Predictive risk modeling


It’s likely every business executive at some time has secretly wished for a crystal ball. After all, knowing what lies ahead would make many business decisions so much easier, especially as they relate to variable costs – like those associated with employee medical needs.

While it may not be a “crystal ball” per se, a tool known as predictive risk modeling (PRM) is a technique some health insurers are using more often to help employers assess future medical costs by quantifying the relative health status of their employees.

Although PRM is not an entirely new approach, it is becoming more sophisticated and more accurate. It can help employers plan their health resource allocation, while helping employees become healthier, because it helps identify medical risks sooner so employers can implement interventions to help prevent sickness.

Two types of risk
All forms of PRM use some combination of age, gender, medical codes and claims histories to calculate medical resource use, also known as a “relative risk score.” This relative risk score assigns “weights” to medical categories. A lower weight is better, because it predicts less utilization of medical resources in the future. The most simplistic models use age and gender to predict costs. The newer models, in addition to age and gender, use diagnosis codes and divide a person’s score into two categories: concurrent risk and prospective (or predictive) risk.

The concurrent risk score looks at the present and is based on all diseases affecting one’s employees “today.” It contains information on chronic and acute conditions. For example, this score takes into account data on employees with chronic diseases such as diabetes, asthma and heart disease as well as data on acute conditions such as broken bones, ear infections and cuts.

The prospective or predictive risk score looks at employees with chronic conditions and co-morbidities (e.g. complications that can result from any condition, whether chronic or acute). It then predicts the risks these conditions will most likely have on future medical resource use. As our health care system evolves and as more employers are looking for alternate ways to control health care costs, this can be an indispensable tool. This is especially true for employers who self-fund, as they assume the financial risk for all medical claims. It helps all employers find ways to tailor their medical benefit offerings to suit the needs of their employee population.

Action plans
For any group, PRM helps predict the health care needs of employees, so an employer can develop an action plan to address them. For example: if an employer identifies a large number of diabetics, he or she might decide to implement more education programs for diabetes. Many health insurers have strong health education and health management programs that focus on chronic illnesses like asthma and heart disease and programs such as smoking cessation that can assist in this regard.

The key is early identification of people at risk. The earlier you detect a high-risk group, the sooner you can put programs in place to keep them healthier and more productive on the job. Like any approach to predicting the future, however, there are limits. PRM is not an exact science, with even the best models accounting for only 40 percent of the variation (e.g. how individual employees compare to the average employee population) in medical resource use. But PRM can certainly help employers plan for better resource allocation, and paint a good picture of the health status of the company as a whole.

As the health care system in the United States continues to change, health insurers are seeing their role evolving from that of plan administrator to health system navigator. They’re in a unique position, because they’re able to see the entire spectrum of the health care system.

Tools like predictive risk modeling are helping insurers assist employers in understanding the overall health status of their workforce and developing ways to meet their specific needs. This helps employers to choose the best route for reaching their destination. A crystal ball would certainly be handy, but predictive risk modeling may be the next best option.

TIM SULLIVAN is director of medical informatics for Care Choices, a non-profit health care organization and a subsidiary of Trinity Health. Care Choices HMO is ranked as No. 7 among 257 commercial plans nationwide and is the top-rated plan in Michigan, according to U.S. News & World Report/NCQA “America’s Best Health Plans, 2005.”