Advanced Machine Learning Models for Predicting Post-Thrombolysis Hemorrhagic Transformation in Acute Ischemic Stroke Patients: A Systematic Review and Meta-Analysis.

Background: Thrombolytic therapy is essential for acute ischemic stroke (AIS) management but poses a risk of hemorrhagic transformation (HT), necessitating accurate prediction to optimize patient care. Methods: A comprehensive search was conducted across PubMed, Web of Science, Scopus, Embase, and Google Scholar, covering studies from inception until July 10, 2024. Studies were included if they used machine learning (ML) or deep learning algorithms to predict HT in AIS patients treated with thrombolysis. Exclusion criteria included studies involving endovascular treatments and those not evaluating model effectiveness. Data extraction and quality assessment were performed following PRISMA guidelines and using the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) and Prediction Model Risk of Bias Assessment Tool (PROBAST) tools. Results: Out of 1943 identified records, 12 studies were included in the final analysis, encompassing 18 007 AIS patients who received thrombolytic therapy. The ML models demonstrated high predictive performance, with pooled area under the curve (AUC) values ranging from 0.79 to 0.95. Specifically, XGBoost models achieved AUCs of up to 0.953 and Artificial Neural Network (ANN) models reached up to 0.942. Sensitivity and specificity varied significantly, with the highest sensitivity at 0.90 and specificity at 0.99. Significant predictors of HT included age, glucose levels, NIH Stroke Scale (NIHSS) score, systolic and diastolic blood pressure, and radiomic features. Despite these promising results, methodological disparities and limited external validation highlighted the need for standardized reporting and further rigorous testing. Conclusion: ML techniques, especially XGBoost and ANN, show great promise in predicting HT following thrombolysis in AIS patients, enhancing risk stratification and clinical decision-making. Future research should focus on prospective study designs, standardized reporting, and integrating ML assessments into clinical workflows to improve AIS management and patient outcomes.

Advancing stroke patient care: a network meta-analysis of dysphagia screening efficacy and personalization.

Introduction: The increasing incidence of stroke globally has led to dysphagia becoming one of the most common complications in stroke patients, with significant impacts on patient outcomes. Accurate early screening for dysphagia is crucial to avoid complications and improve patient quality of life. Methods: Included studies involved stroke-diagnosed patients assessed for dysphagia using bedside screening tools. Data was sourced from Embase, PubMed, Web of Science, Scopus, and CINAHL, including publications up to 10 December 2023. The study employed both fixed-effect and random-effects models to analyze sensitivity, specificity, positive predictive value (PPV), and Negative Predictive Value (NPV), each with 95% confidence intervals. The random-effects model was particularly utilized due to observed heterogeneity in study data. Results: From 6,979 records, 21 studies met the inclusion criteria, involving 3,314 participants from 10 countries. The analysis included six assessment tools: GUSS, MASA, V-VST, BSST, WST, and DNTA, compared against gold-standard methods VFSS and FEES. GUSS, MASA, and V-VST showed the highest reliability, with sensitivity and specificity rates of 92% and 85% for GUSS, 89% and 83% for MASA, respectively. Heterogeneity among studies was minimal, and publication bias was low, enhancing the credibility of the findings. Conclusion: Our network meta-analysis underscores the effectiveness of GUSS, MASA, and V-VST in dysphagia screening for stroke patients, with high sensitivity and specificity making them suitable for diverse clinical settings. BSST and WST, with lower diagnostic accuracy, require more selective use. Future research should integrate patient-specific outcomes and standardize methodologies to enhance dysphagia screening tools, ultimately improving patient care and reducing complications. Systematic review registration: https://www.crd.york.ac.uk/PROSPERO/#recordDetails.

LSTM-Based Prediction Model for Tuberculosis Among HIV-Infected Patients Using Structured Electronic Medical Records: A Retrospective Machine Learning Study.

Background: Both HIV and TB are chronic infectious diseases requiring long-term treatment and follow-up, resulting in extensive electronic medical records. With the exponential growth of health and medical big data, effectively extracting and analyzing these data has become the research hotspot. As a fundamental aspect of artificial intelligence, machine learning has been extensively applied in medical research, encompassing diagnosis, treatment, patient monitoring, drug development, and epidemiological investigations. This significantly enhances medical information systems and facilitates the interoperability of medical data. Methods: In our study, we analyzed longitudinal data from the electronic health records of 4540 patients, gathered from the National Clinical Research Center for Infectious Diseases in Shenzhen, China, spanning from 2017 to 2021. Initially, we employed the fine-tuned ChatGLM to structure the electronic medical records. Subsequently, we utilized a multi-layer perceptron to classify each patient and determined the presence of tuberculosis in HIV patients. Using machine learning-based natural language processing, we structured these records to build a specialized database for HIV and TB co-infection. We studied the epidemiological characteristics, focusing on incidence patterns, patient characteristics, and influencing factors, to uncover the transmission characteristics of these diseases in Shenzhen. Additionally, we used Long Short-Term Memory to create a predictive model for TB co-infection among HIV patients, based on their medical records. This model predicted the risk of TB co-infection, providing scientific evidence for clinical decision-making and enabling early detection and precise intervention. Results: Based on the refined ChatGLM model tailored for structured electronic health records, the accuracy of symptom extraction consistently surpassed 0.95 precision. Key symptoms such as diarrhea and normal showed precision rates exceeding 0.90. High scores were also achieved in recall and F1 scores. Among 4540 HIV patients, 758 were diagnosed with concurrent tuberculosis, indicating a 16.7% co-infection rate, while syphilis co-infection affected 25.1%, underscoring the prevalence of concurrent infections among HIV patients. Utilizing electronic health records, a Multilayer Perceptron classifier was developed as a benchmark against Long Short-Term Memory to predict high-risk groups for HIV and tuberculosis co-infections. The Multilayer Perceptron classifier demonstrated predictive ability with AUROC values ranging from 0.616 to 0.682 on the test set, suggesting opportunities for further optimization and generalization despite its accuracy in identifying HIV-TB co-infections. In tuberculosis intelligent diagnosis based on laboratory results, the Long Short-Term Memory showed consistent performance across 5-fold cross-validation, with AUROC values ranging from 0.827 to 0.850, indicating reliability and consistency in tuberculosis prediction. Furthermore, by optimizing classification thresholds, the model achieved an overall accuracy of 81.18% in distinguishing HIV co-infected tuberculosis from simple HIV infection. Conclusion: Combining the Multilayer Perceptron classifier with Long Short-Term Memory represented an advanced approach for effectively extracting electronic health records and utilizing it for disease prediction. This underscored the superior performance of deep learning techniques in managing both structured and unstructured medical data. Models leveraging laboratory time-series data demonstrated notably better performance compared to those relying solely on electronic health records for predicting tuberculosis incidence. This emphasized the benefits of deep learning in handling intricate medical data and provided valuable insights for healthcare providers exploring the use of deep learning in disease prediction and management.

Analyzing prehospital delays in recurrent acute ischemic stroke: Insights from interpretable machine learning.

OBJECTIVE: This study investigates prehospital delays in recurrent Acute Ischemic Stroke (AIS) patients, aiming to identify key factors contributing to these delays to inform effective interventions. METHODS: A retrospective cohort analysis of 1419 AIS patients in Shenzhen from December 2021 to August 2023 was performed. The study applied the Extreme Gradient Boosting (XGBoost) algorithm and SHapley Additive exPlanations (SHAP) for identifying determinants of delay. RESULTS: Living with others and lack of stroke knowledge emerged as significant risk factors for delayed hospital presentation in recurrent AIS patients. Key features impacting delay times included residential status, awareness of stroke symptoms, presence of conscious disturbance, diabetes mellitus awareness, physical weakness, mode of hospital presentation, type of stroke, and presence of coronary artery disease. CONCLUSION: Prehospital delays are similarly prevalent among both recurrent and first-time AIS patients, highlighting a pronounced knowledge gap in the former group. This discovery underscores the urgent need for enhanced stroke education and management. PRACTICE IMPLICATION: The similarity in prehospital delay patterns between recurrent and first-time AIS patients emphasizes the necessity for public health initiatives and tailored educational programs. These strategies aim to improve stroke response times and outcomes for all patients.

Prevalence and Risk Factors of Diabetes in Patients with Active Pulmonary Tuberculosis: A Cross-Sectional Study in Two Financially Affluent China Cities.

Background: Tuberculosis (TB) and diabetes mellitus (DM) present a dual burden to public health. The screening of DM in TB patients may aid in the early detection and management of diabetes, ultimately improving treatment outcomes for those with the comorbidity of TB-DM. We aim to examine the prevalence and identify risk factors of diabetes in individuals with active pulmonary tuberculosis (PTB) in financially affluent China cities. Methods: A cross-sectional survey was conducted in adult patients with highly suspected TB in two cities of China, spanning from May 9, 2023, to June 30, 2023. We compare the clinical characteristics, nutrition status, fasting blood glucose (FBG) level, living style, and knowledge of TB and DM at admission between patients with and without DM. Univariate and multivariate logistic regression analyses were employed to identify risk factors associated with TB-DM comorbidities. Results: Of the 322 patients diagnosed with pulmonary tuberculosis (PTB), 54 individuals (16.8%) had comorbid diabetes mellitus (DM). This included 43 males (13.4%) and 11 females (3.4%). The average age was 55.44 ± 12.36 in DM patients and 46.09 ± 16.87 in non-DM patients. A multivariate logistic regression analysis revealed that male (adjusted odds ratio [aOR]=3.29, 95% confidence interval [CI]: 1.05-10.30), age older than 47 years (aOR = 1.04, 95% CI: 1.01-1.07), having a family history of diabetes (aOR = 5.09, 95% CI: 1.28-20.32), and an elevated random blood glucose level (aOR = 1.6, 95% CI: 1.38-1.86) were risk factors for DM in patients with PTB. Furthermore, it was found that diabetes awareness (aOR = 0.07, 95% CI: 0.03-0.21) and zero, light to moderate alcohol consumption were associated with a lower risk of diabetes. Conclusion: Diabetes is prevalent in patients with active PTB. Screening and raising awareness of DM are recommended, particularly in men after middle age with a family history of diabetes and elevated random blood glucose. Early diagnosis of diabetes and effective diabetes prevention may reduce the dual burden of TB-DM comorbidity.

Healthcare-Seeking Delays in Acute Ischemic Stroke Patients: The Influence of Gender, Immigrant Status, and Educational Background.

Purpose: Timely medical attention is crucial for patients with Acute Ischemic Stroke (AIS), as delays can significantly impact therapeutic outcomes. These delays are influenced by a combination of socio-cultural, educational, and clinical factors. Patients and Methods: An in-depth analysis was conducted to assess the prevalence and median duration of healthcare-seeking delays in AIS patients. The study specifically investigated the independent impacts of sociocultural and clinical determinants on these delays, with a focus on immigrant status, gender disparities, and educational levels. Multivariate regression analysis was employed to identify these independent effects while controlling for potential confounding factors. Results: Among 1419 AIS patients, 82.52% (n = 1171) experienced delays exceeding 2 hours from symptom onset of symptoms to hospital arrival. The median delay was 12.3 hours. Immigrant populations encountering longer delays compared to native groups. Younger males (<45 years) and elderly females were more prone to delay in healthcare-seeking. Identified independent risk factors for delay included male gender (OR = 1.65 [95% CI:1.14-2.48]), self-acknowledged diabetes (OR = 2.50 [95% CI:1.21-5.17]), small vessel (OR = 2.07 [95% CI:1.27-3.36]), and wake stroke (OR = 7.04 [95% CI:3.69-13.44]). Educational background (high school and above), GCS score with 3-8 points (OR = 0.52 [95% CI:0.09-0.69]), understanding stroke-related knowledge (OR = 0.26 [95% CI:0.09-0.44]), conscious disturbance (OR = 0.25 [95% CI:0.10-0.62]) and limb weakness (OR=0.21[95% CI:0.21-0.49]) are protective factors for timely treatment. Conclusion: Immigrant populations experienced longer delays from symptom onset to hospital arrival. The crucial roles of education and knowledge about stroke underscore the need for enhanced health literacy campaigns and public awareness, with a targeted focus on younger males and elderly females.

Factors associated with loss to follow-up before and after treatment initiation among patients with tuberculosis: A 5-year observation in China.

Background: Loss to follow-up (LTFU) is a significant barrier to the completion of anti-tuberculosis (TB) treatment and a major predictor of TB-associated deaths. Currently, research on LTFU-related factors in China is both scarce and inconsistent. Methods: We collected information from the TB observation database of the National Clinical Research Center for Infectious Diseases. The data of all patients who were documented as LTFU were assessed retrospectively and compared with those of patients who were not LTFU. Descriptive epidemiology and multivariable logistic regression analyses were conducted to identify the factors associated with LTFU. Results: A total of 24,265 TB patients were included in the analysis. Of them, 3,046 were categorized as LTFU, including 678 who were lost before treatment initiation and 2,368 who were lost afterwards. The previous history of TB was independently associated with LTFU before treatment initiation. Having medical insurance, chronic hepatitis or cirrhosis, and providing an alternative contact were independent predictive factors for LTFU after treatment initiation. Conclusion: Loss to follow-up is frequent in the management of patients with TB and can be predicted using patients' treatment history, clinical characteristics, and socioeconomic factors. Our research illustrates the importance of early assessment and intervention after diagnosis. Targeted measures can improve patient engagement and ultimately treatment adherence, leading to better health outcomes and disease control.

Sleep quality and influencing factors and correlation with T-lymphocyte subpopulation counts in patients with pulmonary tuberculosis: a cross-sectional study.

BACKGROUND: Patients diagnosed with pulmonary tuberculosis (TB) have poor sleep quality due to multiple factors. We aimed to assess the sleep status and related factors of TB patients in Shenzhen, China. METHODS: A questionnaire survey was conducted on 461 TB patients hospitalized at Shenzhen Third People's Hospital from March 2021 to January 2022, and sleep quality was assessed using the Pittsburgh sleep quality index (PSQI). RESULTS: A total of 459 valid questionnaires were collected, and 238 of the 459 TB patients had general or poor sleep quality (PSQI > 5). Patients' gender, marriage, nutritional screening score, family atmosphere, fear of discrimination, fear of interactions, and the impact of the disease on their work life had significant effects on sleep quality (P < 0.05); PSQI scores of TB patients were negatively correlated with lymphocyte counts (r = - 0.296, P < 0.01), T-lymphocyte counts (r = - 0.293, P < 0.01), helper T lymphocyte counts (r = - 0.283, P < 0.01), killer T lymphocyte counts (r = - 0.182, P < 0.05), and were positively correlated with depression scores (r = 0.424, P < 0.01). Multivariable logistic regression analysis showed that male (OR = 1.64,95% CI 1.11-2.42, P < 0.05), unmarried (OR = 1.57, 95% CI 1.02-2.42, P < 0.05), NRS score grade 3(OR = 5.35, 95% CI 2.08-15.73, P < 0.01), general family atmosphere (OR = 2.23, 95% CI 1.07-4.93, P < 0.05), and the disease affecting work (OR = 1.66, 95% CI 1.11-2.50, P < 0.05) were factors influencing poor sleep quality. CONCLUSION: Most TB patients had varying degrees of sleep disturbance, which may be affected by their gender, marriage, family atmosphere, nutritional status, the effect of the disease on work life, and, depression, as well as lower absolute T-lymphocyte subpopulation counts. Appropriate interventions should be implemented to improve their sleep quality, when treating or caring for such patients.

Dynamic Predictive Models with Visualized Machine Learning for Assessing the Risk of Lung Metastasis in Kidney Cancer Patients.

Objective: To establish and verify the clinical prediction model of lung metastasis in renal cancer patients. Method: Kidney cancer patients from January 1, 2010, to December 31, 2017, in the SEER database were enrolled in this study. In the first section, LASSO method was adopted to select variables. Independent influencing factors were identified after multivariate logistic regression analysis. In the second section, machine learning (ML) algorithms were implemented to establish models and 10-foldcross-validation was used to train the models. Finally, receiver operating characteristic curves, probability density functions, and clinical utility curve were applied to estimate model's performance. The final model was shown by a website calculator. Result: Lung metastasis was confirmed in 7.43% (3171 out of 42650) of study population. In multivariate logistic regression, bone metastasis, brain metastasis, grade, liver metastasis, N stage, T stage, and tumor size were independent risk factors of lung metastasis in renal cancer patients. Primary site and sequence number were independent protection factors of LM in renal cancer patients. The above 9 impact factors were used to develop the prediction models, which included random forest (RF), naive Bayes classifier (NBC), decision tree (DT), xgboost (XGB), gradient boosting machine (GBM), and logistic regression (LR). In 10-foldcross-validation, the average area under curve (AUC) ranked from 0.907 to 0.934. In ROC curve analysis, AUC ranged from 0.879-0.922. We found that the XGB model performed best, and a Web-based calculator was done according to XGB model. Conclusion: This study provided preliminary evidence that the ML algorithm can be used to predict lung metastases in patients with kidney cancer. This low cost, noninvasive and easy to implement diagnostic method is useful for clinical work. Of course this model still needs to undergo more real-world validation.

Duration and Determinants of Delayed Diagnosis with Tuberculosis in Shenzhen, China: A Cross-Sectional Study.

Background: Early diagnosis and timely treatment of tuberculosis are critical for disease control and management. However, diagnostic delay remains severe around the world. We aim to evaluate the duration and factors associated with diagnostic delay of tuberculosis in Shenzhen, China. Methods: We conducted a face-to-face interview to collect the whole care-seeking process of patients diagnosed with active TB in Shenzhen, China, from April 1 to September 30, 2021. The duration from symptom onset to confirmed diagnosis was recorded. The risk factors of diagnostic delay were identified by binary stepwise logistic regression analysis. Results: Among 288 confirmed TB cases, 170 (59.0%) were delayed diagnosis. The median diagnostic delay was 39.5 days. Median patient delay was 23 days and health system delay was 7 days. Income ≤315USD/month (OR = 2.97 [95% CI: 1.15-7.69]), cough (OR = 3.00 [95% CI: 1.16-7.76]), weight loss (OR = 15.59 [95% CI: 1.85-131.56]), use of traditional Chinese Medicine (OR = 5.03 [95% CI: 1.04-24.31]) and over-the-counter cough syrup (OR = 2.73 [95% CI: 1.10-6.76]) were significant risk factors for patient delay. Fever (OR = 0.13[95% CI: 0.04-0.48]) and hemoptysis (OR = 0.06 [95% CI0.01-0.30]) were protective factors for patient delay. Cough (OR = 2.85 [95% CI: 1.49-5.49]) and availability of chest X-ray (OR = 0.21[CI: 0.11-0.39]) were factors associated with health system delay. Conclusion: Delayed diagnosis of tuberculosis remains an unresolved problem. Patients with low income, self-treatment with over-the-counter medicine and accepting TCM suffered from a higher risk of patient delay. It is important to give more help to the vulnerable people and strengthen tuberculosis knowledge among primary health providers. Keeping all health providers alert to TB symptoms can facilitate earlier TB diagnosis and better disease control.