International Journal of Medical Informatics最新文献

Background: Anomalies in healthcare refer to deviation from the norm of unusual or unexpected patterns or activities related to patients, diseases or medical centres. Detecting these anomalies is crucial for timely interventions and efficient decision-making, helping to identify issues like operational inefficiencies, fraud and emerging health complications.

Objectives: This study presents a novel method for detecting both collective and individual anomalies in healthcare data through time series analysis using unsupervised machine learning. The dual-strategy approach leverages two methodologies: a 'practice centre-based approach' which monitors changes across different practice centres and a 'process-based approach' which focuses on identifying anomalies within individual centres. The former allows for early detection of systemic issues, while the latter highlights specific irregularities within a centre's operations.

Methods: The study utilised a dataset over 500,000 medical records from multiple GP practice centres in the UK collected between 2018-2023. Data are clustered using DBSCAN to identify collective anomalies from deviations from linear trends in consecutive two-month scatterplots. Individual anomalies are identified by examining the SOM-clustered time series of various medical processes within a specific practice centre, where graphs show deviation from the typical pattern.

Findings: Our approach addresses some challenges posed by the complexity and sensitivity of healthcare data by not requiring personal information. The method offers accurate visual representations making the data accessible and interpretable for non-technical users. Unlike traditional methods focusing solely on subsequence anomalies, our technique analyses the collective behaviour across multiple time series providing a more comprehensive perspective.

Conclusion: This study underscores the importance of integrating unsupervised anomaly detection with clinical expertise to ensure that statistically anomalous patterns align with clinical relevance. The dual-strategy clustering method holds significant potential for enabling timely interventions, proactively identifying potential crises, and ultimately contributing to better decision-making and operational efficiency within the healthcare sector.

Background: Stroke recurrence readmission poses an additional burden on both patients and healthcare systems. Risk stratification aims to accurately divide patients into groups to provide targeted interventions at reducing readmission. To accurately predict short and intermediate-term risks of readmission and provide information for further temporal risk stratification, we developed and validated an interpretable machine learning risk scoring tool.

Methods: In this retrospective study, all stroke admission episodes from January 1st 2015 to December 31st 2019 were obtained from the Shanghai Health and Health Development Research Centre database, which covers medical records of all patients hospitalized in 436 medical institutes in Shanghai. The outcome was time to stroke recurrence readmission within 90 days post discharge. The Score for Stroke Recurrence Readmission Prediction (SSRRP) tool was derived via an interpretable machine learning-based system for time-to-event outcomes. SSRRP as six-variable survival score includes sequelae, length of stay, type of stroke, random plasma glucose, medical expense payment, and number of hospitalizations.

Results: A total of 339,212 S admission episodes were finally included in the whole cohort. Among them, 217,393 episodes were included in the training dataset, 54,347 episodes in the internal validation dataset, and 67,472 in the temporal validation dataset. Readmission within 90 days was documented in 33922(9.97 %) episodes, with a median time to emergency readmission of 19 days (Interquartile range: 8-43). In the temporal validation dataset, the SSRRP achieved an integrated area under the curve of 0.730(95 % CI, 0.724-0.737). In addition, SSRRP demonstrated good calibration and clinical benefit rate.

Conclusions: In this retrospective cohort study, the SSRRP, a parsimonious and point-based scoring tool, was developed to predict the risk of recurrent readmission for stroke. It also provided accurate information on the time to stroke readmission, enabling further temporal risk stratification and informed clinical decision-making.

Background: Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that affects social interactions and behavior. Accurate and early diagnosis of ASD is still challenging even with the improvements in neuroimaging technology and machine learning algorithms. It's challenging because of the wide range of symptoms, delayed appearance of symptoms, and the subjective nature of diagnosis. In this study, the aim is to enhance ASD recognition by focusing on brain subcortical regions, which are critical for understanding ASD pathology.

Methodology: First, subcortical structures were extracted from a collection of brain MRI datasets using sophisticated processing steps. Next, a 3D autoencoder was trained on these 3D images to help identify brain regions related to ASD. Two distinct feature selection methods were then applied to the features extracted from the encoder. The highest-ranked features were iteratively selected and increased to reconstruct a specific percentage of the brain that represents the most relevant parts for ASD. Finally, a Siamese Convolutional Neural Network (SCNN) was employed as the classifier model.

Results: The 3D autoencoder stage helped in identifying and reconstructing the significant subcortical regions related to ASD. Based on the studied dataset, high agreement in regions like the Putamen and Pallidum indicated the critical nature of these structures in distinguishing Autism from controls cases. Subsequently, applying SCNN on these selected subcortical regions yielded promising results. For example, using the classifier on the output regions identified by the Mutual Information (MI) features selection method achieved the highest accuracy of 0.66.

Conclusions: This study shows that using a two-stage model involving autoencoder and SCNN can notably improve the classification of ASD from brain MRI volumetric images. Applying an iterative feature extraction approach allowed to achieve a more accurate identification of ASD-related brain areas. This two-stage approach not only improved classification performance but also enhanced the interpretability of the neuroimaging data.

Background: Heart failure with preserved ejection fraction (HFpEF) is associated with elevated rates of readmission and mortality. Accurate prediction of readmission risk is essential for optimizing healthcare resources and enhancing patient outcomes.

Methods: We conducted a retrospective cohort study utilizing HFpEF patient data from two institutions: the First Affiliated Hospital Zhejiang University School of Medicine for model development and internal validation, and the Affiliated Hospital of Xuzhou Medical University for external validation. A machine learning (ML) model was developed and validated using 53 variables to predict the risk of readmission within one year. The model's performance was assessed using several metrics, including the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, F1 score, model training time, model prediction time and brier score. SHAP (SHapley Additive exPlanations) analysis was employed to enhance model interpretability, and a dynamic nomogram was constructed to visualize the predictive model.

Results: Among the 766 HFpEF patients included in the study, 203 (26.5%) were readmitted within one year. The LightGBM model exhibited the highest predictive performance, with an AUC of 0.88 (95% confidence interval (CI):0.84-0.91), an accuracy of 0.79, a sensitivity of 0.81, and a specificity of 0.78. Key predictors included the E/e' ratio, NYHA classification, LVEF, age, BNP levels, MLR, history of atrial fibrillation (AF), use of ACEI/ARB/ARNI, and history of myocardial infarction (MI). External validation also demonstrated strong predictive performance, with an AUC of 0.87 (95%CI:0.83-0.91).

Conclusions: The LightGBM model exhibited robust performance in predicting one-year readmission risk among HFpEF patients, providing a valuable tool for clinicians to identify high-risk individuals and implement timely interventions.

Introduction: In the digital age, electronic health literacy (eHealth literacy) has become crucial for maintaining and improving health outcomes. As the population ages, developing and validating tools that accurately measure eHealth literacy levels among older adults in different cultures is essential.

Objectives: This study aimed to validate the Hebrew version of the electronic Health Literacy scale among Israelis aged 65 and older by assessing its psychometric properties, including content validity, construct validity, age-based convergent validity, internal consistency reliability, and test-retest reliability.

Methods: A sample of 628 Israelis aged 65 and older was recruited using convenience sampling. Participants completed an online survey consisting of the HeHEALS, demographic questions, items related to participants' use of online health information sources, and measures of self-rated health, satisfaction with health, and perceived health compared to others. Psychometric properties were assessed using various statistical analyses.

Results: The HeHEALS demonstrated good content validity, construct validity, age-based convergent validity, internal consistency reliability, and test-retest reliability. Exploratory factor analysis supported a unidimensional structure of the HeHEALS. Significant positive correlations were found between HeHEALS and education, income, and subjective health measures. Users of online health information sources had significantly higher electronic health literacy scores than non-users.

Conclusions: The HeHEALS is a valid and reliable tool for assessing eHealth literacy among older adults in Israel, with potential applications in research and practice to promote digital health equity.

Introduction: Both predictions and performance of clinical predictive models can be presented with various verbal and visual representations. This study aims to investigate how different risk and performance presentations for probabilistic predictions affect clinical users' judgement and preferences.

Methods: We use a clinical Bayesian Network (BN) model that has been developed for predicting the risk of Trauma Induced Coagulopathy (TIC). Three patient scenarios with different levels of TIC risk were shown to trauma care clinicians. The prediction and discriminatory performance of TIC BN were shown with each scenario using different charts in a random order. Bar charts, icon arrays and gauge charts were used for presenting the prediction. Receiver operating characteristic curves, true and false positive rate curves and icon arrays were used for presenting the performance. Risk judgement for patient scenarios, perceived accuracy for the predictions and the model, and preferences for charts were elicited using an online survey.

Results: A total of 25 clinicians evaluated 75 BN predictions. The choice of risk charts was associated with the risk score in the borderline medium-risk scenario. The choice of risk and performance charts interacts with the perceived accuracy of the predictions and model in the high and low-risk scenarios, respectively. The participants had varying but persistent preferences regarding risk presentation charts. Icon arrays were preferred for performance presentations.

Conclusions: The choice of presenting predictions and the performance of predictive models can affect risk and performance interpretation. Clinical predictive models should offer the flexibility of presenting predictions with different illustrations.

Objective: Myasthenic crisis (MC) is a critical progression of Myasthenia gravis (MG), requiring intensive care treatment and invasive therapies. Classifying patients at high-risk for MC facilitates treatment decisions such as changes in medication or the need for mechanical ventilation and helps prevent disease progression by decreasing treatment-induced stress on the patient. Here, we investigated whether it is possible to reliably classify MG patients into groups at low or high risk of MC based entirely on routine medical data using explainable machine learning (ML).

Methods: In this single-center pseudo-prospective cohort study, we investigated the precision of ML models trained with real-world routine clinical data to identify MG patients at risk for MC, and identified explainable distinctive features for the groups. 51 MG patients, including 13 MC, were used for model training based on real-world clinical data available from the hospital management system. Patients were classified to high or low risk for MC using Lasso regression or random forest ML models.

Results: The mean cross-validated AUC classifying MG patients as high or low risk for MC based on simple or compound features derived from real-world clinical data showed a predictive accuracy of 68.8% for a regularized Lasso regression and 76.5% for a random forest model. Studying feature importance across 5100 model runs identified explainable features to distinguish MG patients at high or low risk for MC. Feature importance scores suggested that multimorbidity may play a role in risk classification.

Conclusion: This study establishes feasibility and proof-of-concept for risk classification of MC based on real-world routine clinical data using ML with explainable features and variance control at the point of care. Future research on ML-based prediction of MC should include multi-center, multinational data collection, more in-depth data per patient, more patients, and an attention-based ML model to include free-text.