Application progress of predictive models in the prevention and control of myopia in children and adolescents / 国际眼科杂志(Guoji Yanke Zazhi)

In medical research,predictive models have been widely used to predict disease progression and identify high-risk populations in advance, especially in the prevention and diagnosis of chronic diseases. In ophthalmology, the predictive and diagnostic models for fundus diseases such as age-related macular degeneration and diabetic retinopathy have demonstrated expert-level accuracy. However, the application of predictive models is still in the exploratory stage as for myopia prevention and control. The establishment of a predictive model is helpful to identify the high-risk myopic children in advance, so that preventive measures such as adequate outdoor activities and reducing near work can be taken in time, which is of great significance to prevent or slow down the occurrence and development of myopia. Because the mechanism of myopia has not been fully elucidated, there are still challenges and limitations in the selection of application objects, predictors and predictive outcomes. This paper reviews the research progress of different types of myopia predictive models in order to provide reference for further development and improvement.

Risk prediction models of dangerous behaviors among patients with severe mental disorder in community / 四川精神卫生

BackgroundThe occurrence rate of dangerous behaviors in patients with severe mental disorders is higher than that of the general population. In China， there is limited research on the prediction of dangerous behaviors in community-dwelling patients with severe mental disorders， particularly in terms of predicting models using data mining techniques other than traditional methods. ObjectiveTo explore the influencing factors of dangerous behaviors in community-dwelling patients with severe mental disorders and testing whether the classification decision tree model is superior to the Logistic regression model. MethodsA total of 11 484 community-dwelling patients with severe mental disorders who had complete follow-up records from 2013 to 2022 were selected on December 2023. The data were divided into a training set （n=9 186） and a testing set （n=2 298） in an 8∶2 ratio. Logistic regression and classification decision trees were separately used to establish predictive models in the training set. Model discrimination and calibration were evaluated in the testing set. ResultsDuring the follow-up period， 1 115 cases （9.71%） exhibited dangerous behaviors. Logistic regression results showed that urban residence， poverty， guardianship， intellectual disability， history of dangerous behaviors， impaired insight and positive symptoms were risk factors for dangerous behaviors （OR=1.778， 1.459， 2.719， 1.483， 3.890， 1.423， 2.528， 2.124， P<0.01）. Being aged ≥60 years， educated， not requiring prescribed medication and having normal social functioning were protective factors for dangerous behaviors （OR=0.594， 0.824， 0.422， 0.719， P<0.05 or 0.01）. The predictive effect in the testing set showed an area under curve （AUC） of 0.729 （95% CI： 0.692~0.766）， accuracy of 70.97%， sensitivity of 59.71%， and specificity of 72.05%. The classification decision tree results showed that past dangerous situations， positive symptoms， overall social functioning score， economic status， insight， household registration， disability status and age were the influencing factors for dangerous behaviors. The predictive effect in the testing set showed an AUC of 0.721 （95% CI： 0.705~0.737）， accuracy of 68.28%， sensitivity of 64.46%， and specificity of 68.60%. ConclusionThe classification decision tree does not have a greater advantage over the logistic regression model in predicting the risk of dangerous behaviors in patients with severe mental disorders in the community. ［Funded by Chengdu Medical Research Project （number， 2020052）］

Prognostic predictive model for PONV in cesarean delivery

Introduction: Apfel simplified risk score for postoperative nausea and vomiting (PONV) has shown to be useful in anesthesia; however, since it has not been calibrated in regional anesthesia or in pregnant patients, its use in cesarean section is limited. Objective: To develop a prognostic predictive model for postoperative nausea and vomiting in pregnant patients undergoing cesarean section under spinal anesthesia. Methods: In a cohort of 703 term pregnant patients scheduled of cesarean section, 15 variables were prospectively assessed, to design a prognostic predictive model for the development of postoperative nausea and vomiting. A logistic regression analysis was used to construct the model and its calibration and discrimination were based on the Hosmer-Lemeshow test, the calibration curves, and C statistic. Additionally, the internal calibration was performed with the Bootstrap resampling method. Results: Postoperative nausea and vomiting were experienced by 27% of the patients during the first six hours after surgery. The model included as prognostic variables the development of intraoperative nausea and vomiting, age under 28 years, a history of PONV, the mother's BMI and the weight of the newborn baby. The model showed an adequate calibration (x2: 4.65 p: 0.5888), though a low discrimination (Statistic C = 0.68). Conclusions: A prognostic predictive model was created for the development of PONV in cesarean section. This model was used to build a prognostic scale for the classification of patients into risk groups.

Introducción: La escala de riesgo simplificada de Apfel para náuseas y vómito posoperatorio (NVPO) ha mostrado utilidad en anestesia; sin embargo, al no haber sido calibrada en anestesia regional o en pacientes embarazadas, su utilidad en cesárea es limitado. Objetivo: Desarrollar un modelo de predicción pronóstica para náuseas y vómito posoperatorios en pacientes embarazadas, llevadas a cesárea bajo anestesia espinal. Métodos: En una cohorte de 703 pacientes con embarazo a término programadas para cesárea, se evaluaron 15 variables de forma prospectiva para construir un modelo de predicción pronóstica para el desarrollo de náuseas y vómito posoperatorio. Se utilizó el análisis de regresión logística para la construcción del modelo y se calculó su calibración y discriminación con la prueba de Hosmer-Lemeshow, las curvas de calibración y el estadístico C. Además, se realizó la calibración interna con el método de remuestreo Bootstrap. Resultados: Las náuseas y vómito posoperatorio se presentaron en el 27% de las pacientes durante las primeras seis horas después de la cirugía. El modelo incluyó como variables pro-nósticas el desarrollo de náuseas y vómito en el intraoperatorio, edad menor de 28 años, antecedentes de NVPO, índice de masa corporal (IMC) de la madre y el peso del recién nacido. El modelo mostró una adecuada calibración (x2: 4,65 p: 0,5888), aunque una baja discriminación (Estadístico C = 0,68). Conclusiones: Se construyó un modelo de predicción pronóstica para el desarrollo de NVPO en cirugía cesárea, y con este se construyó una escala pronóstica que permite clasificar a las pacientes por grupos de riesgo.

Modelo predictivo de enfermedad cardiovascular basado en inteligencia artificial en la atención primaria de salud / A Predictive Model for Cardiovascular Disease based on Artificial Intelligence in Primary Health Care

Introducción: En Cuba y en el resto del mundo, las enfermedades cardiovasculares son reconocidas como un problema de salud pública mayúsculo y creciente, que provoca una alta mortalidad. Objetivo: Diseñar un modelo predictivo para estimar el riesgo de enfermedad cardiovascular basado en técnicas de inteligencia artificial. Métodos: La fuente de datos fue una cohorte prospectiva que incluyó 1633 pacientes, seguidos durante 10 años, fue utilizada la herramienta de minería de datos Weka, se emplearon técnicas de selección de atributos para obtener un subconjunto más reducido de variables significativas, para generar los modelos fueron aplicados: el algoritmo de reglas JRip y el meta algoritmo Attribute Selected Classifier, usando como clasificadores el J48 y el Multilayer Perceptron. Se compararon los modelos obtenidos y se aplicaron las métricas más usadas para clases desbalanceadas. Resultados: El atributo más significativo fue el antecedente de hipertensión arterial, seguido por el colesterol de lipoproteínas de alta densidad y de baja densidad, la proteína c reactiva de alta sensibilidad y la tensión arterial sistólica, de estos atributos se derivaron todas las reglas de predicción, los algoritmos fueron efectivos para generar el modelo, el mejor desempeño fue con el Multilayer Perceptron, con una tasa de verdaderos positivos del 95,2 por ciento un área bajo la curva ROC de 0,987 en la validación cruzada. Conclusiones: Fue diseñado un modelo predictivo mediante técnicas de inteligencia artificial, lo que constituye un valioso recurso orientado a la prevención de las enfermedades cardiovasculares en la atención primaria de salud(AU)

Introduction: In Cuba and in the rest of the world, cardiovascular diseases are recognized as a major and growing public health problem, which causes high mortality. Objective: To design a predictive model to estimate the risk of cardiovascular disease based on artificial intelligence techniques. Methods: The data source was a prospective cohort including 1633 patients, followed for 10 years. The data mining tool Weka was used and attribute selection techniques were employed to obtain a smaller subset of significant variables. To generate the models, the rule algorithm JRip and the meta-algorithm Attribute Selected Classifier were applied, using J48 and Multilayer Perceptron as classifiers. The obtained models were compared and the most used metrics for unbalanced classes were applied. Results: The most significant attribute was history of arterial hypertension, followed by high and low density lipoprotein cholesterol, high sensitivity c-reactive protein and systolic blood pressure; all the prediction rules were derived from these attributes. The algorithms were effective to generate the model. The best performance was obtained using the Multilayer Perceptron, with a true positive rate of 95.2percent and an area under the ROC curve of 0.987 in the cross validation. Conclusions: A predictive model was designed using artificial intelligence techniques; it is a valuable resource oriented to the prevention of cardiovascular diseases in primary health care(AU)

Aprendizaje automático aplicado a la predicción de diabetes mellitus, utilizando información socioeconómica y ambiental de usuarios del sistema de salud

Objetivo: Se propuso aplicar modelos basados en técnicas de aprendizaje automático como apoyo para el diagnóstico temprano de la diabetes mellitus, utilizando variables de datos ambientales, sociales, económicos y sanitarios, sin la dependencia de la toma de muestras clínicas. Metodología: Se utilizaron datos de 10 889 usuarios afiliados al régimen subsidiado de salud de la zona suroccidental en Colombia, diagnosticados con hipertensión y agrupados en usuarios sin (74,3 %) y con (25,7 %) diabetes mellitus. Se entrenaron modelos supervisados utilizando k vecinos más cercanos, árboles de decisión y bosques aleatorios, así como modelos basados en ensambles, aplicados a la base de datos antes y después de balancear el número de casos en cada grupo de diagnóstico. Se evalúo el rendimiento de los algoritmos mediante la división de la base de datos en datos de entreno y de prueba (70/30, respectivamente), y se utilizaron métricas de exactitud, sensibilidad, especificidad y área bajo la curva. Resultados: Los valores de sensibilidad aumentaron considerablemente al utilizar datos balanceados, pasando de valores máximos del 17,1 % (datos sin balancear) a valores de hasta 57,4 % (datos balanceados). El valor más alto de área bajo la curva (0,61) fue obtenido con los modelos de ensambles, al aplicar un balance en el número de datos por cada grupo y al codificar las variables categóricas. Las variables de mayor peso estuvieron asociadas con aspectos hereditarios (24,65 %) y con el grupo étnico (5.59 %), además de la dificultad visual, el bajo consumo de agua, una dieta baja en frutas y verduras, y el consumo de sal y azúcar. Conclusiones: Aunque los modelos predictivos, utilizando información socioeconómica y ambiental de las personas, surgen como una herramienta para el diagnóstico temprano de la diabetes mellitus, estos aún deben ser mejorados en su capacidad predictiva.

Objective: The objective was to apply models based on machine learning techniques to support the early diagnosis of diabetes mellitus, using environmental, social, economic and health data variables, without dependence on clinical sample collection. Methodology: Data from 10,889 users affiliated with the subsidized health system in the southwestern area of Colombia, diagnosed with hypertension and grouped into users without (74.3%) and with (25.7%) diabetes mellitus, were used. Supervised models were trained using k-nearest neighbors, decision trees, and random forests, as well as ensemble-based models, applied to the database before and after balancing the number of cases in each diagnostic group. The performance of the algorithms was evaluated by dividing the database into training and test data (70/30, respectively), and metrics of accuracy, sensitivity, specificity, and area under the curve were used. Results: Sensitivity values increased significantly when using balanced data, going from maximum values of 17.1% (unbalanced data) to values as high as 57.4% (balanced data). The highest value of area under the curve (0.61) was obtained with the ensemble models, by applying a balance in the amount of data for each group and by coding the categorical variables. The variables with the greatest weight were associated with hereditary aspects (24.65%) and with the ethnic group (5.59%), in addition to visual difficulty, low water consumption, a diet low in fruits and vegetables, and the consumption of salt and sugar. Conclusions: Although predictive models, using people's socioeconomic and environmental information, emerge as a tool for the early diagnosis of diabetes mellitus, their predictive capacity still needs to be improved.

Objetivo: Propôs-se aplicar modelos baseados em técnicas de aprendizagem automática como apoio para o diagnóstico precoce da diabetes mellitus, utilizando variáveis de dados ambientais, sociais, econômicos e sanitários, sem a dependência da coleta de amostras clínicas. Metodologia: Usaram-se dados de 10.889 usuários filiados ao regime subsidiado de saúde da zona sudoeste da Colômbia, diagnosticados com hipertensão e agrupados em usuários sem (74,3%) e com (25,7%) diabetes mellitus. Foram treinados modelos supervisionados utilizando k vizinhos mais próximos, árvores de decisão e florestas aleatórias, assim como modelos baseados em montagens, aplicados à base de dados antes de depois de equilibrar o número de casos em cada grupo de diagnóstico. Avaliou-se o desempenho dos algoritmos por meio da divisão da base de dados de treino e teste (70/30, respectivamente), e utilizaram-se métricas de exatidão, sensibilidade, especificidade e área sob a curva. Resultados: Os valores de sensibilidade aumentaram de maneira significativa ao utilizar dados equilibrados, passando de valores máximos de 17,1% (dados sem equilibrar) a valores de até 57,4% (dados equilibrados). O valor mais elevado de área sob a curva (0,61) foi obtido com os modelos de montagens, ao aplicar um balanço no número de dados por cada grupo e codificar as variáveis categóricas. As variáveis de maior peso estiveram associadas com fatores hereditários (24,65%) e com o grupo étnico (5,59%), além da dificuldade visual, o baixo consumo de água, um regime baixo em frutas e vegetais e o consumo de sal e açúcar. Conclusões: Embora os modelos preditivos, utilizando informação socioeconômica e ambiental das pessoas, surgem como uma ferramenta para o diagnóstico precoce da diabetes mellitus, ainda devem ser melhorados em sua capacidade preditiva.

Applicability of the acute leukemia (AL) - EBMT score as a prognostic model for allogeneic hematopoietic stem cell transplantation: a single-center, prospective, cohort study at a reference transplant center in Brazil

Abstract Introduction The Acute Leukemia-European Society for Blood and Marrow Transplantation (AL-EBMT) risk score was recently developed and validated by Shouval et al. Objective To assess the ability of this score in predicting the 2-year overall survival (OS-2), leukemia-free survival (LFS-2) and transplant-related mortality (TRM) in acute leukemia (AL) adult patients undergoing a first allogeneic hematopoietic stem cell transplant (HSCT) at a transplant center in Brazil. Methods In this prospective, cohort study, we used the formula published by Shouval et al. to calculate the AL-EBMT score and stratify patients into three risk categories. Results A total of 79 patients transplanted between 2008 and 2018 were analyzed. The median age was 38 years. Acute myeloid leukemia was the most common diagnosis (68%). Almost a quarter of the cases were at an advanced stage. All hematopoietic stem cell transplantations (HSCTs) were human leukocyte antigen-matched (HLA-matched) and the majority used familial donors (77%). Myeloablative conditioning was used in 92% of the cases. Stratification according to the AL-EBMT score into low-, intermediate- and high-risk groups yielded the following results: 40%, 12% and 47% of the cases, respectively. The high scoring group was associated with a hazard ratio of 2.1 (p= 0.007), 2.1 (p= 0.009) and 2.47 (p= 0.01) for the 2-year OS, LFS and TRM, respectively. Conclusion This study supports the ability of the AL-EBMT score to reasonably predict the 2-year post-transplant OS, LFS and TRM and to discriminate between risk categories in adult patients with AL, thus confirming its usefulness in clinical decision-making in this setting. Larger, multicenter studies may further help confirm these findings.

Development and Validation of a Predictive Risk Model for Vision-threatening Diabetic Retinopathy in Patients with Type 2 Diabetes / 中山大学学报(医学科学版)

ObjectiveTo develop and validate a predictive risk model for vision-threatening diabetic retinopathy in patients with type 2 diabetes using readily accessible clinical data， which may provide a convenient and effective prediction tool for early identification and referral of at-risk populations. MethodsA nomogram model was developed using a dataset obtained from patients with T2DM who participated in the Guangzhou Diabetic Eye Study from November 2017 to December 2020. Logistic regression was used to construct the model， and model performance was evaluated using receiver operating characteristic curve， Hosmer-Lemeshow test， calibration curve and decision curve analysis. The model underwent internal validation through the mean AUC of k-fold cross-validation method， and further external validation was conducted in the Dongguan Eye Study. ResultsA total of 2 161 individuals were included in the model development dataset， of whom 135 （6.25%） people were diagnosed with VTDR. Age （P＜0.001，OR=0.927，95%CI：0.898~0.957） and body mass index （P＜0.001，OR =0.845，95%CI：0.821~0.932） were found to be negatively correlated with VTDR， whereas diabetes duration （P＜0.001，OR=1.064，95%CI：1.035~1.094）， insulin use （P =0.045，OR =1.534，95%CI：1.010~2.332）， systolic blood pressure （P＜0.001，OR =1.019，95%CI：1.008~1.029）， glycated hemoglobin （P＜0.001，OR =1.484，95%CI：1.341~1.643）， and serum creatinine （P＜0.001，OR =1.017，95%CI：1.010~1.023） were positively correlated with VTDR. All these variables were included in the model as predictors. The model showed strong discrimination in the development dataset with an area under the receiver operating characteristic curve （AUC） of 0.797 and in the external validation dataset （AUC 0.762）. The Hosmer-Lemeshow test（P＞0.05）and the calibration curve displayed good agreement. Decision curve analysis showed that the nomogram produced net benefit in the two datasets. ConclusionsIndependent factors influencing VTDR include age， duration of diabetes mellitus， insulin use， body mass index， systolic blood pressure， glycosylated hemoglobin， and serum creatinine. The nomogram constructed using these variables demonstrates a high degree of predictive validity. The model can serve as a valuable tool for early detection and referral of VTDR in primary care clinics. Therefore， its application and promotion are highly recommended.

Statistical predictive model nomogram applied in ophthalmology / 国际眼科杂志(Guoji Yanke Zazhi)

The alignment diagram, also known as nomogram, is a statistical prediction model used to predict the risk of events(such as diagnosis or prediction of disease development and consequences)by obtaining the influence power of each related factor on outcome variables through multivariate analysis. Nomogram turns the complex regression equation to a visualized diagram that is intuitive and easy to understand. It is convenient to be used for evaluating the patient's condition and communicating with doctors and patients. With rapid advances of medical science and technology and increasing demands of personalized medicine, nomograms has attracted more and more attention and applied extensively in clinical medicine. This short article introduces the basic concepts of nomogram and examples of its application in ophthalmology.

Analysis of Clinical Data and Construction of A Diagnostic Prediction Model for Metabolic Syndrome after Single-Center Hematopoietic Stem Cell Transplantation / 中国实验血液学杂志

UNLABELLED@#AbstractObjective: To analysis the clinical data of patients after single-center hematopoietic stem cell transplantation (HSCT) and construct a predictive model for metabolic syndrome (MS) diagnosis.@*METHODS@#Ninety-three hematology patients who underwent HSCT at the First Hospital of Lanzhou University from July 2015 to September 2022 were selected to collect basic data, transplantation status and postoperative data, the clinical characteristics of patients with and without MS after transplantation were compared and analyzed. Logistic regression model was used to analyze the influence fators of MS after transplantation, and a predictive model of HSCT-MS diagnosis was constructed under the influence of independent influence factors. The model was evaluated using the ceceiver operating characteristic curve (ROC curve).@*RESULTS@#Metabolic syndrome occurred in 36 of 93 HSCT patients and did not occur in 57. Compared with non-HSCT-MS group, HSCT-MS had significantly higher fasting blood glucose (FBG) levels before transplantation, shorter course before transplantation, and higher bilirubin levels after transplantation (P<0.05). The statistically significant clinical indicators were subjected to multi-factor logistic regression analysis, and the results showed that pre-transplant high FBG, pre-transplant short disease course and post-transplant high bilirubin were independent influence factors for HSCT-MS. The standard error of predicting the occurrence of HSCT-MS based on the clinical model was 0.048, the area under the curve AUC=0.776, 95% CI :0.683-0.869, the optimal threshold was 0.58 based on the Jorden index at maximum, the sensitivity was 0.694, and the specificity was 0.772, which has certain accuracy.@*CONCLUSION@#A clinical prediction model for HSCT-MS based on logistic regression analysis is constructed through the analysis of clinical data, which has certain clinical value.

Development and validation of a predictive model for the assessment of potassium-lowering treatment among hyperkalemia patients / 世界急诊医学杂志（英文）

@#BACKGROUND: Hyperkalemia is common among patients in emergency department and is associated with mortality. While, there is a lack of good evaluation and prediction methods for the efficacy of potassium-lowering treatment, making the drug dosage adjustment quite difficult. We aimed to develop a predictive model to provide early forecasting of treating effects for hyperkalemia patients. METHODS: Around 80% of hyperkalemia patients (n=818) were randomly selected as the training dataset and the remaining 20% (n=196) as the validating dataset. According to the serum potassium (K+) levels after the first round of potassium-lowering treatment, patients were classified into the effective and ineffective groups. Multivariate logistic regression analyses were performed to develop a prediction model. The receiver operating characteristic (ROC) curve and calibration curve analysis were used for model validation. RESULTS: In the training dataset, 429 patients had favorable effects after treatment (effective group), and 389 had poor therapeutic outcomes (ineffective group). Patients in the ineffective group had a higher percentage of renal disease (P=0.007), peripheral edema (P<0.001), oliguria (P=0.001), or higher initial serum K+ level (P<0.001). The percentage of insulin usage was higher in the effective group than in the ineffective group (P=0.005). After multivariate logistic regression analysis, we found age, peripheral edema, oliguria, history of kidney transplantation, end-stage renal disease, insulin, and initial serum K+ were all independently associated with favorable treatment effects. CONCLUSION: The predictive model could provide early forecasting of therapeutic outcomes for hyperkalemia patients after drug treatment, which could help clinicians to identify hyperkalemia patients with high risk and adjust the dosage of medication for potassium-lowering.

Construction and evaluation of an artificial intelligence-based risk prediction model for death in patients with nasopharyngeal cancer / 南方医科大学学报

OBJECTIVE@#To screen the risk factors for death in patients with nasopharyngeal carcinoma (NPC) using artificial intelligence (AI) technology and establish a risk prediction model.@*METHODS@#The clinical data of NPC patients obtained from SEER database (1973-2015). The patients were randomly divided into model building and verification group at a 7∶3 ratio. Based on the data in the model building group, R software was used to identify the risk factors for death in NPC patients using 4 AI algorithms, namely eXtreme Gradient Boosting (XGBoost), Decision Tree (DT), Least absolute shrinkage and selection operator (LASSO) and random forest (RF), and a risk prediction model was constructed based on the risk factor identified. The C-Index, decision curve analysis (DCA), receiver operating characteristic (ROC) curve and calibration curve (CC) were used for internal validation of the model; the data in the validation group and clinical data of 96 NPC patients (collected from First Affiliated Hospital of Bengbu Medical College) were used for internal and external validation of the model.@*RESULTS@#The clinical data of a total of 2116 NPC patients were included (1484 in model building group and 632 in verification group). Risk factor screening showed that age, race, gender, stage M, stage T, and stage N were all risk factors of death in NPC patients. The risk prediction model for NPC-related death constructed based on these factors had a C-index of 0.76 for internal evaluation, an AUC of 0.74 and a net benefit rate of DCA of 9%-93%. The C-index of the model in internal verification was 0.740 with an AUC of 0.749 and a net benefit rate of DCA of 3%-89%, suggesting a high consistency of the two calibration curves. In external verification, the C-index of this model was 0.943 with a net benefit rate of DCA of 3%-97% and an AUC of 0.851, and the predicted value was consistent with the actual value.@*CONCLUSIONS@#Gender, age, race and TNM stage are risk factors of death of NPC patients, and the risk prediction model based on these factors can accurately predict the risks of death in NPC patients.

Study on risk factors of abnormal pulmonary function among dust-exposed workers and prediction model / 中华劳动卫生职业病杂志

Objective: To explore the influencing factors of abnormal pulmonary function in dust-exposed workers and establish the risk prediction model of abnormal pulmonary function. Methods: In April 2021, a total of 4255 dust exposed workers from 47 enterprises in 2020 were included in the study. logistic regression was used to analyze the influencing factors of abnormal pulmonary function in dust-exposed workers, and the corresponding nomogram prediction model was established. The model was evaluated by ROC curve, Calibrationpolt and decision analysis curve. Results: logistic regression analysis showed that age (OR=1.03, 95%CI=1.02~1.05, P<0.001) , physical examination type (OR=4.52, 95%CI=1.69~12.10, P=0.003) , dust type (Comparison with coal dust, Cement dust, OR=3.45, 95%CI=1.45~8.18, P=0.005, Silica dust (OR=2.25, 95%CI=1.01~5.03, P=0.049) , blood pressure (OR=1.63, 95%CI=1.22~2.18, P=0.001) , creatinine (OR=0.08, 95%CI=0.05~0.12, P<0.001) , daily exposure time (OR=1.06, 95%CI=1.10~1.12, P=0.034) and total dust concentration (OR=1.29, 95%CI=1.08~1.54, P=0.005) were the influencing factors of abnormal pulmonary function. The area under the ROC curve of risk prediction nomogram model was 0.764. The results of decision analysis curve showed that the nomogram model had reference value in the prevention and intervention of abnormal pulmonary function when the threshold probability exceeded 0.05. Conclusion: The accuracy ofthe nomogram model constructed by logistic regression werewell in predicting the risk of abnormal lung function of dust-exposed workers.

Application of machine learning in clinical predictive models for infectious diseases: a review / 中国血吸虫病防治杂志

Infectious diseases are one of the major threats to global public health. Inconvenience of diagnosis and treatment frequently causes misdiagnosis, missing diagnosis or overtreatment, resulting in serious clinical outcomes. As an important branch of artificial intelligence, machine learning has been widely used in multiple fields. Predictive models created based on patients’ clinical characteristics, laboratory tests, and imaging examinations are effective for prediction and evaluation of clinical diagnosis, therapeutic efficacy and prognosis, as well as detection of outbreaks. Machine learning modeling has the advantages of high efficiency, high accuracy and interpretability as compared to traditional modeling approaches, which provides a new tool for diagnosis and treatment of infectious diseases. This review summarizes the advances of applications of machine learning in clinical predictive models for infectious diseases.

Risk predictive models of healthcare-seeking delay among imported malaria patients in Jiangsu Province based on the machine learning / 中国血吸虫病防治杂志

Objective To create risk predictive models of healthcare-seeking delay among imported malaria patients in Jiangsu Province based on machine learning algorithms, so as to provide insights into early identification of imported malaria cases in Jiangsu Province. Methods Case investigation, first symptoms and time of initial diagnosis of imported malaria patients in Jiangsu Province in 2019 were captured from Infectious Disease Report Information Management System and Parasitic Disease Prevention and Control Information Management System of Chinese Center for Disease Control and Prevention. The risk predictive models of healthcare-seeking delay among imported malaria patients were created with the back propagation (BP) neural network model, logistic regression model, random forest model and Bayesian model using thirteen factors as independent variables, including occupation, species of malaria parasite, main clinical manifestations, presence of complications, severity of disease, age, duration of residing abroad, frequency of malaria parasite infections abroad, incubation period, level of institution at initial diagnosis, country of origin, number of individuals travelling with patients and way to go abroad, and time of healthcare-seeking delay as a dependent variable. Logistic regression model was visualized using a nomogram, and the nomogram was evaluated using calibration curves. In addition, the efficiency of the four models for prediction of risk of healthcare-seeking delay among imported malaria patients was evaluated using the area under curve (AUC) of receiver operating characteristic curve (ROC). The importance of each characteristic was quantified and attributed by using SHAP to examine the positive and negative effects of the value of each characteristic on the predictive efficiency. Results A total of 244 imported malaria patients were enrolled, including 100 cases (40.98%) with the duration from onset of first symptoms to time of initial diagnosis that exceeded 24 hours. Logistic regression analysis identified a history of malaria parasite infection [odds ratio (OR) = 3.075, 95% confidential interval (CI): (1.597, 5.923)], long incubation period [OR = 1.010, 95% CI: (1.001, 1.018)] and seeking healthcare in provincial or municipal medical facilities [OR = 12.550, 95% CI: (1.158, 135.963)] as risk factors for delay in seeking healthcare among imported malaria cases. BP neural network modeling showed that duration of residing abroad, incubation period and age posed great impacts on delay in healthcare-seek among imported malaria patients. Random forest modeling showed that the top five factors with the greatest impact on healthcare-seeking delay included main clinical manifestations, the way to go abroad, incubation period, duration of residing abroad and age among imported malaria patients, and Bayesian modeling revealed that the top five factors affecting healthcare-seeking delay among imported malaria patients included level of institutions at initial diagnosis, age, country of origin, history of malaria parasite infection and individuals travelling with imported malaria patients. ROC curve analysis showed higher overall performance of the BP neural network model and the logistic regression model for prediction of the risk of healthcare-seeking delay among imported malaria patients (Z = 2.700 to 4.641, all P values < 0.01), with no statistically significant difference in the AUC among four models (Z = 1.209, P > 0.05). The sensitivity (71.00%) and Youden index (43.92%) of the logistic regression model was higher than those of the BP neural network (63.00% and 36.61%, respectively), and the specificity of the BP neural network model (73.61%) was higher than that of the logistic regression model (72.92%). Conclusions Imported malaria cases with long duration of residing abroad, a history of malaria parasite infection, long incubation period, advanced age and seeking healthcare in provincial or municipal medical institutions have a high likelihood of delay in healthcare-seeking in Jiangsu Province. The models created based on the logistic regression and BP neural network show a high efficiency for prediction of the risk of healthcare-seeking among imported malaria patients in Jiangsu Province, which may provide insights into health management of imported malaria patients.

18F-FDG PET/CT Prognostic Role in Diffuse Large B-cell Lymphoma Following Chemotherapy / 中山大学学报(医学科学版)

ObjectiveTo assess the prognostic value of 18F-FDG PET/CT parameters for predicting therapeutic response in diffuse large B-cell lymphoma （DLBCL）. MethodsWe retrospectively analyzed the clinical data and 18F-FDG PET/CT radiomics features of 81 DLBCL patients enrolled between June 2015 and October 2020. Multivariate logistic regression analysis was used to identify the predictive factors for therapeutic response of DLBCL， based on which a predictive model was developed accordingly. The performance of the model was evaluated by receiver operating characteristic （ROC） curves and calibration plots. ResultsDuring the two years after first chemotherapy， 23 patients （28.3%） developed relapse and 58 patients （71.7%） had progression-free survival （PFS）. The analysis for the predictive capability of the binary logistic regression model incorporating the PET/CT features revealed that the imaging features of 18F-FDG PET/CT after chemotherapy were independent prognostic factors for PFS. Among them， SUVTHR-mean2 was the most important factor for predicting therapeutic response in DLBCL patients after chemotherapy， with a cutoff value of 2.00 （AUC=0.81）. Conclusions18F-FDG PET/CT showed a valuable prognostic performance for PFS in DLBCL patients after chemotherapy， with the imaging feature after chemotherapy SUVTLR-mean2 being the optimal independent predictor. Our predictive model of imaging features might have an important prognostic value in assessing the risk of disease progression， guiding the treatment and follow-up protocol， improving therapeutic efficiency and cutting down the medical cost.

Predictive value of the simplified signs scoring system for the severity and prognosis of patients with COVID-19: A multicenter observational study / 中国胸心血管外科临床杂志

@# Objective    To explore the predictive value of a simplified signs scoring system for the severity and prognosis of patients with coronavirus disease 2019 (COVID-19). Methods     Clinical data of 1 605 confirmed patients with COVID-19 from January to May 2020 in 45 hospitals of Sichuan and Hubei Provinces were retrospectively analyzed. The patients were divided into a mild group (n=1 150, 508 males, average age of 51.32±16.26 years) and a severe group (n=455, 248 males, average age of 57.63±16.16 years). Results    Age, male proportion, respiratory rate, systolic blood pressure and mean arterial pressure in the severe group were higher than those in the mild group (P<0.05). Peripheral oxygen saturation (SpO2) and Glasgow coma scale (GCS) were lower than those in the mild group (P<0.05). Multivariate logistic regression analysis showed that age, respiratory rate, SpO2, and GCS were independent risk factors for severe patients with COVID-19. Based on the above indicators, the receiver operating characteristic (ROC) curve analysis showed that the area under the curve of the simplified signs scoring system for predicting severe patients was 0.822, which was higher than that of the quick sequential organ failure assessment (qSOFA) score and modified early warning score (MEWS, 0.629 and 0.631, P<0.001). The ROC analysis showed that the area under the curve of the simplified signs scoring system for predicting death was 0.796, higher than that of qSOFA score and MEWS score (0.710 and 0.706, P<0.001). Conclusion    Age, respiratory rate, SpO2 and GCS are independent risk factors for severe patients with COVID-19. The simplified signs scoring system based on these four indicators may be used to predict patient's risk of severe illness or early death.

Risk Factors and Establishment of Prediction Model for Preoperative Lymph Node Staging Deficiency in Gastric Cancer / 肿瘤防治研究

Objective To analyze the risk factors of preoperative lymph node staging (N-stage) deficiency in gastric cancer and establish a preoperative assessment model to assist in predicting preoperative N-stage. Methods A retrospective method was used to analyze the clinicopathological data of 268 patients with gastric cancer. The patients routinely underwent preoperative thin-section enhanced CT to assess preoperative N-stage. Results The risk factors for preoperative N-stage deficiency were analyzed in combination with postoperative pathological findings. Multifactorial logistic regression analysis was performed to determine influencing factors, and Kaplan-Meier analysis was used to plot the survival curves of preoperative N-stage accurate group and deficiency group. The nomogram plot and ROC curves of the prediction model were drawn using the R package. AUC, 95%CI, sensitivity, and specificity were calculated. Results Age, BMI, poor differentiation, and Lauren's classification as diffuse were independent risk factors for preoperative N-stage deficiency in gastric cancer (P < 0.05). Prognostic survival was significantly worse in the preoperative N stage-inadequate group than that in the accurate group (P=0.041). The AUC area was 0.935, with a sensitivity of 85.9% and specificity of 96.9%. Conclusion Young age, high BMI, poor differentiation, and Lauren's classification as diffuse are independent risk factors for preoperative N-stage deficiency. The established preoperative assessment model based on age, BMI, differentiation degree, and Lauren's classification in this study has relatively high credibility.

Prediction Model of Treatment Effect of Anlotinib on Extensive-stage Small Cell Lung Cancer Based on Combination of Disease and Syndrome Information / 肿瘤防治研究

Objective To construct a nomogram prediction model for the treatment effect of anlotinib with the participation of traditional Chinese medicine syndrome elements on the patients with extensive-stage small cell lung cancer (ES-SCLC) who previously received multiple lines of chemotherapy. Methods The clinical data of 127 patients with ES-SCLC who received at least two cycles of anlotinib treatment were retrospectively studied. Kaplan-Meier method was used to analyze the relationship between each factor and the overall survival time. Cox regression analysis was applied to screen the independent influencing factors of the prognosis of patients with ES-SCLC. R language was employed to build a nomogram prediction model, C-index was used to evaluate the model, and calibration curve was adopted to verify the accuracy of the model. Results Age, PS score, brain metastases, qi deficiency syndrome, yin deficiency syndrome, and blood stasis syndrome were related risk factors for ES-SCLC treated with anlotinib. PS score, brain metastasis, and blood stasis syndrome were independent prognostic factors. On the basis of these three independent influencing factors, a nomogram model was established to predict the prognosis of patients with ES-SCLC treated with anlotinib. The predicted risk was close to the actual risk, showing a high degree of coincidence. Conclusion The nomogram model established with PS score, blood stasis syndrome elements, and brain metastasis as independent factors can predict the prognosis of patients with ES-SCLC receiving second- and third-line treatment of anlotinib.

Construction of predictive model for pregnancy-related anxiety risk in pregnant women based on LASSO regression / 环境与职业医学

Background Pregnancy-related anxiety has a negative impact on the physical and mental health of pregnant women and the normal growth and development of the fetus. Establishing prediction models for pregnancy-related anxiety to screen associated predictive factors may provide important opportunities for prenatal intervention. Objective To establish a prediction model of pregnancy-related anxiety risk of pregnant women. Methods From January to July 2021, a questionnaire survey on pregnancy-related anxiety and predictors was conducted among pregnant women having routine prenatal check-ups provided by an obstetrics clinic of a tertiary grade A hospital in Ningxia. The socio-demographic characteristics of the subjects were collected, and the pregnant women were evaluated by the Life Event Scale (LES), Perceived Social Support Scale (PSSS), Family APGAR Index (APGAR), and Pregnancy-related Anxiety Questionnaire (PAQ). R 4.2.0 software was used to fit all selected variables by least absolute shrinkage and selection operator (LASSO) regression to identify predictors of pregnancy-related anxiety in the second and third trimesters. On the basis of logistic regression analysis, prediction models of pregnancy-related anxiety in the second and third trimesters were constructed, and the model nomogram and receiver operating characteristic curve (ROC) were drawn. The prediction effect of the model was evaluated by area under the curve (AUC). A calibration chart was drawn to evaluate the calibration of the model. Results A total of 1500 questionnaires were distributed, and 1448 valid questionnaires were recovered, with an effective rate of 96.53%. Among the 1448 pregnant women, the overall positive rate of pregnancy-related anxiety was 28.80% (417/1448), and the positive rates in the second and third trimesters were 29.21% (276/935) and 27.49% (141/513), respectively. The predictors entering the the second trimester model were age of marriage, family care, social support, family expectations for the fetus, physical condition during pregnancy, and whether experiencing life stressful events during pregnancy. The predictors entering the the third trimester model were pregnancy intention, physical discomfort, and whether experiencing life stress during pregnancy. A risk prediction model of pregnancy-related anxiety for the second trimester was established: risk of pregnancy-related anxiety=−0.07× marriage age +0.12× family care −0.03× social support −0.65× family expectation of fetal sex +0.42× physical condition during pregnancy +0.47× whether experiencing life stressful events during pregnancy. A risk prediction model of pregnancy-related anxiety for the third trimester was established: risk of pregnancy-related anxiety=−5.69+0.82× pregnancy intention +1.06× physical discomfort +0.94× whether experiencing life stressful events during pregnancy. The ROC curves of the two models were drawn. The AUC of the second trimester model was 0.71, and the AUC of related validation model was 0.68. The AUC of the third trimester model was 0.72, and the AUC of related validation model was 0.66. Conclusion The risk prediction models of pregnancy-related anxiety constructed based on LASSO regression and logistic regression have good prediction ability, and they suggest that pregnant women in the second trimester with short marriage age, high family care, low social support, family expectations for fetal sex, average physical condition, and experiencing life stress during pregnancy, and pregnant women in the third trimester with spontaneous pregnant intention, unintended pregnancy, physical discomfort, and experiencing life stress during pregnancy are high-risk groups for pregnancy-related anxiety.

Construction of a predictive model for performing bronchoalveolar lavage in children with Mycoplasma pneumoniae pneumonia and pulmonary consolidation / 中国当代儿科杂志

OBJECTIVES@#To investigate the risk factors for performing bronchoalveolar lavage (BAL) in children with Mycoplasma pneumoniae pneumonia (MPP) and pulmonary consolidation, and to construct a predictive model for performing BAL in these children.@*METHODS@#A retrospective analysis was performed for the clinical data of 202 children with MPP who were hospitalized in the Department of Pediatrics, Changzhou No. 2 People's Hospital Affiliated to Nanjing Medical University, from August 2019 to September 2022. According to whether BAL was performed, they were divided into BAL group with 100 children and non-BAL group with 102 children. A multivariate logistic regression analysis was used to identify the risk factors for performing BAL in MPP children with pulmonary consolidation. Rstudio software (R4.2.3) was used to establish a predictive model for performing BAL, and the receiver operator characteristic (ROC) curve, C-index, and calibration curve were used to assess the predictive performance of the model.@*RESULTS@#The multivariate logistic regression analysis demonstrated that the fever duration, C-reactive protein levels, D-dimer levels, and presence of pleural effusion were risk factors for performing BAL in MPP children with pulmonary consolidation (P<0.05). A nomogram predictive model was established based on the results of the multivariate logistic regression analysis. In the training set, this model had an area under the ROC curve of 0.915 (95%CI: 0.827-0.938), with a sensitivity of 0.826 and a specificity of 0.875, while in the validation set, it had an area under the ROC curve of 0.983 (95%CI: 0.912-0.996), with a sensitivity of 0.879 and a specificity of 1.000. The Bootstrap-corrected C-index was 0.952 (95%CI: 0.901-0.986), and the calibration curve demonstrated good consistency between the predicted probability of the model and the actual probability of occurrence.@*CONCLUSIONS@#The predictive model established in this study can be used to assess the likelihood of performing BAL in MPP children with pulmonary consolidation, based on factors such as fever duration, C-reactive protein levels, D-dimer levels, and the presence of pleural effusion. Additionally, the model demonstrates good predictive performance.