Intelligence Unit Special Reports Special Events Subscribe Sponsored Departments Follow Us

Twitter Facebook LinkedIn RSS

Survey-based predictive modeling goes beyond claims

Most DM organizations use the rearview mirror approach when trying to intervene with at-risk members.

This is done when DM companies review medical and pharmacy claims as a way to catch at-risk patients before they become costly, but Alfred Lewis, JD, executive director at Disease Management Purchasing Consortium International, Inc., in Wellesley, MA, says the claims-based approach “simply does not work.”

“Generally speaking, I am of the opinion that you need to know much more about people than what’s in the claims,” he says. Reviewing medical claims data alone indicates who has received services, but not who is at risk of costly healthcare services in the near future. Ariel Linden, DrPH, MS, president of Linden Consulting Group in Hillsboro, OR, says relying solely on claims data is insufficient as an identifying methodology because:

Those who were hospitalized in 2007 are unlikely to be rehospitalized this year

Those who were not hospitalized in 2007 could wind up having a lengthy stay this year

New members in a health plan do not have any claims data to feed into the predictive model

As a result, these models significantly underpredict high-cost individuals and overpredict low-cost individuals. Linden, instead, suggests a hybrid approach that includes survey-based predictive modeling coupled with claims data review.

“There are several [health risk assessment] tools that have good accuracy at identifying people who are at risk of a near-term hospitalization,” says Linden, pointing to the probability of repeated admissions (PRA) tool, which identifies seniors at risk of hospitalization. “People think [claims-data analysis] using a predictive model is a good standard, and it’s not. It’s far from it. That’s one of the major reasons why disease management does not prevent avoidable hospitalization. They miss most of the people who are actually going to get admitted.”

One of the pioneering programs that developed a survey-based predictive model is One Care Street, which was developed in the mid-1990s by Julie Meek, DNS, founder of The Haelan Group in Indianapolis, and now chief science officer at CareGuide in Coral Springs, FL. CareGuide now runs One Care Street, which includes a health perception survey, health coaching, and claims data review.

The goal of the program is to gauge whether a person is going to need expensive medical care within the next six to 12 months.

Finding those most at risk of expensive medical care benefits everyone. “Everybody wins [when healthcare costs are under control]. Society wins, the employer wins, and certainly the person and the family win,” says Jim Kerr, vice president of business development at CareGuide.

The multiple-choice questions in One Care Street’s health perception surveys (one geared to preretirees and one for the senior population) go beyond asking about a person’s condition and diseases. They touch upon their emotional and mental state, health habits, and feelings toward their health and status.

Although basic health risk assessments predict early morbidity and mortality, health perception surveys ask the person about current health issues and focus on clinical risk factors and disease conditions. A properly worded health perception survey determines a person’s emotional state and response to illness and stresses. One Care Street also gauges whether a person is a good candidate for coaching, says Kerr.

“A typical [health risk assessment] is going to find what people’s health issues are. That’s not nearly as important as determining how concerned they are about those issues,” says Lewis. “If you can figure out who is starting to panic about the health habits that they have developed, you’ll be more likely to find people who are willing to change them but don’t know how to go about changing them.”

According to CareGuide, One Care Street’s 45-question health perception surveys capture 63%–67% of at-risk people rather than 10%–15% using the standard DM approach of analyzing claims.

After the respondents complete the survey, CareGuide uses its predictive algorithm to find out whether the person is at risk and would be receptive to coaching. In a matter of a few weeks, coaches contact those deemed at risk (those who take the survey are also informed that they can utilize coaching if they believe they need it). Kerr says the quick turnaround allows CareGuide to reach out to at-risk people before their condition deteriorates.

The coaching goes beyond the usual tenets of healthy living; it touches upon real-life topics. For instance, the coaches are trained in problem-solving skills and motivational interviewing, an intervention technique that helps people change their behavior.

Linden says implementing motivational interviewing and hiring the right coaches with the proper training are important for a successful program. Lewis concurs, adding that motivation is more important than dissemination.

“What has been learned the last several years is it’s not about the dissemination of the information. It’s about motivation, and motivation is much harder than disseminating information,” says Lewis.

CareGuide’s health coaches spend an average of two to two and a half hours on the phone over a two- to four-month period in an attempt to improve the at-risk person’s level of perceived health.

Rather than mailing everyone the same letter, giving them identical refrigerator magnets, and reading from a script about how to live healthy, the coaches teach the at-risk people how to react in different situations. For example, rather than eating a candy bar when stressed, the coaches may suggest listening to a favorite song or taking a walk and revisiting the problem in 10–15 minutes.

Kerr says training is continuous for nurses, who know the different benefit structures for each person so they can refer people to available services if needed.

The combination of health perception surveys and coaching appears to be working. CareGuide trumpets a 74% engagement rate versus the industry standard of 15%–20%. “Our engagement is so high because the predictive model is so good. We’re calling people who want the help and need the help,” says Kerr.

According to CareGuide, employers who have used One Care Street have cut healthcare costs. For instance, MacAllister Machinery, a heavy equipment supplier in Indianapolis, saw healthcare costs drop 7% from 2005–2006 and another 4% from 2006–2007. In the year before signing up for One Care Street, MacAllister faced a 14.5% increase from 2004–2005, according to CareGuide.

In early 2007, CareGuide also reported that a review of 10 clients showed an overall ROI of 8:1. The clients chosen had two full years worth of data, so CareGuide was able to properly analyze the results. CareGuide plans additional studies for the first and second quarters of 2008.

Reasons for success

Here are four examples as to why health perception surveys boost predictive accuracy:

Surveys reveal not only those needing interventions but also which members are willing to make changes

Information comes directly from members, which leads to more stable and reliable data

Results and reporting are immediate, and populations can be profiled across demographic, geographic location, age, gender, etc., so that additional programs and services can be developed

High engagement rates are achieved because the right people are targeted

Source: The Haelan Group’s Predictive Modeling and Finding and Intervening with the High-Cost Healthcare Consumer white paper.

Keys to success

Jim Kerr, vice president of business development at CareGuide in Coral Springs, FL, says there are three elements needed to create a successful survey-based program:

Incentive design

Communication

Leadership support

Kerr says 90%–95% of companies that partner with Care-Guide offer incentives to take health perception surveys. Ariel Linden, DrPH, MS, president of Linden Consulting Group in Hillsboro, OR, says it’s important to develop incentives depending on the population. Some of the successful incentives have included discounted copays and deductibles, gift cards, and drawings for a Hawaiian trip, a big-screen TV, and iPods. “No two populations respond the same to incentives,” says Linden. Communication can include mailings, on-site talks with employees, videos, meetings to spark interest, and question-and-answer sessions.

The importance of getting leadership support cannot be overstated. A leadership’s backing is more than just getting an okay. Kerr says the clients that have enjoyed the best results were companies that embraced the idea from top to bottom. Not getting a company’s leadership behind a program is a recipe for disappointing results, says Kerr. Linden adds one more key to success: communicating with PCPs.

Unless commercial DM makes a serious attempt to engage with physicians, Linden says, the industry will miss a critical piece in properly identifying people at risk of a near hospitalization. Physicians know better than anyone who can benefit from DM, but vendors have historically chosen the path of least resistance and relied on the patient’s claims to identify them rather than get physician buy-in, says Linden.