Development and validation of a machine learning-augmented algorithm for diabetes screening in community and primary care settings: A population-based study

Abstract

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.

Development and validation of machine learning-augmented algorithm for insulin sensitivity assessment in the community and primary care settings: a population-based study in China

“…Our results indicated that the inclusion of laboratory variables in the prediction models significantly improved the performances compared to the models developed exclusively with non-laboratory variables, suggesting the considerable role of laboratory variables in the insulin sensitivity assessment models developed by machine learning. These results seemed consistent with the findings in previous research that the inclusion of laboratory variables such as urinary glucose, urinary vitamin C, and FPG improves the accuracy of diabetes prediction models developed by GBM or LightGBM ( 42 , 45 ). Noteworthy, the laboratory features (e.g., fasting glucose, serum lipids, liver enzymes, etc.)…”

“…Additionally, our results showed that the streamlined LightGBM models utilizing the top-20 ranked variables exhibited comparable performances to the models constructed with all features. Likewise, our previous study indicated that the streamlined diabetes prediction model utilizing the top 20 variables developed by the LightGBM algorithm, exhibited great performances similar to that of the model using all variables (45), which is consistent with the findings in the present study. Moreover, it is reported that the simplified model for gestational diabetes prediction using 9 variables based on LightGBM demonstrates only a modest reduction in predictive accuracy compared to its full variable model (38).…”

“…), can be promptly obtained through easy-to-use electronic questionnaires or open-access web pages. Moreover, the ML models could be further developed and presented in open and accessible web pages to make them easier and more available for residents in communities or clinical practitioners in the setting of primary care to evaluate the risk of low insulin sensitivity, and the information increasingly inputted in the ML model would be extremely helpful to improve the performance of the prediction models in turn, which is one of the biggest advantages of ML ( 45 ).…”

“…), can be promptly obtained through easy-to-use electronic questionnaires or open-access web pages. Moreover, the ML models could be further developed and presented in open and accessible web pages to make them easier and more available for residents in communities or clinical practitioners in the setting of primary care to evaluate the risk of low insulin sensitivity, and the information increasingly inputted in the ML model would be extremely helpful to improve the performance of the prediction models in turn, which is one of the biggest advantages of ML (45). However, it should be noted that although HOMA-IR is commonly used as a surrogate indicator for insulin sensitivity assessment in clinical and epidemiological studies (16), we were unable to use the gold standard -the hyperinsulinemic-euglycemic clamp, to determine insulin sensitivity in this study, so the findings of the current study should be interpreted with caution and further research would be necessary.…”

Artificial Intelligence in the Health Sector: Key Skills for Future Health Professionals. (Preprint)

“…Healthcare intrinsically involves significant human interaction, where empathy and active listening are paramount. Implementing AI technologies complements human interaction by mitigating cognitive biases such as unconscious reasoning, gut feelings, and heuristic shortcuts through the application of AI algorithms (7).…”