1.Explainable Machine Learning Model for Predicting Prognosis in Patients with Malignant Tumors Complicated by Acute Respiratory Failure: Based on the eICU Collaborative Research Database in the United States
Zihan NAN ; Linan HAN ; Suwei LI ; Ziyi ZHU ; Qinqin ZHU ; Yan DUAN ; Xiaoting WANG ; Lixia LIU
Medical Journal of Peking Union Medical College Hospital 2026;17(1):98-108
To develop and validate a model for predicting intensive care unit (ICU) mortality risk in patients with malignant tumors complicated by acute respiratory failure (ARF) based on an explainable machine learning framework. Clinical data of patients with malignant tumors and ARF were extracted from the eICU Collaborative Research Database in the United States, including demographic characteristics, comorbidities, vital signs, laboratory test indicators, and major interventions within the first 24 hours after ICU admission.The study outcome was ICU death.Enrolled patients were randomly divided into a training set and a validation set at a ratio of 7:3.Predictor variables were selected using least absolute shrinkage and selection operator (LASSO) regression.Five machine learning algorithms-extreme gradient boosting (XGBoost), support vector machine (SVM), Logistic regression, multilayer perceptron (MLP), and C5.0 Decision Tree-were employed to construct predictive models.Model performance was evaluated based on the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and other metrics.The optimal model was further interpreted using the Shapley additive explanations (SHAP) algorithm. A total of 3196 patients with malignant tumors complicated by ARF were included.The training set comprised 2, 261 patients and the validation set 935 patients; 683 patients died during ICU stay, while 2513 survived.LASSO regression ultimately selected 12 variables closely associated with patient ICU outcomes, including sepsis comorbidity, use of vasoactive drugs, and within the first 24 hours after ICU admission: minimum mean arterial pressure, maximum heart rate, maximum respiratory rate, minimum oxygen saturation, minimum serum bicarbonate, minimum blood urea nitrogen, maximum white blood cell count, maximum mean corpuscular volume, maximum serum potassium, and maximum blood glucose.After model evaluation, the XGBoost model demonstrated the best performance.The AUCs for predicting ICU mortality risk in the training and validation sets were 0.940 and 0.763, respectively; accuracy was 88.3% and 81.2%;sensitivity was 98.5% and 95.9%.Its predictive performance also remained optimal in sensitivity analyses.SHAP analysis indicated that the top five variables contributing to the model's predictions were minimum oxygen saturation, minimum serum bicarbonate, minimum mean arterial pressure, use of vasoactive drugs, and maximum white blood cell count. This study successfully developed a mortality risk prediction model for ICU patients with malignant tumors complicated by ARF based on a large-scale dataset and performed explainability analysis.The model aids clinicians in early identification of high-risk patients and implementing individualized interventions.
2.Association between short-term exposure to air pollution and outpatient and emergency visits for neurological diseases in Conghua District, Guangzhou from 2015 to 2022
Lu LUO ; Zhi LI ; Yanmei CAI ; Chunming HE ; Yi ZHENG ; Sirong WANG ; Ruijun XU ; Yuewei LIU ; Qinqin JIANG
Journal of Environmental and Occupational Medicine 2025;42(11):1307-1314
Background Exposure to air pollutants increases the risk of diseases in multiple systems, including respiratory and cardiovascular systems, yet its association with neurological diseases remains unclear. Objective To quantitatively evaluate the association between short-term exposure to air pollutants and outpatient and emergency visits for neurological diseases, identify potential susceptible populations, and quantify associated disease burden. Methods Daily 24-hour average concentrations of fine particulate matter (PM2.5), inhalable particulate matter (PM10), sulfur dioxide (SO2), nitrogen dioxide (NO2), and carbon monoxide (CO), daily maximum 8-hour average concentration of ozone (O3), daily meteorological data (24-hour average temperature, 24-hour average relative humidity), and data on daily outpatient and emergency department visits for neurological diseases from two hospitals in Conghua District, Guangzhou, China, were collected from 2015 to 2022. A time-stratified case-crossover design was adopted, and a conditional Poisson regression model was constructed to analyze the association between air pollution exposure and neurological disease visits. Two-pollutant models and sensitivity analysis were used to validate model stability. Stratified analyses by season (cold season: from November to March; warm season: from April to October), sex (male, female), and age (≤45 years, 46–60 years, ≥61 years) were performed to identify vulnerable group. Additionally, the number and proportion of neurological disease visits attributable to short-term air pollutant exposure were calculated. Results A total of 72 673 outpatient and emergency department visits for neurological diseases were included during the study period. Most of the patients were middle-aged and elderly individuals (69.89% were over 45 years old) and females (60.25%). The results of single-pollutant models showed that for each interquartile range (IQR) increase in exposure to PM2.5, PM10, SO2, NO2, CO, and O3, the risk of outpatient and emergency department visits for neurological diseases increased by 7.54% (95%CI: 4.69%, 10.46%), 6.66% (95%CI: 3.92%, 9.46%), 16.72% (95%CI: 10.58%, 23.19%), 8.12% (95%CI: 4.82%, 11.53%), 5.60% (95%CI: 2.34%, 8.97%), and 6.11% (95%CI: 2.91%, 9.40%), respectively. The results of the two-pollutant model showed that the association between PM2.5 and SO2 exposure and outpatient and emergency department visits for neurological diseases were relatively stable. The stratified analyses showed that the effect of SO2 was stronger in the cold season. It was estimated that 8.32% (95%CI: 5.55%, 10.96%) and 6.65% (95%CI: 4.27%, 8.96%) of the outpatient and emergency department visits were attributable to short-term exposure to SO2 and PM2.5, respectively. Conclusion Exposure to PM2.5 and SO2 is associated with increased risks of outpatient and emergency visits for neurological diseases. SO2 shows stronger effects during the cold season, and exposure to air pollution contributes to up to 8.32% of neurological disease visits.
3.Potential utility of albumin-bilirubin and body mass index-based logistic model to predict survival outcome in non-small cell lung cancer with liver metastasis treated with immune checkpoint inhibitors.
Lianxi SONG ; Qinqin XU ; Ting ZHONG ; Wenhuan GUO ; Shaoding LIN ; Wenjuan JIANG ; Zhan WANG ; Li DENG ; Zhe HUANG ; Haoyue QIN ; Huan YAN ; Xing ZHANG ; Fan TONG ; Ruiguang ZHANG ; Zhaoyi LIU ; Lin ZHANG ; Xiaorong DONG ; Ting LI ; Chao FANG ; Xue CHEN ; Jun DENG ; Jing WANG ; Nong YANG ; Liang ZENG ; Yongchang ZHANG
Chinese Medical Journal 2025;138(4):478-480
4.Microbial population distribution on outer surface and in contents of SPF chicken embryos for vaccine production
Chinese Journal of Biologicals 2025;38(09):1025-1028+1034
Objective To research the microbial population distribution on the outer surface and in the contents of SPF chicken embryos, so as to standardize the production of SPF chicken embryos, guide the selection of disinfectants for vaccine production, and improve the sterile guarantee level of vaccine production.Methods The cotton swab method was used to isolate the external surface microorganisms of three batches of SPF chicken embryos before and after treatment with 70%isopropanol. The microorganisms of SPF chicken embryo contents were isolated with tryptose soya agar(TSA) medium from the urine sac fluid of three batches of SPF chicken embryos that died after 3 d inoculation with attenuated yellow fever 17D strain. The colony was purified using the three-zone streaking method, and the microorganisms were identified using 16s rRNA/ITS sequencing.Results After disinfection with 70% isopropanol, the number of microorganisms detected on the outer surface of SPF chicken embryos significantly decreased(t = 2. 67, P = 0. 011). The microorganisms on the outer surface of SPF chicken embryos were mainly Bacillus, with Bacillus genus accounting for 52. 6% and other types of bacillus accounting for36. 8%. There were significant differences in the detection rates of microorganisms in the contents of three batches of SPF chicken embryos. Proteobacteria, Enterococcus and Escherichia were detected the most frequently in the contents of SPF chicken embryos, and their detection numbers were significantly higher than other microorganisms.Conclusion The dominant microorganisms on the outer surface of SPF chicken embryos are Bacillus, and the dominant microorganisms in the contents are Proteobacteria, Enterococcus and Escherichia. The microbial population distribution provides basis for the selection of disinfectants for vaccine production, suggesting that the control measures should be carried out at multiple stages such as SPF chicken embryos production, storage, transportation, and selection to improve the sterile assurance level of vaccine production.
5.Association of short-term exposure to polycyclic aromatic hydrocarbons in ambient fine particulate matter with resident mortality: a case-crossover study
Sirong WANG ; Zhi LI ; Yanmei CAI ; Chunming HE ; Huijing LI ; Yi ZHENG ; Lu LUO ; Ruijun XU ; Yuewei LIU ; Huoqiang XIE ; Qinqin JIANG
Journal of Public Health and Preventive Medicine 2025;36(6):6-11
Objective To quantitatively assess the association of short-term exposure to polycyclic aromatic hydrocarbons (PAHs) in ambient fine particulate matter (PM2.5) with residents mortality. Methods A time-stratified case-crossover study was conducted from 2020 to 2022 among 10606 non-accidental residents by using the Guangzhou Cause of Death Surveillance System in Conghua District, Guangzhou. Exposure levels of PAHs in PM2.5 and meteorological data during the study period were obtained from the Center for Disease Control and Prevention in Conghua District and the China Meteorological Administration Land Data Assimilation System (CLDAS-V2.0), respectively. Conditional Poisson regression model was used to estimate the exposure-response association between PAHs and the mortality risk. Results Fluoranthene, chrysene, benzo[k]fluoranthene, benzo[a]pyrene, and indeno[1,2,3-cd]pyrene were significantly associated with an increased risk of mortality. For every one interquartile range increase in exposure levels, the non-accidental mortality risks increased by 8.33% (95% CI: 1.80%, 15.27%), 4.67% (95% CI: 1.86%, 7.57%), 6.07% (95% CI: 2.08%, 10.21%), 4.62% (95% CI: 1.85%, 7.47%), and 4.70% (95% CI: 0.53%, 9.03%), respectively. The estimated non accidental deaths attributable to exposure to fluoranthene, chrysene, benzo[k]fluorine, benzo[a]pyrene and indine[1,2,3-cd]pyrene were 5.91%, 6.08%, 6.51%, 6.46%, and 4.21%, respectively. Conclusions Short-term exposure to PAHs in PM2.5, including fluoranthene, chrysene, benzo[k]fluoranthene, benzo[a]pyrene and indine[1,2,3-cd]pyrene, was significantly associated with an increased risk of mortality among residents.
6.Interpretation of the group standard of " Humanistic Caring Management Standards for Patients in the Operating Room"
Ruiying YU ; Xinyue MIAO ; Qingmin ZHANG ; Yilan LIU ; Shujie GUO ; Huiling LI ; Guo CHEN ; Chunlan ZHOU ; Ting LIU ; Shuhua DENG ; Hongzhen XIE ; Yu CHENG ; Yinglan LI ; Yanlan MA ; Xia XIN ; Yanjin LIU ; Yongyi CHEN ; Gendi LU ; Xiaoqin GAN ; Feng XU ; Zuwei XIA ; Li HE ; Qinqin CHEN ; Fukang ZHANG ; Songmei WU ; Yi LI ; Wenjuan ZHOU
Chinese Journal of Hospital Administration 2025;41(7):512-517
Humanistic caring for patients in the operating room refers to providing the whole process of caring medical services for patients in the operating room. In order to standardize humanistic caring services for patients in the operating room of medical institutions, improve the comprehensive service level of the operating room, and enhance the surgical experience of patients, the Chinese Association for Life Care released the group standard " Humanistic Caring Management Standards for Patients in the Operating Room" in December 2023. This article interpreted the basic requirements for humanistic caring of patients in the operating room, the environment and facilities for humanistic caring, the procedures and measures for humanistic caring, and the quality management framework, aiming to assist administrators and clinical practitioners across various levels of medical institutions in accurately understanding and effectively implementing the standard, and to provide essential textual reference and practical guidance for promoting the application of the standard.
7.Construction of a new predictive score for severe fever with thrombocytopenia syndrome combined with bacterial/fungal infections based on clinical data
Ran WANG ; Yan DAI ; Qinqin PU ; Nannan HU ; Ke JIN ; Jun LI
Chinese Journal of Infectious Diseases 2025;43(4):202-209
Objective:To study the risk factors for combined bacterial/fungal infections in patients with severe fever with thrombocytopenia syndrome (SFTS) and to develop a novel and validated prediction model.Methods:The basic data and the results of the first laboratory examination after admission were retrospectively collected from patients diagnosed with SFTS who were hospitalized in the First Affiliated Hospital, Nanjing Medical University from January 2018 to December 2022. The patients were categorized into co-infected and non-co-infected groups according to whether they had co-infections with bacterial/fungal infections or not.Independent risk factors were screened by multivariate logistic regression analyses. A novel prediction model was constructed, and the predictive value of the model was assessed using receiver operating characteristic curve. Non-parametric tests and chi-square test were used for statistical analysis.Results:A total of 294 patients were included, and 62 cases were in the combined infection group including 39 cases of simple respiratory tract infections, 11 cases of simple bloodstream infections, four cases of simple urinary tract infections, four cases of respiratory tract combined with bloodstream infection, and four cases of respiratory tract combined with urinary tract infection. Acinetobacter baumannii was mostly found in bacterial infections, with a total of 19 strains, followed by Escherichia coli and Pseudomonas aeruginosa, both with seven strains. Aspergillus were mostly common in fungi, with a total of 16 strains which were all collected from patients with pulmonary infections. Compared with the non-co-infected group, patients in the co-infected group had longer hospital stays, with statistically significant differences ( Z=-6.18, P<0.001). The patients also had higher frequencies of bleeding symptoms, neurological symptoms, severe illness, and death, with statistically significant differences ( χ2=23.91, 16.37, 15.51 and 15.58, respectively, all P<0.001). The aspartate transaminase-to-platelet ratio index (APRI) was also higher in patients with coinfection, with a statistically significant difference ( Z=-4.64, P<0.001). Multivariate binary logistic regression showed that severe illness (odds ratio ( OR)=2.567, 95% confidence interval ( CI) 1.344 to 4.904, P=0.004), blood glucose level higher than 7.782 mmol/L ( OR=4.766, 95% CI 2.493 to 9.109, P<0.001), procalcitonin level higher than 0.228 μg/L ( OR=2.487, 95% CI 1.289 to 4.799, P=0.007), and APRI value higher than 6.268 ( OR=3.032, 95% CI 1.404 to 6.548, P=0.005) were the independent risk factors for co-infections in SFTS patients. Disease severity, blood glucose, procalcitonin, and APRI were combined to construct a novel predictive model: Infect-risk score=-3.331+ 0.654×severity (severe=1, non-severe=0)+ 0.160×blood glucose+ 0.066×procalcitonin+ 0.013×APRI. The AUC for this score was 0.764 (95% CI 0.698 to 0.830, P<0.001), with Youden index of 0.416, sensitivity of 0.839, and specificity of 0.578. Conclusions:Severe illness, blood glucose levels higher than 7.782 mmol/L, procalcitonin levels above 0.228 μg/L, and APRI values above 6.268 are independent risk factors for bacterial/fungal coinfection in SFTS patients. The constructed Infect-risk score model has good predictive value for bacterial/fungal coinfection in SFTS patients.
8.Research on the current status and influencing factors of post ICU syndrome in severe patients
Tingshu WANG ; Li YAO ; Yan LIU ; Bei JING ; Qinqin LI ; Tingrui WANG
Chinese Journal of Nursing 2025;60(19):2356-2363
Objective To explore the current situation and influencing factors of post-intensive care syndrome(PICS)in critically ill patients a month after their transfer out,and to provide a basis for formulating individualized intervention measures.Methods By the convenience sampling method,550 patients who were transferred out of the ICU of a tertiary A hospital in Guizhou Province for a month from April to November 2024 were selected as the survey subjects.The general information of the patients was collected and sorted out.During their stay in the ICU,the Barthel Index,the Richards Campbell Sleep Scale,and the Perceived Social Support Scale were employed for evaluation.A month after the patients were transferred out of the ICU,the assessment was completed through telephone follow-up using the Chinese version of the Brain Care Monitoring Questionnaire for Healthy Aging.According to the assessment a month after the patients were transferred out,they were divided into a PICS group and a non-PICS group.Univariate analysis and binary Logistic regression analysis were completed based on the data of the 2 groups.Results A total of 550 questionnaires were distributed,and 442 valid questionnaires were ultimately collected,with a valid questionnaire collection rate of 80.3%.Among them,194 cases(43.9%)were high-risk patients with PICS.Binary logistic regression analysis showed that advanced age,pulmonary infection,total hospitalization time>15 days,and internal environment disorder were the risk factors for PICS in ICU patients(P<0.05).Conclusion The risk of PICS is relatively high in critically ill patients a month after the transfer.Medical staff need to pay attention to patients who are older,have pulmonary infections and internal environment disorders,and formulate personalized intervention measures for different disease types to improve the long-term health prognosis of patients.
9.A systematic review of quality assessment tools for pediatric palliative care based on COSMIN guidelines
Sishan JIANG ; Qinqin CHENG ; Tingwei LUO ; Na ZHANG ; Junchen GUO ; Dongya LI ; Dandan LI ; Lihui ZHU
Chinese Journal of Nursing 2025;60(5):611-618
Objective To evaluate the methodological quality and measurement attribute quality of the evaluation tool for pediatric palliative care quality assessment tools,and to provide references for medical staff to select the best assessment tools.Methods The PubMed,Embase,Cochrane Library,Web of Science,CINAHL,Scopus,China National Knowledge Infrastructure(CNKI),Wanfang Database,VIP Database,Chinese Biomedical Literature Database,GIN,NGC,NICE,NRAO,medlive,WHO,AAHPM,WHPCA,APHN were searched from inception to March 28,2024.Data were screened and extracted independently by 2 researchers.The consensus-based standards for the selection of health measurement instruments(COSMIN)checklist and quality criteria were employed to evaluate the methodological quality and psychometric properties of the included pediatric palliative care quality assessment tools.Finally,recommendations were formulated based on these evaluations.Results A total of 13 articles were included,involving 9 pediatric palliative care quality assessment tools.Among them,the PICU-QODD,PaPEQu and QCPCI demonstrated good content validity and internal consistency,and are recommended as Grade A.The remaining assessment tools are recommended as Grade B or C.Conclusion The PICU-QODD,PaPEQu and QCPCI are recommended for use,but further validation of their psychometric properties is still needed.
10.Application of 18F-FDG PET/MR and its derived parameters in the diagnosis and staging of bladder cancer
Qinqin YOU ; Fei YU ; Rushuai LI ; Fengjiao YANG ; Shuyue AI ; Jun TANG ; Feng WANG
Chinese Journal of Nuclear Medicine and Molecular Imaging 2025;45(7):405-410
Objective:To investigate the application of 18F-FDG PET/MR and its derived parameters in the diagnosis and staging of bladder cancer. Methods:Forty patients (32 males, 8 females; age (66.8±11.2) years) with suspected bladder cancer between December 2019 and March 2022 were retrospectively included and underwent 18F-FDG PET/MR in Nanjing First Hospital. Parameters including SUV max, SUV mean, maximum tumor diameter and mean of apparent diffusion coefficient (ADC mean) were obtained, and bladder cancer muscle invasiveness and lymph node involvement were determined. The efficacy of 18F-FDG PET/MR and its derived parameters for tumor diagnosis and staging was analyzed using transurethral resection of bladder tumor (TUR-BT) or radical cystectomy (RC) and extended pelvic lymph node dissection (ePLND) histopathology as the " gold standard". Independent-sample t test, Mann-Whitney U test or χ2 test was used to analyze the data, and Delong test was used to compare different AUCs. Results:Of 40 patients, 8 were non-muscle invasive bladder cancer (NMIBC), 32 were muscle invasive bladder cancer (MIBC), and 5 were pathologically confirmed to have lymph node metastasis. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of 18F-FDG PET/MR for identifying MIBC were 96.9%(31/32), 7/8, 96.9%(31/32), 7/8, 95.0%(38/40), respectively, and those for lymph node metastasis were 4/5, 90.0%(18/20), 4/6, 18/19, 88.0%(22/25), respectively. For pathological tumor (pT) staging, significant differences were observed between pT2-3 and pT1 groups in maximum tumor diameter ( t=-2.37, P=0.024), SUV mean( Z=-2.11, P=0.035), and ADC mean( t=2.91, P=0.006). The AUCs of maximum tumor diameter, SUV mean and ADC mean in distinguishing MIBC were 0.781, 0.746, and 0.825, respectively. The sensitivity, specificity, PPV, NPV, and accuracy of MRI alone in identifying MIBC were 87.5%(28/32), 1/8, 80.0%(28/35), 1/5 and 72.5%(29/40), respectively, with the AUC of 0.500. The AUC of 18F-FDG PET/MR in identifying MIBC was 0.796, which was better than MRI alone ( Z=5.54, P<0.001), and the accuracy of PET/MR was also higher than MRI alone ( χ2=7.44, P=0.006). Conclusion:Compared with MRI alone, 18F-FDG PET/MR significantly improves the diagnostic efficacy of bladder cancer and the accuracy of pT staging.


Result Analysis
Print
Save
E-mail