Using electronic health records to develop and validate a machine-learning tool to predict type 2 diabetes outcomes: a study protocol

Machine learning techniques to predict the risk of developing diabetic nephropathy: a literature review

Purpose

Diabetes is a major public health challenge with widespread prevalence, often leading to complications such as Diabetic Nephropathy (DN)-a chronic condition that progressively impairs kidney function. In this context, it is important to evaluate if Machine learning models can exploit the inherent temporal factor in clinical data to predict the risk of developing DN faster and more accurately than current clinical models.

Methods

Three different databases were used for this literature review: Scopus, Web of Science, and PubMed. Only articles written in English and published between January 2015 and December 2022 were included.

Results

We included 11 studies, from which we discuss a number of algorithms capable of extracting knowledge from clinical data, incorporating dynamic aspects in patient assessment, and exploring their evolution over time. We also present a comparison of the different approaches, their performance, advantages, disadvantages, interpretation, and the value that the time factor can bring to a more successful prediction of diabetic nephropathy.

Conclusion

Our analysis showed that some studies ignored the temporal factor, while others partially exploited it. Greater use of the temporal aspect inherent in Electronic Health Records (EHR) data, together with the integration of omics data, could lead to the development of more reliable and powerful predictive models.

Subtle variation in sepsis-III definitions markedly influences predictive performance within and across methods

Early detection of sepsis is key to ensure timely clinical intervention. Since very few end-to-end pipelines are publicly available, fair comparisons between methodologies are difficult if not impossible. Progress is further limited by discrepancies in the reconstruction of sepsis onset time. This retrospective cohort study highlights the variation in performance of predictive models under three subtly different interpretations of sepsis onset from the sepsis-III definition and compares this against inter-model differences. The models are chosen to cover tree-based, deep learning, and survival analysis methods. Using the MIMIC-III database, between 867 and 2178 intensive care unit admissions with sepsis were identified, depending on the onset definition. We show that model performance can be more sensitive to differences in the definition of sepsis onset than to the model itself. Given a fixed sepsis definition, the best performing method had a gain of 1-5% in the area under the receiver operating characteristic (AUROC). However, the choice of onset time can cause a greater effect, with variation of 0-6% in AUROC. We illustrate that misleading conclusions can be drawn if models are compared without consideration of the sepsis definition used which emphasizes the need for a standardized definition for sepsis onset.

Extracting patient-level data from the electronic health record: Expanding opportunities for health system research

BACKGROUND

Epidemiological studies of interstitial lung disease (ILD) are limited by small numbers and tertiary care bias. Investigators have leveraged the widespread use of electronic health records (EHRs) to overcome these limitations, but struggle to extract patient-level, longitudinal clinical data needed to address many important research questions. We hypothesized that we could automate longitudinal ILD cohort development using the EHR of a large, community-based healthcare system.

STUDY DESIGN AND METHODS

We applied a previously validated algorithm to the EHR of a community-based healthcare system to identify ILD cases between 2012-2020. We then extracted disease-specific characteristics and outcomes using fully automated data-extraction algorithms and natural language processing of selected free-text.

RESULTS

We identified a community cohort of 5,399 ILD patients (prevalence = 118 per 100,000). Pulmonary function tests (71%) and serologies (54%) were commonly used in the diagnostic evaluation, whereas lung biopsy was rare (5%). IPF was the most common ILD diagnosis (n = 972, 18%). Prednisone was the most commonly prescribed medication (911, 17%). Nintedanib and pirfenidone were rarely prescribed (n = 305, 5%). ILD patients were high-utilizers of inpatient (40%/year hospitalized) and outpatient care (80%/year with pulmonary visit), with sustained utilization throughout the post-diagnosis study period.

DISCUSSION

We demonstrated the feasibility of robustly characterizing a variety of patient-level utilization and health services outcomes in a community-based EHR cohort. This represents a substantial methodological improvement by alleviating traditional constraints on the accuracy and clinical resolution of such ILD cohorts; we believe this approach will make community-based ILD research more efficient, effective, and scalable.

Predicting the 2-Year Risk of Progression from Prediabetes to Diabetes Using Machine Learning among Chinese Elderly Adults

Identifying people with a high risk of developing diabetes among those with prediabetes may facilitate the implementation of a targeted lifestyle and pharmacological interventions. We aimed to establish machine learning models based on demographic and clinical characteristics to predict the risk of incident diabetes. We used data from the free medical examination service project for elderly people who were 65 years or older to develop logistic regression (LR), decision tree (DT), random forest (RF), and extreme gradient boosting (XGBoost) machine learning models for the follow-up results of 2019 and 2020 and performed internal validation. The receiver operating characteristic (ROC), sensitivity, specificity, accuracy, and F1 score were used to select the model with better performance. The average annual progression rate to diabetes in prediabetic elderly people was 14.21%. Each model was trained using eight features and one outcome variable from 9607 prediabetic individuals, and the performance of the models was assessed in 2402 prediabetes patients. The predictive ability of four models in the first year was better than in the second year. The XGBoost model performed relatively efficiently (ROC: 0.6742 for 2019 and 0.6707 for 2020). We established and compared four machine learning models to predict the risk of progression from prediabetes to diabetes. Although there was little difference in the performance of the four models, the XGBoost model had a relatively good ROC value, which might perform well in future exploration in this field.