Narrow Results

Diabetes is a chronic disease that poses a serious threat to health, and its early risk prediction has been a hot research topic in the field of medical artificial intelligence. Routine medical checkups are the most common way to monitor people’s health status, and the data from medical checkups contain rich diagnostic information, which is valuable for diabetes risk prediction. Currently, most of the available studies on diabetes risk prediction are based on publicly available datasets, and the models and algorithms do not work well on real clinical datasets. Real routine checkup data are characterized by complex information, diverse features, high redundancy and poor balance, which pose great challenges for diabetes risk prediction. To address this problem, this paper proposes a multi-strategy data augmentation-based diabetes risk prediction method, after completing data pre-processing and feature selection, a counterfactual-based data balancing strategy is used to augment a minority class of instances, and a density clustering-based supplemental counterfactual data augmentation strategy is proposed to address the problem of insufficient representation of generated instances in the counterfactual method. Moreover, the uncertainty-weighted method is used in the model training phase. Based on the real checkup dataset, five machine learning methods including Logistic Regression (LR), SVM, Decision Tree, Random Forest and Gradient Boosting are used to model and use 5-fold cross-validation to carry out diabetes risk assessment and prediction. The experimental results showed that the sensitivity and precision of the models were significantly improved compared with the existing methods, and the sensitivity of the LR model for diabetes risk prediction on the real routine checkup dataset reached more than 90$\%$, which meet the requirements of clinical application.

Diabetes is a global health concern that affects people of all races. With different uncertainties in human lifestyles, it is difficult to predict diabetes while assuming that the risk patterns are the same for all. The likelihood of diabetes in a patient is mostly predicted using machine learning (ML) models on features explicitly available in datasets, while the intrinsic relationship between features viz-a-viz their potential relevance to the presence of diabetes is oftentimes neglected. In this work, we explored feature importance and correlation to derive the top 15 feature pairs from a dataset of 263,882 samples of anonymized patient information. These top-15 feature pairs were fed into five different ML models (decision tree (DT), neural networks (NN), random forest (RF), support vector machine (SVM) and extreme gradient boosting (XGB)) for predicting the likelihood of diabetes, while also feeding the direct features (without correlated pairing) separately into the same 5ML models. The models’ performances were evaluated using accuracy, precision, recall and F1-score and NN presented the best performance overall achieving an F1-score of 85% for the correlated feature pairs (CF) and 75% for the direct feature pairs. The results confirm the importance of the correlation/relationship between features in predicting the likelihood of diabetes in patients more accurately.

This PSB 2023 session discusses challenges in clinical implication and application of risk prediction models, which includes but is not limited to: implementation of risk models, responsible use of polygenic risk scores (PGS), and other risk prediction strategies. We focus on the development and use of new, scalable methods for harmonizing and refining risk prediction models by incorporating genetic and non-genetic risk factors, applying new phenotyping strategies, and integrating clinical factors and biomarkers. Lastly, we will discuss innovation in expanding the utility of these prediction models to underrepresented populations. This session focuses on the overarching theme of enabling early diagnosis, and treatment and preventive measures related to complex diseases and comorbidities.

Polygenic risk scores (PRS) have led to enthusiasm for precision medicine. However, it is well documented that PRS do not generalize across groups differing in ancestry or sample characteristics e.g., age. Quantifying performance of PRS across different groups of study participants, using genome-wide association study (GWAS) summary statistics from multiple ancestry groups and sample sizes, and using different linkage disequilibrium (LD) reference panels may clarify which factors are limiting PRS transferability. To evaluate these factors in the PRS generation process, we generated body mass index (BMI) PRS (PRS_BMI) in the Electronic Medical Records and Genomics (eMERGE) network (N=75,661). Analyses were conducted in two ancestry groups (European and African) and three age ranges (adult, teenagers, and children). For PRS_BMI calculations, we evaluated five LD reference panels and three sets of GWAS summary statistics of varying sample size and ancestry. PRS_BMI performance increased for both African and European ancestry individuals using cross-ancestry GWAS summary statistics compared to European-only summary statistics (6.3% and 3.7% relative R² increase, respectively, p_African=0.038, p_European=6.26x10^-4). The effects of LD reference panels were more pronounced in African ancestry study datasets. PRS_BMI performance degraded in children; R² was less than half of teenagers or adults. The effect of GWAS summary statistics sample size was small when modeled with the other factors. Additionally, the potential of using a PRS generated for one trait to predict risk for comorbid diseases is not well understood especially in the context of cross-ancestry analyses – we explored clinical comorbidities from the electronic health record associated with PRS_BMI and identified significant associations with type 2 diabetes and coronary atherosclerosis. In summary, this study quantifies the effects that ancestry, GWAS summary statistic sample size, and LD reference panel have on PRS performance, especially in cross-ancestry and age-specific analyses.

The primary efforts of disease and epidemiological research can be divided into two areas: identifying the causal mechanisms and utilizing important variables for risk prediction. The latter is generally perceived as a more obtainable goal due to the vast number of readily available tools and the faster pace of obtaining results. However, the lower barrier of entry in risk prediction means that it is easy to make predictions, yet it is incredibility more difficult to make sound predictions. As an ever-growing amount of data is being generated, developing risk prediction models and turning them into clinically actionable findings is crucial as the next step. However, there are still sizable gaps before risk prediction models can be implemented clinically. While clinicians are eager to embrace new ways to improve patients’ care, they are overwhelmed by a plethora of prediction methods. Thus, the next generation of prediction models will need to shift from making simple predictions towards interpretable, equitable, explainable and ultimately, casual predictions.