New AI model uses readily available healthcare data to predict type 2 diabetes

Researchers have developed an advanced AI algorithm that predicts the onset of type 2 diabetes (T2D) up to five years in advance using routinely collected healthcare data, sharing their findings in JAMA Network Open.

The study’s authors explained that their AI model stands out because it was specifically designed to make its assessments at the population level.

“The main purpose of our model was to inform population health planning and management for the prevention of diabetes that incorporates health equity,” wrote lead author Mathieu Ravaut, MSc, of the University of Toronto, and colleagues. “It was not our goal for this model to be applied in the context of individual patient care.”

Ravaut et al. created their machine learning model using data from more than 2.1 million patients treated at a single health system in Ontario, Canada, from 2006 to 2016. While all patients were technically from the same region, the authors emphasized that Ontario is known for its especially diverse population.

The team’s algorithm was trained with data from more than 1.6 million patients, validating with data from more than 243,000 patients and tested with data from more than 236,000 patients.

The two-year medical histories of each patient—including their demographic information, prescription medication history and any lab results—were all used to fine-tune the algorithm.

Overall, the model achieved a test area under the ROC curve of 80.26 when predicting the onset of T2D within five years.

“Our model showed consistent calibration across sex, immigration status, racial/ethnic and material deprivation, and a low to moderate number of events in the health care history of the patient,” the authors wrote. “The cohort was representative of the whole population of Ontario, which is itself among the most diverse in the world. The model was well calibrated, and its discrimination, although with a slightly different end goal, was competitive with results reported in the literature for other machine learning–based studies that used more granular clinical data from electronic medical records without any modifications to the original test set distribution.”

Also, by continuously assessing the population’s risk of T2D, countries with thorough administrative databases could potentially boost patient care and target specific subgroups that may face an especially high risk of poor outcomes.

“Because our machine learning model included social determinants of health that are known to contribute to diabetes risk, our population-wide approach to risk assessment may represent a tool for addressing health disparities,” the authors added.

The full study from JAMA Network Open can be read here.

Around the web

Healthcare lags other industries in funding programs to combat cyberattacks, and successful breeches can leave imaging departments reeling.

In late May, the CDC warned of several dozen cases of heart Inflammation, typically occurring within four days after the second vaccine dose.

There was a sevenfold difference between the lowest and highest price for imaging of the abdomen, according to an analysis of data from more than 2,000 hospitals. 

Moffitt Cancer Center researchers used PET/CT features to help develop their deep learning tool, sharing their process on Wednesday.