Pleural Fluid Analytes Collected at an Academic Medical Center, February 2011 – January 2026

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.本数据集配套一项针对成人胸腔穿刺标本（thoracentesis specimens）中胸腔积液pH值（pleural fluid pH）的回顾性队列研究。本研究共设定三项研究目标：（1）估算符合预先指定非病理学标准的标本的胸腔积液pH值；（2）单独评估pH值对渗出性胸腔积液（exudative effusions）与漏出性胸腔积液（transudative effusions）的诊断效能；（3）借助线性混合效应模型（linear mixed-effects models），明确胸腔积液pH值的生化预测因子，包括胸腔积液的病因。 数据采集自2011年2月至2026年1月期间，于加利福尼亚州萨克拉门托市加州大学戴维斯分校医学中心（UC Davis Medical Center）接受胸腔穿刺术的（以成人为主）患者。胸腔积液pH值采用血气分析仪（blood gas analyzer）检测，胸腔积液生化指标[胆固醇、乳酸脱氢酶（lactate dehydrogenase, LDH）、总蛋白、葡萄糖]则通过标准自动化临床实验室分析仪完成检测。胸腔积液主要依据Light标准（Light’s criteria）进行分类，敏感性分析中采用Costa标准（Costa’s criteria）及改良LDH-胆固醇联合定义。若标本满足以下全部条件，则被归为非病理性标本：（1）胸腔积液pH＞7.45；（2）胸腔穿刺液量＜500 mL；（3）临床病历中未发现明确的病理性病因。 本数据集包含来自1953名独立患者的2315份胸腔穿刺标本。每条记录涵盖胸腔积液pH值、胸腔积液中胆固醇（单位：mg/dL）、LDH（单位：U/L）、总蛋白（单位：g/dL）及葡萄糖（单位：mg/dL）浓度、血清LDH正常上限、患者人口学资料[采集时年龄、性别、身高、体重、体质量指数（Body Mass Index, BMI）]，以及pH值与生化指标检测的样本送检至出结果时间（单位：分钟）。设置二元状态变量，将每份标本分类为非病理性（0）或病理性（1）。所有识别信息均已去除，患者被分配为连续的去标识化受试者编号。部分患者提供了多份标本，按采集日期的先后顺序，以样本编号对每名受试者内的标本进行排序。 主要研究结果显示，非病理性胸腔积液pH值（中位数为7.54）低于既往文献引用的参考值。胸腔积液pH值对渗出性胸腔积液与漏出性胸腔积液具有中等区分效能（曲线下面积（Area Under the Curve, AUC）=0.72；95%置信区间（Confidence Interval, CI）：0.70~0.74）。类肺炎性胸腔积液（parapneumonic effusions）、肝性胸水（hepatic hydrothorax）、心脏疾病及胸腔积液中LDH与葡萄糖浓度是胸腔积液pH值的主要预测因子（边际R²=0.27）。 使用本数据集的研究人员需考虑受试者内相关性，因为15%的患者提供了多份标本，推荐采用线性混合效应模型或广义估计方程（generalized estimating equations）进行分析。部分生化及人体测量学变量存在缺失数据，且并非完全随机缺失（Little's MCAR检验，P＜0.001）；基于随机缺失（missing-at-random）假设的多重插补（multiple imputation）方法适用于此类缺失数据的处理。配套Excel文件中的变量字典提供了变量名称、标签、数据类型及编码值说明。