: To analyze the prevalence of anemia and iron deficiency among primary and secondary school students in Rural Nutrition Improvement Program areas of Guizhou Province in 2023, and to explore the related factors, so as to provide evidence for Rural Nutrition Improvement Program optimization. Methods: In September 2023, a stratified random cluster sampling strategy was used to select 40 rural compulsory education schools with rural nutrition improvement program in five counties of Guizhou Province. School level questionnaire was employed to collect information of basic characteristics and school meal implementation. A total of 7 826 primary and secondary school students aged 6-16 underwent anthropometry and hemoglobin (Hb) determination; serum ferritin (SF) was additionally measured in a random subsample of 1 795 pupils. Students in Grade 3 and above also completed a questionnaire covering demographic characteristics, dietary behaviours and nutrition knowledge. Group comparisons were conducted by Chi square test or Fisher s exact test, and multivariable Logistic regression models were constructed to identify factors associated with anemia and iron deficiency. Results: The overall Hb level was (133.21±12.95)g/L, with an anemia prevalence of 7.17%. The overall SF level was (69.58±59.01)μg/L, with an iron deficiency prevalence of 2.73%. Multivariable analysis showed that stunting ( OR =1.88), school menus without nutrient calculation ( OR =1.61) and absence of menu planning software in the current semester ( OR =2.34) independently increased anemia risk, whereas obesity reduced it ( OR =0.54) (all P <0.05). Girls ( OR =4.16) and Grades 7-9 ( OR =5.93) increased iron deficiency risk (both P <0.05). Compared with rarely eating fresh vegetables, students with consuming <3 kinds per day ( OR =0.08) or exactly 3 kinds per day ( OR =0.06) had lower iron deficiency risks (both P <0.05). Conclusions Anemia and iron deficiency are prevalent among primary and secondary school students in Guizhou. Targeted intervention measures should be implemented for key populations to enhance the effectiveness of nutrition improvement program.

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.

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.

ObjectiveTo evaluate the clinical efficacy and safety of Qidan Tangshen Granules （芪丹糖肾颗粒， QTG） in the treatment of early diabetic kidney disease （DKD） with qi deficiency， blood stasis， and kidney deficiency syndrome， and to explore its mechanism. MethodsA double-blind， double-simulated method was used to enroll 200 patients with early DKD and qi deficiency， blood stasis， and kidney deficiency syndrome. Patients were randomly assigned in a 1∶1 ratio to the treatment group （100 cases） and the control group （100 cases）. The treatment group received QTG plus a valsartan capsule simulant， while the control group received valsartan capsules plus a QTG simulant， both for 12 weeks. The primary outcome was the urinary albumin-to-creatinine ratio （UACR）. Secondary outcomes included estimated glomerular filtration rate （eGFR）， fasting blood glucose （FBG）， 2-hour postprandial blood glucose （PBG）， glycated hemoglobin （HbA1c）， and traditional Chinese medicine （TCM） syndrome scores （including individual symptom scores for fatigue， dull complexion， soreness and weakness of the waist and knees， headache and chest pain， irritability， spontaneous sweating， thirst and polydipsia， polyphagia， polyuria， numbness of the limbs， and the total TCM syndrome score）. Oxidative stress markers including serum 8-hydroxy-2'-deoxyguanosine （8-OHDG）， 3-nitrotyrosine （3-NT）， and superoxide dismutase （SOD） were also assessed. Clinical efficacy and TCM syndrome efficacy were evaluated after treatment， and routine blood tests， urinalysis， and liver function tests were conducted and adverse reaction during the tria was recorded to assess safety. ResultsA total of 191 patients completed the study （95 in the treatment group and 96 in the control group）. The treatment group showed significant reductions in UACR， FBG， PBG， and HbA1c levels after treatment （P＜0.05 or P＜0.01）. The single TCM symptom scores except for polyphagia and total TCM syndrome scores significantly decreased （P＜0.05 or P＜0.01）. Compared to the control group， the treatment group had signi-ficantly lower UACR， FBG， PBG levels， and total TCM syndrome scores， sinlge symptoms scores except for polyphagia and limb numbness （P＜0.05 or P＜0.01）. Among 40 randomly selected patients （21 cases in the treatment group and 19 cases in the control group） for oxidative stress analysis， there were no significant differences in SOD， 3-NT， and 8-OHDG levels before and after treatment within or between groups （P＞0.05）. The overall effective rate in the treatment group was 64.2% （61/95） and 39.6% （38/96） in the control group， while the TCM syndrome efficacy rates were 80.0% （76/95） and 24.0% （23/96）， respectively， with the treatment group showing superior efficacy （P＜0.01）. No significant differences were observed in routine blood tests， urinalysis， or liver function indices before and after treatment in either group （P＞0.05）. The incidence of adverse reactions was 8.4% （8/95） in the treatment group and 9.4% （9/96） in the control group， with no statistically significant difference （P＞0.05）. ConclusionQTG can effectively reduce UACR and blood glucose levels， alleviate clinical symptoms， and improve clinical efficacy in patients with early DKD with qi deficiency， blood stasis， and kidney deficiency syndrome. The treatment is well-tolerated and safe， with no significant impact on oxidative stress markers.

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 Kurtosis reflecting noise's temporal structure is an effective metric for evaluating noise-induced hearing loss (NIHL), and its threshold is still unclear. Objective To explore the energy range of kurtosis and the threshold of NIHL induced by kurtosis in this energy rangeMethods Using cross-sectional design, 4631 noise-exposed workers in manufacturing industry were selected. The hearing loss and noise exposure data of each worker were collected. Logistic regression model was used to establish the dose-response relationship between noise exposure and hearing loss and estimate the range of energy effects. Restricted cubic spline model was utilized to analyze the dose-response relationship curves of kurtosis to noise-induced hearing loss (NIHI) and high-frequency noise-induced hearing loss (HFNIHL) under different 8 h equivalent sound levels (L_{EX,8 h}), as well as to estimate the kurtosis effect threshold. Results The logistic regression model analysis showed that the prevalence of NIHI and HFNIHL increased significantly with increase of L_{EX,8 h} for steady noise; when L_{EX,8 h} of non-steady noise was 80-100 dB(A), the risk of hearing loss increased with an increase in kurtosis (P<0.05). The restricted cubic spline model showed that when L_{EX,8 h} of non-steady noise was 0-80 dB(A) or >100 dB(A), there was no significant relationship between kurtosis and NIHI or HFNIHL. When L_{EX,8 h} was 80-100 dB(A), there was a significant nonlinear dose-response relationship between kurtosis and NIHL or HFNIHL (P <0.001), and kurtosis > 28.08 exerted effect in elevating NIHI or HFNIHL. Conclusion The effect threshold of kurtosis on NIHL is 28.08 when non-steady noise levels are in the range of 80-100 dB(A). The effect threshold of kurtosis needs further verification.

Background Temporal kurtosis (without frequency weighting, i.e., Z-weighted kurtosis) can evaluate noise-induced hearing loss (NIHL). However, few studies have considered the function of frequency weighting (A- or C-weighted) kurtosis on NIHL. Objective To study the significance of A- and C-weighted kurtosis adjustment for equivalent sound level (L'_{EX,8 h}) in evaluating occupational hearing loss. Methods A cross-sectional survey was used to select 973 noise-exposed workers in seven industries as the subjects. The noise exposure of all workers was assessed by distributions of A-, C-, and Z-weighted kurtosis (e.g., K_A, K_C, and K_Z) and respective adjusted equivalent sound level (e.g., L'_{EX,8 h}-K_A, L'_{EX,8 h}-K_C, and L'_{EX,8 h}-K_Z). The significance of A- and C-weighted kurtosis in evaluating NIHL was evaluated by correlations between three types of L'_{EX,8 h} and NIHL, and improvement of noise-induced permanent threshold shift (NIPTS) underestimation predicted by the ISO prediction model (Acoustics—Estimation of noise-induced hearing loss, ISO 1999-2013). Results The median K_A, K_C, and K_Z were 68.33, 28.22, and 19.82, respectively. The binary logistic regression showed that L_{EX, 8 h}-K_A, L_{EX, 8 h}-K_C, and L'_{EX, 8 h}-K_Z were risk factors for NIHL (OR>1, P<0.001). The receiver operating characteristic (ROC) curve showed that when the outcome variable was noise-induced hearing impairment (NIHI), the areas under the curves corresponding to L'_{EX,8 h}-K_A, L'_{EX,8 h}-K_C, and L'_{EX,8 h}-K_Z were 0.625, 0.628, and 0.625, respectively. When the outcome variable was high-frequency noise-induced hearing loss (HFNIHL), the areas under the curves corresponding to L'_{EX,8 h}-K_A, L'_{EX, 8 h}-K_C, and L'_{EX,8 h}-K_Z were 0.624, 0.623, and 0.622, respectively (P<0.05). The order of underestimation improvement values predicted by L'_{EX,8 h} for NIPTS₁₂₃₄ was: L'_{EX,8 h}-K_A (4.68 dB HL)>L'_{EX,8 h}-K_C (4.38 dB HL)>L'_{EX,8 h}-K_Z (4.28 dB HL) (P<0.001). The order of underestimation improvement values predicted by L'_{EX,8 h}-K for NIPTS₃₄₆ was: L'_{EX,8 h}-K_A (7.20 dB HL)>L'_{EX,8 h}-K_C (6.83 dB HL)>L'_{EX,8 h}-K_Z (6.71 dB HL) (P<0.001). Conclusion The adjustment of A- and C-weighted kurtosis to equivalent sound level L_{EX,8 h} can effectively improve the accuracy of the ISO 1999 prediction model in NIPTS prediction, and compared with the C-weighted, the A-weighted kurtosis can improve the result of the ISO 1999 prediction model in terms of underestimating NIPTS.

background Current assessment of noise-induced hearing loss relies on the hearing threshold statistical distribution table of ISO 7029-2017 standard (ISO 7029), which is based on foreign population data and lacks a hearing threshold distribution table derived from pure-tone audiometry data of the Chinese population, hindering accurate evaluation of hearing loss in this group. Objective To establish a statistical distribution table of hearing threshold level (HTL) for otologically normal Chinese adults and to provide a scientific basis for revising the diagnostic criteria of occupational noise-induced deafness in China. Methods A total of 1271 otologically normal Chinese adults aged 18-60 years were divided into four groups with 10-year age intervals. Based on the pure tone audiometry data and power function operation formula, the relevant parameters of the median and percentile calculation formula of ISO 7029 were adjusted. Based on the adjusted parameters, a statistical distribution table of HTL of normal Chinese adults in otology was preliminarily constructed according to the calculation formula of ISO 7029. A nonparametric rank sum test was used to compare the difference between the adjusted median deviation value and the ISO 7029 calculation. Results Except for a few frequencies (2000, 3000, and 8000 Hz), the value of parameter α adjusted for the auditory threshold deviation of the Chinese population (α_CHN-md) increased with the increase of frequency, while the value of parameter β adjusted for the auditory threshold deviation of the Chinese population (β_CHN-md) decreased with the increase of frequency. The preliminary distribution table of otologically normal Chinese adults' median hearing threshold deviation (ΔH_CHN-md) showed that the hearing threshold deviation increased with age and frequency. At 8000 Hz, the increase in ΔH_CHN-md with age was the greatest for both men and women, from 0 dB to 23 dB and 21 dB respectively. The minimal threshold elevation was noted at 500, 1000, and 2000 Hz, which increased from 0 dB to 2 dB. In otologically normal Chinese adults, the maximum increase of ΔH_CHN-md with frequency was observed at age of 50 years, which increased from 2 dB at 500 Hz to 23 dB at 8000 Hz in men and from 2 dB at 500 Hz to 21 dB at 8000 Hz in women. The comparison results with the ISO 7029 calculation showed that the trend was consistent. The median hearing threshold deviation value of the adjusted Chinese population was lower than that of the ISO 7029 calculation (ΔH_md). However, they had no significant statistical difference (P>0.05). For otologically normal normal Chinese men and women, the maximum difference of ΔH_md between the 50-year-old group and the ISO calculation at 8000 Hz was 7 dB and 9 dB, respectively. Conclusion This preliminary analysis of hearing threshold deviations in otologically normal Chinese adults suggests that current standards may potentially underestimate hearing loss at frequencies below 8000 Hz while possibly overestimating it at 8000 Hz. These findings require further validation through expanded sample sizes to more accurately assess the characteristics of hearing thresholds in the Chinese population and their discrepancies with existing standards.