Objective To explore the influencing factors of antimicrobial drug-induced thrombocy-topenia and construct a predictive model based on machine learning (ML).

Methods Patients who received antimicrobial therapy at Qingbaijiang District People's Hospital in Chengdu from January 2020 to December 2022 were selected as study subjects. Antimicrobial drug-induced thrombocytopenia was set as the outcome variable. Independent risk factors for antimicrobial-induced thrombocytopenia were identified using least absolute shrinkage and selection operator regression and multivariate Logistic regression (LR) analysis. Synthetic Minority Over-sampling Technique-Nominal Continuous was applied for oversampling, and 5 different ML models were constructed based on feature variables to identify the optimal model. The performance of the predictive model for antimicrobial-induced thrombocytopenia was evaluated using area under the receiver operating characteristic curve (AUC), calibration curves, the Hosmer-Lemeshow test, and decision curve analysis (DCA). The Delong test was used to compare differences in AUC among models, and the integrated discrimination improvement index (IDI) and net reclassification improvement index (NRI) were calculated with conventional LR as the reference model. The best predictive model was comprehensively selected. The contribution and impact of key features on model predictions were interpreted using SHapley Additive exPlanations (SHAP).

Results A total of 701 patients were included, among whom 41 (5.85%) developed antibiotic-induced thrombocytopenia. Multivariate LR analysis revealed that elevated total bilirubin (TBIL) [OR=1.285, 95%CI (1.118, 1.477)] was an independent risk factor for antibiotic-induced thrombocytopenia, while elevated albumin (ALB) [OR=0.954, 95%CI (0.912, 0.998)] and higher creatinine clearance (CCr) [OR=0.856, 95%CI (0.749, 0.978)] were protective factors (P < 0.05). The eXtreme gradient boosting (XGBoost) model demonstrated the best predictive performance. Hosmer-Lemeshow test and the calibration curve of the XGBoost model showed a high consistency between predicted and observed risks. The DCA curve revealed that the XGBoost model achieved the highest net benefit across the entire threshold range (0-1.0). Based on bootstrap internal validation, the DeLong test showed statistically significant differences between the XGBoost model and the LR, random forest, decision tree, and support vector machine models (all P < 0.05). Compared with LR as the reference, the XGBoost model significantly outperformed other models in terms of IDI and NRI (all P < 0.05). SHAP analysis indicated that TBIL, ALB, and CCr levels played important roles in predicting antibiotic-induced thrombocytopenia.

Conclusion TBIL levels, ALB levels and CCr levels are the influencing factors of antimicrobial drug-induced thrombocytopenia. The XGBoost model has a good predictive performance for predicting platelet reduction caused by antibacterial drugs, providing a reference for screening high-risk patients with antimicrobial drug-induced thrombocytopenia.

1. KuwabaraG, TazoeK, ImotoW, et al. Isoniazid-induced immune thrombocytopenia[J]. Intern Med, 2021, 60(22): 3639-3643. DOI: 10.2169/internalmedicine.6520-20.

2. 周国威, 张宏岩, 矫艳艳, 等. 基于FAERS数据库的药物相关血小板减少症信号挖掘[J]. 药物流行病学杂志, 2025, 34(12): 1382-1389.ZhouGW, ZhangHY, JiaoYY, et al. Signal mining and analysis of drug-related thrombocytopenia based on FAERS database[J]. Chinese Journal of Pharmacoepidemiology, 2025, 34(12): 1382-1389. DOI: 10.12173/j.issn.1005-0698.202507057.

3. KunyuL, ShupingS, ChangS, et al. An updated comprehensive pharmacovigilance study of drug-induced thrombocytopenia based on fda adverse event reporting system data[J]. J Clin Pharmacol, 2024, 64(4): 478-489. DOI: 10.1002/JCPH.2389.

4. 李雪莲, 朱庆东, 马怡静, 等. 利奈唑胺血液系统不良反应发生率及危险因素分析: 一项多中心队列研究[J]. 中国防痨杂志, 2025, 47(6): 719-726.LiXL, ZhuQD, MaYJ, et al. Analysis of incidence and risk factorsfor linezolid-related hematological side effects: a multicenter cohort study[J]. Chinese Journal of Antituberculosi, 2025, 47(6): 719-726. DOI: 10.19982/j.issn.1000-6621.20240539.

5. 蔡乐, 汤智慧, 邱子涵, 等. 药源性血小板减少症的危险因素分析与风险预测模型的构建[J]. 临床药物治疗杂志, 2024, 22(8): 22-28.CaiL, TangZH, QiuZH, et al. Analysis of risk factors and construction of a risk prediction model for drug-induced thrombocytopenia[J]. Clinical Medication Journal, 2024, 22(8): 22-28. DOI: 10.3969/j.issn.1672-3384.2024.08.005.

6. AndrèsE, El Hassani HajjamA, MaloiselF, et al. Artificial intelligence (AI) and drug-induced and idiosyncratic cytopenia: the role of ai in prevention, prediction, and patient participation[J]. Hematol Rep, 2025, 17(3): 24. DOI: 10.3390/hematolrep17030024.

7. 廖茹, 崔祎, 程晓亮, 等. 基于机器学习算法的利奈唑胺相关血小板减少预测[J]. 医药导报, 2025, 44(4): 676-681.LiaoR, CuiY, ChengXL, et al. Prediction of linezolid-induced thrombocytopenia based on machine learning algorithm[J]. Herald of Medicine, 2025, 44(4): 676-681. DOI: 10.3870/j.issn.1004-0781.2025.04.028.

8. 黄翠丽, 高奥, 王嘉熙, 等. 抗菌药物相关血小板减少症的自动监测与评价研究[J]. 中国药物警戒, 2023, 20(7): 807-811.HuangCL, GaoA, WangJX, et al. Automatic monitoring and assessment of antibiotics-related thrombocytopenia[J]. Chinese Journal of Pharmacovigilance, 2023, 20(7): 807-811. DOI: 10.19803/j.1672-8629.20220348.

9. NietschKS, RoachTM, WilsonZD, et al. Principles and considerations in the early identification and prehospital treatment of thrombocytopenia[J]. J Spec Oper Med, 2022, 22(2): 75-79. DOI: 10.55460/333T-XIYF.

10. LiJ, YanZ. Machine learning model predicting factors for incisional infection following right hemicolectomy for colon cancer[J]. BMC Surg, 2024, 24(1): 279. DOI: 10.1186/s12893-024-02543-8.

11. 王梅英, 杨敏, 刘佳微, 等. 基于SMOTE算法的化疗肿瘤患者下呼吸道感染预警模型构建[J]. 中国感染控制杂志, 2021, 20(12): 1094-1101.WangMY, YangM, LiuJW, et al. Construction of early warning model of lower respiratory tract infection in chemotherapy tumor patients based on SMOTE algorithm[J]. Chinese Journal of Infection Control, 2021, 20(12): 1094-1101. DOI: 10.12138/j.issn.1671-9638.20211135.

12. YuanH, XuH. Deep multi-modal fusion network with gated unit for breast cancer survival prediction[J]. Comput Methods Biomech Biomed Engin, 2024, 27(7): 883-896. DOI: 10.1080/10255842.2023.2211188.

13. Gozukara BagHG, YaginFH, GormezY, et al. Estimation of obesity levels through the proposed predictive approach based on physical activity and nutritional habits[J]. Diagnostics (Basel), 2023, 13(18): 2949. DOI: 10.3390/diagnostics13182949.

14. LiuHQ, LinSY, SongYD, et al. Machine learning on MRI radiomic features: identification of molecular subtype alteration in breast cancer after neoadjuvant therapy[J]. Eur Radiol, 2023, 33(4): 2965-2974. DOI: 10.1007/S00330-022-09264-7.

15. 甘文思, 黄一睿, 王笑青, 等. 基于机器学习的ICU老年患者呼吸机相关肺炎风险预测模型的构建及评价[J]. 中华医院感染学杂志, 2025, 35(2): 290-296.GanWS, HuangYR, WangXQ, et al. Establishment of risk prediction model for ventilator-associated pneumoniain elderly ICU patients based on machine learning and its performance[J]. Chinese Journal of Nosocomiology, 2025, 35(2): 290-296. DOI: 10.11816/cn.ni.2025-240990.

16. 马佳佳, 刘晓芯, 薛蓓, 等. 基于机器学习算法的肺癌四级胸腔镜手术后肺部感染风险预测模型构建[J]. 实用临床医药杂志, 2025, 29(6): 111-117.MaJJ, LiuXX, XueB, et al. Risk prediction model construction of postoperativepulmonary infection in lung cancer patients undergoingfour-level thoracoscopic surgery based on machinelearning algorithms[J]. Journal of Clinical Medicine in Practice, 2025, 29(6): 111-117. DOI: 10.7619/jcmp.20245679.

17. ShiC, XiaJ, YeJ, et al. Effect of renal function on the risk of thrombocytopaenia in patients receiving linezolid therapy: a systematic review and meta-analysis[J]. Br J Clin Pharmacol, 2022, 88(2): 464-475. DOI: 10.1111/bcp.14965.

18. GouJ, LiQ, FanN, et al. High accumulation of linezolid and its major metabolite in the serum of patients with hepatic and renal dysfunction is significantly associated with thrombocytopenia and anemia[J]. Microbiol Spectr, 2025, 13(7): 0249324. DOI: 10.1128/spectrum.02493-24.

19. LiuY, HuangJ, LiL, et al. Regulatory effect of PDGF/PDGFR on hematopoiesis[J]. Semin Thromb Hemost, 2025, 51(5): 572-577. DOI: 10.1055/S-0044-1796630.

20. BukaRJ, MontagueSJ, MoranLA, et al. PF4 activates the c-Mpl-Jak2 pathway in platelets[J]. Blood, 2024, 143(1): 64-69. DOI: 10.1182/blood.2023020872.

21. 马勇, 高伟波, 朱继红. 细菌性肝脓肿患者发生血小板减少影响因素研究[J]. 中国全科医学, 2023, 26(17): 2120-2124.MaY, GaoWB, ZhuJH. Risk factors of thrombocytopenia caused by pyogenic liver abscess[J]. Chinese General Practice, 2023, 26(17): 2120-2124. DOI: 10.12114/j.issn.1007-9572.2022.0742.

22. 赵资德, 吴令杰, 张海生, 等. 纠正低蛋白血症减少抗结核药物性肝损伤发生临床研究[J]. 实用肝脏病杂志, 2022, 25(6): 792-795.ZhaoZD, WuLJ, ZhangHS, et al. Could the supplement of human blood albumin product reduce the incidence of drug-induced liver injury in patients with pulmonary tuberculosis and hypoproteinemia[J]. Journal of Practical Hepatology, 2022, 25(6): 792-795. DOI: 10.3969/j.issn.1672-5069.2022.06.009.

23. 崔大广, 肖玲燕, 史东阳, 等. 特比澳对肝病合并血小板减少症患者的疗效分析[J]. 肝脏, 2022, 27(12): 1335-1339.CuiDG, XiaoLY, ShiDY, et al. Analysis of application of recombinant human thrombopoietin in patients with liver disease complicated with thrombocytopenia[J]. Chinese Hepatology, 2022, 27(12): 1335-1339. DOI: 10.3969/j.issn.1008-1704.2022.12.022.

24. HaoY, SunJ, WangX, et al. Difference in hematocrit and plasma albumin levels as an early biomarker of severity and prognosis in patients with severe fever and thrombocytopenia syndrome[J]. J Med Virol, 2024, 96(10): 29941. DOI: 10.1002/jmv.29941.

25. ChenC, LiY, YuJ, et al. Linezolid-induced thrombocytopenia in patients with acute myeloid leukemia: a matched case-control study[J]. Clin Transl Oncol, 2022, 24(3): 540-545. DOI: 10.1007/s12094-021-02711-9.

26. LimHI, CukerA. Thrombocytopenia and liver disease: pathophysiology and periprocedural management[J]. Hematology Am Soc Hematol Educ Program, 2022, 2022(1): 296-302. DOI: 10.1182/hematology.2022000408.