By using a predictive modeling tool to identify members at risk of becoming high-cost healthcare users, an Oklahoma health plan is reducing per member per month costs while increasing member satisfaction.


GlobalHealth ImageLike many healthcare organizations, Tulsa, Oklahoma-based GlobalHealth, Inc., has used traditional care management to reduce costs and improve care for plan members with acute and chronic conditions who are frequent users of services. Recently, however, the health maintenance organization (HMO) has also started to focus on members who are not currently high users of services—but may soon be.

“In the past, we’ve been very good at managing care after a disease or condition is identified,” says J. David Thompson, GlobalHealth’s vice president of health plan operations. “But we haven’t had the data necessary to see the opportunity for intervention before a health episode may occur.”

Last year, GlobalHealth implemented a predictive modeling software that identifies a hidden cohort of members—those who are not high users of healthcare services, but are at high risk for an acute event. These members do not currently receive routine care—in the form of primary care visits, health screenings, and diagnostic tests, for example—that would prevent the need for more acute and costlier care.

Early results show that the tool, coupled with extensive member outreach, has enabled the HMO to reduce the number of emergency department (ED) visits and inpatient admissions. Preliminary results indicate a per member per month (PMPM) savings of about $4, but Thompson says the goal is to at least quadruple that figure, while helping to ensure that high-risk members are receiving the appropriate care at the appropriate time.

Realizing the Risk

GlobalHealth provides health insurance coverage for approximately 40,000 plan members in Oklahoma. The majority of members are government employees who work in state and federal offices and school districts.

“We had a select group of members—in the range of 1,200 to 1,500—for which we had active care management in place,” Thompson says. These members, for example, receive education on managing their conditions and assistance with transitioning to non-acute levels of care.

Another cohort of members includes low users of services; however, some of these members experience a sudden acute event that requires a hospitalization and, consequently, become high users. When GlobalHealth’s clinical and administrative leaders began reviewing statistics on daily inpatient admissions (e.g., reasons for admissions, diagnosis, health and claims history), they found that there were clear indicators of these acute events.

“There was increasing consensus that a large percentage of those hospitalizations could have been prevented. We believed that was important not just from a cost standpoint, but from a quality of care standpoint and from a service standpoint,” Thompson says.

To confirm its theory, a GlobalHealth consultant used a predictive modeling tool to conduct a retrospective review of member data. The review identified approximately 4,000 members at the end of 2011 who were low-cost healthcare users but had a high risk of an acute event. These members began using a lot more health services between 2012 and 2013, with costs tripling in 2012 and then again in 2013. “Every year, about 12 to 15 percent of members in the “hidden” category moved to the high-cost category,” Thompson says.

The challenge became preventing acute incidents by identifying at-risk patients.

Identifying the Risk

The predictive modeling tool analyzes clinical and nonclinical data to create a profile for each plan member. Various data is extracted from internal and external sources, including the following:

  • GlobalHealth’s electronic health record
  • Utilization data from claims
  • Medication data from pharmacy systems
  • Scheduling data (e.g., completed and missed appointments) from practice management systems
  • Demographic data from publicly available records

The tool then calculates a member’s risk for clinical chronic conditions, including congestive heart disease, diabetes, asthma, and hypertension. It also assesses five nonclinical factors that can affect a member’s risk of experiencing an acute event:

  • Utilization: Patterns of care are indicators of costs and whether members are using too few services given their conditions.
  • Compliance: Members are measured for their adherence to evidence-based care protocols, such as months since a patient with diabetes missed a scheduled appointment for regular A1C test, foot exam, etc.
  • Access to care: Members’ ease of access to appropriate care services tends to affect whether they obtain those services.
  • Socioeconomic: Research shows that demographics (e.g., education levels, household incomes, native languages) may affect health status.
  • Perceived well-being: Studies have shown that patients’ perceptions of their health may affect their actual health outcomes.

“The diagnosis codes can indicate risk, but then there are all the behavior and utilization patterns that can impact risk, as well,” Thompson says.

Based on clinical and non-clinical indicators, members are categorized into four cohorts:

  • High risk, high cost
  • Low risk, high cost
  • Low risk, low cost
  • High risk, low cost

The last category is the hidden cohort that represents an opportunity for cost savings if a newly identified high-risk member can be prevented from becoming a high-cost patient through more effective care management.

Managing the Risk

GlobalHealth care managers receive a report from the predictive modeling tool of high-risk members, including the hidden cohort. The report identifies problem areas based on the member’s clinical and nonclinical scores, prioritizes the members who should receive outreach first, and recommends care management actions. For example, if a member has a high risk for diabetes, a low compliance score, and trouble accessing needed care, the report would suggest home health or transportation services.

Sample Predictive Report

The care managers use these recommendations to inform the outreach process—to both members and physicians.

Member outreach. “We use different member outreach interventions depending on the scenario,” says Thompson. A member who is not compliant with medications, for example, would be contacted by a pharmacist to determine the issue. If the member’s case is found to require a greater level of care management, the case may then be assigned to a disease management or complex care management team.

The data gathered on the members prompts care managers to ask specific questions. For example, if a member with diabetes is past due on an A1C-level check, is it because the member is concerned about his or her financial responsibility for the test? Does the member have transportation to get to the physician office? “This information helps us to categorize what the specific issues are and what we need to do to ‘move the needle’ with that member,” Thompson says.

The predominant diagnoses found within the hidden cohort include heart disease, hypertension, diabetes, and anxiety disorders, says Thompson. Many of these “hidden” members are assigned to the disease management team. “What we’re doing there is basically high telephonic outreach where we coordinate with specialists to get our members the tests they need to stay on top of their diseases,” he says.

The addition of the hidden cohort of members significantly increased the case management case load. “We went from very high-touch management of about 1,500 members to about 4,000 to 5,000 members.” Currently, a staff of approximately seven FTE care managers provides telephonic outreach to high-risk members, with some calls lasting as long as 45 to 60 minutes. “The care managers are really establishing a relationship with these members,” says Thompson.

Physician outreach. Care managers also contact physicians and others involved in a member’s care processes. For example, a GlobalHealth pharmacist may contact a member’s primary care physician or retail pharmacist to gain a better understanding of why a member is not taking her medicine as prescribed, and discuss how to best address the issue.

Sometimes physician outreach is about resolving access issues. The GlobalHealth care managers will help members schedule or coordinate care with a physician’s office. If the issue is more clinically oriented or related to noncompliance with treatment recommendations, then the care managers will contact physicians directly. Physicians often are unaware that their patients are noncompliant, Thompson says.

To address compliance challenges across entire patient panels, GlobalHealth is developing committees comprised of HMO clinicians and physicians in the HMO’s network. Such committee-level meetings will enable each side to gain the other’s perspective on care management strategies that best meet the needs of members.

Calculating Early Results

Because the predictive modeling tool has only been in use since late 2013, it is too early to declare GlobalHealth’s approach a success. However, Thompson is pleased with the preliminary results. GlobalHealth has seen an approximate 20 percent reduction in ED encounters and readmissions among all members between January 2014 and May 2014. At the same time, member satisfaction has increased 3.5 percent among all members.

In addition, hospital admissions have declined by 5 percent; however, a change in case mix, resulting in a healthier population, accounts for about half of that improvement.

By identifying at-risk members and ensuring they get needed care to prevent acute events, GlobalHealth expects quality scores to improve as well. “We think we’re on the right track,” says Thompson.

The early results, he says, have already translated into cost reductions.

  • ED visits represent about 10 percent of PMPM costs, so a 20 percent reduction equals approximately $2 to $3 in PMPM cost savings.
  • Inpatient admissions represent about one-third of PMPM costs, so a 5 percent reduction equals a savings of $3 to $4 PMPM.

The goal is a $15 to $20 savings PMPM over a 12- to 18-month period, Thompson says.

Allocating Resources

To accommodate the additional member outreach, GlobalHealth is also expanding its care management staff. Another pharmacist and two case managers are in the process of being hired. “Based on the preliminary results of this program we anticipate doubling our team within one year,” he says.

So far, the improvements in both medical cost savings and member satisfaction exceed the cost of adding staff, Thompson says.

The results of the program will become more apparent in August or September 2014, but the overall trend is promising. “Right now it is indicating very positively that we’re putting our resources in the right spot,” he says.


Karen Wagner is a freelance healthcare writer in Forest Lake, Ill., and frequent contributor to HFMA publications.

Interviewed for this article:
J. David Thompson is vice president, health plan operations, for GlobalHealth, Inc., located in Tulsa and Oklahoma City, Okla., and a member of HFMA’s Oklahoma Chapter.

Publication Date: Monday, June 16, 2014