Search

This paper gives a performance analysis of multiple vote classifiers based on metaclassification methods for estimating the risk of diabetes. The study's dataset includes a number of biological and clinical risk variables that can result in the development of diabetes. In the analysis, classifiers like Random Forest, Logistic Regression, Gradient Boosting, Support Vector Machines, and Artificial Neural Networks were used. In the study, each classifier was trained and evaluated separately, and the outcomes were compared to those attained using meta-classification methods. Some of the meta-classifiers used in the analysis included Majority Voting, Weighted Majority Voting, and Stacking. The effectiveness of each classifier was evaluated using a number of measures, including accuracy, precision, recall, F1-score, and Area under the Curve (AUC). The results show that meta-classification techniques often outperform solo classifiers in terms of prediction precision. Random Forest and Gradient Boosting, two different classifiers, had the highest accuracy, while Logistic Regression performed the worst. The best performing meta-classifier was stacking, which achieved an accuracy of 84.25%. Weighted Majority Voting came in second (83.86%) and Majority Voting came in third (82.95%). [ABSTRACT FROM AUTHOR]

Abstract Background Diabetes mellitus (DM) is a chronic metabolic disease that could produce severe complications threatening life. Its early detection is thus quite important for the timely prevention and treatment. Normally, fasting blood glucose (FBG) by physical examination is used for large-scale screening of DM; however, some people with normal fasting glucose (NFG) actually have suffered from diabetes but are missed by the examination. This study aimed to investigate whether common physical examination indexes for diabetes can be used to identify the diabetes individuals from the populations with NFG. Methods The physical examination data from over 60,000 individuals with NFG in three Chinese cohorts were used. The diabetes patients were defined by HbA1c ≥ 48 mmol/mol (6.5%). We constructed the models using multiple machine learning methods, including logistic regression, random forest, deep neural network, and support vector machine, and selected the optimal one on the validation set. A framework using permutation feature importance algorithm was devised to discover the personalized risk factors. Results The prediction model constructed by logistic regression achieved the best performance with an AUC, sensitivity, and specificity of 0.899, 85.0%, and 81.1% on the validation set and 0.872, 77.9%, and 81.0% on the test set, respectively. Following feature selection, the final classifier only requiring 13 features, named as DRING (diabetes risk of individuals with normal fasting glucose), exhibited reliable performance on two newly recruited independent datasets, with the AUC of 0.964 and 0.899, the balanced accuracy of 84.2% and 81.1%, the sensitivity of 100% and 76.2%, and the specificity of 68.3% and 86.0%, respectively. The feature importance ranking analysis revealed that BMI, age, sex, absolute lymphocyte count, and mean corpuscular volume are important factors for the risk stratification of diabetes. With a case, the framework for identifying personalized risk factors revealed FBG, age, and BMI as significant hazard factors that contribute to an increased incidence of diabetes. DRING webserver is available for ease of application ( http://www.cuilab.cn/dring ). Conclusions DRING was demonstrated to perform well on identifying the diabetes individuals among populations with NFG, which could aid in early diagnosis and interventions for those individuals who are most likely missed.

Background: Non-invasive diabetes risk models are a cost-effective tool in large-scale population screening to identify those who need confirmation tests, especially in resource-limited settings. Aims: This study aimed to evaluate the ability of six non-invasive risk models (Cambridge, FINDRISC, Kuwaiti, Omani, Rotterdam, and SUNSET model) to identify screen-detected diabetes (defined by HbA1c) among Ghanaian migrants and non-migrants. Study design: A multicentered cross-sectional study. Methods: This analysis included 4843 Ghanaian migrants and non-migrants from the Research on Obesity and Diabetes among African Migrants (RODAM) Study. Model performance was assessed using the area under the receiver operating characteristic curves (AUC), Hosmer-Lemeshow statistics, and calibration plots. Results: All six models had acceptable discrimination (0.70 ≤ AUC

Background: Diabetes mellitus (DM) is a chronic metabolic disease that could produce severe complications threatening life. Its early detection is thus quite important for the timely prevention and treatment. Normally, fasting blood glucose (FBG) by physical examination is used for large-scale screening of DM; however, some people with normal fasting glucose (NFG) actually have suffered from diabetes but are missed by the examination. This study aimed to investigate whether common physical examination indexes for diabetes can be used to identify the diabetes individuals from the populations with NFG. Methods: The physical examination data from over 60,000 individuals with NFG in three Chinese cohorts were used. The diabetes patients were defined by HbA1c ≥ 48 mmol/mol (6.5%). We constructed the models using multiple machine learning methods, including logistic regression, random forest, deep neural network, and support vector machine, and selected the optimal one on the validation set. A framework using permutation feature importance algorithm was devised to discover the personalized risk factors. Results: The prediction model constructed by logistic regression achieved the best performance with an AUC, sensitivity, and specificity of 0.899, 85.0%, and 81.1% on the validation set and 0.872, 77.9%, and 81.0% on the test set, respectively. Following feature selection, the final classifier only requiring 13 features, named as DRING (diabetes risk of individuals with normal fasting glucose), exhibited reliable performance on two newly recruited independent datasets, with the AUC of 0.964 and 0.899, the balanced accuracy of 84.2% and 81.1%, the sensitivity of 100% and 76.2%, and the specificity of 68.3% and 86.0%, respectively. The feature importance ranking analysis revealed that BMI, age, sex, absolute lymphocyte count, and mean corpuscular volume are important factors for the risk stratification of diabetes. With a case, the framework for identifying personalized risk factors revealed FBG, age, and BMI as significant hazard factors that contribute to an increased incidence of diabetes. DRING webserver is available for ease of application (http://www.cuilab.cn/dring). Conclusions: DRING was demonstrated to perform well on identifying the diabetes individuals among populations with NFG, which could aid in early diagnosis and interventions for those individuals who are most likely missed. [ABSTRACT FROM AUTHOR]

Background: Non-invasive diabetes risk models are a cost-effective tool in large-scale population screening to identify those who need confirmation tests, especially in resource-limited settings., Aims: This study aimed to evaluate the ability of six non-invasive risk models (Cambridge, FINDRISC, Kuwaiti, Omani, Rotterdam, and SUNSET model) to identify screen-detected diabetes (defined by HbA1c) among Ghanaian migrants and non-migrants., Study Design: A multicentered cross-sectional study., Methods: This analysis included 4843 Ghanaian migrants and non-migrants from the Research on Obesity and Diabetes among African Migrants (RODAM) Study. Model performance was assessed using the area under the receiver operating characteristic curves (AUC), Hosmer-Lemeshow statistics, and calibration plots., Results: All six models had acceptable discrimination (0.70 ≤ AUC <0.80) for screen-detected diabetes in the overall/combined population. Model performance did not significantly differ except for the Cambridge model, which outperformed Rotterdam and Omani models. Calibration was poor, with a consistent trend toward risk overestimation for screen-detected diabetes, but this was substantially attenuated by recalibration through adjustment of the original model intercept., Conclusion: Though acceptable discrimination was observed, the original models were poorly calibrated among populations of African ancestry. Recalibration of these models among populations of African ancestry is needed before use., Competing Interests: By letter the authors declare no competing interests. And that all funding sources for this work have been declared., (© 2023 The Authors.)

Aims: Predicting incident diabetes could inform treatment strategies for diabetes prevention, but the incremental benefit of recalculating risk using updated risk factors is unknown. We used baseline and 1-year data from the Nateglinide and Valsartan in Impaired Glucose Tolerance Outcomes Research (NAVIGATOR) Trial to compare diabetes risk prediction using historical or updated clinical information.Methods: Among non-diabetic participants reaching 1year of follow-up in NAVIGATOR, we compared the performance of the published baseline diabetes risk model with a "landmark" model incorporating risk factors updated at the 1-year time point. The C-statistic was used to compare model discrimination and reclassification analyses to demonstrate the relative accuracy of diabetes prediction.Results: A total of 7527 participants remained non-diabetic at 1year, and 2375 developed diabetes during a median of 4years of follow-up. The C-statistic for the landmark model was higher (0.73 [95% CI 0.72-0.74]) than for the baseline model (0.67 [95% CI 0.66-0.68]). The landmark model improved classification to modest (<20%), moderate (20%-40%), and high (>40%) 4-year risk, with a net reclassification index of 0.14 (95% CI 0.10-0.16) and an integrated discrimination index of 0.01 (95% CI 0.003-0.013).Conclusions: Using historical clinical values to calculate diabetes risk reduces the accuracy of prediction. Diabetes risk calculations should be routinely updated to inform discussions about diabetes prevention at both the patient and population health levels. [ABSTRACT FROM AUTHOR]