Predictive modeling helps identify CM candidates
Predictive modeling helps identify CM candidates
Drill down to find those who will benefit most
Predictive modeling is an invaluable tool in early identification of people who can benefit from case management, but using the score alone doesn’t go far enough, says Kay Sherwin, RN, CCM, director of client services for Integrated Healthcare Information Services (IHCIS), a Waltham, MA-based predictive modeling and information solutions firm.
"Predictive modeling is a very effective tool that can be used with other tools to identify potential case management clients, but it is not the be-all, end-all by itself," she adds.
The risk score shows how likely members are to need care or to utilize health care services, but to be truly effective, you need to look at the clinical details to see what is driving the risk, she adds.
For instance, if a patient has cancer, diabetes, and depression, knowing that diabetes is driving 60% of his risk may help you decide to refer him to a diabetes disease management program, Sherwin says.
Predictive modeling data can help you identify gaps in care, such as which diabetics are not getting their A1c tests or which women haven’t been screened for breast cancer. These gaps in care, or care opportunities, are more for identifying members for disease management or wellness activities, Sherwin adds.
"We know that around 1% of any given population will need case management services. But which 1% is the right group to manage?" she says.
Using predictive modeling only to identify members who are at highest risk will identify a lot of members who already are in care management.
Depending on how often you run the data, it could identify members who already have died of the disease or those who are following their treatment plan, and case management would have limited impact.
"The data also can give false negatives and false positives. Just getting the members who are at risk for health care events isn’t going far enough," Sherwin explains.
Your predictive modeling tool may identify members who have a high risk for utilization of services but are managing their health and navigating the system very well on their own. These members don’t need case management intervention.
"You want to find those members who have difficulty in managing the system or who have problems making decisions about their own health. They are not identified from predictive modeling but from a case management assessment," Sherwin says.
The ideal situation is to use risk modeling in combination with other tools and attributes to identify the members you want to reach, she adds.
For instance, by bringing in information from your organization’s health risk assessment into the predictive modeling tool, you can use specific responses to add additional criteria as a filter that will help identify specific members, Sherwin says.
For instance, the health risk assessment could look at potential obstacles to compliance, such as whether there is a reliable caregiver at home.
Considering a member’s response to rating his or her health status today in comparison to last year may help identify members more likely to benefit from case management, she adds.
Become familiar with member population
"Every organization I’ve worked with and every group of case managers has a different idea of where to find the sweet spot, those members are most appropriate to assess for case management. To be effective, you need to explore multiple ways to filter the information from your predictive modeling, such as looking at comorbidities in conjunction with high pharmacy utilization, abnormal test results, or any combination of factors," she says.
Sherwin encourages case managers to use their risk-modeling data to become intimately familiar with all the attributes of their health plan’s member population.
"My advice is to play with the data. Pull up different criteria, and see what you yield," she adds.
For instance, you may want to target frequent fliers in the ED, regardless of diagnosis; members who are receiving more than three home care services; anyone with multiple chronic conditions; or members with postoperative infections.
"You’re limited only by your imagination in determining where you can look to find the opportunities to have an impact," Sherwin adds.
No matter how good a predictive modeling system is, it’s still based on historical claims data, she cautions.
"There are so many pieces of information about a human that can’t be quantified from claims information alone," Sherwin says.
Look for numbers in terms of diseases, utilization, providers, complications, or gaps in care to identify opportunities, she says.
Look at whether the member needs intensive case management or episodic case management.
It’s not enough just to get risk assessment information. You need to drill down and find the opportunities for intervention and where you can deliver actions, she says.
For instance, you might look at the ICD-9 codes on claims to determine people with asthma with multiple prescriptions for rescue medicine, a sign that they aren’t managing their asthma well.
Or, if you’re examining data on cancer patients, look for patients who have advanced markers, who have a port for chemotherapy, or who show excessive use of pain medications and antidepressants.
Creating a case definition
"A predictive modeling tool is simply that. It’s a tool, and the value you get depends on how much you drill down and identify members where you can take action and make a difference," she says.
Sherwin recommends creating a case definition to define the people you want to consider for case management.
Here is an example: All members who have diabetes, live in a particular region, and belong to a particular network product or employee group. The members have a 50% probability of an inpatient admission, have a relative risk score of greater than 10, and are currently not in case management or disease management. They have diabetic peripheral vascular disease, have not had a hemoglobin A1c in the past 12 months, and are being treated by Dr. X, whom you have determined is not your most effective provider.
By using this definition, you can identify diabetics with significant risk who are not currently in a program, who already have diabetic complications, and who are not being monitored appropriately, she adds.
"Case management is very appropriate for people who have a combination of problems, including multiple psychosocial issues. Case managers can help these patients get the right care at the right time by the right clinicians at the right site but must start by identifying the members who can benefit most from case management," she says.
In the past, hospitalizations and illnesses have been the primary triggers for identifying members for case management. These are reactive and may be too late for effective interventions, she says.
"When I started as a case manager 15 years ago, the only referral system we had was patient who had been admitted to the hospital.’ We had no history or demographics to work with. Now, we have a lot more data to help us identify people we can assist," she says.
In her early days as a case manager, Sherwin took a proactive approach to care, making outreach telephone calls to her clients.
"This is a very resource-intensive way to manage your members with the challenge of documenting the return on investment, and the approach has faltered in recent years," she says.
However, by using predictive modeling and other tools to identify the members who will benefit most from interventions, case managers use their resources where they will make the most difference.
"With the advent of predictive modeling and the ability to identify opportunities to improve patient care, I see a paradigm shift back to proactive case management," Sherwin adds.
When you identify members with the potential for case management, an outreach telephone call is the most effective method of conducting a health assessment, she says.
Today’s challenge is that many nurse case managers haven’t done much proactive care and they’re not accustomed to making cold calls to members who may or may not be feeling sick at the moment.
Remember other tools
"Because of this, a lot of organizations are engaging their case managers in additional training, including motivational interviewing," Sherwin says.
Don’t let predictive modeling take the place of concurrent review to identify members who could benefit from case management, she cautions.
"Predictive modeling and evidence-based medicine are wonderful tools, but they won’t identify members who will have a spinal cord injury, or a be high-risk newborn, or become a newly diagnosed diabetic," Sherwin says.
[Kay Sherwin, RN, CCM, is director of client services at Integrated Healthcare Information Services in Waltham, MA, which offers a predictive modeling system called Impact Pro. Web site: www.ihcis.com. Phone: (781) 895-9950.]
Predictive modeling is an invaluable tool in early identification of people who can benefit from case management, but using the score alone doesnt go far enough, says Kay Sherwin, RN, CCM, director of client services for Integrated Healthcare Information Services (IHCIS), a Waltham, MA-based predictive modeling and information solutions firm.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.