MIMIC数据库研究套路汇总

官方
阅读 22 0 收藏 0 2025-06-05 19:01:00

套路1

生物标志物对某类患者不良结局的预后的影响

生物标志物包括:血糖相关、炎症相关等指标

患者包括:脓毒症、心血管疾病、脑血管疾病等

 案例文章 

标题:Association between stress hyperglycemia ratio index and all-cause mortality in critically ill patients with atrial fibrillation: a retrospective study using the MIMIC-IV database

期刊:Cardiovascular diabetologyIF=8.5

PMID:39402588

标题:Risk analysis of the association between different hemoglobin glycation index and poor prognosis in critical patients with coronary heart disease-A study based on the MIMIC-IV database 

期刊:Cardiovascular diabetologyIF=8.5

PMID:38555454

标题:The relationship between hemoglobin, albumin, lymphocyte, and platelet (HALP) score and 28-day mortality in patients with sepsis: a retrospective analysis of the MIMIC-IV database 

期刊:BMC infectious diseasesIF=3.4

PMID:40065235

套路总结:采用Kaplan-Meier生存曲线探讨不同4分组患者死亡结局的差异。采用Cox模型探究自变量和结局之间的关系,并根据多种因素进行调整。采用限制性立方样条回归模型来研究自变量和结局间的剂量反应关系或潜在非线性关联。根据年龄、性别等亚组分析验证结果的稳定性。

套路2

列线图+训练集和验证集做疾病不良预后的预测价值

训练集和验证集:

  • MIMIC IV数据库同时作为训练集和验证集;

  • MIMIC IV数据库作为训练集,MIMIC III或eICU数据库或自有数据库作为验证集;

  • MIMIC IV数据库作为训练集,同时Bootstrap分析用于内部验证。此外,来自 MIMIC III数据库的患者被进一步安置为验证队列。

 案例文章 

标题:The predictive values of admission characteristics for 28-day all-cause mortality in septic patients with diabetes mellitus: a study from the MIMIC database

期刊:Frontiers in endocrinologyIF=3.9

PMID:37608790

标题:Development and validation of a prediction model for the early occurrence of acute kidney injury in patients with acute pancreatitis

期刊:Renal failureIF=3.0

PMID:36999227

标题:Development and validation of a predictive model for pulmonary infection risk in patients with traumatic brain injury in the ICU: a retrospective cohort study based on MIMIC-IV

期刊:BMJ open respiratory researchIF=3.6

PMID:39089740

套路总结:使用LASSO回归以选择最佳预后因素。进行多因素COX回归分析以确定独立的预后因素和列线图构建。列线图通过自举法进行内部验证,并通过MIMIC III数据库进行外部验证,包括受试者工作特征(ROC)表示模型的敏感性、特异性以及模型的预测能力;校准曲线评估模型预测概率的可靠性;决策曲线分析(DCA)量化不同阈值概率下模型的净收益。

套路3

药物对某类患者人群不良结局的影响(改善或者加剧)

注意:这里的药物一定不能是指南中推荐或者禁止使用的

不良结局:包括死亡、ICU期间出现新发疾病(主要是急性肾损伤和谵妄)

 案例文章 

标题:Association between statin administration and Clostridium difficile-induced enteritis: a retrospective analysis of the MIMIC-IV database

期刊:Frontiers in pharmacologyIF=4.4

PMID:40066339

标题:The Relationship Between the Average Infusion Rate of Propofol and the Incidence of Delirium During Invasive Mechanical Ventilation: A Retrospective Study Based on the MIMIC IV Database

期刊:

CNS neuroscience & therapeuticsIF=4.8

PMID:40018993

标题:Association of early dexmedetomidine administration with the risk of ARDS in sepsis patients during ICU stay: results from MIMIC-IV

期刊:The American journal of the medical sciencesIF=2.3

PMID:40324522

标题:Association between acetaminophen and risk of mortality in patients with sepsis-associated acute kidney injury: A retrospective cohort study from the MIMIC-IV database

期刊:J. Investig. MedIF=2.5

PMID:39324182

套路总结:参与者根据药物使用情况分为两组:一组接受治疗,另一组未接受;使用倾向评分匹配(PSM)进行调整,以最大限度地减少两组之间的不平衡。Kaplan-Meier (KM)分析来确定药物使用对患者不良结局的影响(如果结局是死亡);用Cox(结局为死亡及死亡时间)或者logistics回归(结局是新发疾病)探究自变量和结局之间的关系,并根据多种因素进行调整。亚组分析验证结果的稳定性。

套路4:机器学习

 案例文章 

标题:Predicting Mortality in Sepsis-Associated Acute Respiratory Distress Syndrome: A Machine Learning Approach Using the MIMIC-III Database

期刊:

Journal of intensive care medicineIF=3.0

PMID:39234770

标题:Machine learning prediction models for multidrug-resistant organism infections in ICU ventilator-associated pneumonia patients: Analysis using the MIMIC-IV database

期刊:

Computers in biology and medicineIF=7.0

PMID:40154202

标题:Stress hyperglycemia ratio and machine learning model for prediction of all-cause mortality in patients undergoing cardiac surgery

期刊:Cardiovascular diabetologyIF=8.5

PMID:39955587

标题:Machine learning prediction models for mortality risk in sepsis-associated acute kidney injury: evaluating early versus late CRRT initiation

期刊:Frontiers in medicineIF=3.1

PMID:39911678

标题:Machine learning-based 28-day mortality prediction model for elderly neurocritically Ill patients

期刊:Computer methods and programs in biomedicineIF=4.9

PMID:39799642

套路总结:生成训练集和验证集(参考列线图的几种情况),用lasso或者boruta算法筛选变量;在训练集中应用多种机器学习算法(常见的包括逻辑回归、支持向量机、CatBoost、随机森林、XGBoost、LightGBM、AdaBoost 和 MLP等),根据之前筛选的变量进行建模;受试者工作特征(ROC)表示模型的敏感性、特异性以及模型的预测能力;校准曲线评估模型预测概率的可靠性;决策曲线分析(DCA)量化不同阈值概率下模型的净收益;使用Net Reclassification Improvement (NRI)和Integrated Discrimination Improvement (IDI)来量化比较模型与常见的评分系统的预测能力。

套路5:某指标动态轨迹对结局的影响

 案例文章 

标题:Influence of the mean airway pressure trajectory on the mortality and AKI occurrence in septic shock patients with mechanical ventilation: insights from the MIMIC-IV database

期刊:Frontiers in medicineIF=3.1

PMID:40109722

标题:Analysis of the correlation between the group-based trajectory modeling of serum osmolality and prognosis in patients with sepsis-associated encephalopathy at 72 h after admission

期刊:BMC infectious diseasesIF=3.4

PMID:39849352

套路总结:首先采用基于组的轨迹模型(GBTM)来识别指标的动态轨迹;使用 Kaplan-Meier(K-M)生存曲线比较轨迹组之间的生存差异;用Cox(结局为死亡及死亡时间)或者logistics回归(结局是新发疾病)探究这些轨迹和结局之间的关系,亚组分析用于评估结果在不同人群间的稳定性。


评论列表

发表评论