How Better Intervention Targeting Improves Care Management

Transcription

1 How Better Intervention Targeting Improves Care Management Care Management Fundamentals Care Management organizations employ numerous intervention strategies to help their members lead healthier lives and reduce medical utilization. These interventions are often resource intensive, ranging from long term substance abuse treatment to in-home visits. To maximize the effectiveness of care management interventions, it s critical to deliver each intervention only to the specific members who are most likely to benefit. For example, a home nurse visit is likely to help identify lifestyle changes that can improve outcomes for members with asthma, while other members may require the expertise of a mental health professional or substance abuse counselor. Traditional Risk Scores Often Fail to Identify the Right People For years, care management organizations have used Risk Scores to identify members for care management interventions. However, these traditional risk scores, which rely on simple statistics and point based systems such as LACE, often mis-direct resources to those unlikely to benefit. For example, a LACE based readmission risk score is often only 20% predictive, meaning that the care management team must provide five interventions to reach one person who is truly headed for a readmission. This poor targeting frustrates care management teams because 4 out of the 5 people they reach do not need their help. By only considering 5-7 claims based indicators, traditional risk scores ignore hundreds of potentially valuable indicators that are hidden in free text such as Health Risk Assessments (HRAs), physician & nurse s notes, care management systems and call center transcripts that can dramatically improve targeting. Why Risk Scores Struggle Limited to claims data Only direct utilization 3 month lag, incomplete Horribly inaccurate Blind to previous non-utilizers 50% of admissions have no prior admission in last 90 days Unable to target subgroups Treats high-risk pregnancy and geriatric patient the same. Cannot improve over time Encoded judgements about care grow stale Overwhelmed by chronic/complex populations Not actionable when the entire population is high risk 1

2 Total Medical Expense (TME) Misses New Utilization Many care management teams also target interventions using metrics incorporating Total Medical Expense (TME) as a key indicator. Unfortunately, targeting the highest cost members fails to get ahead of upcoming utilization and focuses on people who may not be able to benefit. The three-month lag in claims data is just the beginning of the problem with TME derived metrics. Even more troublingly, high utilization often occurs during acute episodes and utilization reverts to lower levels by the time the care management team acts. The following figure shows the magnitude of this effect - at an urban safety-net integrated delivery system, fewer than half of super utilizers remained in the super utilizer cohort 7 months later and only 28% remained a year later. Most super utilizers at the end of the year were not in the original cohort. For Many Patients Who Use Large Amounts Of Health Care Services, The Need Is Intense Yet Temporary By Tracy L. Johnson, D. J. (2015). Health Affairs, 34: Using TME to direct care management teams is like sending out firetrucks after a home insurance claim is filed the fire is long gone and often little can be done. 2

3 New Approaches Dramatically Improve Targeting Many care management teams are now improving outcomes and generating high ROI by transcending traditional risk scores with new approaches to target interventions. These organizations start by using all available data to generate highly accurate rank-ordered lists of the individuals most likely to benefit from specific interventions. These systems move beyond a one-size-fits-all risk score with close targeting of different interventions. Instead of treating everyone the same, the care team is now able to provide substance use programs, behavioral health, fall prevention, and acute episode care to the members who can benefit from each of these different interventions. The best systems require minimal IT resources by taking the organization s data in its native format, including free text. These systems then employ machine learning with built-in natural language processing to generate actionable guidance for care management teams. New Data Sources Improve Targeting Health Risk Assessments Activities of Daily Living Care Management Notes Electronic Health Records Pharmacy Labs Customer Support Call Center Transcripts To close the loop on care management, it is also critical to measure each intervention s effectiveness and use this information to improve. The best systems offer real-time dashboards showing all individuals identified, interventions performed, and outcomes. Even the best technology does little to improve care if it is not aligned to the organization s people, processes, and priorities. To drive real results, care management teams must find a dedicated partner who understands their challenges and works together to operationalize and evolve their intervention programs over time. Better targeting is not just a more accurate risk score. The biggest improvements come from partnering with care management to design, implement, and monitor increasingly personalized intervention programs. 3

4 Better Targeting Improves Care and Reduces Cost By reaching the right people more often, care management teams can develop more personalized and intensive interventions. More people are helped by each intervention being performed when interventions are better targeted. For a care management team with the capacity to deliver 1,000 interventions, improving targeting accuracy from the traditional 20% to an achievable 50% enables the team to benefit 300 additional people. Members Benefioted Members Benefited by 1,000 Interventions 200 Traditional Risk Score 500 Machine Learning with Free Text Costs decrease when fewer interventions are performed on those who cannot benefit. For example, when using 20% accurate risk scores, a plan with 500 members who can benefit from a $1,000 intervention need to spend $2.5M. Raising the intervention accuracy to 50% enables the team to reach these 500 members who can benefit by performing only 1,000 interventions. This increase in accuracy saves the care management team $1.5M that would otherwise have been wasted providing interventions to people who cannot benefit. Program Cost $2 M $1 M $ M Program Cost to Benefit 100 Members $2.5 M Traditional Risk Score $1. M Machine Learning with Free Text Enabling 5X ROI From Care Management Assuming 500 successful interventions are each able to prevent an average hospital admission, targeting care management interventions with 50% accuracy saves the plan $6.3M in claims with a $1M investment. 4

5 Better Targeting Often Leads to New Intervention Programs As we have seen, improving intervention targeting provides immediate benefits to the existing intervention programs. By quickly providing interventions are reaching the right members, the care management team also often begins to employ more focused and sometimes more intensive interventions. The biggest opportunities for new intervention programs often appear in chronic / complex populations. Whereas traditional risk scores typically identified the entire population as high risk, new intervention targeting provides actionable guidance on who can benefit from intervention. These programs include: Spending more time better educating patients with high risk of a preventable readmission at hospital discharge. Investing in home monitoring devises for patients at risk of medicine non-compliance. Making home visits and even modifications for patients with high risk of falling. These programs continue to evolve with feedback from real time monitoring on intervention accuracy and outcomes. By partnering with a vendor that understands care management, the team is able to develop increasingly personalized intervention strategies to further decrease utilization and improve member s health. Case Study: Cyft is helping a health plan predict admissions and direct care management interventions to those members most likely to have a preventable admission. All available data is used for targeting, including free text in nurses notes and health risk assessments, as well as derived data from call center notes and pharmacy data. New interventions were focused on members with no prior admissions that were identified as likely to benefit - a population typically ignored by traditional approaches. Cyft improved care management outreach accuracy by 221%, moving their care management program from a 1.3x ROI to a potential 4x return. 5

6 Key Takeaway: Not all patients are the same, nor is all risk. Unfortunately, most approaches to finding members that require intervention treat people and disease as one size fits all. New approaches that capitalize on all available data, including free text, to target specific interventions to those most likely to benefit are dramatically improving care management. About Cyft Cyft partners with leading care management organizations to optimize care management for chronic / complex, behavioral health, and dual-eligible populations. Technology enabled, Cyft uses all available data in its native format (including free text) to inform care management strategies, populate programs, and measure progress. This high tech + high touch approach has helped 20+ plans capitalize on their existing data and talent to realize >$100M in cost savings and new revenue and improved member retainment and satisfaction. Cyft was founded by Harvard Medical School Professor, Dr. Leonard D'Avolio. The company builds on experience learning from over 10 million chronic / complex cases, engagements with over 250 hospitals, and 25 health plans. For more information please visit Written by Brendon Kellner bkellner@cyft.com 6