Healthcare analytics (New York, N.Y.)最新文献_第7页

A clustering-based federated deep learning approach for enhancing diabetes management with privacy-preserving edge artificial intelligence 一种基于聚类的联合深度学习方法，用于增强糖尿病管理与隐私保护边缘人工智能

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-04-01 DOI: 10.1016/j.health.2025.100392

Xinyi Yang, Juan Li

The increasing prevalence of diabetes necessitates innovative glucose prediction methods that prioritize patient privacy. While edge artificial intelligence (AI) offers potential, its limitations in resource-constrained devices can be mitigated through federated learning (FL). However, challenges remain in accounting for patient variability and optimizing FL for glucose prediction. This research introduces a novel personalized clustering-based federated deep learning (Clu-FDL) model to address these challenges. We develop tailored models that enhance prediction accuracy by clustering patients based on carbohydrate (CHO) intake patterns. Utilizing Simple Recurrent Neural Network (SimpleRNN) and Gated Recurrent Unit (GRU) methods, the study evaluates the performance of local patients who contribute to training the cluster and global (non-cluster) models. The results show that the Clu-FDL approach achieves high precision (0.93), recall (0.96), and F1 scores (0.95), along with low Root Mean Square Error (RMSE) values (11.08 ± 1.77 mg/dL). Additionally, for new patients with different data durations, analysis based on 0.25–3 days of data indicates that Clu-FDL models exhibit greater stability, with smaller RMSE and higher precision, recall, and F1 scores compared to non-clustering models. The study identifies that SimpleRNN and GRU models are most effective for new patients with 9 and 6 days of data. This privacy-preserving, clustering-based personalized approach empowers patients to manage their diabetes effectively.

糖尿病患病率的增加需要创新的血糖预测方法，优先考虑患者隐私。虽然边缘人工智能（AI）提供了潜力，但它在资源受限设备中的局限性可以通过联邦学习（FL）来缓解。然而，在考虑患者的可变性和优化血糖预测的FL方面仍然存在挑战。本研究引入了一种新的基于个性化聚类的联邦深度学习（clul - fdl）模型来解决这些挑战。我们开发了量身定制的模型，通过基于碳水化合物（CHO）摄入模式的患者聚类来提高预测准确性。利用简单递归神经网络（SimpleRNN）和门控递归单元（GRU）方法，该研究评估了有助于训练聚类和全局（非聚类）模型的局部患者的表现。结果表明，该方法具有较高的精密度（0.93）、召回率（0.96）和F1评分（0.95），均方根误差（RMSE）值（11.08±1.77 mg/dL）低。此外，对于不同数据持续时间的新患者，基于0.25-3天数据的分析表明，与非聚类模型相比，clul - fdl模型具有更大的稳定性，RMSE更小，精度、召回率和F1分数更高。该研究确定SimpleRNN和GRU模型对9天和6天的新患者最有效。这种保护隐私、基于聚类的个性化方法使患者能够有效地管理他们的糖尿病。

{"title":"A clustering-based federated deep learning approach for enhancing diabetes management with privacy-preserving edge artificial intelligence","authors":"Xinyi Yang, Juan Li","doi":"10.1016/j.health.2025.100392","DOIUrl":"10.1016/j.health.2025.100392","url":null,"abstract":"<div><div>The increasing prevalence of diabetes necessitates innovative glucose prediction methods that prioritize patient privacy. While edge artificial intelligence (AI) offers potential, its limitations in resource-constrained devices can be mitigated through federated learning (FL). However, challenges remain in accounting for patient variability and optimizing FL for glucose prediction. This research introduces a novel personalized clustering-based federated deep learning (Clu-FDL) model to address these challenges. We develop tailored models that enhance prediction accuracy by clustering patients based on carbohydrate (CHO) intake patterns. Utilizing Simple Recurrent Neural Network (SimpleRNN) and Gated Recurrent Unit (GRU) methods, the study evaluates the performance of local patients who contribute to training the cluster and global (non-cluster) models. The results show that the Clu-FDL approach achieves high precision (0.93), recall (0.96), and F1 scores (0.95), along with low Root Mean Square Error (RMSE) values (11.08 ± 1.77 mg/dL). Additionally, for new patients with different data durations, analysis based on 0.25–3 days of data indicates that Clu-FDL models exhibit greater stability, with smaller RMSE and higher precision, recall, and F1 scores compared to non-clustering models. The study identifies that SimpleRNN and GRU models are most effective for new patients with 9 and 6 days of data. This privacy-preserving, clustering-based personalized approach empowers patients to manage their diabetes effectively.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100392"},"PeriodicalIF":0.0,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143760594","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A comparative study of explainable machine learning models with Shapley values for diabetes prediction 可解释机器学习模型与Shapley值用于糖尿病预测的比较研究

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-03-11 DOI: 10.1016/j.health.2025.100390

Keona Pang

Over the years, numerous machine learning models have been developed, leading to successful applications across various fields. This study uses a large dataset related to type 2 diabetes prediction from the Centers for Disease Control and Prevention (CDC) in the United States. The dataset with 70692 samples has 21 input features and one output (non-diabetes or diabetes). In addition to health indicators like Body Mass Index (BMI), blood pressure, and cholesterol level, the features include socioeconomic factors (e.g., income, education) and lifestyle factors such as diet and physical activity. This paper aims to study how these features influence diabetes risk. 80 % of the dataset is used for training and 20 % for testing. Six machine learning models, as well as the Multivariate Adaptive Regression Splines (MARS) model, were used in the investigation. A detailed comparison of the performance of these models is given. Shapley values explain the nature of various machine learning models using visualization by color graphs to demonstrate the reliability of different machine learning models. This paper shows how Shapley values can improve their explainability and interpretability on diabetes prediction. We leverage the SHapley Additive exPlanations (SHAP) scores to provide information about the relative importance of each predictive feature, and these results shed light on the relationship between the features and the risk of developing type 2 diabetes.

多年来，人们开发了许多机器学习模型，并在各个领域取得了成功的应用。本研究使用了来自美国疾病控制与预防中心（CDC）的与2型糖尿病预测相关的大型数据集。具有70692个样本的数据集有21个输入特征和一个输出（非糖尿病或糖尿病）。除了身体质量指数（BMI）、血压和胆固醇水平等健康指标外，这些特征还包括社会经济因素（如收入、教育）和生活方式因素（如饮食和体育活动）。本文旨在研究这些特征如何影响糖尿病风险。80%的数据集用于训练，20%用于测试。研究中使用了六种机器学习模型以及多元自适应回归样条（MARS）模型。对这些模型的性能进行了详细的比较。Shapley值通过彩色图形的可视化来解释各种机器学习模型的本质，以证明不同机器学习模型的可靠性。本文展示了Shapley值如何提高其在糖尿病预测中的可解释性和可解释性。我们利用SHapley加性解释（SHAP）评分来提供关于每个预测特征的相对重要性的信息，这些结果揭示了特征与患2型糖尿病风险之间的关系。

{"title":"A comparative study of explainable machine learning models with Shapley values for diabetes prediction","authors":"Keona Pang","doi":"10.1016/j.health.2025.100390","DOIUrl":"10.1016/j.health.2025.100390","url":null,"abstract":"<div><div>Over the years, numerous machine learning models have been developed, leading to successful applications across various fields. This study uses a large dataset related to type 2 diabetes prediction from the Centers for Disease Control and Prevention (CDC) in the United States. The dataset with 70692 samples has 21 input features and one output (non-diabetes or diabetes). In addition to health indicators like Body Mass Index (BMI), blood pressure, and cholesterol level, the features include socioeconomic factors (e.g., income, education) and lifestyle factors such as diet and physical activity. This paper aims to study how these features influence diabetes risk. 80 % of the dataset is used for training and 20 % for testing. Six machine learning models, as well as the Multivariate Adaptive Regression Splines (MARS) model, were used in the investigation. A detailed comparison of the performance of these models is given. Shapley values explain the nature of various machine learning models using visualization by color graphs to demonstrate the reliability of different machine learning models. This paper shows how Shapley values can improve their explainability and interpretability on diabetes prediction. We leverage the SHapley Additive exPlanations (SHAP) scores to provide information about the relative importance of each predictive feature, and these results shed light on the relationship between the features and the risk of developing type 2 diabetes.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100390"},"PeriodicalIF":0.0,"publicationDate":"2025-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143629245","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A machine learning model for automated contact tracing during disease outbreaks 疾病暴发期间用于自动追踪接触者的机器学习模型

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-03-08 DOI: 10.1016/j.health.2025.100389

Zeyad Aklah , Amean Al-Safi , Marwa H. Abdali , Khalid Al-jabery

This study aims to develop and evaluate a conceptual model for assessing the Risk of Infection (ROI) within the context of automated digital contact tracing during pandemics. The proposed model incorporates five input parameters: distance, overlap time, contamination interval, incubation time, and contact facility size. These parameters capture various aspects of disease transmission dynamics. The model employs logistic functions to quantify the influence of each parameter on the overall ROI. The evaluation of the model involves two methods: a partial evaluation to observe the impact of parameter pairs on ROI, and a full evaluation, which is trained on a dataset of 24,000 simulated scenarios to identify central clusters for high, medium, and low-risk categories using K-means and the Hidden Markov Model. Additionally, the model is tested on another 16,000 simulated scenarios to assess its overall performance. Results indicate that the Hidden Markov Model categorizes 63.8% of the testing dataset as low risk, 20.7% as medium risk, and 15.5% as high risk. In contrast, K-means classifies 44.3% as low risk, 30.7% as medium risk, and 25% as high risk. The evaluation metrics favor the Hidden Markov Model, which demonstrates higher performance in terms of Log-Likelihood, with a value of 50,688, as well as in the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), with values of -101,365.6430 and -101,319.5609, respectively. In both evaluations, the results validate the model’s ability to automate digital contact tracing based on the input parameters. Future studies could explore classification accuracy using real contact tracing datasets. The proposed approach enhances the efficiency of public health authorities by directing their efforts toward individuals with the highest risk of infection, rather than applying the same level of intervention indiscriminately to everyone.

本研究旨在开发和评估在大流行期间自动数字接触者追踪背景下评估感染风险（ROI）的概念模型。提出的模型包含五个输入参数：距离、重叠时间、污染间隔、孵化时间和接触设施大小。这些参数反映了疾病传播动力学的各个方面。该模型采用logistic函数来量化各参数对整体ROI的影响。模型的评估包括两种方法：部分评估，观察参数对对ROI的影响；全面评估，在24,000个模拟场景的数据集上进行训练，使用K-means和隐马尔可夫模型识别高、中、低风险类别的中心聚类。此外，该模型还在另外16,000个模拟场景中进行了测试，以评估其整体性能。结果表明，隐马尔可夫模型将测试数据集的63.8%分类为低风险，20.7%为中等风险，15.5%为高风险。相比之下，K-means将44.3%分类为低风险，30.7%为中风险，25%为高风险。评价指标倾向于隐马尔可夫模型，它在对数似然方面表现出更高的性能，其值为50,688，而赤池信息准则（AIC）和贝叶斯信息准则（BIC）的值分别为-101,365.6430和-101,319.5609。在这两个评估中，结果验证了模型基于输入参数自动数字接触跟踪的能力。未来的研究可以利用真实的接触追踪数据集来探索分类的准确性。拟议的方法提高了公共卫生当局的效率，将工作重点放在感染风险最高的个人身上，而不是不分青红皂白地对所有人实施同样水平的干预。

{"title":"A machine learning model for automated contact tracing during disease outbreaks","authors":"Zeyad Aklah , Amean Al-Safi , Marwa H. Abdali , Khalid Al-jabery","doi":"10.1016/j.health.2025.100389","DOIUrl":"10.1016/j.health.2025.100389","url":null,"abstract":"<div><div>This study aims to develop and evaluate a conceptual model for assessing the Risk of Infection (ROI) within the context of automated digital contact tracing during pandemics. The proposed model incorporates five input parameters: distance, overlap time, contamination interval, incubation time, and contact facility size. These parameters capture various aspects of disease transmission dynamics. The model employs logistic functions to quantify the influence of each parameter on the overall ROI. The evaluation of the model involves two methods: a partial evaluation to observe the impact of parameter pairs on ROI, and a full evaluation, which is trained on a dataset of 24,000 simulated scenarios to identify central clusters for high, medium, and low-risk categories using K-means and the Hidden Markov Model. Additionally, the model is tested on another 16,000 simulated scenarios to assess its overall performance. Results indicate that the Hidden Markov Model categorizes 63.8% of the testing dataset as low risk, 20.7% as medium risk, and 15.5% as high risk. In contrast, K-means classifies 44.3% as low risk, 30.7% as medium risk, and 25% as high risk. The evaluation metrics favor the Hidden Markov Model, which demonstrates higher performance in terms of Log-Likelihood, with a value of 50,688, as well as in the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), with values of -101,365.6430 and -101,319.5609, respectively. In both evaluations, the results validate the model’s ability to automate digital contact tracing based on the input parameters. Future studies could explore classification accuracy using real contact tracing datasets. The proposed approach enhances the efficiency of public health authorities by directing their efforts toward individuals with the highest risk of infection, rather than applying the same level of intervention indiscriminately to everyone.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100389"},"PeriodicalIF":0.0,"publicationDate":"2025-03-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143619509","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A machine learning and neural network approach for classifying multidrug-resistant bacterial infections 多药耐药细菌感染分类的机器学习和神经网络方法

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-02-22 DOI: 10.1016/j.health.2025.100388

Preeda Mengsiri , Ratchadaporn Ungcharoen , Sethavidh Gertphol

Antimicrobial resistance (AMR) represents a major public health challenge, significantly complicating infection prevention and treatment. This study employs machine learning and neural network techniques to classify multidrug-resistant Gram-negative bacterial (MDR-GNB) infections using electronic health records from 624 patients at Thatphanom Crown Prince Hospital in Thailand. We compared several algorithms, including Logistic Regression, Random Forest, Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN), Multilayer Perceptron (MLP), and Light Gradient Boosting Machine (LightGBM), with the MLP model exhibiting the highest accuracy and specificity. Performance was further enhanced by integrating feature selection methods such as Sequential Forward Selection (SFS), Recursive Feature Elimination with Cross-Validation (RFE-CV), and SelectKBest with data augmentation techniques, including ADASYN and SMOTE variants. Utilizing SHapley Additive exPlanations (SHAP) provided valuable insights into the most influential predictors for MDR-GNB. Notably, the MLP model achieved an AUC of 0.70, surpassing prior studies and highlighting its potential to advance clinical decision-making in managing MDR-GNB infections.

抗菌素耐药性（AMR）是一项重大的公共卫生挑战，使感染预防和治疗严重复杂化。本研究采用机器学习和神经网络技术，利用泰国Thatphanom王储医院624名患者的电子健康记录，对耐多药革兰氏阴性细菌（MDR-GNB）感染进行分类。我们比较了几种算法，包括逻辑回归、随机森林、支持向量机（SVM）、极端梯度增强（XGBoost）、k近邻（KNN）、多层感知器（MLP）和光梯度增强机（LightGBM），其中MLP模型具有最高的准确性和特异性。通过将特征选择方法(如顺序前向选择（SFS）、递归特征消除与交叉验证（RFE-CV）、SelectKBest与数据增强技术（包括ADASYN和SMOTE变体）集成在一起，性能得到了进一步提高。利用SHapley加性解释（SHAP）为耐多药- gnb最有影响力的预测因子提供了有价值的见解。值得注意的是，MLP模型的AUC达到了0.70，超过了先前的研究，并突出了其在管理耐多药gnb感染方面推进临床决策的潜力。

{"title":"A machine learning and neural network approach for classifying multidrug-resistant bacterial infections","authors":"Preeda Mengsiri , Ratchadaporn Ungcharoen , Sethavidh Gertphol","doi":"10.1016/j.health.2025.100388","DOIUrl":"10.1016/j.health.2025.100388","url":null,"abstract":"<div><div>Antimicrobial resistance (AMR) represents a major public health challenge, significantly complicating infection prevention and treatment. This study employs machine learning and neural network techniques to classify multidrug-resistant Gram-negative bacterial (MDR-GNB) infections using electronic health records from 624 patients at Thatphanom Crown Prince Hospital in Thailand. We compared several algorithms, including Logistic Regression, Random Forest, Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN), Multilayer Perceptron (MLP), and Light Gradient Boosting Machine (LightGBM), with the MLP model exhibiting the highest accuracy and specificity. Performance was further enhanced by integrating feature selection methods such as Sequential Forward Selection (SFS), Recursive Feature Elimination with Cross-Validation (RFE-CV), and SelectKBest with data augmentation techniques, including ADASYN and SMOTE variants. Utilizing SHapley Additive exPlanations (SHAP) provided valuable insights into the most influential predictors for MDR-GNB. Notably, the MLP model achieved an AUC of 0.70, surpassing prior studies and highlighting its potential to advance clinical decision-making in managing MDR-GNB infections.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100388"},"PeriodicalIF":0.0,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143510183","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An exploration of the interplay between treatment and vaccination in an Age-Structured Malaria Model using non-linear ordinary differential equations 利用非线性常微分方程探索年龄结构疟疾模型中治疗和疫苗接种之间的相互作用

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-02-22 DOI: 10.1016/j.health.2025.100386

Mahmudul Bari Hridoy, Angela Peace

Malaria continues to be a significant global health challenge, particularly in tropical regions. Resistance to key antimalarial drugs is spreading, complicating treatment efforts. While progress toward eradication has been slow, the development and introduction of novel malaria vaccines offer hope for reducing the disease burden in endemic areas. To address these challenges, we develop an extended Susceptible–Exposed–Infected–Recovered (SEIR) age-structured model incorporating malaria vaccination for children, drug-sensitive and drug-resistant strains, and interactions between human hosts and mosquitoes. Our research evaluates how malaria vaccination coverage influences disease prevalence and transmission dynamics. We derive both strains’ basic, intervention, and invasion reproduction numbers and conduct sensitivity analysis to identify key parameters affecting infection prevalence. Our findings reveal that model outcomes are primarily influenced by scale factors that reduce transmission and natural recovery rates for the resistant strain, as well as by drug treatment and vaccination efficacies and mosquito death rates. Numerical simulations indicate that while treatment reduces the malaria disease burden, it also increases the proportion of drug-resistant cases. Conversely, higher vaccination efficacy correlates with lower infection cases for both strains. These results suggest that a synergistic approach involving vaccination and treatment could effectively decrease the overall proportion of the infected population.

疟疾仍然是一个重大的全球卫生挑战，特别是在热带地区。对主要抗疟疾药物的耐药性正在蔓延，使治疗工作复杂化。虽然在消灭疟疾方面进展缓慢，但开发和采用新型疟疾疫苗为减轻流行地区的疾病负担带来了希望。为了应对这些挑战，我们开发了一个扩展的易感-暴露-感染-恢复（SEIR）年龄结构模型，将儿童疟疾疫苗接种、药物敏感和耐药菌株以及人类宿主和蚊子之间的相互作用纳入其中。我们的研究评估了疟疾疫苗接种覆盖率如何影响疾病流行和传播动态。我们得出了菌株的基本、干预和入侵繁殖数，并进行了敏感性分析，以确定影响感染流行的关键参数。我们的研究结果表明，模型结果主要受降低耐药菌株传播和自然恢复率的规模因素、药物治疗和疫苗接种效果以及蚊子死亡率的影响。数值模拟表明，虽然治疗减轻了疟疾疾病负担，但也增加了耐药病例的比例。相反，较高的疫苗接种效力与两种菌株的较低感染病例相关。这些结果表明，涉及疫苗接种和治疗的协同方法可以有效地降低感染人口的总体比例。

{"title":"An exploration of the interplay between treatment and vaccination in an Age-Structured Malaria Model using non-linear ordinary differential equations","authors":"Mahmudul Bari Hridoy, Angela Peace","doi":"10.1016/j.health.2025.100386","DOIUrl":"10.1016/j.health.2025.100386","url":null,"abstract":"<div><div>Malaria continues to be a significant global health challenge, particularly in tropical regions. Resistance to key antimalarial drugs is spreading, complicating treatment efforts. While progress toward eradication has been slow, the development and introduction of novel malaria vaccines offer hope for reducing the disease burden in endemic areas. To address these challenges, we develop an extended Susceptible–Exposed–Infected–Recovered (SEIR) age-structured model incorporating malaria vaccination for children, drug-sensitive and drug-resistant strains, and interactions between human hosts and mosquitoes. Our research evaluates how malaria vaccination coverage influences disease prevalence and transmission dynamics. We derive both strains’ basic, intervention, and invasion reproduction numbers and conduct sensitivity analysis to identify key parameters affecting infection prevalence. Our findings reveal that model outcomes are primarily influenced by scale factors that reduce transmission and natural recovery rates for the resistant strain, as well as by drug treatment and vaccination efficacies and mosquito death rates. Numerical simulations indicate that while treatment reduces the malaria disease burden, it also increases the proportion of drug-resistant cases. Conversely, higher vaccination efficacy correlates with lower infection cases for both strains. These results suggest that a synergistic approach involving vaccination and treatment could effectively decrease the overall proportion of the infected population.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100386"},"PeriodicalIF":0.0,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143480475","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A data-driven approach to pricing models for balanced public–private healthcare systems 以数据为导向的公私平衡医疗保健系统定价模式

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-02-17 DOI: 10.1016/j.health.2025.100385

Aydin Teymourifar , Onur Kaya , Gurkan Ozturk

This study focuses on a real-world healthcare system with coexisting public and private hospitals with distinct characteristics. While public hospitals have lower costs, they also suffer from long waiting times and diminishing patients’ perceived quality of care. Conversely, despite their higher fees, private hospitals offer shorter waiting times, leading to a more favorable perception of quality. A balanced healthcare system could provide societal benefits. Pricing strategies greatly influence a patient’s hospital selection. For instance, reduced fees in private hospitals attract more patients, consequently reducing overcrowding in public facilities and elevating the overall quality of services provided. This study aims to develop pricing models to foster a balanced and socially advantageous healthcare system. This system determines private hospital pricing through contract mechanisms with the government. Thus, we delve into the ramifications of various contract models between the government and private hospitals on social utility. Our findings underscore the communal advantages of contract mechanisms. Furthermore, we generalize the proposed models to apply to similar systems.

本研究以现实世界中公立医院与私立医院并存、各具特色的医疗体系为研究对象。虽然公立医院的成本较低，但它们也面临着等待时间过长和患者对护理质量的感知下降的问题。相反，尽管收费较高，但私立医院的等待时间较短，因此对质量的看法更有利。一个平衡的医疗体系可以带来社会效益。定价策略极大地影响了患者对医院的选择。例如，私立医院收费的降低吸引了更多的病人，从而减少了公共设施的拥挤，提高了所提供服务的总体质量。本研究旨在发展定价模式，以建立一个平衡且对社会有利的医疗保健系统。这一制度通过与政府的合同机制来决定民营医院的价格。因此，我们深入研究了政府与民营医院之间的各种合同模式对社会效用的影响。我们的研究结果强调了契约机制的共同优势。此外，我们将所提出的模型推广到类似的系统。

{"title":"A data-driven approach to pricing models for balanced public–private healthcare systems","authors":"Aydin Teymourifar , Onur Kaya , Gurkan Ozturk","doi":"10.1016/j.health.2025.100385","DOIUrl":"10.1016/j.health.2025.100385","url":null,"abstract":"<div><div>This study focuses on a real-world healthcare system with coexisting public and private hospitals with distinct characteristics. While public hospitals have lower costs, they also suffer from long waiting times and diminishing patients’ perceived quality of care. Conversely, despite their higher fees, private hospitals offer shorter waiting times, leading to a more favorable perception of quality. A balanced healthcare system could provide societal benefits. Pricing strategies greatly influence a patient’s hospital selection. For instance, reduced fees in private hospitals attract more patients, consequently reducing overcrowding in public facilities and elevating the overall quality of services provided. This study aims to develop pricing models to foster a balanced and socially advantageous healthcare system. This system determines private hospital pricing through contract mechanisms with the government. Thus, we delve into the ramifications of various contract models between the government and private hospitals on social utility. Our findings underscore the communal advantages of contract mechanisms. Furthermore, we generalize the proposed models to apply to similar systems.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100385"},"PeriodicalIF":0.0,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143430002","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An interpretable machine learning study for developing a binary classifier for predicting rehospitalization from skilled nursing facilities 一项可解释的机器学习研究，用于开发用于预测熟练护理机构再住院的二元分类器

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-02-15 DOI: 10.1016/j.health.2025.100387

Zhouyang Lou , Zachary Hass , Nan Kong

Reducing hospital readmissions for older adults discharged to a skilled nursing facility (SNF) is important to the Unites States (U.S.) both from financial and care quality perspectives. To identify potential risk factors, researchers have used data from claims, national surveys, and administrative databases to train models that predict hospital readmissions that occur within 30 days of discharge. Machine learning techniques hold promise for this binary classification task. However, analysis pipelines are underdeveloped in data balancing, feature selection, and model interpretability. In this paper, we utilized individual resident-level data from the Long-Term Care Minimum Data Set (MDS) collected from SNFs in a midwestern U.S. state (n = 93,058). We further triangulated this data with publicly available facility quality and staffing data from the Nursing Home Compares tool of the Medicare.gov and facility neighborhood data from the National Neighborhood Data Archive. We compared several machine learning models, data balancing techniques, and feature selection methods, for the prediction task. We found that XGBoost, with Synthetic Minority Oversampling Edited Nearest Neighbor (SMOTE-ENN) to balance the data, and hierarchical clustering based on spearman correlation to select the features that produces the best prediction performance. We then used SHapley Additive exPlanations (SHAP) values to identify features that contribute most to the performance and used partial dependence plots to examine curvilinear and moderating relationships between features and the risk of 30-day rehospitalization.

从财务和护理质量的角度来看，减少老年人出院到专业护理机构（SNF）的再入院率对美国（U.S.）都很重要。为了确定潜在的风险因素，研究人员使用来自索赔、国家调查和行政数据库的数据来训练模型，预测出院后30天内再次住院的情况。机器学习技术为这种二元分类任务带来了希望。然而，分析管道在数据平衡、特征选择和模型可解释性方面还不发达。在本文中，我们利用了从美国中西部一个州的snf收集的长期护理最低数据集（MDS）中的个体居民水平数据（n = 93,058）。我们进一步将这些数据与Medicare.gov的养老院比较工具中公开的设施质量和人员配备数据以及国家社区数据档案中的设施社区数据进行三角分析。为了预测任务，我们比较了几种机器学习模型、数据平衡技术和特征选择方法。我们发现XGBoost使用合成少数派过采样编辑最近邻（SMOTE-ENN）来平衡数据，并基于spearman相关的分层聚类来选择产生最佳预测性能的特征。然后，我们使用SHapley加性解释（SHAP）值来识别对表现贡献最大的特征，并使用部分依赖图来检验特征与30天再住院风险之间的曲线关系和调节关系。

{"title":"An interpretable machine learning study for developing a binary classifier for predicting rehospitalization from skilled nursing facilities","authors":"Zhouyang Lou , Zachary Hass , Nan Kong","doi":"10.1016/j.health.2025.100387","DOIUrl":"10.1016/j.health.2025.100387","url":null,"abstract":"<div><div>Reducing hospital readmissions for older adults discharged to a skilled nursing facility (SNF) is important to the Unites States (U.S.) both from financial and care quality perspectives. To identify potential risk factors, researchers have used data from claims, national surveys, and administrative databases to train models that predict hospital readmissions that occur within 30 days of discharge. Machine learning techniques hold promise for this binary classification task. However, analysis pipelines are underdeveloped in data balancing, feature selection, and model interpretability. In this paper, we utilized individual resident-level data from the Long-Term Care Minimum Data Set (MDS) collected from SNFs in a midwestern U.S. state (n = 93,058). We further triangulated this data with publicly available facility quality and staffing data from the Nursing Home Compares tool of the Medicare.gov and facility neighborhood data from the National Neighborhood Data Archive. We compared several machine learning models, data balancing techniques, and feature selection methods, for the prediction task. We found that XGBoost, with Synthetic Minority Oversampling Edited Nearest Neighbor (SMOTE-ENN) to balance the data, and hierarchical clustering based on spearman correlation to select the features that produces the best prediction performance. We then used SHapley Additive exPlanations (SHAP) values to identify features that contribute most to the performance and used partial dependence plots to examine curvilinear and moderating relationships between features and the risk of 30-day rehospitalization.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100387"},"PeriodicalIF":0.0,"publicationDate":"2025-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143480474","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A recommender system with multi-objective hybrid Harris Hawk optimization for feature selection and disease diagnosis 基于多目标混合Harris Hawk优化的特征选择和疾病诊断推荐系统

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-01-31 DOI: 10.1016/j.health.2025.100384

Madhusree Kuanr, Puspanjali Mohapatra

This study proposes a health recommender system to analyze health risk and disease prediction by identifying the most responsible disease-causing factors using a hybrid Genetic–Harris Hawk optimization multi-objective feature selection approach. The proposed recommender system uses the Tree-based Pipeline Optimization Tool (TPOT) automated machine learning model to recommend the most suitable machine learning prediction model with the best classifier in terms of classification accuracy for a disease with the selected features. It also recommends the top three disease-causing features for a particular disease that can be utilized to analyze a person’s health risk. The proposed system has also been compared with the competing prediction approaches using Principal Component Analysis (PCA), Singular Vector Decomposition (SVD), and Autoencoders. We show that the proposed system outperforms competing approaches in terms of classification accuracy.

本研究提出了一个健康推荐系统，通过混合遗传-哈里斯鹰优化多目标特征选择方法识别最主要的致病因素，分析健康风险和疾病预测。提出的推荐系统使用基于树的管道优化工具（TPOT）自动化机器学习模型，根据所选特征的分类精度，推荐最适合的机器学习预测模型和最佳分类器。它还推荐了一种特定疾病的三大致病特征，可以用来分析一个人的健康风险。该系统还与使用主成分分析（PCA）、奇异向量分解（SVD）和自编码器的竞争预测方法进行了比较。我们表明，该系统在分类精度方面优于竞争方法。

引用次数: 0

An application of natural language processing for hypoglycemic event identification in patients with diabetes mellitus 自然语言处理在糖尿病患者低血糖事件识别中的应用

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-01-21 DOI: 10.1016/j.health.2024.100381

J.E. Camacho-Cogollo , Cristhian Felipe Patiño Zambrano , Christian Lochmuller , Claudia C. Colmenares-Mejia , Nicolas Rozo , Mario A. Isaza-Ruget , Paul Rodriguez , Andrés García

The therapeutic goal for diabetes mellitus is to maintain normal blood glucose levels, but in some cases, hypoglycemia may occur as a consequence of treatment. Identifying patients with hypoglycemia is critical to preventing adverse events and mortality. However, hypoglycemic events are often not accurately documented in electronic health records (EHRs). This study presents a retrospective analysis of the EHRs of patients with diabetes mellitus. We hypothesize that text analytics and machine learning can identify possible hypoglycemic incidents from unstructured physician notes in electronic health records. Our analysis applies these techniques using the Python programming language as a tool. It also considers words that describe symptoms related to hypoglycemia. The analysis involves searching physicians' notes for keywords and applying supervised classification methods to 146,542 records. Natural language processing (NLP) and machine learning algorithms are used to identify possible hypoglycemic events and related symptoms in physicians’ notes. A multi-layer perceptron (MLP) model produces the best classification performance among all the models tested in this study, with an obtained accuracy of 0.87. We show that the NLP approach can effectively identify and automate the text-based detection process of potential hypoglycemic events, and can subsequently be used to make informed decisions about potential patient risks.

糖尿病的治疗目标是维持正常的血糖水平，但在某些情况下，治疗后可能出现低血糖。识别低血糖患者对于预防不良事件和死亡率至关重要。然而，低血糖事件通常不能准确地记录在电子健康记录（EHRs）中。本研究对糖尿病患者的电子病历进行回顾性分析。我们假设文本分析和机器学习可以从电子健康记录中的非结构化医生笔记中识别可能的低血糖事件。我们的分析使用Python编程语言作为工具来应用这些技术。它还考虑描述与低血糖相关症状的单词。该分析包括搜索医生笔记中的关键词，并对146,542条记录应用监督分类方法。自然语言处理（NLP）和机器学习算法用于识别医生记录中可能的低血糖事件和相关症状。在本研究测试的所有模型中，多层感知器（MLP）模型的分类性能最好，获得的准确率为0.87。我们表明，NLP方法可以有效地识别和自动化基于文本的潜在低血糖事件检测过程，并随后可用于对潜在的患者风险做出明智的决定。

{"title":"An application of natural language processing for hypoglycemic event identification in patients with diabetes mellitus","authors":"J.E. Camacho-Cogollo , Cristhian Felipe Patiño Zambrano , Christian Lochmuller , Claudia C. Colmenares-Mejia , Nicolas Rozo , Mario A. Isaza-Ruget , Paul Rodriguez , Andrés García","doi":"10.1016/j.health.2024.100381","DOIUrl":"10.1016/j.health.2024.100381","url":null,"abstract":"<div><div>The therapeutic goal for diabetes mellitus is to maintain normal blood glucose levels, but in some cases, hypoglycemia may occur as a consequence of treatment. Identifying patients with hypoglycemia is critical to preventing adverse events and mortality. However, hypoglycemic events are often not accurately documented in electronic health records (EHRs). This study presents a retrospective analysis of the EHRs of patients with diabetes mellitus. We hypothesize that text analytics and machine learning can identify possible hypoglycemic incidents from unstructured physician notes in electronic health records. Our analysis applies these techniques using the Python programming language as a tool. It also considers words that describe symptoms related to hypoglycemia. The analysis involves searching physicians' notes for keywords and applying supervised classification methods to 146,542 records. Natural language processing (NLP) and machine learning algorithms are used to identify possible hypoglycemic events and related symptoms in physicians’ notes. A multi-layer perceptron (MLP) model produces the best classification performance among all the models tested in this study, with an obtained accuracy of 0.87. We show that the NLP approach can effectively identify and automate the text-based detection process of potential hypoglycemic events, and can subsequently be used to make informed decisions about potential patient risks.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100381"},"PeriodicalIF":0.0,"publicationDate":"2025-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143172047","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An automated information extraction model for unstructured discharge letters using large language models and GPT-4 基于大语言模型和GPT-4的非结构化离职信自动信息提取模型

Healthcare analytics (New York, N.Y.)

Pub Date : 2025-01-10 DOI: 10.1016/j.health.2024.100378

Robert M. Siepmann , Giulia Baldini , Cynthia S. Schmidt , Daniel Truhn , Gustav Anton Müller-Franzes , Amin Dada , Jens Kleesiek , Felix Nensa , René Hosch

The administrative burden of manually extracting clinical information from discharge letters is a common challenge in healthcare. This study aims to explore the use of Large Language Models (LLMs), specifically Generative Pretrained Transformer 4 (GPT-4) by OpenAI, for automated extraction of diagnoses, medications, and allergies from discharge letters. Data for this study were sourced from two healthcare institutions in Germany, comprising discharge letters for ten patients from each institution. The first experiment is conducted using a standardized prompt for information extraction. However, challenges were encountered, and the prompt was fine-tuned in a second experiment to improve the results. We further tested whether open-source LLMs can achieve similar results. In the first experiment, primary diagnoses were identified with 85% accuracy and secondary diagnoses with 55.8%. Medications and allergies were extracted with 85.9% and 100% accuracy, respectively. The International Classification of Diseases, 10th revision (ICD-10) codes for the identified diagnoses achieved an accuracy of 85% for primary diagnoses and 60.7% for secondary diagnoses. Anatomical Therapeutic Chemical (ATC) codes were identified with an accuracy of 78.8%. On the other hand, open-source LLMs did not provide similar levels of accuracy and could not consistently fill the template. With prompt fine-tuning in the second experiment, the primary diagnoses, secondary diagnoses, and medications could be predicted with 95%, 88.9%, and 92.2% accuracy, respectively. GPT-4 shows excellent potential for automated extraction of crucial diagnostic and medication information from discharge letters, presumably lowering the administrative burden for healthcare professionals and improving patient outcomes.

手动从出院信中提取临床信息的管理负担是医疗保健领域的一个常见挑战。本研究旨在探索使用大型语言模型（llm），特别是OpenAI的生成预训练转换器4 (GPT-4)，从出院信中自动提取诊断、药物和过敏。本研究的数据来自德国的两家医疗机构，包括每家机构10名患者的出院信。第一个实验是使用标准化提示进行信息提取。然而，遇到了挑战，在第二次实验中对提示进行了微调，以改善结果。我们进一步测试了开源llm是否可以达到类似的结果。在第一次实验中，原发性诊断的准确率为85%，继发性诊断的准确率为55.8%。药物和过敏反应的提取准确率分别为85.9%和100%。国际疾病分类第10版（ICD-10）对已确定诊断的编码，原发性诊断的准确率为85%，继发性诊断的准确率为60.7%。解剖治疗化学（ATC）编码的识别准确率为78.8%。另一方面，开源法学硕士没有提供类似的准确性，也不能始终如一地填充模板。在第二次实验中，通过及时的微调，初步诊断、二次诊断和药物预测的准确率分别为95%、88.9%和92.2%。GPT-4显示了从出院信中自动提取关键诊断和药物信息的巨大潜力，可能会降低医疗保健专业人员的管理负担，并改善患者的治疗效果。

{"title":"An automated information extraction model for unstructured discharge letters using large language models and GPT-4","authors":"Robert M. Siepmann , Giulia Baldini , Cynthia S. Schmidt , Daniel Truhn , Gustav Anton Müller-Franzes , Amin Dada , Jens Kleesiek , Felix Nensa , René Hosch","doi":"10.1016/j.health.2024.100378","DOIUrl":"10.1016/j.health.2024.100378","url":null,"abstract":"<div><div>The administrative burden of manually extracting clinical information from discharge letters is a common challenge in healthcare. This study aims to explore the use of Large Language Models (LLMs), specifically Generative Pretrained Transformer 4 (GPT-4) by OpenAI, for automated extraction of diagnoses, medications, and allergies from discharge letters. Data for this study were sourced from two healthcare institutions in Germany, comprising discharge letters for ten patients from each institution. The first experiment is conducted using a standardized prompt for information extraction. However, challenges were encountered, and the prompt was fine-tuned in a second experiment to improve the results. We further tested whether open-source LLMs can achieve similar results. In the first experiment, primary diagnoses were identified with 85% accuracy and secondary diagnoses with 55.8%. Medications and allergies were extracted with 85.9% and 100% accuracy, respectively. The International Classification of Diseases, 10th revision (ICD-10) codes for the identified diagnoses achieved an accuracy of 85% for primary diagnoses and 60.7% for secondary diagnoses. Anatomical Therapeutic Chemical (ATC) codes were identified with an accuracy of 78.8%. On the other hand, open-source LLMs did not provide similar levels of accuracy and could not consistently fill the template. With prompt fine-tuning in the second experiment, the primary diagnoses, secondary diagnoses, and medications could be predicted with 95%, 88.9%, and 92.2% accuracy, respectively. GPT-4 shows excellent potential for automated extraction of crucial diagnostic and medication information from discharge letters, presumably lowering the administrative burden for healthcare professionals and improving patient outcomes.</div></div>","PeriodicalId":73222,"journal":{"name":"Healthcare analytics (New York, N.Y.)","volume":"7 ","pages":"Article 100378"},"PeriodicalIF":0.0,"publicationDate":"2025-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143172049","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0