & pleural fluid chemistry (cholesterol, lactate dehydrogenase [LDH], total protein, glucose) was analyzed using standard automated clinical laboratory analyzers. Effusions were classified primarily by Light’s criteria, with Costa’s criteria and a modified LDH-plus-cholesterol definition used in sensitivity analyses. A subcategory of specimens were classified as non-pathological if they met all the following: (1) pleural fluid pH > 7.45; (2) thoracentesis volume < 500 mL; (3) no identifiable pathological etiology in the clinical record. The dataset comprises 2,315 pleural fluid specimens from 1,953 unique patients. Each record includes the pleural fluid pH, concentrations of cholesterol (mg/dL), LDH (U/L), total protein (g/dL), and glucose (mg/dL) in pleural fluid, the upper limit of normal for serum LDH, patient demographics (age at collection, sex, height, weight, BMI), & collection-to-result times (minutes) for pH and the fluid chemistry panel. A binary condition variable classified each specimen as non-pathological (0) or pathological (1). Identifying information has been removed. Patients have been assigned sequential de-identified subject numbers. Some patients contributed multiple specimens & are ordered chronologically by collection date within each subject via the sample number. Key findings include that non-pathological pleural fluid pH (median 7.54) was lower than historically cited values. Pleural fluid pH provided moderate discrimination between exudative and transudative effusions (AUC 0.72; 95% CI: 0.70–0.74). Parapneumonic effusions, hepatic hydrothorax, cardiac issues, & LDH + glucose concentrations in pleural fluid were major predictors of pleural fluid pH (marginal R² = 0.27). Researchers using this dataset should account for within-subject correlation, as 15% of patients contributed more than one specimen. Mixed-effects models or generalized estimating equations are recommended. Missing data are present for some biochemical and anthropometric variables and were not missing completely at random (Little's MCAR test p < 0.001); multiple imputation under the missing-at-random assumption is appropriate. The Variable Dictionary in the accompanying Excel file provides variable names, labels, data types, and coded value descriptions.
Files
Steps to reproduce
This retrospective cohort study used EHR data from UC Davis Medical Center. The population included adult patients who underwent thoracentesis from Feb 2011 to Jan 2026. Demographics (age, sex, height, weight, race, ethnicity) came from the EHR. BMI was calculated from height and weight. Living status and dates of death came from institutional records for survival analyses. Pleural fluid specimens were collected during clinically indicated thoracentesis. Pleural fluid pH was measured using a blood gas analyzer. Pleural fluid chemistry — cholesterol, lactate dehydrogenase (LDH), total protein, glucose, albumin, amylase, triglycerides, and hematocrit — was analyzed using automated chemistry analyzers. The upper limit of normal for serum LDH was recorded for LDH ratio calculations. Collection-to-result times were recorded in minutes for each assay. Effusion classification followed a two-step approach. The primary classification used cholesterol and LDH criteria: effusions were exudative if pleural fluid LDH exceeded 141 IU/L or cholesterol exceeded 55 mg/dL, and transudative if neither criterion was met. Light's criteria classification was also recorded. A subset was classified as non-pathological if pH > 7.45, thoracentesis volume was <500 mL, and no pathological etiology was identified. Two reviewers independently evaluated candidate specimens; inclusion required concordance. Statistical analyses were performed in R. Linear mixed-effects models (lme4 and lmerTest) with random intercepts for patient identifier accounted for multiple specimens per patient. Missing biochemical data (25–71% across analytes) were addressed using multiple imputation with predictive mean matching (mice), generating 20 datasets pooled using Rubin's rules. Little's MCAR test confirmed data were not MCAR, supporting imputation under the missing-at-random assumption. Diagnostic performance was evaluated using ROC analysis (pROC) with cluster bootstrap confidence intervals (2,000-10,000 iterations) resampled at the patient level. Model fit was quantified using marginal and conditional R² via the Nakagawa-Schielzeth method (MuMIn). Linear mixed-effects models identified predictors of pleural fluid pH. Kaplan-Meier curves compared survival between effusion types under Light's criteria. A Cox model adjusted for age, sex, and BMI estimated the effect of effusion type on mortality. PH assumption was evaluated with scaled Schoenfeld residuals; a time-varying-coefficient Cox model was fit when PH violations were detected. Statistical significance was set at 0.05. Specimens were excluded for technical assay error, insufficient clinical documentation, or missing key demographic variables. To reproduce these analyses, researchers should use this de-identified dataset, employ mixed-effects models for within-subject correlation, and apply multiple imputation for missing biochemical variables. The Variable Dictionary Tab defines all variable names, labels, and coded values.
Institutions
- University of California, DavisCalifornia, Davis