Meeting the ‘gold standard
Meeting the gold standard’
Predictive models not only identify members at future risk for disease complications. They rank plan members from highest to lowest risk making it easier for plans to decide which patients to target for intense case management and which to provide with less-costly disease management mailouts, notes Jeffrey J. Rice, MD, JD, chief executive officer for Axonal, based in the company’s Dallas office.
"Different organizations have different goals. Some organizations know they only have the resources to manage 100 patients. They can look at the predictive model and target the top 100 to receive case management intervention."
Predictive modeling requires an artificial intelligence (AI) neural network. The size of an AI expert system database and the need to constantly update it make it impractical for all but the very largest health care organizations to develop and implement their own AI systems, note Ridinger and Rice. (A definition of neural networks and other terms vital to this new disease management era appear in the box, above.)
"It takes a large and knowledgeable information technology staff and clinical staff to run the type of neural network necessary for predictive modeling. It’s fairly expensive and time-consuming to get through the learning curve," says Rice. "We spent more than two years on pure development. In addition, it takes a sophisticated operational environment and the appropriate expertise, including a good understanding of the clinical problems of the disease."
Some questions Rice suggests you ask as you work through the risk assessment vs. predictive modeling and then the "buy or build" decision include:
• What is the return on your investment?
• What are the parameters for the models?
• What are the pros and cons of the different statistical and methodologic applications that are available?
• What advantages do advanced statistical tools such as artificial neural networks provide over linear regression techniques?
• What source or sources of data are most appropriate?
• Are the necessary outcomes data available at this time to build a predictive model?
If you do decide to outsource your predictive modeling needs, Ridinger urges that you carefully research the available medical predictive modeling products to see if they meet your "gold standard."
All that glitters . . .
Questions he suggests you ask predictive modeling vendors include:
• How long has the company existed, and what is its financial status?
• What is the background of the company’s leadership? Is all the necessary information technology and clinical and health care industry experience represented among the company’s officers?
• Has the predictive model been scientifically validated? What is its accuracy?
• How timely is information returned and reported?
• How user-friendly are the reports generated?
• Is the model a robust one capable of generalizing to my population?
• Does the predictive model stratify individuals by risk, so that different levels of intervention can be matched with the appropriate risk categories?
• Is the model patient-specific rather than actuarial? (For a list of companies marketing AI predictive modeling products, see box, above right.)
"The continuum of care is a perfect arena for predictive modeling," says Ridinger. "But the true potential of AI systems may not be reached until other barriers fall away and we have clinical data in a standardized and integrated electronic record, and a business philosophy which accepts that dollars spent now are worth thousands saved later."
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.