RESUMEN
Objective: To develop a model using a Chinese ICU infection patient database to predict long-term health-related quality of life (HRQOL) in survivors. Methods: A patient database from the ICU of the Fourth People's Hospital in Zigong was analyzed, including data from 2019 to 2020. The subjects of the study were ICU infection survivors, and their post-discharge HRQOL was assessed through the SF-36 survey. The primary outcomes were the physical component summary (PCS) and mental component summary (MCS). We used artificial intelligence techniques for both feature selection and model building. Least absolute shrinkage and selection operator regression was used for feature selection, extreme gradient boosting (XGBoost) was used for model building, and the area under the receiver operating characteristic curve (AUROC) was used to assess model performance. Results: The study included 917 ICU infection survivors. The median follow-up was 507.8 days. Their SF-36 scores, including PCS and MCS, were below the national average. The final prognostic model showed an AUROC of 0.72 for PCS and 0.63 for MCS. Within the sepsis subgroup, the predictive model AUROC values for PCS and MCS were 0.76 and 0.68, respectively. Conclusions: This study established a valuable prognostic model using artificial intelligence to predict long-term HRQOL in ICU infection patients, which supports clinical decision making, but requires further optimization and validation.
RESUMEN
Tremor, defined as an "involuntary, rhythmic, oscillatory movement of a body part", is a key feature of many neurological conditions including Parkinson's disease and essential tremor. Clinical assessment continues to be performed by visual observation with quantification on clinical scales. Methodologies for objectively quantifying tremor are promising but remain non-standardized across centers. Our center performs full-body behavioral testing with 3D motion capture for clinical and research purposes in patients with Parkinson's disease, essential tremor, and other conditions. The objective of this study was to assess the ability of several candidate processing pipelines to identify the presence or absence of tremor in kinematic data from patients with confirmed movement disorders and compare them to expert ratings from movement disorders specialists. We curated a database of 2272 separate kinematic data recordings from our center, each of which was contemporaneously annotated as tremor present or absent by a movement physician. We compared the ability of six separate processing pipelines to recreate clinician ratings based on F1 score, in addition to accuracy, precision, and recall. The performance across algorithms was generally comparable. The average F1 score was 0.84±0.02 (mean ± SD; range 0.81-0.87). The second highest performing algorithm (cross-validated F1=0.87) was a hybrid that used engineered features adapted from an algorithm in longstanding clinical use with a modern Support Vector Machine classifier. Taken together, our results suggest the potential to update legacy clinical decision support systems to incorporate modern machine learning classifiers to create better-performing tools.
Asunto(s)
Algoritmos , Trastornos del Movimiento , Temblor , Humanos , Temblor/diagnóstico , Temblor/fisiopatología , Trastornos del Movimiento/diagnóstico , Trastornos del Movimiento/fisiopatología , Enfermedad de Parkinson/diagnóstico , Enfermedad de Parkinson/fisiopatología , Fenómenos Biomecánicos , Temblor Esencial/diagnóstico , Temblor Esencial/fisiopatología , Masculino , Femenino , Persona de Mediana Edad , AncianoRESUMEN
Load recognition remains not comprehensively explored in Home Energy Management Systems (HEMSs). There are gaps in current approaches to load recognition, such as enhancing appliance identification and increasing the overall performance of the load-recognition system through more robust models. To address this issue, we propose a novel approach based on the Analysis of Variance (ANOVA) F-test combined with SelectKBest and gradient-boosting machines (GBMs) for load recognition. The proposed approach improves the feature selection and consequently aids inter-class separability. Further, we optimized GBM models, such as the histogram-based gradient-boosting machine (HistGBM), light gradient-boosting machine (LightGBM), and XGBoost (extreme gradient boosting), to create a more reliable load-recognition system. Our findings reveal that the ANOVA-GBM approach achieves greater efficiency in training time, even when compared to Principal Component Analysis (PCA) and a higher number of features. ANOVA-XGBoost is approximately 4.31 times faster than PCA-XGBoost, ANOVA-LightGBM is about 5.15 times faster than PCA-LightGBM, and ANOVA-HistGBM is 2.27 times faster than PCA-HistGBM. The general performance results expose the impact on the overall performance of the load-recognition system. Some of the key results show that the ANOVA-LightGBM pair reached 96.42% accuracy, 96.27% F1, and a Kappa index of 0.9404; the ANOVA-HistGBM combination achieved 96.64% accuracy, 96.48% F1, and a Kappa index of 0.9434; and the ANOVA-XGBoost pair attained 96.75% accuracy, 96.64% F1, and a Kappa index of 0.9452; such findings overcome rival methods from the literature. In addition, the accuracy gain of the proposed approach is prominent when compared straight to its competitors. The higher accuracy gains were 13.09, 13.31, and 13.42 percentage points (pp) for the pairs ANOVA-LightGBM, ANOVA-HistGBM, and ANOVA-XGBoost, respectively. These significant improvements highlight the effectiveness and refinement of the proposed approach.
RESUMEN
Otitis media (OM), a highly prevalent inflammatory middle-ear disease in children worldwide, is commonly caused by an infection, and can lead to antibiotic-resistant bacterial biofilms in recurrent/chronic OM cases. A biofilm related to OM typically contains one or multiple bacterial species. OCT has been used clinically to visualize the presence of bacterial biofilms in the middle ear. This study used OCT to compare microstructural image texture features from bacterial biofilms. The proposed method applied supervised machine-learning-based frameworks (SVM, random forest, and XGBoost) to classify multiple species bacterial biofilms from in vitro cultures and clinically-obtained in vivo images from human subjects. Our findings show that optimized SVM-RBF and XGBoost classifiers achieved more than 95% of AUC, detecting each biofilm class. These results demonstrate the potential for differentiating OM-causing bacterial biofilms through texture analysis of OCT images and a machine-learning framework, offering valuable insights for real-time in vivo characterization of ear infections.
RESUMEN
Necrophagous flies, particularly blowflies, serve as vital indicators in forensic entomology and ecological studies, contributing to minimum postmortem interval estimations and environmental monitoring. The study investigates variations in the predominant cuticular hydrocarbons (CHCs) viz. n-C25, n-C27, n-C28, and n-C29 of empty puparia of Calliphora vicina Robineau-Desvoidy, 1830, (Diptera: Calliphoridae) across diverse environmental conditions, including burial, above-ground and indoor settings, over 90 days. Notable trends include a significant decrease in n-C25 concentrations in buried and above-ground conditions over time, while n-C27 concentrations decline in buried and above-ground conditions but remain stable indoors. Burial conditions show significant declines in n-C27 and n-C29 concentrations over time, indicating environmental influences. Conversely, above-ground conditions exhibit uniform declines in all hydrocarbons. Indoor conditions remain relatively stable, with weak correlations between weathering time and CHC concentrations. Additionally, machine learning techniques, specifically Extreme Gradient Boosting (XGBoost), are employed for age estimation of empty puparia, yielding accurate predictions across different outdoor and indoor conditions. These findings highlight the subtle responses of CHC profiles to environmental stimuli, underscoring the importance of considering environmental factors in forensic entomology and ecological research. The study advances the understanding of insect remnant degradation processes and their forensic implications. Furthermore, integrating machine learning with entomological expertise offers standardized methodologies for age determination, enhancing the reliability of entomological evidence in legal contexts and paving the way for future research and development.
RESUMEN
Introduction: Individuals with diabetes mellitus (DM) are at an increased risk of Mycobacterium tuberculosis (Mtb) infection and progressing from latent tuberculosis (TB) infection to active tuberculosis disease. TB in the DM population is more likely to go undiagnosed due to smear-negative results. Methods: Exhaled breath samples were collected and analyzed using high-pressure photon ionization time-of-flight mass spectrometry. An eXtreme Gradient Boosting (XGBoost) model was utilized for breathomics analysis and TB detection. Results: XGBoost model achieved a sensitivity of 88.5%, specificity of 100%, accuracy of 90.2%, and an area under the curve (AUC) of 98.8%. The most significant feature across the entire set was m106, which demonstrated a sensitivity of 93%, specificity of 100%, and an AUC of 99.7%. Discussion: The breathomics-based TB detection method utilizing m106 exhibited high sensitivity and specificity potentially beneficial for clinical TB screening and diagnosis in individuals with diabetes.
RESUMEN
In China, population growth and aging have partially negated the public health benefits of air pollution control measures, underscoring the ongoing need for precise PM2.5 monitoring and mapping. Despite its prevalence, the satellite-derived Aerosol Optical Depth (AOD) method for estimating PM2.5 concentrations often encounters significant spatial data gaps. Additionally, current research still needs better representation of PM2.5 spatiotemporal heterogeneity. Addressing these challenges, we developed a two-stage model employing the Extreme Gradient Boosting (XGBoost) algorithm. By incorporating improved spatiotemporal factors, we achieved high-precision and full-coverage daily 1-km PM2.5 mappings across China for the year 2020 without utilizing AOD products. Specifically, Model 1 develops improved temporal encodings and a terrain classification factor (DC), while Model 2 constructs an enhanced spatial autocorrelation term (Ps) by integrating observed and estimated values. Notably, Model 2 excelled in 10-fold sample-based cross-validation, achieving a coefficient of determination of 0.948, a mean absolute error of 3.792 µg/m³, a root mean square error of 7.144 µg/m³, and a mean relative error of 14.171%. Feature importance and Shapley Additive exPlanations (SHAP) analyses determined the relative importance of predictors in model training and outcome prediction, while correlation analysis identified strong links between improved temporal encodings, PM2.5 concentrations, and significant meteorological factors. Two-way Partial Dependence Plots (PDPs) further explored the interactions among these factors and their impact on PM2.5 levels. Compared to traditional methods, improved temporal encodings align more closely with seasonal variations and synergize more effectively with meteorological factors. Besides, the structured nature of DC aids in model training, while the improved Ps more effectively captures PM2.5's spatial autocorrelation, outperforming traditional Ps. Overall, this study effectively represents spatiotemporal information, thereby boosting model accuracy and enabling seamless large-scale PM2.5 estimations. It provides deep insights into variables and models, providing significant implications for future air pollution research.
RESUMEN
As artificial intelligence (AI) becomes widespread, there is increasing attention on investigating bias in machine learning (ML) models. Previous research concentrated on classification problems, with little emphasis on regression models. This paper presents an easy-to-apply and effective methodology for mitigating bias in bagging and boosting regression models, that is also applicable to any model trained through minimizing a differentiable loss function. Our methodology measures bias rigorously and extends the ML model's loss function with a regularization term to penalize high correlations between model errors and protected attributes. We applied our approach to three popular tree-based ensemble models: a random forest model (RF), a gradient-boosted model (GBT), and an extreme gradient boosting model (XGBoost). We implemented our methodology on a case study for predicting road-level traffic volume, where RF, GBT, and XGBoost models were shown to have high accuracy. Despite high accuracy, the ML models were shown to perform poorly on roads in minority-populated areas. Our bias mitigation approach reduced minority-related bias by over 50%.
RESUMEN
The primary objective of this study was to develop a risk-based readmission prediction model using the EMR data available at discharge. This model was then validated with the LACE plus score. The study cohort consisted of about 310,000 hospital admissions of patients with cardiovascular and cerebrovascular conditions. The EMR data of the patients consisted of lab results, vitals, medications, comorbidities, and admit/discharge settings. These data served as the input to an XGBoost model v1.7.6, which was then used to predict the number of days until the next readmission. Our model achieved remarkable results, with a precision score of 0.74 (±0.03), a recall score of 0.75 (±0.02), and an overall accuracy of approximately 82% (±5%). Notably, the model demonstrated a high accuracy rate of 78.39% in identifying the patients readmitted within 30 days and 80.81% accuracy for those with readmissions exceeding six months. The model was able to outperform the LACE plus score; of the people who were readmitted within 30 days, only 47.70 percent had a LACE plus score greater than 70, and, for people with greater than 6 months, only 10.09 percent had a LACE plus score less than 30. Furthermore, our analysis revealed that the patients with a higher comorbidity burden and lower-than-normal hemoglobin levels were associated with increased readmission rates. This study opens new doors to the world of differential patient care, helping both clinical decision makers and healthcare providers make more informed and effective decisions. This model is comparatively more robust and can potentially substitute the LACE plus score in cardiovascular and cerebrovascular settings for predicting the readmission risk.
RESUMEN
Sepsis triggers a harmful immune response due to infection, causing high mortality. Predicting sepsis outcomes early is vital. Despite machine learning's (ML) use in medical research, local validation within the Medical Information Mart for Intensive Care IV (MIMIC-IV) database is lacking. We aimed to devise a prognostic model, leveraging MIMIC-IV data, to predict sepsis mortality and validate it in a Chinese teaching hospital. MIMIC-IV provided patient data, split into training and internal validation sets. Four ML models logistic regression (LR), support vector machine (SVM), deep neural networks (DNN), and extreme gradient boosting (XGBoost) were employed. Shapley additive interpretation offered early and interpretable mortality predictions. Area under the ROC curve (AUROC) gaged predictive performance. Results were cross verified in a Chinese teaching hospital. The study included 27,134 sepsis patients from MIMIC-IV and 487 from China. After comparing, 52 clinical indicators were selected for ML model development. All models exhibited excellent discriminative ability. XGBoost surpassed others, with AUROC of 0.873 internally and 0.844 externally. XGBoost outperformed other ML models (LR: 0.829; SVM: 0.830; DNN: 0.837) and clinical scores (Simplified Acute Physiology Score II: 0.728; Sequential Organ Failure Assessment: 0.728; Oxford Acute Severity of Illness Score: 0.738; Glasgow Coma Scale: 0.691). XGBoost's hospital mortality prediction achieved AUROC 0.873, sensitivity 0.818, accuracy 0.777, specificity 0.768, and F1 score 0.551. We crafted an interpretable model for sepsis death risk prediction. ML algorithms surpassed traditional scores for sepsis mortality forecast. Validation in a Chinese teaching hospital echoed these findings.
RESUMEN
As cities continue to grow globally, characterizing the built environment is essential to understanding human populations, projecting energy usage, monitoring urban heat island impacts, preventing environmental degradation, and planning for urban development. Buildings are a key component of the built environment and there is currently a lack of data on building height at the global level. Current methodologies for developing building height models that utilize remote sensing are limited in scale due to the high cost of data acquisition. Other approaches that leverage 2D features are restricted based on the volume of ancillary data necessary to infer height. Here, we find, through a series of experiments covering 74.55 million buildings from the United States, France, and Germany, it is possible, with 95% accuracy, to infer building height within 3 m of the true height using footprint morphology data. Our results show that leveraging individual building footprints can lead to accurate building height predictions while not requiring ancillary data, thus making this method applicable wherever building footprints are available. The finding that it is possible to infer building height from footprint data alone provides researchers a new method to leverage in relation to various applications.
RESUMEN
BACKGROUND: Intracranial haemorrhage (ICH) poses a significant threat to patients on Direct Oral Anticoagulants (DOACs), with existing risk scores inadequately predicting ICH risk in these patients. We aim to develop and validate a predictive model for ICH risk in DOAC-treated patients. METHODS: 24,794 patients treated with a DOAC were identified in a province-wide electronic medical and health data platform in Tianjin, China. The cohort was randomly split into a 4:1 ratio for model development and validation. We utilized forward stepwise selection, Least Absolute Shrinkage and Selection Operator (LASSO), and eXtreme Gradient Boosting (XGBoost) to select predictors. Model performance was compared using the area under the curve (AUC) and net reclassification index (NRI). The optimal model was stratified and compared with the DOAC model. RESULTS: The median age is 68.0 years, and 50.4% of participants are male. The XGBoost model, incorporating six independent factors (history of hemorrhagic stroke, peripheral artery disease, venous thromboembolism, hypertension, age, low-density lipoprotein cholesterol levels), demonstrated superior performance in the development dateset. It showed moderate discrimination (AUC: 0.68, 95% CI: 0.64-0.73), outperforming existing DOAC scores (ΔAUC = 0.063, P = 0.003; NRI = 0.374, P < 0.001). Risk categories significantly stratified ICH risk (low risk: 0.26%, moderate risk: 0.74%, high risk: 5.51%). Finally, the model demonstrated consistent predictive performance in the internal validation. CONCLUSION: In a real-world Chinese population using DOAC therapy, this study presents a reliable predictive model for ICH risk. The XGBoost model, integrating six key risk factors, offers a valuable tool for individualized risk assessment in the context of oral anticoagulation therapy.
Asunto(s)
Hemorragias Intracraneales , Humanos , Masculino , Femenino , Anciano , Hemorragias Intracraneales/inducido químicamente , Persona de Mediana Edad , Administración Oral , Anticoagulantes/efectos adversos , Anticoagulantes/uso terapéutico , Anticoagulantes/administración & dosificación , Factores de Riesgo , Medición de Riesgo/métodosRESUMEN
Asthma is a chronic inflammatory airway disease with significant burden; exacerbations can severely affect quality of life and healthcare costs. Advances in big data analysis and artificial intelligence have made it easier to predict future exacerbations more accurately. This study used an integrated dataset of Korean National Health Insurance, meteorological, air pollution, and viral data from national public databases to develop a model to predict asthma exacerbations on a daily basis in South Korea. We merged these sources and applied random forest, AdaBoost, XGBoost, and LightGBM machine learning models to compare their performances at predicting future exacerbations. Of the models, XGBoost (AUROC of 0.68 and accuracy of 0.96) and LightGBM (AUROC of 0.67 and accuracy of 0.96) were the most promising. Common important variables were the number of visits and exacerbations per year, and medical resource utilization, including the prescription of asthma medications. Comorbid diabetes, hypertension, gastroesophageal reflux, arthritis, metabolic syndrome, osteoporosis, and ischemic heart disease were also associated with elevated exacerbation risk. The models examined in this study highlight the importance of previous exacerbations, use of medical resources, and comorbidities in the prediction of future exacerbations in patients with asthma.
RESUMEN
Accurately mapping ground-level ozone concentrations at high spatiotemporal resolution (daily, 1 km) is essential for evaluating human exposure and conducting public health assessments. This requires identifying and understanding a proxy that is well-correlated with ground-level ozone variation and available with spatiotemporal high-resolution data. This study introduces a high-resolution ozone modeling method utilizing the XGBoost algorithm with satellite-derived land surface temperature (LST) as the primary predictor. Focusing on China in 2019, our model achieved a cross-validation R2 of 0.91 and a root-mean-square error (RMSE) of 13.51 µg/m3. We provide detailed maps highlighting ground-level ozone concentrations in urban areas, uncovering spatial variations previously unresolved, along with time series aligning with established understandings of ozone dynamics. Our local interpretation of the machine learning model underscores the significant contribution of LST to spatiotemporal ozone variations, surpassing other meteorological, pollutant, and geographical predictors in its influence. Validation results indicate that model performance decreases as spatial resolution becomes coarser, with R2 decreasing from 0.91 for the 1 km model to 0.85 for the 25 km model. The methodology and data sets generated by this study offer new insights into ground-level ozone variability and mapping and can significantly aid in exposure assessment and epidemiological research related to this critical environmental challenge.
RESUMEN
Intravenous ganciclovir and oral valganciclovir display significant variability in ganciclovir pharmacokinetics, particularly in children. Therapeutic drug monitoring currently relies on the area under the concentration-time (AUC). Machine-learning (ML) algorithms represent an interesting alternative to Maximum-a-Posteriori Bayesian-estimators for AUC estimation. The goal of our study was to develop and validate an ML-based limited sampling strategy (LSS) approach to determine ganciclovir AUC0-24 after administration of either intravenous ganciclovir or oral valganciclovir in children. Pharmacokinetic parameters from four published population pharmacokinetic models, in addition to the World Health Organization growth curve for children, were used in the mrgsolve R package to simulate 10,800 pharmacokinetic profiles of children. Different ML algorithms were trained to predict AUC0-24 based on different combinations of two or three samples. Performances were evaluated in a simulated test set and in an external data set of real patients. The best estimation performances in the test set were obtained with the Xgboost algorithm using a 2 and 6 hours post dose LSS for oral valganciclovir (relative mean prediction error [rMPE] = 0.4% and relative root mean square error [rRMSE] = 5.7%) and 0 and 2 hours post dose LSS for intravenous ganciclovir (rMPE = 0.9% and rRMSE = 12.4%). In the external data set, the performance based on these two sample LSS was acceptable: rMPE = 0.2% and rRMSE = 16.5% for valganciclovir and rMPE = -9.7% and rRMSE = 17.2% for intravenous ganciclovir. The Xgboost algorithm developed resulted in a clinically relevant individual estimation using only two blood samples. This will improve the implementation of AUC-targeted ganciclovir therapeutic drug monitoring in children.
RESUMEN
Current medical research has been demonstrating the roles of miRNAs in a variety of cellular mechanisms, lending credence to the association between miRNA dysregulation and multiple diseases. Understanding the mechanisms of miRNA is critical for developing effective diagnostic and therapeutic strategies. miRNA-mRNA interactions emerge as the most important mechanism to be understood despite their experimental validation constraints. Accordingly, several computational models have been developed to predict miRNA-mRNA interactions, albeit presenting limited predictive capabilities, poor characterisation of miRNA-mRNA interactions, and low usability. To address these drawbacks, we developed PRIMITI, a PRedictive model for the Identification of novel miRNA-Target mRNA Interactions. PRIMITI is a novel machine learning model that utilises CLIP-seq and expression data to characterise functional target sites in 3'-untranslated regions (3'-UTRs) and predict miRNA-target mRNA repression activity. The model was trained using a reliable negative sample selection approach and the robust extreme gradient boosting (XGBoost) model, which was coupled with newly introduced features, including sequence and genetic variation information. PRIMITI achieved an area under the receiver operating characteristic (ROC) curve (AUC) up to 0.96 for a prediction of functional miRNA-target site binding and 0.96 for a prediction of miRNA-target mRNA repression activity on cross-validation and an independent blind test. Additionally, the model outperformed state-of-the-art methods in recovering miRNA-target repressions in an unseen microarray dataset and in a collection of validated miRNA-mRNA interactions, highlighting its utility for preliminary screening. PRIMITI is available on a reliable, scalable, and user-friendly web server at https://biosig.lab.uq.edu.au/primiti.
RESUMEN
Falls among the elderly population pose significant health risks, often leading to morbidity and decreased quality of life. Traditional fall detection methods, namely wearable devices and cameras, have limitations such as lighting conditions and privacy concerns. Radar-based fall detection has emerged as a promising alternative, offering unobtrusive technique. In this study, an attempt has been made to classify fall detection using smoothed pseudo wigner-ville distribution (SPWVD) images and XGBoost learning. For this, online publicly available radar database (N=15) is considered. Radar signals is employed to SPWVD for time-frequency representation images. Ten features are extracted and applied to XGBoost learning. Experiments are performed and performance is evaluated using 10-fold cross validation. The proposed approach is able to discriminate elderly fall. Using XGBoost learning, the approach yields a maximum average classification accuracy, f1-score, precision, sensitivity, specificity, and kappa scores of 87.47%, 87.38%, 88.12%, 86.81%, 88.31% and 74.94% respectively. The combination of conventional features with concentration measures and median frequency obtained the second best performance. Thus, the proposed framework could be utilized for accurate and efficient detection of falls among the elderly population in their private spaces.
Asunto(s)
Accidentes por Caídas , Radar , Humanos , Anciano , Aprendizaje Automático , Anciano de 80 o más Años , Algoritmos , Monitoreo Ambulatorio/métodosRESUMEN
The overwhelming volume of patients in emergency departments (EDs) is a significant problem that hinders the delivery of high quality healthcare. Despite their great value, triage protocols are challenging to put into practice. This paper examines the utility of prediction models as a tool for clinical decision support, with a focus on medium-severity patients as defined by the ESI algorithm. 689 cases of medium-risk patients were gathered from the AHEPA hospital, evaluated, and their data fed three classifiers: XGBoost (XGB), Random Forest (RF) and Logistic Regression (LR), with the prediction goal being the outcome of their visit, i.e. admission or discharge. Essential features for the prediction task were determined using feature importance and distribution analysis. Despite having many missing values or high sparsity datasets, several symptoms and metrics are recommended as crucial for outcome prediction. When fed the patients' vital signs, XGB achieved an accuracy score of 91.30%. Several chief complaints were also proven beneficial. Prediction models can, in general, not only lessen the drawbacks of triage implementation, but also enhance its delivery.
Asunto(s)
Sistemas de Apoyo a Decisiones Clínicas , Servicio de Urgencia en Hospital , Triaje , Humanos , Inteligencia Artificial , Algoritmos , Alta del PacienteRESUMEN
BACKGROUND: Diabetic ketoacidosis (DKA) is the leading cause of morbidity and mortality in pediatric type 1 diabetes (T1D), occurring in approximately 20% of patients, with an economic cost of $5.1 billion/year in the United States. Despite multiple risk factors for postdiagnosis DKA, there is still a need for explainable, clinic-ready models that accurately predict DKA hospitalization in established patients with pediatric T1D. OBJECTIVE: We aimed to develop an interpretable machine learning model to predict the risk of postdiagnosis DKA hospitalization in children with T1D using routinely collected time-series of electronic health record (EHR) data. METHODS: We conducted a retrospective case-control study using EHR data from 1787 patients from among 3794 patients with T1D treated at a large tertiary care US pediatric health system from January 2010 to June 2018. We trained a state-of-the-art; explainable, gradient-boosted ensemble (XGBoost) of decision trees with 44 regularly collected EHR features to predict postdiagnosis DKA. We measured the model's predictive performance using the area under the receiver operating characteristic curve-weighted F1-score, weighted precision, and recall, in a 5-fold cross-validation setting. We analyzed Shapley values to interpret the learned model and gain insight into its predictions. RESULTS: Our model distinguished the cohort that develops DKA postdiagnosis from the one that does not (P<.001). It predicted postdiagnosis DKA risk with an area under the receiver operating characteristic curve of 0.80 (SD 0.04), a weighted F1-score of 0.78 (SD 0.04), and a weighted precision and recall of 0.83 (SD 0.03) and 0.76 (SD 0.05) respectively, using a relatively short history of data from routine clinic follow-ups post diagnosis. On analyzing Shapley values of the model output, we identified key risk factors predicting postdiagnosis DKA both at the cohort and individual levels. We observed sharp changes in postdiagnosis DKA risk with respect to 2 key features (diabetes age and glycated hemoglobin at 12 months), yielding time intervals and glycated hemoglobin cutoffs for potential intervention. By clustering model-generated Shapley values, we automatically stratified the cohort into 3 groups with 5%, 20%, and 48% risk of postdiagnosis DKA. CONCLUSIONS: We have built an explainable, predictive, machine learning model with potential for integration into clinical workflow. The model risk-stratifies patients with pediatric T1D and identifies patients with the highest postdiagnosis DKA risk using limited follow-up data starting from the time of diagnosis. The model identifies key time points and risk factors to direct clinical interventions at both the individual and cohort levels. Further research with data from multiple hospital systems can help us assess how well our model generalizes to other populations. The clinical importance of our work is that the model can predict patients most at risk for postdiagnosis DKA and identify preventive interventions based on mitigation of individualized risk factors.
RESUMEN
In this study, 80 carbapenem-resistant Klebsiella pneumoniae (CR-KP) and 160 carbapenem-susceptible Klebsiella pneumoniae (CS-KP) strains detected in the clinic were selected and their matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) peaks were collected. K-means clustering was performed on the MS peak data to obtain the best "feature peaks", and four different machine learning models were built to compare the area under the ROC curve, specificity, sensitivity, test set score, and ten-fold cross-validation score of the models. By adjusting the model parameters, the test efficacy of the model is increased on the basis of reducing model overfitting. The area under the ROC curve of the Random Forest, Support Vector Machine, Logistic Regression, and Xgboost models used in this study are 0.99, 0.97, 0.96, and 0.97, respectively; the model scores on the test set are 0.94, 0.91, 0.90, and 0.93, respectively; and the results of the ten-fold cross-validation are 0.84, 0.81, 0.81, and 0.85, respectively. Based on the machine learning algorithms and MALDI-TOF MS assay data can realize rapid detection of CR-KP, shorten the in-laboratory reporting time, and provide fast and reliable identification results of CR-KP and CS-KP.