Two of the possible complications of type 2 diabetes are heart failure as well as diabetic kidney disease as the leading causes of disability or death for people with diabetes. While promising new medical interventions are available, identifying high-risk patients has been challenging. We’ll discuss in more detail below the latest developments from artificial intelligence (AI), machine learning aided by analytics and EHR that can help predict potential risks with far more accuracy.
Predicting diabetic heart failure risk with machine learning
As previously stated, one life-threatening complication that may be faced by patients with type 2 diabetes is the possibility of heart failure. However, recently a machine learning-based model risk score called WATCH-DM has been developed by Brigham and Women’s Hospital with UT Southwestern Medical Center according to a recent study reported in Diabetes Care. The scores were designed to help physicians identify top predictors of heart failure and have shown a high degree of accuracy and success in predicting risk.
- The research process and findings
The researchers followed 8,756 diabetic patients, from whom data was leveraged, including laboratory data, clinical EHR evaluations and population demographic data, finding 147 variables accurately predicting the risk of heart failure.
Over the next five years, 319 participating patients developed heart failure, allowing researchers to identify ten main risk factors for heart failure. These included age, weight, degree of diabetes control, hypertension, as well as other risk factors used to develop the WATCH-DM risk scores. Not surprisingly, it was found that patients with the highest scores had a five-year 20 percent heart failure risk.
Thanks to machine learning and analytics, data was extrapolated that did not require clinical cardiovascular biomarkers or even advanced imaging, according to one of the study’s authors.
- Advantages and limitations of WATCH-DM scores
One of the advantages of the WATCH-DM risk scores is there is no requirement for advanced imaging or specific cardiovascular biomarkers. Instead, the tool is available online for providers and other clinicians for convenient implementation for bedside practice as well as electronic health record systems to help identify at-risk patients who will benefit from earlier therapeutic interventions.
The risk score tool is designed to aid primary care physicians, nephrologists, endocrinologists, cardiologists and others who treat diabetic patients and deciding which strategies to best use in providing care. The healthcare industry spends billions annually on prevention services and treating complications and other conditions related to diabetes, so giving providers and their patients a chance to get ahead of any possible complications can not only improve outcomes but lower care costs as well. Because this risk score study provides essential information to diabetes patients wishing to effectively manage their condition, taking better control over their lifestyles and diet, among other things, can help improve their healthcare outcomes while lowering costs.
For example, BMI was one of the top indicators of risk of heart failure, which confirms the long-held position that excess weight increases future heart failure risk. By focusing on interventions involving lifestyle such as weight loss programs as well as using SGLT2 inhibitors, physicians can help patients be more proactive in self-managing their care.
One drawback, however, was that while the scores accurately predicted one type of heart failure, they were less accurate in predicting another kind. Researchers will therefore have to focus of more type-specific scores to better predict all forms of heart failure among diabetics.
Artificial Intelligence, machine learning and predicting diabetic kidney disease
Type 2 diabetes mellitus (T2DM) is a serious health concern affecting both developed and developing countries. The accumulation of advanced glycation end-products from chronic hyperglycemia can result in a multitude of complications including imcro- and macrovascular diseases, such as diabetic kidney disease (DKD).
Managing risks for complications such as DKD, including diabetic nephropathy, is another concern for providers with diabetic patients. DKD is a common cause for hemodialysis as well as proven connected to cardiovascular disease. While there are currently some good predictors already in use, the need for more precise models toward earlier intervention and improved outcomes, is critical, especially in patients not showing obvious signs or symptoms but who are nonetheless at risk.
- Methodology and findings of DKD research with big data
Artificial intelligence (AI) is already supporting providers’ clinical judgment and experience in many areas of medicine. The developers of a recent study wanted to create a more predictive model for identifying and managing risk factors in DKD and prevent attendant complications, particularly in those without outward signs or other indicators.
The researchers used three AI approaches in processing natural language (text data), structural data as well as longitudinal data combined with big data machine learning which was in turn based on electronic health records (EHR) involving more than 64,059 type 2 diabetic patients.
Utilizing the above approaches, the researchers extracted data for clinical features for a modeling six months aggravation of DKD from 858,660 medical records. During this process, structural factors including laboratory tests, prescription, ICD-10 and diagnosis codes were extracted by AI as well as previous and current disease history and prescriptions from EMR texts via natural language processing.
Final results over a 6-month period showed that AI could detect progression of DKD before patients showed symptoms of microalbuminuria or other clinical signs allowing earlier, more successful intervention.
Partner with a trusted medical billing and practice management company
Providers who partner with an experienced medical claims service such as M-Scribe will find it easier and faster to get their AI and related claims out in an accurate, timely manner, regardless of payer. Nephrology, dialysis and chronic care conditions can be especially challenging to bill, so having a trusted billing and practice management partner in your corner is invaluable. Since 2002, we’ve been helping practices of all sizes and specialties take better control of their revenue cycles to thrive and grow. Call us at 770-666-0470 or email us to learn more about we can help you identify and reach your practice’s unique goals for revenue and growth.