Crystal ball of health care? Can predictive modeling help lower costs?

• Author: Share, Michael R.
• Position: ADVICE / EMPLOYEE BENEFITS

MANAGED-CARE companies and insurers have long used some form of experience rating (claim dollars paid out/premium dollars paid in) to help price their products for large companies. Life insurers have used mortality tables and "the law of large numbers" to price policies for years.

Every decade or so, the health-care community comes up with the latest and greatest "silver bullet" to address the rising cost of health care. Most of them have familiar acronyms: PPO (preferred provider organization), FSA (flexible spending account), HMO (health maintenance organization), POS (point of service), MSA (medical savings account), HRA (health reimbursement account), and most recently, HDHP (high deductible health plan) and HSA (health savings account).

The industry has also tried self-funding, wellness programs, disease- and case-management programs, health-risk assessments, provider profiling and more.

All of these ideas and strategies helped ... for a while ... then typically wore out their welcome as costs once again started to escalate.

Statistics show that approximately 80 percent of health-care costs are being generated by less than 20 percent of the population. In addition, a high percentage of those costs tend to come from individuals with comorbidities (people with two or more disease states; i.e., a person with heart disease and diabetes).

The greatest deficiency of disease-management programs is timing. They are reactive versus proactive. The claims data used have a lag time and are often incomplete. Most of the participants have already had a major episode resulting in a large claim. That is how they were identified for the program.

Logic dictates that if we were able to stratify the population to identify people who might incur significant medical claims in the future, we could potentially reduce costs, lower health-care trends and improve the quality of life of those who may be afflicted.

That is where predictive modeling comes into play. By...