Senior Design Projects

ECS193 A/B Winter & Spring 2023

Predicting Complications of Diabetes Using Electronic Health Records and Ocular Imaging

Email **********
Parisa Emami, MD, MPH
Department of Ophthalmology

Project's details

Predicting Complications of Diabetes Using Electronic Health Records and Ocular Imaging
Diabetes is one of the most common conditions causing disability, including heart attack, visual impairment, etc., among children and adults. It has been estimated that ~10% of the US population have diabetes and more than 60% of these patients develop various degrees of diabetes complications, including eye disease (retinopathy), cardiovascular disease, renal failure, neuropathy, and foot ulcers. Various guidelines have been developed to identify individuals at risk of complications and encourage lab testing and regular exam in these individuals. However, these guidelines are generally very broad and fail to predict individual patient’s risk for developing each of these conditions. Moreover, none of the screening guidelines takes into consideration social determinants of health and other factors that has shown to affect risk of complications. Given that these guidelines do not personalize recommendations for individual patient’s risk level, adherence has been limited. More personalized guidelines and risk stratifications based on individual patients’ risk factors and social factors are needed to improve outcomes and prevent long-term complications.
UC Davis electronic medical records contain longitudinal data (lab results, physician notes, exam findings, ocular imaging) on a large population of patients referred to UCD Health. This data will be used to predict 3-year and 5-year risk of developing various complications of diabetes. In this project we are aiming to use this large database and extract variables including demographics, lab findings, physician notes (structures and unstructured free-text notes), ocular imaging, etc. to develop a machine learning model to predict patient’s risk of developing each of the above mentioned complications (hundreds of input variables). This project requires a multidisciplinary approach. Team will consist of physicians (to extract and curate the database) and computer science students to develop the machine learning algorithm. Extracting data, curating database and labelling data is well underway at this time and will be ready to use by CS students in January.
N/A
Final product for this project will be a machine learning algorithm that can predict the risk of developing various complications of diabetes, including severe diabetic retinopathy, renal failure, amputation, and heart attack at 3 and 5 years follow up. This algorithm will later be used to predict risk in a larger database of patients with diabetes (including administrative claims databases and biobanks).
**********
30-60 min weekly or more
Client wishes to keep IP of the project
Attachment N/A
No
Team members N/A
N/A
N/A