Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

In this study, lipopolysaccharide(LPS), ovalbumin(OVA), and compound 48/80(C48/80) were administered to establish non-infectious pneumonia models under simulated clinical conditions, and the correlation between their pathological characteristics and traditional Chinese medicine(TCM) syndromes was compared, providing the basis for the selection of appropriate animal models for TCM efficacy evaluation. An acute pneumonia model was established by nasal instillation of LPS combined with intraperitoneal injection for intensive stimulation. Three doses of OVA mixed with aluminum hydroxide adjuvant were injected intraperitoneally on days one, three, and five and OVA was administered via endotracheal drip for excitation on days 14-18 to establish an OVA-induced allergic pneumonia model. A single intravenous injection of three doses of C48/80 was adopted to establish a C48/80-induced pneumonia model. By detecting the changes in peripheral blood leukocyte classification, lung tissue and plasma cytokines, immunoglobulins(Ig), histamine levels, and arachidonic acid metabolites, the multi-dimensional analysis was carried out based on pathological evaluation. The results showed that the three models could cause pulmonary edema, increased wet weight in the lung, and obvious exudative inflammation in lung tissue pathology, especially for LPS. A number of pyrogenic cytokines, inclading interleukin(IL)-6, interferon(IFN)-γ, IL-1β, and IL-4 were significantly elevated in the LPS pneumonia model. Significantly increased levels of prostacyclin analogs such as prostaglandin E2(PGE2) and PGD2, which cause increased vascular permeability, and neutrophils in peripheral blood were significantly elevated. The model could partly reflect the clinical characteristics of phlegm heat accumulating in the lung or dampness toxin obstructing the lung. The OVA model showed that the sensitization mediators IgE and leukotriene E4(LTE4) were increased, and the anti-inflammatory prostacyclin 6-keto-PGF2α was decreased. Immune cells(lymphocytes and monocytes) were decreased, and inflammatory cells(neutrophils and basophils) were increased, reflecting the characteristics of "deficiency", "phlegm", or "dampness". Lymphocytes, monocytes, and basophils were significantly increased in the C48/80 model. The phenotype of the model was that the content of histamine, a large number of prostacyclins(6-keto-PGE1, PGF2α, 15-keto-PGF2α, 6-keto-PGF1α, 13,14-D-15-keto-PGE2, PGD2, PGE2, and PGH2), LTE4, and 5-hydroxyeicosatetraenoic acid(5S-HETE) was significantly increased, and these indicators were associated with vascular expansion and increased vascular permeability. The pyrogenic inflammatory cytokines were not increased. The C48/80 model reflected the characteristics of cold and damp accumulation. In the study, three non-infectious pneumonia models were constructed. The LPS model exhibited neutrophil infiltration and elevated inflammatory factors, which was suitable for the efficacy study of TCM for clearing heat, detoxifying, removing dampness, and eliminating phlegm. The OVA model, which took allergic inflammation as an index, was suitable for the efficacy study of Yiqi Gubiao formulas. The C48/80 model exhibited increased vasoactive substances(histamine, PGs, and LTE4), which was suitable for the efficacy study and evaluation of TCM for warming the lung, dispersing cold, drying dampness, and resolving phlegm. The study provides a theoretical basis for model selection for the efficacy evaluation of TCM in the treatment of pneumonia.
Animals ; Disease Models, Animal ; Mice ; Pneumonia/genetics* ; Medicine, Chinese Traditional ; Male ; Humans ; Cytokines/immunology* ; Female ; Lipopolysaccharides/adverse effects* ; Lung/drug effects* ; Drugs, Chinese Herbal ; Ovalbumin ; Mice, Inbred BALB C

OBJECTIVE: To analyze the incidence of abnormal hemoglobin (Hb) in neonates in Guangzhou area, as well as the results of quantitative analysis of Hb in neonatal umbilical cord blood and genetic diagnosis of thalassemia in neonates with abnormal Hb; And to explore the hematological phenotypes and clinical characteristics of neonates with abnormal Hb and their parents, providing a reference for eugenics and childcare. METHODS: 650 neonates born at Guangdong Provincial People's Hospital who underwent Hb electrophoresis were included in this study. The results of routine blood test of umbilical cord blood , Hb electrophoresis and α-, β-thalassemia gene detection of the neonates were collected. The genotype distribution of thalassemia in the neonates was analyzed. Additionally, the abnormal Hb content of α and β variants was studied. Furthermore, the differences in hematological parameters between abnormal Hb neonates and normal neonates and α-thalassemia neonates, as well as between the parents of abnormal Hb neonates and normal adults were compared. RESULTS: Among the 650 neonates, 332 (51.08%) were diagnosed with thalassemia, including 235 cases of α-thalassemia (36.15%), 79 cases of β-thalassemia (12.15%), and 18 cases of compound αβ-thalassemia (2.77%). Among all the α-thalassemia genotypes, the most prevalent one was -- ^SEA/αα (48.94%), followed by -α^3.7/αα (20.00%), -α^4.2/αα (11.06%), and αα^CS/αα (8.94%). The four most common genotypes of β-thalassemia were β^CD41-42 (32.91%), β^IVS-Ⅱ-654 (26.58%), β^-28 (21.52%), and β^E (10.13%), respectively. 275 cases of abnormal bands were found in Hb electrophoresis of umbilical cord blood, with a detection rate of 42.31%. The abnormal Hb content of α-variant in the neonates was significantly higher than that of β-variant (P < 0.001). The levels of Hb, MCV, MCH, Hb A, and Hb F in neonates with abnormal Hb were lower than those in normal neonates, while the RDW-CV was higher than that in normal neonates, with statistical significantce (P < 0.05). The levels of RBC and Hb A in neonates with abnormal Hb were lower than those in neonates with α-thalassemia, while the level of MCH was higher than that in neonats with α-thalassemia, with statistical significance (P < 0.05). The levels of Hb, MCV, MCH, and Hb A in parents of neonates with abnormal Hb were lower than those in normal adults, while the RDW-CV was higher than that in normal adults, and the differences were statistically significant (P < 0.05). CONCLUSION The abnormal Hb content of α-variant in the neonates is significantly higher than that of β-variant in the neonates in Guangzhou, which can help to presume whether it is α chain or β chain based on the abnormal Hb content, providing a reference for globin gene sequencing. Meanwhile, analysis of various hematological screening-related indicators in neonates in the early stage is beneficial for early warning of the occurrence of abnormal Hb combined with thalassemia, reducing missed diagnoses to a certain extent.
Humans ; Infant, Newborn ; Genotype ; Hemoglobins, Abnormal/genetics* ; China/epidemiology* ; alpha-Thalassemia/epidemiology* ; beta-Thalassemia/genetics* ; Parents ; Female ; Male ; Fetal Blood

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

BACKGROUND: The American Heart Association (AHA) introduced the concept of cardiovascular-kidney-metabolic (CKM) health and stage, reflecting the interaction among metabolism, chronic kidney disease (CKD), and the cardiovascular system. However, the association between CKM stage and the long-term risk of cardiovascular disease (CVD) has not been validated. This study aimed to evaluate the long-term CVD risk associated with CKM health metrics and CKM stage using data from a population-based cohort study. METHODS: In total, 5293 CVD-free participants were followed up to around 13 years in the Chinese Multi-provincial Cohort Study (CMCS). Considering the pathophysiologic progression of CKM health metrics abnormalities (comprising obesity, central adiposity, prediabetes, diabetes, hypertriglyceridemia, CKD, and metabolic syndrome), participants were divided into CKM stages 0, 1, and 2. The time-dependent Cox regression models were used to estimate the cardiovascular risk associated with CKM health metrics and stage. Additionally, broader CVD outcomes were examined, with a specific assessment of the impact of stage 3 in 2581 participants from the CMCS-Beijing subcohort. RESULTS: Among participants, 91.2% (4825/5293) had at least one abnormal CKM health metric, 8.8% (468/5293), 13.3% (704/5293), and 77.9% (4121/5293) were in CKM stages 0, 1, and 2, respectively; and 710 incident CVD cases occurred during a median follow-up time of 13.3 years (interquartile range: 12.1 to 13.6 years). Participants with each poor CKM health metric exhibited significantly higher CVD risk. Compared with stage 0, the hazard ratio (HR) (95% confidence interval [CI]) for CVD incidence was 1.31 (0.84-2.04) in stage 1 and 2.27 (1.57-3.28) in stage 2. Significant interactive impacts existed between CKM stage and age or sex, with higher CVD risk related to increased CKM stages in participants aged <60 years or females. CONCLUSION These findings highlight the contribution of CKM health metrics and CKM stage to the long-term risk of CVD, suggesting the importance of multi-component recognition and management of poor CKM health in CVD prevention.
Humans ; Female ; Male ; Cardiovascular Diseases/etiology* ; Middle Aged ; Adult ; Cohort Studies ; Renal Insufficiency, Chronic/metabolism* ; Aged ; Risk Factors ; Metabolic Syndrome/metabolism* ; China ; East Asian People

The Healthy China 2030 Plan set the goal of reducing premature deaths from diabetes by 30% by 2030. However, there has been a lack of assessment of premature mortality for diabetes since the action plan was issued. This study used data from the Global Burden of Disease Study 2021, calculated the premature deaths for diabetes by sex, provinces, and subtypes from 1990 to 2021. We explored the temporal trend of premature mortality using the average annual percent change (AAPC) for different sexes, provinces, and subtypes from 1990 to 2021. Furthermore, we predicted premature mortality for diabetes through 2030 for China and its provinces according to the average annual change rate from 2010 to 2021. There was a first slow upward trend in premature mortality for diabetes from 0.5% in 1990 to 0.6% in 2004, and then a decline until 2021 with premature mortality of 0.4%. By 2030, only Fujian (30.3%) will achieve the desired level of reduction, with only seven provinces meeting the target for females and none for males. There is a large range in the degree of decline between inland and coastal regions, showing obvious geographic differences, and there should be a focus on balancing medical resources.
Humans ; China/epidemiology* ; Female ; Male ; Mortality, Premature/trends* ; Diabetes Mellitus/mortality* ; Goals ; Middle Aged ; Adult

In-depth study of the components of polymyxins is the key to controlling the quality of this class of antibiotics. Similarities and variations of components present significant analytical challenges. A two-dimensional (2D) liquid chromatography-mass spectrometr (LC-MS) method was established for screening and comprehensive profiling of compositions of the antibiotic colistimethate sodium (CMS). A high concentration of phosphate buffer mobile phase was used in the first-dimensional LC system to get the components well separated. For efficient and high-accuracy screening of CMS, a targeted method based on a self-constructed high resolution (HR) mass spectrum database of CMS components was established. The database was built based on the commercial MassHunter Personal Compound Database and Library (PCDL) software and its accuracy of the compound matching result was verified with six known components before being applied to genuine sample screening. On this basis, the unknown peaks in the CMS chromatograms were deduced and assigned. The molecular formula, group composition, and origins of a total of 99 compounds, of which the combined area percentage accounted for more than 95% of CMS components, were deduced by this 2D-LC-MS method combined with the MassHunter PCDL. This profiling method was highly efficient and could distinguish hundreds of components within 3 h, providing reliable results for quality control of this kind of complex drugs.