How Artificial Intelligence and Real World Data Can Improve the Management of Type 2 Diabetes: A Case Study

November 18, 2019

This is a sponsored blog, underwitten by Veradigm.

“A giant opportunity.” That’s how Forbes described the transformative power of artificial intelligence (AI) in healthcare.1

The revolutionary impact of mining “big data” in electronic health records (EHR) and other complex data sets – then using various analytic tools to make sense of the data – has opened exciting new avenues in outcomes research and medical care. Read on to see how Veradigm, a business unit of Allscripts, focused on generating data-driven healthcare insights applying artificial intelligence to EHRs to improve therapeutic management of patients with concordant comorbidities who are living with type 2 diabetes.

Interested in this white paper? Email

A Million Markets of One

One of the most intriguing new applications of AI is expanding its use from macro-level insights (e.g., population analytics) to individual patient-level insights. This refocus to patient-level insight has been characterized as a “a million markets of one”3, meaning a hyper-personalized experience within the construct of a large group. Regardless of its application, AI offers a greater access to speed and precision in decision-making and healthcare organizations are using AI to improve outcomes and reduce costs for individuals, as well as for entire populations.

Within healthcare, any technology that promises benefit from analyzing health records is frequently met with a healthy dose of skepticism – most commonly, from the provider on the front lines of patient care. Historically, providers have had to weigh the administrative burden of completing EHRs against the benefits of having consistency with data recorded for patient, ease of data submission for population health and quality measures, and access to clinical support tools. While EHRs impose fields to routinely capture and standardize fields such as diagnosis, biometrics and test results, providers always have, and often exercise, the option to record data as free text notes. Thus, the rich yet disjointed real-world data (RWD) pool contained within EHRs are a fertile environment for employing AI. As such, EHRs used in combination with AI may then be more useful, flexible, and impactful for both the clinician treating an individual patient and for health systems managing populations.

Specialty Pharmaceutical Fuel AI Growth

We are seeing expanded efforts in the use of AI with new pharmaceuticals, particularly specialty drugs, because the interventions are expensive and there is intense scrutiny on who can and should get the drug. The legacy approaches to treatment access, such as prescribing guidelines or clinical pathways, have limitations because they are neither nimble nor specific enough, particularly for orphan diseases or life-threatening conditions. For example, many of the new specialty products are targeted therapies and appropriate only for patients with specific genetic characteristics. Simply put, prescribing guidelines are not designed to predict individual patient outcomes, and cannot be reasonably formulated to accommodate for the multitudes of patient types. Lastly, stakeholders have become more numerous, multi-dimensional and vocal, resulting in greater demand for transparency around patient access, and a process that increases clarity on how to achieve optimal patient outcomes is welcomed.

These drivers are resulting in application of AI to efficiently process large data sets, typically based on real-world events, and identify the most effective outcomes at the individual- and subgroup levels.

Application of AI in Diabetes

How are organizations actually implementing AI to improve patient outcomes? Recently, Veradigm applied AI to its proprietary cloud-based EHR research platform Practice Fusion to assess treatment for individuals living with Type 2 diabetes (T2D) who also had concordant comorbidities. Natural language processing (NLP) was used. NLP is an artificial intelligence machine learning process to analyze free text, such as physicians’ notes in medical records and real-world evidence (RWE) derived from patient-level data.

Diabetes was chosen because it is prevalent (affecting 30 million Americans, 95% with T2D), confers substantial independent risk of high-risk disease and/or death, is costly, and has a complex treatment paradigm made even more challenging by the effort to individualize treatment. Moreover, the complexity of interactions between T2D, concordant comorbidities, and downstream complications requires a clinical approach that manages risk while maintaining guideline-specified therapeutic targets.

The business and the research challenge is described below.

Business Challenge: Are there patients living with T2D and cardiovascular and/or renal comorbidities who are struggling to meet glycemic control and cardiovascular risk targets, and can an individualized diabetes management plan be developed to improve outcomes? How can the EHR be better leveraged to identify patients at greater risk and support individualized management plan, without significant additional burden to the provider’s workflow.

Research Challenge: Using AI applied to real-world data from EHRs, the research challenge is to identify the impact of concordant comorbidities on patient responsiveness to treatment intensification among those living with T2D who are using new glucose-lowering therapy. At both the sub-group and individual patient level, the results of this case study and cohort analysis demonstrate how to utilize AI and NLP to identify new prescribing options and develop individualized diabetes management plans.

As described in the White Paper, Veradigm used evolving natural language processing to efficiently and consistently capture multiple data elements stored within free text on the interactive EHR platform. Important data that was embedded in provider notes, consultation notes, discharge summaries and descriptive reports associated with medical testing was explored, categorized, and analyzed. These elements were then combined with other real-world data on medication use, laboratory values, and treatment outcomes.

Potential to Impact Patient Care

Veradigm’s research demonstrates how EHRs can further evolve to assist HCPs in care coordination and patient support. A cloud-based digital health information system such as Practice Fusion that collects real world data bi-directionally, and combines this with longitudinal analyses of medication use, laboratory values and patient-reported outcomes may be offered to HCPs at the point-of-care. For example, algorithms embedded in EHR platforms may be used to monitor laboratory values occurring outside of normal ranges and offer tools for clinical support to enable accurate diagnosis and implementation of evidence-based, guideline-recommended adjustments to care plans. Informed as to how a patient’s progress may align with care plan goals and treatment guidelines, HCPs may offer recommendations for adjustments to therapy and lifestyle to enable shared decision-making for implementing individualized care management.


This white paper summarizes a cohort analysis and case study among individuals with concordant comorbidities living with type 2 diabetes and further explores how real-world data from the EHR platform Practice Fusion, a Veradigm offering, and natural language processing may be used to offer insight regarding individualized care management. Interactive EHR learning platforms have the potential to support patient care and therapeutic outcomes in a variety of real-world scenarios.

To obtain a copy of the white paper that was the basis of this blog, please email


1 “AI and Healthcare: A Giant Opportunity”. Forbes; Feb 11, 2019. Accessed on 16 SSeptember 2019 from

2 Liu R, Hemaraj Y, Boxe A. “Digital Transformation of Healthcare – State of the Union”. BGV Forum; Aug 2, 2018. Accessed on 16 September 2019 from

3 Artificial Intelligence. Accenture. Accessed on 16 September 2019 from

Share This Story!