OBJECTIVE To investigate a dynamic monitoring of drug consumption （DMDC） model based on the seasonal Mann-Kendall trend test， aiming to provide scientific evidence for the efficient and macroscopic monitoring of drug use. METHODS A monitoring list of key outpatient drugs was established based on the top 20% of drugs ranked by sales volume in the outpatient pharmacy in October 2024. A DMDC model based on the Mann-Kendall trend test was constructed using the monthly usage data of key outpatient drugs from November 2021 to October 2024， aiming to eliminate the impact of seasonal fluctuations and analyze the temporal trends in drug consumption. Taking mucolytic expectorants， triazole derivatives for dermatophytosis， and single-agent hydroxymethylglutaryl coenzyme A （HMG-CoA） reductase inhibitors as examples， the monitoring effectiveness of the DMDC model was demonstrated， and its performance was compared with that achieved by the traditional sequential growth rate ranking method. RESULTS A total of 215 drug varieties were included in the monitoring list， and DMDC models were successfully established for all of them. Among these， 119 showed a significant increasing trend （P＜0.05， S′＞0）. The model successfully monitored the monthly consumption of mucolytic expectorants， triazole derivatives for dermatophytosis， and single- agent HMG-CoA reductase inhibitors. The precision and recall rates of the DMDC model for identifying abnormal drug use were 60.7% and 85.0%， respectively， both significantly higher than those of the sequential growth rate ranking method （8.3% and 15.0%， respectively） （χ2＝20.114， P＜0.001； χ2＝19.600， P＜0.001）. CONCLUSIONS DMDC model based on the seasonal Mann-Kendall trend test can effectively identify long-term trends in drug consumption， eliminate seasonal interference， enhance monitoring accuracy and management efficiency， and is suitable for the dynamic monitoring of drug consumption.

OBJECTIVE To provide scientific evidence and practical references for refined antimicrobial stewardship in multi- branch medical institutions. METHODS Data of antibiotic use density （AUD） of inpatients as well as physician-level case mix index （CMI） were collected from the Liji Road main campus and Panlongcheng branch campus of our hospital from August 2023 to March 2026. The two-factor decomposition method was adopted to decompose total AUD variation into a level effect and structural effect. Grey relational analysis （GRA） was performed to quantify the correlation degree between ward CMI， physician CMI and physician AUD， so as to identify wards and physicians requiring targeted key intervention in different campuses. A refined antimicrobial management framework for multi-branch hospitals was constructed and implemented based on the above analytical results. An interrupted time series （ITS） model incorporating seasonal dummy variables was applied. The research period was divided into pre-intervention stage （August 2023 to March 2024） and post-intervention stage （April 2024 to March 2026）. The temporal changing trends of AUD in two campuses were compared to evaluate management efficacy. RESULTS Driving factors for AUD variation presented significant heterogeneity between the two campuses. AUD variation in the main campus was dominated by level effect， while AUD variation in the branch campus was jointly affected by structural effect and level effect. The grey relational degrees of ward CMI and physician CMI with physician AUD were 0.901 and 0.882， respectively. After refined management implementation， AUD decreased by 7.48 DDDs/（100 bed·days） and 20.54 DDDs/（100 bed·days） in the main campus and branch campus， respectively； the rising trends of AUD in both campuses reversed to declining trends after intervention. CONCLUSIONS Driving factors of AUD variation under the multi-branch hospital model show obvious inter-campus heterogeneity. CMI is highly correlated with AUD. The refined management system developed via AUD attribution decomposition and CMI correlation analysis matches the differentiated clinical characteristics of multi- branch hospitals and effectively improves the precision of antimicrobial stewardship.

ObjectiveTo construct a risk prediction model for carotid artery plaques in high-risk populations of stroke based on five machine learning methods. MethodsThe clinical information of the high-risk population of stroke was collected. Factor analysis and statistical analysis of syndrome elements were conducted on their traditional Chinese medicine （TCM） symptoms and tongue and pulse manifestations. On the basis of the results of factor analysis and variable screening， five machine learning methods-classification and regression tree （CART） decision tree， support vector machine （SVM）， back propagation（BP） neural network， logistic regression， and random forest-were used to construct the risk prediction model for carotid artery plaques. ResultsThe most common TCM syndrome elements in the high-risk population of stroke was Qi deficiency. The scores of Qi deficiency， Yin deficiency， and Yang deficiency in the population with carotid artery plaques were higher than those without carotid artery plaques （P<0.05）. The CART decision tree， SVM， logistic regression， BP neural network， and random forest models showed the areas under the receiver operating characteristic （ROC） curves of 0.71， 0.75， 0.76， 0.76， and 0.75， the accuracy rates of 68.94%， 69.27%， 69.44%， 70.10%， and 69.60%， the precision rates of 68.56%， 68.53%， 68.75%， 69.74%， and 69.42%， the recall rates of 68.91%， 67.93%， 67.92%， 68.10%， and 67.31%， and the F₁ values of 0.69， 0.68， 0.68， 0.68， and 0.68， respectively. ConclusionAmong the high-risk population of stroke， the most frequently distributed TCM syndrome element is Qi deficiency， with the rest mainly being fire heat， Yin deficiency， Yang deficiency， phlegm dampness， blood stasis， and Qi stagnation. Deficiency syndrome may be a major factor leading to carotid artery plaques in the high-risk population of stroke. The BP neural network model demonstrates better performance in predicting the risk of carotid artery plaques in the high-risk population of stroke. People with carotid artery plaques are more likely to present with symptoms such as a heavy head， dizziness， headache， and thready pulse. The primary community benefits more widely when the BP neural network model is adopted to predict the risk of carotid artery plaques in the high-risk population of stroke over 40 years old.

ObjectiveTo construct a risk prediction model for carotid artery plaques in high-risk populations of stroke based on five machine learning methods. MethodsThe clinical information of the high-risk population of stroke was collected. Factor analysis and statistical analysis of syndrome elements were conducted on their traditional Chinese medicine （TCM） symptoms and tongue and pulse manifestations. On the basis of the results of factor analysis and variable screening， five machine learning methods-classification and regression tree （CART） decision tree， support vector machine （SVM）， back propagation（BP） neural network， logistic regression， and random forest-were used to construct the risk prediction model for carotid artery plaques. ResultsThe most common TCM syndrome elements in the high-risk population of stroke was Qi deficiency. The scores of Qi deficiency， Yin deficiency， and Yang deficiency in the population with carotid artery plaques were higher than those without carotid artery plaques （P<0.05）. The CART decision tree， SVM， logistic regression， BP neural network， and random forest models showed the areas under the receiver operating characteristic （ROC） curves of 0.71， 0.75， 0.76， 0.76， and 0.75， the accuracy rates of 68.94%， 69.27%， 69.44%， 70.10%， and 69.60%， the precision rates of 68.56%， 68.53%， 68.75%， 69.74%， and 69.42%， the recall rates of 68.91%， 67.93%， 67.92%， 68.10%， and 67.31%， and the F₁ values of 0.69， 0.68， 0.68， 0.68， and 0.68， respectively. ConclusionAmong the high-risk population of stroke， the most frequently distributed TCM syndrome element is Qi deficiency， with the rest mainly being fire heat， Yin deficiency， Yang deficiency， phlegm dampness， blood stasis， and Qi stagnation. Deficiency syndrome may be a major factor leading to carotid artery plaques in the high-risk population of stroke. The BP neural network model demonstrates better performance in predicting the risk of carotid artery plaques in the high-risk population of stroke. People with carotid artery plaques are more likely to present with symptoms such as a heavy head， dizziness， headache， and thready pulse. The primary community benefits more widely when the BP neural network model is adopted to predict the risk of carotid artery plaques in the high-risk population of stroke over 40 years old.

Objective To measure radiation filed distribution in the treatment room of the Varian Halcyon medical linear accelerator, and to provide a basis for shielding design and potential exposure analysis of treatment rooms for this type of accelerator. Methods Under the 6 MV X-ray (FFF) mode at a maximum dose rate of 800 MU/min and a maximum irradiation field of 28.00 cm × 28.00 cm, a total of 540 MU was delivered during gantry rotation. Radiation field distribution was measured using thermoluminescence dosimeters located at multiple points in the room. The measured data were then applied to shielding calculations, and the results were compared with those obtained using empirical formulas. Results The overall radiation levels in the treatment room were in the range of 12.2 µGy/540 MU to 5.520 Gy/540 MU, with the highest dose (5.520 Gy/540 MU) observed at the isocenter, and the lowest dose (12.2 µGy/540 MU) recorded at approximately 6.5 m from the gantry head. The radiation levels at most points were within the range of 100-1000 µGy/540 MU. At heights of 0.5, 1.1, and 1.7 m, the radiation field exhibited a rapid attenuation from the beam center outward, with a faster decay along the treatment couch direction (Y-axis) than along the perpendicular direction (X-axis). Shielding calculations based on the measured radiation field yielded required wall thicknesses of 1219 mm (east wall), 949 mm (south wall), 1235 mm (west wall), and 1252 mm (north wall), all of which were smaller than those calculated using empirical formulas. Conclusion Thermoluminescence dosimeter is suitable for multi-point in situ measurement of radiation field distribution in Halcyon accelerator treatment rooms. The measurement results can be used to optimize the shielding design of Halcyon accelerator treatment rooms.

Objective To explore the CT diagnosis and differential diagnosis of perivascular epithelioid cell neoplasms (PEComa), improve the accuracy of PEComa diagnosis, and reduce misdiagnosis. Methods CT findings of 8 cases of PEComa confirmed by pathology in Jining First People’s Hospital from January 2020 to April 2024 were retrospectively analyzed for the location, shape, size, boundary, plain scan density, and enhancement characteristics of the lesions. Results All 8 tumors were solitary, with 5 located in the kidney, 1 in the liver, 1 in the extraperitoneal space, and 1 in the retroperitoneal space. The tumors were round in 3 cases, oval in 1 case, and irregular in 4 cases. Seven cases were benign with clear boundaries and 1 case was malignant with unclear boundaries. On plain CT, 2 cases showed slightly low density, 3 cases showed slightly high density, and 3 cases showed low density. One tumor had uniform density, and 7 tumors had nonuniform density with internal necrosis and cystic changes. Contrast-enhanced CT revealed diverse enhancement patterns. Four cases showed a “fast in and fast out” enhancement pattern, with significant arterial-phase enhancement and reduced portal venous-phase enhancement. Three cases showed a “fast-in and slow-out” enhancement pattern, with significant enhancement in the arterial phase, persistent enhancement in the portal venous phase, and slightly reduced density in the delayed phase. One case showed mild enhancement in the arterial phase and significant enhancement in the portal venous phase. In 3 cases, multiple tortuous and thickened blood vessels were observed around the tumors, while 3 cases showed tortuous vascular shadows within the tumors. Conclusion PEComa demonstrates characteristic CT features, predominantly with “fast in and fast out” or “fast in and slow out” enhancement patterns. When thickened and tortuous blood vessels are observed within or around the tumor, PEComa should be considered in combination with clinical findings.

Objective:To investigate the short-term efficacy, safety and factors affecting recurrence of intermediate-risk and high-risk non-muscle invasive bladder cancer (NMIBC) treated with intravesical instillation of domestic Bacillus Calmette-Guérin (BCG) infusion.Methods:This study was a retrospective cohort study. We collected the data of 163 patients with NMIBC treated with domestic BCG after transurethral resection of bladder tumor (TURBT) from October 2016 to October 2020 in the Department of Urology, the Affiliated Hospital of Qingdao University. There were 140 males and 23 females, the age was(67.3±10.3) years old. 23 cases had only been received BCG and 140 cases received other chemotherapy drugs before. The induction scheme of instillation was started after the TURBT at once a week for 6 consecutive weeks, continue instillation at every two weeks for 3 doses, then maintenance instillation once a month for 10 consecutive times, for a total of 19 instillations. Kaplan-Meier analysis was used to calculate recurrence-free survival. Binary Logistic regression was used to analyze factors and stepwise regression (backward method) was employed to identify independent risk factors for recurrence after BCG instillation. The incidence of adverse reactions was recorded. Measurement data with normal distribution were expressed as mean±standard deviation( ± s), measurement data with skewed distribution were expressed as M( Q1, Q3), and count data were expressed as frequency and percentage(%). Results:A total of 23 cases experienced recurrence within the 13-month of instillation, of which 7 cases were found to have progressed by pathological biopsy, and the cumulative recurrence-free rate was 85.9%. The results of binary logistic regression analysis showed that the history of re-TURBT ( P=0.010) was an independent predictor for recurrence after BCG intravesical instillation. Adverse events occurred in 110 cases. The main symptoms of the 65 cases were urgency of urination, pollakiuria and dysuria with urinary irritation, urinary tract infection in 15 cases, hematuria in 10 cases, and other symptoms in 20 cases. Instillation was terminated in 7 cases due to side effects, no serious adverse events such as spread of tuberculous bacteria were observed in the cases. Conclusions:Patients with NMIBC treated with intravesical instillation of domestic BCG have significant short-term efficacy, while patients who have previously received re-TURBT is an independent predictor for recurrence. The domestic BCG have the characteristic of slight side effect, good efficacy and safety.

Cardiotoxicity is a critical issue in drug development that poses serious health risks,including potentially fatal arrhythmias.The human ether-à-go-go related gene(hERG)potassium channel,as one of the pri-mary targets of cardiotoxicity,has garnered widespread attention.Traditional cardiotoxicity testing methods are expensive and time-consuming,making computational virtual screening a suitable alter-native.In this study,we employed machine learning techniques utilizing molecular fingerprints and descriptors to predict the cardiotoxicity of compounds,with the aim of improving prediction accuracy and efficiency.We used four types of molecular fingerprints and descriptors combined with machine learning and deep learning algorithms,including Gaussian naive Bayes(NB),random forest(RF),support vector machine(SVM),K-nearest neighbors(KNN),eXtreme gradient boosting(XGBoost),and Trans-former models,to build predictive models.Our models demonstrated advanced predictive performance.The best machine learning model,XGBoost Morgan,achieved an accuracy(ACC)value of 0.84,and the deep learning model,Transformer_Morgan,achieved the best ACC value of 0.85,showing a high ability to distinguish between toxic and non-toxic compounds.On an external independent validation set,it achieved the best area under the curve(AUC)value of 0.93,surpassing ADMETlab3.0,Cardpred,and CardioDPi.In addition,we explored the integration of molecular descriptors and fingerprints to enhance model performance and found that ensemble methods,such as voting and stacking,provided slight improvements in model stability.Furthermore,the SHapley Additive exPlanations(SHAP)explanations revealed the relationship between benzene rings,fluorine-containing groups,NH groups,oxygen in ether groups,and cardiotoxicity,highlighting the importance of these features.This study not only improved the predictive accuracy of cardiotoxicity models but also promoted a more reliable and scientifically interpretable method for drug safety assessment.Using computational methods,this study facilitates a more efficient drug development process,reduces costs,and improves the safety of new drug candidates,ultimately benefiting medical and public health.

Objective:To develop a blockchain-based clinical configuration management model for medical equipment and evaluate its application value in optimizing clinical configuration management of hospital emergency and life-support devices.Methods:A data traceability management model was implemented.The Spatial Durbin Model(SDM)was used to identify issues in equipment configuration,and a blockchain-enabled review framework was established for procurement management of emergency and life-support devices.From January 2019 to December 2022,57 emergency and life-support devices deployed in Shaanxi Provincial People's Hospital were retrospectively analyzed.Among them,26 devices(January 2019-December 2020)were managed via conventional expert evaluation,while 31 devices(January 2021-December 2022)were managed using blockchain-based review.In the conventional mode,four traceability parameters-procurement declaration,supplier qualification,transaction records,and after-sales support-were randomly sampled 128,85,119,and 100 times,respectively;in the blockchain mode,these parameters were sampled 145,94,124,and 105 times.Procurement process compliance was evaluated across device categories:emergency,monitoring,therapeutic,and others required 25,40,30,and 35 review steps(conventional mode)versus 30,45,45,and 35 steps(blockchain mode).Comparative metrics included data traceability rates,process compliance rates,and procurement performance target achievement rates.Results:The blockchain mode demonstrated superior traceability rates:92.41%(134/145)for declarations,100.00%(94/94)for suppliers,97.58%(121/124)for transactions,and 97.14%(102/105)for after-sales support-all significantly higher than the conventional mode(x2=5.898,4.525,9.185,8.362,P＜0.05).Process compliance rates reached 100.00%(30/30)for emergency devices,95.56%(43/45)for monitoring devices,97.78%(44/45)for therapeutic devices,and 97.14%(34/35)for others,with statistically significant improvements(x2=5.176,4.936,5.103,3.968,P＜0.05).Procurement performance targets for progress,benefit,quality,and satisfaction were achieved at 96.77%(30/31),100.00%(31/31),100.00%(31/31),and 93.55%(29/31),respectively,surpassing the conventional mode(x2=6.581,6.535,5.129,5.780,P＜0.05).Conclusion:The blockchain-based clinical configuration management model enhances data traceability,standardizes procurement workflows,and improves performance goal attainment in hospital emergency and life-support device deployment.

Poly(ADP-ribosyl)ation (PARylation) is a specific form of post-translational modification (PTM) predominantly triggered by the activation of poly-ADP-ribose polymerase 1 (PARP1). However, the role and mechanism of PARylation in the advancement of acute kidney injury (AKI) remain undetermined. Here, we demonstrated the significant upregulation of PARP1 and its associated PARylation in murine models of AKI, consistent with renal biopsy findings in patients with AKI. This elevation in PARP1 expression might be attributed to trimethylation of histone H3 lysine 4 (H3K4me3). Furthermore, a reduction in PARylation levels mitigated renal dysfunction in the AKI mouse models. Mechanistically, liquid chromatography-mass spectrometry indicated that PARylation mainly occurred in receptor for activated C kinase 1 (RACK1), thereby facilitating its subsequent phosphorylation. Moreover, the phosphorylation of RACK1 enhanced its dimerization and accelerated the ubiquitination-mediated hypoxia inducible factor-1α (HIF-1α) degradation, thereby exacerbating kidney injury. Additionally, we identified a PARP1 proteolysis-targeting chimera (PROTAC), A19, as a PARP1 degrader that demonstrated superior protective effects against renal injury compared with PJ34, a previously identified PARP1 inhibitor. Collectively, both genetic and drug-based inhibition of PARylation mitigated kidney injury, indicating that the PARylated RACK1/HIF-1α axis could be a promising therapeutic target for AKI treatment.