Comprehensive Summary
Lockee et. al built a database named the D-data dock that uses data from multiple sources to make individual and population-level predictions for T1D patients, organize data, and support future research in diabetes. D-data dock was built in the Microsoft Azure cloud platform and was made to ingest and store health data, refine raw data, and report insights about the data to clinicians. Data was ingested, manually uploaded using an Extract-Load-Transform pattern, reformed into 1 format that conforms to T1D Exchange specifications, and visualized in the form of dashboards that showcase both individual and population metrics. The D-data dock took in and de-duplicated more than 10 million continuous glucose monitor (CGM) records of 1900 children from 5 different data sources, Clarit, Glooko, Libreview, TConnect, and Carelink, and now contains millions of CGM measurements dating back to 2014. Several case studies demonstrated the usefulness of this tool. The first was performed with an organization, Timely Interventions for Diabetes Excellence (TIDE), in which they used an open-source algorithm-enabled care model looking at CGM measurements to identify children at risk for deterioration. Using the D-data dock infrastructure, the TIDE model was able to be adopted within just two weeks demonstrating the advantage of the infrastructure’s use of raw CGM data. A second organization, Coin2Dose, provided patients with financial incentives based on the frequency of pre-meal insulin doses. This data was able to be easily uploaded to D-data dock, increased automation, and allowed for wider recruitment to meet study goals. Lastly, the D-data dock has helped create, implement, and rapidly test two machine learning models to predict patients’ 90-day change in A1C and 180-day risk of diabetic ketoacidosis (DKA). Overall, the study uses these examples to show how D-data dock can improve patient care, implement interventions efficiently, and drive innovation to improve the lives of patients with T1D.
Outcomes and Implications
Over the last 2 decades, there has been an uptick in the use of electronic health records (EHR) systems to record patient data. CGMs are noninvasive and provide incredibly useful information, including the need for changes to medication, diet, and automatic insulin delivery (AID) systems. The number of patients wearing CGMs has also increased the amount of available health data, but this data remains poorly integrated into a single database. Integrating data from CGMs into EHR systems would be beneficial for clinicians but has been limited by the number of different manufacturers and device types, immense amount of data in EHRs record, outdated technology, and inflexibility of EHRs in managing population health data, all of which the D-data dock improves on in some way. This study shows how D-data dock can further research in T1D and allow clinicians to make informed decisions about patient care from compiled population and individual data. The D-data dock serves as a blueprint for data integration in many medical fields, such as in adult patients with diabetes, and in the clinical setting of other conditions too. As it becomes implemented in medical practices, it can be an example of the standardization of clinical information between hospitals with profound impacts on the care and monitoring of patients with T1D.