BackgroundOpportunely screening for diabetes is crucial to reduce its related morbidity, mortality, and socioeconomic burden. Machine learning (ML) has excellent capability to maximize predictive accuracy. We aim to develop ML-augmented models for diabetes screening in community and primary care settings.Methods8425 participants were involved from a population-based study in Hubei, China since 2011. The dataset was split into a development set and a testing set. Seven different ML algorithms were compared to generate predictive models. Non-laboratory features were employed in the ML model for community settings, and laboratory test features were further introduced in the ML+lab models for primary care. The area under the receiver operating characteristic curve (AUC), area under the precision-recall curve (auPR), and the average detection costs per participant of these models were compared with their counterparts based on the New China Diabetes Risk Score (NCDRS) currently recommended for diabetes screening.ResultsThe AUC and auPR of the ML model were 0·697and 0·303 in the testing set, seemingly outperforming those of NCDRS by 10·99% and 64·67%, respectively. The average detection cost of the ML model was 12·81% lower than that of NCDRS with the same sensitivity (0·72). Moreover, the average detection cost of the ML+FPG model is the lowest among the ML+lab models and less than that of the ML model and NCDRS+FPG model.ConclusionThe ML model and the ML+FPG model achieved higher predictive accuracy and lower detection costs than their counterpart based on NCDRS. Thus, the ML-augmented algorithm is potential to be employed for diabetes screening in community and primary care settings.

Predicting all-cause mortality using available or conveniently modifiable risk factors is potentially crucial in reducing deaths precisely and efficiently. Framingham risk score (FRS) is widely used in predicting cardiovascular diseases, and its conventional risk factors are closely pertinent to deaths. Machine learning is increasingly considered to improve the predicting performances by developing predictive models. We aimed to develop the all-cause mortality predictive models using five machine learning (ML) algorithms (decision trees, random forest, support vector machine (SVM), XgBoost, and logistic regression) and determine whether FRS conventional risk factors are sufficient for predicting all-cause mortality in individuals over 40 years. Our data were obtained from a 10-year population-based prospective cohort study in China, including 9143 individuals over 40 years in 2011, and 6879 individuals followed-up in 2021. The all-cause mortality prediction models were developed using five ML algorithms by introducing all features available (182 items) or FRS conventional risk factors. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of the predictive models. The AUC and 95% confidence interval of the all-cause mortality prediction models developed by FRS conventional risk factors using five ML algorithms were 0.75 (0.726–0.772), 0.78 (0.755–0.799), 0.75 (0.731–0.777), 0.77 (0.747–0.792), and 0.78 (0.754–0.798), respectively, which is close to the AUC values of models established by all features (0.79 (0.769–0.812), 0.83 (0.807–0.848), 0.78 (0.753–0.798), 0.82 (0.796–0.838), and 0.85 (0.826–0.866), respectively). Therefore, we tentatively put forward that FRS conventional risk factors were potent to predict all-cause mortality using machine learning algorithms in the population over 40 years.