HIV/AIDS合并社区获得性肺炎临床预后评分系统的建立及验证

背景　社区获得性肺炎（CAP）是人类免疫缺陷病毒（HIV）/艾滋病（AIDS）患者最常见的机会性感染疾病。对于HIV/AIDS合并CAP患者，最重要的是评估患者的病情严重程度，制订预测短期死亡率的严重程度评分，以便在患者初步就诊时做出更客观的决策，对患者的预后至关重要。目的　建立HIV/AIDS合并CAP患者临床预后评分系统，并验证其预测效能。方法　回顾性分析2016-2019年在柳州市人民医院住院的615例HIV/AIDS合并CAP患者的临床资料，随机分为建模组455例和验模组160例，根据预后评估将建模组分为好转亚组和恶化亚组，收集患者入院24 h内的一般信息、合并基础疾病、生命体征、实验室检查（血常规、电解质、肝肾功能、血气分析等）及影像学检查等资料，并对建模组临床资料数据进行多因素Logistic回归分析，筛选影响患者临床预后的危险因素，根据危险因素建立HIV/AIDS合并CAP患者临床预后评分系统。使用Kaplan-Meier生存分析比较不同风险组的恶化死亡率，并利用验模组数据验证该评分系统的预测效能。结果　好转亚组进入重症加强护理病房（ICU）、意识障碍、呼吸>30次/min、低血压、血小板计数（PLT）<100×109/L、红细胞比容（HCT）<35%、pH值<7.35或>7.45、血氧分压（PaO2）<60 mm Hg、血氧饱和度（SaO2）<93%、尿素氮（BUN）>7 mmol/L、乳酸脱氢酶（LDH）>230 U/L、血清白蛋白（ALB）<30 g/L、总胆红素（TBil）>34.2 μmol/L、丙氨酸氨基转移酶（ALT）>40 U/L、天冬氨酸氨基转移酶（AST）>40 U/L、血钠（Na）<135 mmol/L或>145 mmol/L、CD4淋巴细胞计数<50个/mm3比例与恶化亚组比较，差异均有统计学意义（P<0.05）。多因素Logistic回归分析，结果显示，进入ICU、意识障碍、低血压、PLT<100×109/L、HCT<35%、SaO2<93%、LDH>230 U/L、ALT>40 U/L、Na<135 mmol/L或>145 mmol/L及CD4淋巴细胞计数<50个/mm3共10个变量是影响患者预后的独立危险因素（P<0.05）。建立评分系统包括进入ICU（6分）、意识障碍（2分）、低血压（1分）、PLT<100×109/L（1分）、HCT<35%（2分）、SaO2<93%（3分）、LDH>230 U/L（1分）、 ALT>40 U/L（2分）、Na<135 mmol/L或>145 mmol/L（2分）、CD4淋巴细胞计数<50个/mm3（1分），根据评分系统得分分为低风险组（0~6分）、中风险组（7~12分）、高风险组（12分以上），Kaplan-Meier生存分析显示，HIV/AIDS合并CAP患者不同风险组间的恶化死亡率比较，差异有统计学意义（χ2=87.634，P<0.001）。建模组的受试者工作特征（ROC）曲线下面积（AUC）为0.858，其灵敏度、特异度分别为77.9%、78.4%。验模组的AUC为0.820，其灵敏度、特异度分别为73.7%、77.6%。结论　根据HIV/AIDS合并CAP患者临床预后独立危险因素建立的评分系统具有良好的预测效应能力。

Development and Validation of a Clinical Prognosis Scoring System for HIV/AIDS Patients with Community-acquired Pneumonia

Background　Community-acquired pneumonia（CAP） is the most common opportunistic infection among people with human immunodeficiency virus（HIV）/acquired immune deficiency syndrome（AIDS）. Evaluating disease severity and scoring short-term mortality will greatly help physicians to make objective decisions during the initial visit of HIV/AIDS patients with CAP，which is crucial to patient prognosis. Objective　To develop and verify a clinical prognosis scoring system for HIV/AIDS patients with CAP. Methods　Clinical data of 615 HIV/AIDS patients with CAP（455 in modeling group，and 160 in validation group） recruited from Liuzhou People's Hospital from 2016 to 2019 were retrospectively analyzed，including demographics，underlying diseases，24-hour post-admission vital signs，clinical laboratory results（routine blood test，liver and kidney function tests，blood gas analysis），imaging examination and so on. Multivariate Logistic regression analysis of the clinical data of modeling group was conducted to screen for independent risk factors of clinical prognosis to develop a clinical prognosis scoring system. Kaplan-Meier survival analysis was used to compare the worsening mortality between different risk groups rated by the scoring system. And the predictive performance of the scoring system was tested by using the data of validation group. Results　Patients with improved and deteriorated conditions had significant differences in the prevalence of admission to the ICU，disturbance of consciousness，respiratory rate>30 breaths per minute，hypotension，platelet blood count（PLT）<100×109/L，hematocrit （HCT） <35%，pH<7.35 or >7.45，partial pressure of oxygen （PaO2） <60 mm Hg，oxygen saturation（SaO2）<93%，urea nitrogen（BUN）>7 mmol/L，lactate dehydrogenase（LDH）>230 U/L，seralbumin（ALB）<30 g/L，total bilirubin（TBil）>34.2 μmol/L，alanine aminotransferase（ALT）>40 U/L，aspartate aminotransferase（AST）>40 U/L，serum sodium（Na）<135 mmol/L or>145 mmol/L，and CD4 lymphocyte count<50 cells/mm3 （P<0.05）. Multivariate Logistic regression analysis identified 10 risk factors for prognosis: admission to the ICU，disturbance of consciousness，hypotension，PLT<100×109/L，HCT<35%，SaO2<93%，LDH>230 U/L，ALT>40 U/L，Na<135 mmol/L or >145 mmol/L and CD4 lymphocyte count <50 cells/mm3. These 10 factors were included in the prognosis scoring system and assigned the value of 6，2，1，1，2，3，1，2，2，1，respectively，and mortality risks assessed using the system were stratified into low risk（0-6 points），medium risk（7-12 points）and high risk（more than 12 points）. Kaplan- Meier survival analysis showed that the worsening mortality varied significantly across different risk groups（χ2=87.634，P<0.001）. In predicting the mortality risk，the scoring system had an AUC of 0.858 with 77.9% sensitivity and 78.4% specificity in modeling group，and had an AUC of 0.820 with 73.7% sensitivity and 77.6% specificity in validation group. Conclusion　Our clinical prognosis scoring system based on with risk factors for CAP in HIV/AIDS has been confirmed with good predictive ability.