Tuberculous meningitis (TM) is a common intracranial infection with high misdiagnosis rate due to slow onset, atypical symptoms and difficult etiological diagnosis. There are few available effective diagnostic tools for TM. Developing a diagnostic scoring system based on the common symptoms and examination findings of TM may help to improve the diagnostic accuracy and reduce the misdiagnosis rate regarding TM.

To establish a clinical diagnostic scoring system for TM and to evaluate its application value.

One hundred and eighty-seven inpatients with a diagnosis of meningitis were selected from Department of Infectious Diseases, Liuzhou People's Hospital from November 2011 to September 2021, and randomly divided into a model group (n=147, including 71 with TM, and 76 with non-TM) and a validation group (n=40) by a SPSS (version 21.0) algorithm. General data of all cases were collected, mainly including sex, age, clinical symptoms (fever, headache, disturbance of consciousness, cervical resistance) , laboratory and imaging examination results〔including prevalence of HIV infection, CD₄⁺ T lymphocyte count, C-reactive protein, intracranial pressure and routine cerebrospinal fluid (CSF) biochemical markers (glucose, chlorine, protein and cell number) 〕. The influencing factors of TM in the model group were identified using Multivariate Logistic regression, and were used to develop a diagnostic scoring system for TM with each factor rated according to its β coefficient. Then its predictive value for TM was tested using the receiver operating characteristic (ROC) curve.

TM and non-TM patients in the model group had statistically significant differences in the prevalence of headache and HIV infection, CD₄⁺ T lymphocyte count less than 200/μl, elevated C-reactive protein, intracranial pressure greater than 200 mm H₂O, decreased glucose and chlorine in CSF, elevated protein and monocytes in CSF (P<0.05) . Multivariate Logistic regression analysis revealed that headache, CD₄⁺ T lymphocyte cell count less than 200/ μl, elevated C-reactive protein, decreased CSF glucose, and elevated CSF protein were associated with TM (P<0.05) . Based on our clinical practical experience, CSF chlorine and monocytes were added in the multivariate Logistic regression, and further analysis found that headache, CD₄⁺ T lymphocyte count less than 200/μl, elevated C-reactive protein, decreased CSF glucose, and elevated CSF protein were still the associated factors of TM (P<0.05) . The clinical diagnostic scoring system for TM was developed using the aforementioned 7 factors with their values assigned based on the β coefficient value for scoring, among which the assigned value for decreased CSF was 1 point, and the assigned value for each of the other six factors was calculated using that of decreased CSF multiplying the corresponding multiples. For the ease of clinical use, the assigned values for both headache and CD₄⁺ T lymphocyte count less than 200/μl were increased by 2.5 points since the original assigned values of them were negative. The area under the curve (AUC) of the clinical diagnostic scoring system for TM in predicting TM in model group was 0.807〔95%CI (0.735, 0.879) , standard error=0.037, P<0.001〕with 21.50 points as the optimal cut-off value. And that of the system in predicting TM in the validation group was 0.766〔95%CI (0.610, 0.921) , standard error=0.079, P=0.004〕, with sensitivity of 0.789 and specificity of 0.667.

The clinical diagnostic scoring system for TM developed using seven factors is simple and easy-to-use, which has proven to be effective in early diagnosis of TM.

结核性脑膜炎（TM）是临床常见的中枢性感染的一种，其起病较慢，症状不典型，病原学诊断困难，误诊率高。目前有效的TM诊断工具较少。利用常见的临床症状、检查指标等建立诊断评分系体可提高诊断准确率，减少误诊。

建立TM临床诊断评分体系（TMCDS），并对其应用价值进行初步评价。

选取2011年11月至2021年9月在柳州市人民医院感染病科住院并诊断为脑膜炎的患者187例为研究对象，采用SPSS 21.0统计软件将患者随机分成建模组（147例）和验模组（40例）。根据是否为TM将建模组分为非TM亚组（76例）和TM亚组（71例）。收集患者的一般资料，主要包括性别、年龄、临床症状（发热、头痛、意识障碍、颈抵抗），实验室及影像学检查结果，包括人类免疫缺陷病毒（HIV）感染情况、CD₄⁺ T淋巴细胞计数、C反应蛋白、颅内压、脑脊液常规生化检查（糖、氯、蛋白、细胞数）。建模组采用多因素Logistic回归分析探讨TM的影响因素；根据每个因素的β值所占比重设立相应分值，建立TMCDS；采用受试者工作特征曲线（ROC曲线）分析TMCDS诊断TM的价值。

两亚组头痛、HIV感染、CD₄⁺ T淋巴细胞计数<200/μl、C反应蛋白升高、颅内压>200 mm H₂O（1 mm H₂O=0.009 8 kPa）、脑脊液糖降低、脑脊液氯降低、脑脊液蛋白升高、脑脊液单核细胞升高情况比较，差异有统计学意义（P<0.05）。多因素Logistic回归分析结果显示，头痛、CD₄⁺ T淋巴细胞<200/μl、C反应蛋白升高、脑脊液糖降低、脑脊液蛋白升高均是TM的影响因素（P<0.05）。将以上5个影响因素同时结合临床经验纳入脑脊液氯、脑脊液细胞数再次进行多因素Logistic回归分析，结果显示，头痛、CD₄⁺ T淋巴细胞<200/μl、C反应蛋白升高、脑脊液糖降低、脑脊液蛋白升高均是TM的影响因素（P<0.05）。根据上述7个因素β值建立评分系统，将脑脊液氯降低β值设定为1分，其他因素β值与其的倍数即为该因素所对应的分值，因2个影响因素评分为负值，为方便临床，每个因素对应分值增加2.5分，最终建立TMCDS。TMCDS诊断建模组TM的ROC曲线下面积（AUC）为0.807〔95%CI（0.735，0.879），标准误=0.037，P<0.001〕，最佳诊断界值为21.50分。TMCDS诊断验模组TM的AUC为0.766〔95%CI（0.610，0.921），标准误=0.079，P=0.004〕，灵敏度为0.789，特异度为0.667。

通过7个变量建立的TMCDS简单易行，对于早期TM具有较高的临床诊断价值。