A Managed Care Organization (MCO) was experiencing increasing costs from diabetic patients that were not appropriately managing their disease.  Not only was the MCO concerned about their rising costs, they were also concerned about the general health and care of these members.  The MCO’s solution was to develop a comprehensive patient care program where they could enroll "at-risk" diabetic patients to teach them how to correctly manage and care for their disease. The program included education about appropriate eating, exercise, glucose monitoring, and drug therapy methods.  The challenges to this approach, however, were two-fold:

1. They needed to identify high-risk patients for enrollment into the program before they experienced a life-threatening episode.

2. The program was too comprehensive and expensive to aggressively promote to every diabetic patient; therefore they needed a method to accurately predict which patients were high-risk diabetics that would benefit from the patient care program.

The MCO called upon Peak Data Solutions to see if a predictive modeling system could be developed and implemented that would identify "at-risk" diabetic patients before the patient suffered a life-threatening inpatient hospital visit due to poor disease management. 

Peak examined all data sources available to the MCO that could be used as dependent variables within a modeling system (e.g. medical, eligibility, provider, pharmacy data, and customer care), and evaluated the feasibility of creating a predictive modeling system. After carefully reviewing the client’s goals, business requirements, and data sources, Peak concluded that the usage of medical data, although very predictive, could not be utilized for the plan because there was generally a two to three month lag in receiving patient medical data. This lag would prevent the achievement of one of the major program goals: to catch and enroll diabetic patients early, before a life threatening episode occurred. However, even without the availability of patient medical data, the remaining data sources proved quite predictive, while also being available in a timely fashion.

Peak then developed and implemented a predictive modeling system.  It successfully identified diabetic members within the MCO that are significantly more likely to have an inpatient hospital visit. Once the modeling system was in place, targeted plan members were then invited to an array of diabetic disease management programs from newsletters to one-to-one tutorials to specialized educational mailings.

Results

The modeling system and disease management program was a tremendous success.  With the help of the modeling system, the MCO experienced a 19% decrease in inpatient hospital visits due to a diabetic episode from poor disease management, saving over $650,000 in costs in the first year alone.