CareSteps TO USE NEURAL NETWORK PREDICTIVE MODELING

Techniques used by energy companies to predict the demand for electricity and by financial institutions to manage stock portfolios are now helping healthcare organizations proactively identify people at risk for healthcare problems, according to Mark H.T. Ridinger, M.D., president and chief medical officer of CareSteps, Inc. The physician-founded company, based in Dallas, was launched today at the Disease Management Association of America's Annual Leadership Forum.

The technique, called neural network predictive modeling, is an artificial intelligence tool that can accurately identify a future event based on statistical links among known data. For example, financial institutions use predictive modeling to determine their customers' likelihood of purchasing a particular product or service such as home equity loans and money market accounts. Neural networks are more accurate than other statistical techniques because they can recognize and learn patterns in data, similar to the neurons in the human brain.

CareSteps is using these models to help physicians, hospitals, health plans, employers and other organizations more accurately identify people who are at risk for certain health problems so that healthcare providers can proactively intervene. A key success factor in this effort is the company's ability to link these models to a comprehensive database of proven standards of care for more than 30 specific healthcare conditions, from cancer screenings to depression.

"Our healthcare system still focuses on addressing acute healthcare needs rather than on preventing healthcare problems from occurring in the first place," Ridinger said. "As a result many Americans aren't getting the care they need.

"For example, high blood pressure and diabetes are two conditions where we have ample evidence that people aren't receiving proper treatment," he said. "We formed CareSteps to address this very problem. By combining neural network predictive modeling with a database of nationally recognized care standards, we have created a powerful and more accurate way to help healthcare payers and providers deliver the right care at the right time to the right patient, preventing potential health problems from emerging or becoming more complex than they have to be."

Carey E. Floyd, Jr., Ph.D., Duke University and member of CareSteps' Academic Advisory Board, explained that while predictive modeling tools have been used in healthcare for several years, most are static and rules-based, and therefore frequently inaccurate.

"With current tools available, a patient with asthma may be identified as being at risk for a hospitalization because he or she may have many visits to his or her primary care physician, but the reality may be that the patient is actually doing well because he or she is receiving appropriate preventive service, thus reducing the chance for a costly emergency room and other information, including the patient's total health status, into account."

The information used for CareSteps' tools comes from medical and pharmaceutical claims data and demographic information supplied by health plans. CareSteps supplements this data with information gleaned directly from health plan members through a secured Web application called CareSteps.com.

Here's how it works. A consumer enters a medical profile. CareSteps.com produces a personalized health assessment and customized care plan based on the latest medical guidelines and clinical protocols. Unlike other Web-based health assessments, the care plan takes into account the person's total health status, including gender, age, and other factors, not just a specific disease or condition. While this information is an important piece of the predictive model, consumers do have the option as to whether or not they share this information with their health plan.

"Consumers who use CareSteps.com have complete control over who sees and uses their healthcare information," said Ridinger. "We firmly believe that a person's health information is private and should only be shared in ways that support the ability for doctors and nurses to provide care for their patients."

CareSteps offers its services to physician groups, hospitals, health plans, disease management organizations, employers and others. Consumers who aren't able to access the Web application through these channels can use an abbreviated version of the program by visiting www.CareSteps.com.

CareSteps is already growing quickly. The company recently acquired Axonal Health Solutions, a specialized healthcare informatics company, and has strategic relationships with companies like LabCorp, Glaxo Wellcome, Willis, Affiliated Computer Services (ACS) and PeopleMed.com. The company currently works with numerous clients such as the Hawaii Medical Service Association (HMSA) and Midway Airlines.

CareSteps was founded in 1999 by a team of physicians. Through their National Institutes of Health funded research at Duke University, these physicians found that neural networks could help radiologists more accurately predict the presence of breast cancer. The results showed that the number of breast biopsies performed each year can be reduced without losing the ability to detect cancerous tumors.

Contact Gayla Daugherty, 314-982-7773, pager, 888-528-3016 or daugherg@fleishman.com.