Catching up with Rong Yi


Published on

Published in: Economy & Finance, Business
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Catching up with Rong Yi

  1. 1. Volume 2, Number 5 May 2009 Hospital Contracts: The Critical In This Issue Elements of Financial Performance Modeling 1 Hospital Contracts: The Critical Elements of Reconciliation between different data sets minimizes misunderstandings Financial Performance Modeling by Terri L. Welter, Principal, and Charles A. Brown, Senior Manager, ECG Management Consultants Inc., Arlington, VA 1 Care Management, Not ospitals and managed care organizations are taking a more collaborative DM, Needed to Control H approach to ensuring that negotiated terms produce the financial results that they expect. In the current reimbursement environment, it is common for new 3 Costs, Improve Quality American Health’s contracts to include significant methodology changes as the parties move toward Population episode-based payments. The development of an accurate and flexible contract Management Solution model is critical to understanding the expected financial performance of negotiated Uses PM to Target terms, particularly as reimbursement methodologies change. Modeling errors can Interventions result in reimbursement differences when compared to projections – with the potential impact of millions of dollars in either direction under a single contract. 8 Thought Leader’s Corner In order to successfully negotiate a new agreement and create financial predictability, the parties must accurately model the impact of proposed terms. That typically 9 Industry News involves applying the proposed methodology and terms to historic patient claims data and comparing the projected payments to expected reimbursement under the current 11 Subscriber’s Corner arrangement. While the concept is simple in nature, in practice, the complexities of hospital billing and data capture result in significant differences between data sets 12 Catching Up With… used by MCOs and hospitals. Rong Yi Failure to appreciate the intricacies and subtle differences between the data sets and adjudication interpretations can result in unintentional but material misrepresentations of the financial impact of a proposed agreement. When those problems are uncovered, they complicate negotiations by inserting a measure of mistrust. continued on page 6 Care Management, Not Disease Management, Needed to Control Costs, Improve Quality How to meet patient needs while avoiding unwarranted treatment variations is the question by Russell A. Jackson isease management has the power to keep a lid on spiraling healthcare costs by keeping patients with chronic D conditions as close to well as possible. But proving that disease management programs can do so cost-effectively has been difficult, especially in the Medicare fee-for-service population. The predictive modeling solutions that populate the tools that target patients for disease management intervention have a stake, of course, in the resolution of that dilemma. The problem, says Health Dialog -- a leading provider of care management and analytic services and a wholly owned subsidiary of Bupa, a global provider of healthcare services based in London -- is traditional DM programs only address 12% of the causes of unwarranted health cost and quality variations in the US healthcare system. That won’t be enough for the Obama Administration, the company points out, which has upped the emphasis on cost containment and quality improvement. Care management, the company says in a recent white paper called “Care Management: What Works,” has evolved into an enhanced solution to the problem. It, according to the white paper, addresses “the other 88%.” continued on page 4 Published by Health Policy Publishing, LLC ● 209-577-4888 ●
  2. 2. 2 Predictive Modeling News May 2009 Predictive Modeling News Care Management, Not DM, Needed to Control Costs… continued Editorial Advisory Board Swati Abbott The company offers a care management solution it says does just that. Here President, MEDai Inc., Orlando, FL are excerpts from the March 2009 white paper, which was written by David Veroff MPP, Health Dialog’s vice president for evaluation services. Ian Duncan FSA FIA FCIA MAAA President, Solucia Inc., Farmington, CT Despite the debate [over the cost-effectiveness of disease management programs] and often because of it, care management has evolved into an Peter N. Grant JD PhD enhanced set of services that includes disease management as one piece. President, Health Care Conference Though the form varies widely, almost every major health plan offers a care Administrators LLC, Partner and Co-Chair, management program and most employers pay directly or indirectly for those Health Law Group, Davis Wright Tremaine services. Deciding which of the models is ideal is a key public policy and LLP, Editorial Board, Health Affairs, Advisory Board, Harvard Health Policy business issue. A growing body of evidence indicates that well-designed, Review, Seattle, WA analytically driven programs that facilitate shared decision-making between doctor and patient can deliver success, particularly in addressing health cost Soyal Momin MBA concerns. Manager, Research and Development and Consulting, BlueCross BlueShield of Some of the wide variation in medical practice and treatment patterns is linked Tennessee, Chattanooga, TN to real differences in patient needs, but a considerable portion is unwarranted – that is, it cannot be explained by illness, medical need or the dictates of Jeremy J. Nobel MD MPH evidence-based medicine. In 2003, researchers at Dartmouth College produced Adjunct Lecturer, Department of Health a seminal research paper that suggested that if unwarranted variation in the Policy and Management, Harvard University Medicare system could be eliminated, the quality of care for Medicare School of Public Health, Boston, MA participants would be dramatically improved and Medicare costs would be 30% Seth Serxner PhD MPH lower. Principal, Mercer, Los Angeles, CA Unwarranted variation drives cost and quality problems in three categories of care, and can be addressed by these means: Jonathan P. Weiner DrPH Professor of Health Policy & Management, • More effective care. All too often, patients do not get enough effective care Johns Hopkins University Bloomberg – care for which evidence of ability to produce better outcomes is clear and School of Public Health, Baltimore, MD uncontested. While in the long run, effective care services improve health, they may actually increase costs in the short and medium terms. Publisher • Less supply-sensitive care. Care that is driven by health system capacity is Clive Riddle, President, MCOL known as supply-sensitive care. Evidence suggests that reducing unnecessary healthcare services dramatically reduces costs and improves Editor health outcomes. Russell A. Jackson • The right amount of preference-sensitive care. For ailments that do not Predictive Modeling News is published have a clear, evidence-based, singular treatment choice, patients often monthly by Health Policy Publishing, LLC. receive care that is not in line with their values and preferences. Health Newsletter publication administration is service choices that should be determined by patient preferences are provided by MCOL. known as preference-sensitive care. Predictive Modeling News The average healthcare dollar is spent on all three types of care, with 63 cents 1101 Standiford Avenue, Suite C-3 going for supply-sensitive care, 25 cents for preference-sensitive care and 12 Modesto, CA 95350 cents for effective care. Disease management has thus typically focused on just Phone: 209.577.4888 12% of the average dollar, as it focuses on getting appropriate tests and Fax: 209.577.3557 treatment for individuals with chronic conditions – the effective care piece of the puzzle. What about the other 88%? A care management program should address all three sources of unwarranted variation with a special emphasis on care that is preference-sensitive or supply-sensitive. Copyright © 2009 by Health Policy Ideal care management provides a broad range of support to help individuals Publishing, LLC. All rights reserved. No participate in their care choices so they can make informed decisions with their part of this publication may be reproduced or transmitted by any means, electronic or physicians and lead healthier lives. Individuals can influence their destinies mechanical including photocopy, fax, or when they perform these important functions: electronic delivery without the prior written • Participate in key decisions at key junctures of care. Support to help them permission of the publisher. get well-informed and know how to discuss preferences with providers is critical. For conditions for which there is no indicated treatment choice, informed patients will often choose less-costly and less-risky courses of action. • Become more informed about health conditions to reduce uncertainty about how much to respond to symptoms and warning signs. continued on page 5 To subscribe: visit or call 209-577-4888 page 2
  3. 3. May 2009 Predictive Modeling News 3 American Health’s Population Management Solution Uses Predictive Modeling to Target Interventions Integrating a health assessment and biometric values with PM to keep people healthy By Russell A. Jackson merican Health Holding Inc., Worthington, OH, has unveiled a new population health management program aimed at A helping payers and employers address skyrocketing health costs and at assisting individuals in achieving optimal health. A key component of the program is a robust predictive modeling application that’s populated both with daily and weekly downloads of medical and pharmacy data and with biometric data transmitted automatically from disease-specific devices provided to patients to AHH through the Internet. In a down economy, the company says, customers appreciate every opportunity available to them to reduce healthcare spending by keeping people healthier. Comprised of both disease management and lifestyle & wellness components, AHH’s population health management suite of applications is designed to help individuals become more involved in managing their health, a statement from the company says. Program highlights include disease management and lifestyle coaching; interactive online tools; at-home and onsite biometric testing; onsite and online educational programs; and disease-specific equipment, such as glucose meters, weight scales and blood pressure cuffs, which are equipped to send electronic readings to American Health’s nurse coaches. “Traditionally, disease management has placed an emphasis on helping to minimize health complications that result from the presence of chronic conditions,” says Michael J. Reidelbach, CEO and president at American Health. “However, population health management addresses the health needs of everyone, helping individuals achieve personalized goals to effectively manage a chronic condition, establish healthy living habits or maintain a healthy lifestyle.” American Health’s proprietary i-Suite software provides “superior integration of data” via leading-edge predictive modeling capabilities, biometric screenings and a health assessment, the statement adds. That facilitates early identification of individuals who can benefit from the company’s disease management and lifestyle & wellness resources. To promote program participation, each eligible household receives a welcome kit, which details the program benefits and options. The kit also contains healthy living information, product samples to help individuals effectively manage their health and an at-home biometric test. The test, which measures HbA1c and cholesterol levels, can help patients determine whether they have -- or are at risk for developing -- diabetes or heart disease. “One of the unique elements of the disease management program is the disease-specific equipment that is offered to participants,” notes Jan Marie Reed, vice president of marketing and client services at AHH. “The equipment, which can electronically send results to our nurse coaches, includes an online consumer portal that allows participants to track their results and easily print their testing history to share with their physicians. That promotes successful self-management and enables our nurse coaches to monitor participant progress and provide appropriate coaching.” American Health’s disease management coaching is “taking aim at the self-management competencies that drive behaviors and patient activation,” the firm says in the statement. The company has adopted the Patient Activation Measure used by many leading healthcare organizations, a clinically proven behavioral evaluation tool that assesses an individual’s confidence, knowledge, skills and ability to self-manage a chronic condition. Results of the assessment are used to help coaches collaborate with participants to set behavior change goals and pursue customized action steps. Participants in the company’s lifestyle & wellness programs have a variety of educational tools available to them, including a health web portal, which provides these components: • healthy living information; • a drug dictionary; • health and fitness trackers; “Assessment results are used to stratify participants • online health coaching programs; based on their level of risk and to identify those who can • a disease and condition center; benefit most from working one-on-one with an AHH lifestyle coach. Participants who meet the criteria for • a heart attack risk factor assessment tool; participating in lifestyle coaching receive education and • a menu planner; support to develop personalized plans for improving • quizzes; nutrition, implementing and adhering to a fitness • glossaries; program, managing stress and modifying other lifestyle • self-care tools; behaviors that impact health and well-being.” • a prescription drug guide; • a symptom checker; • live, 24/7 nurse chat; and • a health assessment. The portal also includes an Incentive Tracking and Reward Fulfillment Center that enables employers to provide gift card rewards to employees who engage in healthy behaviors. The health assessment is a critical component of the lifestyle & wellness program, AHH emphasizes, as it helps participants identify ways in which they are effectively managing their health and ways they can reduce their modifiable personal health risks. Assessment results are used to stratify participants based on their level of risk and to identify those who can benefit most from working one-on-one with an AHH lifestyle coach. continued on page 4 © 2009, Health Policy Publishing, LLC. All rights reserved. No reproduction or electronic forwarding without permission. page 3
  4. 4. 4 Predictive Modeling News May 2009 American Health’s Population Management Solution Uses PM…continued Participants who meet the criteria for participating in lifestyle coaching receive education and support to develop personalized plans for improving nutrition, implementing and adhering to a fitness program, managing stress and modifying other lifestyle behaviors that impact health and well-being. Predictive Modeling News talked to Reed about the disease management and lifestyle & wellness programs – and about predictive modeling’s key role in both. Predictive Modeling News: Is the predictive modeling component of the program tied directly to the health assessment tool? Or does it also process data from the home monitoring devices and the personal tracking services available through the online portal? Jan Marie Reed: AHH relies on daily and weekly downloads of medical and prescription drug claims data into the predictive modeling tool to identify persons with chronic conditions who may benefit from the disease management program. The health assessment tool is also used to identify persons with DM-manageable conditions by virtue of the member indicating on the health assessment that he or she has one or more of the targeted chronic diseases. The value of relying on both sources -- healthcare claims and the health assessment -- is it allows AHH to cast the widest net possible for identifying potential participants for the DM program and to do so in a timely manner. The predictive modeling tool also allows AHH’s nurse coaches to monitor DM participants’ compliance with nationally recognized evidence-based care guidelines. For example, a nurse coach can assess whether a diabetic patient has undergone a semi-annual glycated hemoglobin test and can counsel the participant accordingly. Data from the home monitoring devices, personal tracking services and at-home biometric test results are downloaded into AHH’s proprietary software, i-Suite, and are readily accessible to DM nurse coaches. The coaches reference each participant’s data values from the disease-specific equipment and incorporate that information into their coaching sessions with the participant. At any point in time, the nurse coach can not only assess the participant’s biometric readings, such as blood glucose levels or daily weights, but can also determine the extent to which the participant is compliant with the testing frequency recommended by his or her doctor. That additional information further enhances the education and clinical support provided to participants. PMN: Are the data used to populate a predictive modeling solution fully valid if they’re entered, unsupervised, by patients? Are doctor-derived data more reliable? JMR: We have learned through experience that self-reported results are not reliable and, therefore, the AHH program has addressed that shortcoming. The biometric data used to populate i-Suite are automatically downloaded via the disease- specific equipment DM participants receive from AHH. The equipment calculates the applicable readings and sends those electronic readings automatically to AHH; therefore, there is no opportunity to falsify the information. The data supplied through the disease-specific equipment are as reliable as physician-derived data. PMN: What kind of response are you seeing to the new product launch? What kinds of companies are clients of the service? Do you get more employers or more health plans as clients? JMR: AHH is successfully moving its current client base to the new program. In these economic times, employers in all arenas realize the need to improve the health status of their populations, which will ultimately yield lower healthcare costs. AHH has third-party administrator customers, health plan customers and direct employer groups. The program incorporates many components, including webinars, “Lunch & Learns” and coaching for disease management and lifestyle challenges. Various components are appropriate for different groups. The components are “building block” components that allow our customers to expand the programs to offer cost-containment initiatives to the full population of employers. PMN: Does your company also offer the disease management and wellness programs that the predictive modeling component indicates that patients might benefit from? If not, how do you direct patients to those programs? JMR: AHH offers a full range of products under its population health management umbrella. AHH’s PHM product line offers resources and tools to all persons, regardless of where they are on the health continuum – from healthy persons with no health risks to members identified as “at-risk” for developing chronic conditions to individuals already diagnosed with diseases. AHH’s lifestyle & wellness product serves all members, regardless of health status, but is most effective at keeping healthy people healthy and preventing those at risk from developing diseases. AHH’s disease management product provides comprehensive education, support, tools and resources to individuals with one or more targeted chronic conditions, such as diabetes and heart disease. The predictive modeling tool, the health assessment and the biometric values play a role in identifying members for the appropriate health management products and services. AHH is a multi-URAC-accredited medical management firm that provides case management, utilization management, disease management, lifestyle & wellness programs, maternity management, medical review, hospital bill auditing and pharmacy benefits management. Offerings also include case management specialty services that encompass neonates, organ transplants and oncology services. Visit To subscribe: visit or call 209-577-4888 page 4
  5. 5. May 2009 Predictive Modeling News 5 Care Management, Not DM, Needed…continued • Learn appropriate self-care for instances when self-care is appropriate, thereby avoiding unnecessary exposure to the forces that drive supply-sensitive care. • Coordinate care directly through a loved one or advocate. Individuals who understand how the mix of providers and treatments fit together are less likely to have disconnected care and complications that result in increased costs and worse outcomes. • Understand how to handle transitions in care to reduce the chance of ending up in the hospital again. • Change behaviors that put health at risk, such as poor nutrition, low physical activity, smoking, excessive alcohol use and other key behaviors. • Advocate for clinically indicated care to ensure receipt of key health treatments and tests. Care management must consider the spectrum of healthcare needs across individuals and their families and take into consideration not only their conditions, but their lives, values and preferences as well -- and their ability to take an active role in their care. Large opportunities can be missed by not understanding all of a population’s cost and health drivers. For example, a singular focus on people with a high risk for asthma may miss other key cost and quality drivers in the population and may, in fact, be money poorly spent if other individuals have higher risk. Care management programs must have these key components to be effective in developing the right skills in individuals: Tailored analytics. Care management programs need innovative analytics that uncover cost drivers at a population level and then drill down to individual interventions. Opportunities for impact can be identified by feeds of hospital discharge notices or specialty referrals, but truly effective analytics must utilize more data sources, including medical, pharmacy and disability claims; biometric and laboratory data; and self-reported information such as health risk assessments. Analytics should support rapid assessment and improvement of program components. Structured and timely assessment of the effect of interventions – measured in formal studies as well as quality monitoring systems – should enable learning, innovation and sensible resource usage. Innovative engagement tactics. Sophisticated and tailored outreach practices, which are fundamentally dependent on insightful analytics, are critical for success. The right set of analytics enables identification of the highest-risk individuals, but can also reveal the channel and content most appropriate for a particular individual. What gets shown, on paper, in email or on the web, or said, through personal interaction with a health coach or with an interactive voice response system, can dramatically affect an individual’s reactions. Varying the form and content of outreach to match each individual’s profile and preferences demonstrates sensitivity to the unique situation of each person and can considerably increase the likelihood of engagement. Individualized support. Interventions are far less likely to be effective with a one-size-fits-all approach. Individuals respond far better to support that meets them where they are, rather than support that assumes a single, linear course of reasoning with a narrow goal, such as taking a particular test or getting a prescription refilled. Flexible information systems. Information systems should simplify opportunity assessment, enable tailored production of outreach, deliver rapid evaluation of program effectiveness and, perhaps most importantly, facilitate understanding and prioritization for front-line clinicians who are supporting engaged individuals. In repeated assessments, across a broad range of populations, the core Health Dialog program has demonstrated an ability to reduce healthcare cost increases by $1.50 to $2.50 per dollar spent in the first year of the program, with rapid escalation in savings in subsequent years. The methods utilized for assessing those savings must, and do, meet these criteria: • They must be credible. • They must be conservative. Health Dialog’s adaptations to the basic savings methodology reduce systematic bias that inflates savings results. For example, Health Dialog does not take credit for large natural cost reductions that occur in the brief period after a major claim indicates an individual has a chronic condition. Not adjusting methodologies for the regression-to-the-mean time period can dramatically inflate savings. • They must be transparent. • They must be plausible. Take the case of several populations within a large regional health plan. In the first three years of the Health Dialog program, the focus of interventions was on individuals with specific chronic conditions and decision support needs. In the past year, services have expanded to a much broader set of individuals with high risk of future costs -- supply-sensitive and preference- sensitive care -- across the spectrum of conditions. The services are available to enrollees in HMO and PPO plans, for both administrative-services-only and fully insured groups, and to Medicare Advantage plans. The health plan validated the following first- and second-year savings, which greatly exceeded fees paid: • First-year savings exceeded fees by 60% for HMO enrollees, 80% for PPO enrollees and 20% for Medicare Advantage plan enrollees. • Second-year savings exceeded fees by 280% for HMO enrollees, 260% for PPO enrollees and 190% for Medicare Advantage plan enrollees. continued on page 6 © 2009, Health Policy Publishing, LLC. All rights reserved. No reproduction or electronic forwarding without permission. page 5
  6. 6. 6 Predictive Modeling News May 2009 Care Management, Not DM, Needed…continued Care Management, Not DM, Needed…continued Those results are well-supported by rigorous plausibility inconsistencies and lags in data, most are predictable testing of Health Dialog practices. Those plausibility tests, and can be factored into analytic approaches. In intended to support internal learning and quality some instances, including the Medicare fee-for- improvement, utilize formal study processes with service population, data problems are tremendous randomization, and each demonstrates the powerful impact and work-arounds complex. Examples of potential on medical costs of robust support for individuals with key problems that restrict program performance include risks. massive, unstructured errors in member identification data; large amounts of missing claims data; and A series of tests of different types of outreach for individuals significant claims lag problems. at risk for back and joint surgery, for example, demonstrated that populations with much higher levels of health coaching • Lack of real-time data feeds for urgent needs. had significantly lower surgical rates and significantly lower overall health costs. A second test of different outreach • Limited or poor pharmacy information. Pharmacy data approaches for individuals with chronic conditions, who had are extremely useful for identifying individuals with a high likelihood of future costs but who had not yet been conditions, for developing appropriate targeting reached by traditional outreach processes, demonstrated strategies and for enabling coaches to support that populations with much higher levels of health coaching individuals in need. With poor pharmacy data, had significantly lower overall health costs. program effectiveness is constrained. Visit Certainly there are care management programs that do not work, which has been most widely discussed around the Medicare fee-for-service population, but which can occur in Hospital Contracts: The Critical Elements of other populations as well. Problems that care management Financial Performance Modeling… continued programs can face in such instances include program design When the problems are not uncovered prior to contract constraints and systemic data limitations. Health Dialog’s execution, the negative financial consequences can be analysis shows that programs that do not reduce medical significant. Recognizing the risks and planning for costs are subject to one or more of these constraints: appropriate data reconciliation thus become important parts • Enrollment or opt-in requirements. Individuals with of the negotiation process. progressive chronic conditions are poor at predicting future needs. Enrollment or opt-in requirements bar Case Study: Overview access to initially unreceptive people who might pose This case study is based on an actual negotiation between tremendous opportunities for impact in the future. a large Midwestern hospital and an MCO. The MCO • Limited or no provider touch points. proposed a new agreement that changed the inpatient payment methodology from a percentage-of-charge • Narrow focus on clinical quality improvement. reimbursement scheme to All Patient Refined-Diagnosis Improving clinical quality as currently measured may Related Group case rates. The data and adjudication rules have little to no impact on short-term medical costs used by the MCO in its initial modeling resulted in a because it focuses on effective care, where just 12% material misstatement of expected contract performance. of dollars are spent, and because its initial goal is to The hospital was able to perform a reconciliation that increase use of that care. A common mistake is to resulted in a mutually acceptable data set that both parties overemphasize those types of measures at the agreed resulted in a reasonable projection of the contract’s expense of ensuring that people with chronic and financial outcome. other high-risk conditions avoid excess supply- sensitive care and get the right preference-sensitive Key Actions Taken care – where 88% of the money is spent. The main reason that people with chronic conditions can be so Because the hospital did not have the ability to group costly is their conditions cause them to interact claims into APR-DRGs, the MCO provided a model using frequently with the medical system. Thus, they are historic claims data and calculated a transitional base rate more at risk to be drawn into the other 88% than are estimated to be revenue-neutral. After an initial high-level healthy individuals. analysis, it became apparent that there were significant differences between the hospital’s and MCO’s data. A logic • Single- or limited-condition programs. The largest test compared total charges modeled by the health plan opportunities for cost savings can be discovered with the hospital’s claims data for the same time period. when the entire population is available for potential The hospital’s data included $4 million in charges – 14% -- intervention. A focus on one or two diseases more than what the MCO’s model contained. An in-depth constrains the overall impact significantly. reconciliation of the hospital and MCO data was needed • Data sources that are inconsistent or untimely. Data before a base rate could be calculated that both parties are the life’s blood of a care management program. agreed would yield revenue-neutral reimbursement. In Without reliable and timely data sources, addition, the initial review of the data and modeled programmatic focus is off, front-line support payments compared to the contract terms resulted in processes are ill-advised and course corrections are identification of errors in the way the MCO model treated difficult. While it is expected to have small some of the claims continued continued on page 7 To subscribe: visit or call 209-577-4888 page 6
  7. 7. May 2009 Predictive Modeling News 7 Hospital Contracts … continued A quick visual scan of the data showed that there were payments in excess of charges and that the short-stay methodologies had not been applied. Those red flags suggested the need to perform an in-depth reconciliation. Data/Modeling Reconciliation Process The problem: The $4 million difference between the hospital’s and the MCO’s data sets. The first-pass comparison involved identifying the cases that were common to the two databases. Working with the hospital and health plan, we identified a key -- unique reference number -- that would enable us to create a crosswalk of individual cases. Combining the reference number with the account number allowed us to match individual cases with the health plan’s data. The ability to crosswalk cases between the two data sets enabled us to identify discrepancies on a line-by-line basis. They included: * Cases in the hospital’s data that were not in the MCO’s, and vice versa * Single cases in the hospital’s data that were treated as multiple cases by the MCO * Professional fees inadvertently included * Babies treated as separate episodes of care by the hospital but combined with the mother by the MCO * Patients registered with the MCO as secondary but later determined to be primary * Interim bills from a percentage-of-charges methodology not combined into a single DRG case for the proposed scenario * Short-stay rules not applied properly The Financial Impact The final revenue-neutral base rate was 21% higher than the initial MCO-proposed base rate -- an impact of more than $4 million a year. The root cause was that the health plan initially modeled its proposed conversion from a percentage-of-charges to an APR-DRG methodology by using a model designed to compare a conversion from AP-DRGs to APR-DRGs. The magnitude of the error was 3% of the allowed amount. While tedious, the data and model reconciliation process can identify variances that may have an impact of millions of dollars. Failure to recognize the issue would likely erase any of the increases gained during the negotiations or, even worse, leave the hospital in a less favorable financial condition than when it started. Issues Identified During the Reconciliation Process The table below includes a summary of the key issues identified during the data/modeling reconciliation process. Key Issues List – Data/Model Reconciliation Type of Issue Explanation Impact “Lesser of” Application The health plan model indicated Overstated the yield of the proposal. reimbursement at the DRG rate that was higher than charges. Annual Charge Master The health plan modeled the impact of current Understated the yield of the current Increases charges compared to the proposed DRG situation by not accounting for payments. contractually allowed increases. Registration Issues The MCO was originally indicated as Data missing from the hospital database secondary and then was found to be primary. but included in the MCO database. Differences in Charges The MCO model did not include all charges Understated charges compared to tied to a specific claim. expected payments. Interim Billing The health plan model treated interim bills for Overstated the yield of the proposal. individual cases as separate claims and projected a DRG payment for each. Refunds Claims were included that had been refunded Overstated payments in both current and or retracted. proposed scenarios. Babies Charges for babies were included with the Understated expected payments by mothers. missing the additional DRG. Inclusion of Inappropriate Data Hospital data inadvertently included Overstated both current total charges and professional fees. reimbursement. Conclusions Current economic conditions are placing stress on hospital and MCO finances, and there is little or no leeway for unanticipated reimbursement outcomes. Creating a financial model to understand the impact of a proposed contract is critical. continued on page 8 © 2009, Health Policy Publishing, LLC. All rights reserved. No reproduction or electronic forwarding without permission. page 7
  8. 8. 8 Predictive Modeling News May 2009 Hospital Contracts … continued Differences in the data sets and modeling assumptions can result in a material misstatement of financial performance. It is paramount that MCOs and hospitals reconcile their data sets, assumptions and contract models during negotiations so that the intended financial consequences are realized. ECG offers a broad range of strategic, financial, operational and technology-related consulting services to healthcare providers. The firm has extensive experience assisting hospitals and medical groups in different regions of the country in commercial health plan contract analysis, negotiation and revenue capture and boasts expertise in every step of the commercial health plan process, including defining the contracting strategy, modeling the revenue impact of contract terms, negotiating contracts and resolving payment disputes. Visit Contact Welter at or Brown at or either at 703-522-8450. Thought Leader’s Corner Each month, Predictive Modeling News asks a panel of industry experts to discuss a topic suggested by a subscriber. To suggest a topic, send it to us at Here’s this month’s question: Q: “What categories of high-risk patients have you found predictive modeling applications the most successful, and least successful, at identifying?” “Employer groups want to understand how they can use predictive modeling to reduce healthcare costs as well as efficiently provide appropriate healthcare to their employees at the optimal point in time. Health plan clients are using predictive modeling to inform employer groups regarding medical cost drivers and to facilitate coordination of care for subsets of employees. Examples include: • Describing member population demographics by condition or diagnosis predicted with high forecasted cost compared to the prior time period -- e.g., since renewal or the prior calendar year. • Members who have gaps in care. Describing subsets of the employee population compliant with preventive healthcare gap measures. • Quantifying employee subsets that have currently low healthcare expenditures but that are predicted to have higher costs in the next 12 months. That empowers employer groups and payers alike to appropriately identify members for case or disease management programs. Health plans use the predictive modeling risk stratification to place members in risk categories or levels for appropriate care management interventions. • Structure customized benefit plans for employees. Employer groups may agree to a health plan’s recommendation to focus more intensive clinical resources on predictive modeling-identified employees with diabetes-related care gaps who are expected to have higher inpatient or emergency room utilization within the next 12 months. • Demonstrate to employer groups that predictive modeling provides a means to identify their ‘well associates’ who require educational/informational health literature outreach or a ‘light’ touch by a case or disease manager.” Swati Abbott President, MEDai, Inc. Orlando, FL “We have found that predictive modeling is effective at predicting future high-cost patients. However, high-cost patients are frequently unmanageable by the type of programs that a plan has in place. For that reason, Solucia has developed a predictive model ‘overlay’ that we incorporate into our financial predictions that takes into account additional factors: intervenability, the type of care management program and the program cost. Thus, a member’s risk score is adjusted to incorporate the intervenability score, resulting in high-risk predictions that are more likely to be amenable to management. For the purpose of re-scoring members, we define ‘intervenability’ as the percentage of a member’s predicted costs that are likely to be amenable to management by the program. Thus, a member who is high-risk because of, say, cancer, may have a low intervenability score, because there may be few things that a program can do to reduce the cost of normal care. On the other hand, a patient with multiple chronic conditions may be intervenable because a program may be able to impact costs. “ Ian Duncan FSA FIA FCIA MAAA President, Solucia Inc. Farmington CT continued on page 9 To subscribe: visit or call 209-577-4888 page 8
  9. 9. May 2009 Predictive Modeling News 9 Thought Leader’s Corner…continued “Categories of members we have found predictive modeling applications to be most successful at identifying include moderate risk with propensity to progress toward high risk with high impactability. That includes conditions that are impactable by lifestyle changes/compliance with treatment or therapy. We have not had much success with identifying members with catastrophic conditions/rare events. That may be due to the unavailability of a diverse but integrated data source, timeliness of data and embedded experience.” Soyal Momin MS MBA Manager, R&D and Consulting, BlueCross BlueShield of Tennessee Chattanooga, TN “Chronic conditions are the simplest to identify and are a good tool to assess the involvement of members in different wellness programs or initiatives. The easiest correlation has been with those conditions that are coordinated to disease management programs, such as diabetes, asthma, COPD, heart disease and CHF. Cancer can be identified, but the stage and treatment can impact the ability of the models. What we have observed is that for employers utilizing the services of a disease or case management vendor, there is easier identification and understanding of the care of individuals with chronic diseases. By understanding the risk scores generated by the predictive models, the vendors are able to stratify the individuals to find the ones on whom they can have the most impact. We can also see how well higher risk scores and compliance with treatment correlate to other improvements, such as increased medication adherence, lower ER use and reduced inpatient stays.” Russell D. Robbins MD MBA Principal & Senior Clinical Consultant, Mercer Norwalk, CT “One aspect of the answer has to do with data availability; that is, the more risk information you have on the member, the more likely the PM predictions are to be accurate. For example, a long-time enrollee with full diagnosis and pharmacy -- and maybe survey -- data will have a more complete predictive picture than a new enrollee with only partial data. Also, PM models will be particularly useful in identifying persons who have multiple morbidities, none of which alone would be enough to put the patient on the plan’s radar screen.” Jonathan Weiner DrPH Professor, Health Policy and Management; Director, PhD Program in Health Services Research and Policy; Deputy Director, Health Services Research & Development Center, Johns Hopkins University Baltimore, MD INDUSTRY NEWS Verisk Health Appoints Gunn Chief Medical Officer Verisk Health Inc., Waltham, MA, reports the appointment of Nathan Gunn MD as chief medical officer, with responsibility for managing scientific and clinical research operations, developing and maintaining clinical content and intellectual property and overseeing the clinical content of Verisk Health software products. Prior to leading the clinical team at Verisk Health, Gunn was senior vice president of clinical solutions at D2Hawkeye, which merged with Verisk Health, a subsidiary of ISO, in January. Before joining D2Hawkeye, he worked at McKinsey and Company, where he counseled industry and government leaders nationally and internationally on provider operations and growth strategies, private equity and international health system reform and design. Visit continued on page 10 © 2009, Health Policy Publishing, LLC. All rights reserved. No reproduction or electronic forwarding without permission. page 9
  10. 10. 10 Predictive Modeling News May 2009 INDUSTRY NEWS Carefx Reports 200% Year-Over-Year Revenue Growth Rehab Firm to Study Predictive Tools for Nursing Carefx Corp., Scottsdale, AZ, reports 200% revenue growth Home Falls in 2008 and increased penetration of the hospital, health Northbrook, IL’s Simply Rehab LLC has launched a multi- system, integrated delivery network and health information state project to help operators of skilled nursing facilities exchange markets, with an 80% increase in client bookings and assisted living facilities improve their ability to predict and the addition of some 200 hospitals to its roster. “We see which residents are at risk for falling. The program also our strong growth as continued validation that organizations includes measurements that lead to appropriate can no longer settle for proprietary or monolithic systems preventive actions for the at-risk residents, the company that do not support clinical workflow or cognitive tasks and says. Falls among community-dwelling older adults are a force clinicians to sift through enormous bundles of data significant problem in the US, it adds, with Medicare’s cost from disparate sources,” says Andrew Hurd, chair and CEO of treating the injuries resulting from falls approaching $3 at Carefx. “We built the Fusionfx solution suite to address billion a year. Statistics gathered by the Centers for the information gaps that occur between disparate systems Medicare and Medicaid Services indicate that for every 10 to simplify clinical and business workflow and to ultimately residents of nursing homes in the United States, five falls improve the quality of care delivery.” are reported each year. In the study, nursing homes in Hurd’s comments are echoed, according to the company, in Illinois, Colorado and North Carolina will use a predictive a January 9, 2009, National Research Council report, which methodology that incorporates “a unique collection of offers a critical evaluation of current healthcare IT solutions’ elements” from the popular Berg Balance Scale Test and performance in the areas of data sharing, integration and “improved predictive techniques” that include the Alternate management. The NRC’s “Computational Technology for Step Test, the Sit-to-Stand-5-Times Test and the Six- Effective Health Care: Immediate Steps and Strategic Meter Walk Test. In one study, those methodologies Directions” advises organizations to “develop the necessary proved to be superior to other popular techniques at data infrastructure for healthcare improvement by predicting who would be a multiple faller. Robert Kunio, aggregating data concerning people, processes and project manager and chief operating officer at Simply outcomes from all sources” and to “organize incentives, Rehab, says skilled nursing facilities and assisted living roles, workflow, processes and supporting infrastructure to facilities are under enormous pressure to reduce the encourage, support and respond to opportunities for clinical frequency and cost of falls among their residents. “Rather performance gains.” Visit than focusing our efforts on helping our clients better react to falls,” he comments, “we have decided to concentrate our efforts on better predicting who is at risk and then taking appropriate preventive measures.” Adds Elizabeth Roth, president there: “We have now assembled and begun testing a collection of improved predictive Mayo Study Shows Finger Device May Help Predict techniques for both the ambulating population and the Future Heart Events propelling population. Our database is giving us a good Results of a Mayo Clinic study show that a simple, non- correlation between quarterly predictive scores and actual invasive finger sensor test is “highly predictive” of a major falls, which then allows us to fine-tune the protocol on a cardiac event, such as a heart attack or stroke, for people regular basis for various segments of the geriatric who are considered at low or moderate risk, according to population. The outcome of the effort should be an overall researchers there. The device, called the EndoPAT, by reduction of falls and the cost of falls to our clients, to Itamar Medical, measures the health of endothelial cells by Medicare and ultimately to taxpayers.” Simply Rehab measuring blood flow, a statement from Mayo explains. provides physical, occupational and speech therapy Those cells line blood vessels and regulate normal blood services to nursing homes and exercise programs to flow, and research has shown that if they don’t function assisted living facilities. Visit properly, they can set the stage for atherosclerosis and lead to major cardiovascular health problems. In the study, 49% of patients whose EndoPAT tests indicated poor endothelial function had a cardiac event Connextions Taps Auerbach as President during the seven-year study period. Researchers at Orlando’s Connextions Inc. has named Steven G. Rochester, MN-based Mayo and at Tufts-New England Auerbach president. He’ll take on overall management -- Medical Center, Boston, used the device to test 270 patients including strategy development, client services, operations between the ages of 42 and 66 and followed their progress and sales -- for the company’s sites in Florida, North from August 1999 to August 2007. Some of their risk factors Carolina and Indiana. He was most recently executive vice were high blood pressure, high cholesterol, obesity and a president of provider/member operations at UnitedHealth family history of heart disease. The test can be used in an Group (NYSE:UNH). Connextions is part of the New individualized medicine model of risk assessment of Mountain Partners II LP portfolio. Visit patients. Visit To subscribe: visit or call 209-577-4888 page 10
  11. 11. May 2009 Predictive Modeling News 11 Subscribers’ Corner Delivery Options Remember, you can receive each issue of Predictive Modeling News via email in an electronic pdf version, via regular mail in print version, or both. There is no additional charge for whichever option you select. Should you wish to confirm or change your delivery option, feel free to contact us anytime. Subscriber Web Site Subscribers can access an archive of current and past issues of Predictive Modeling News, view added features, change account information and more from the Subscriber web site. To access the site, click the “Subscribers” link at, and then click the Subscriber Login link. Predictive Modeling Web Summit 2009 Positioning the Impact of Healthcare Predictive Analytics in 2009 and beyond Wednesday June 3rd, 2009 1PM Eastern Webinar with Jonathan P. Weiner, DrPH, Johns Hopkins, Leslee J. Budge, Kaiser and Brian Wolf, MD, BCBS Rhode Island, plus five additional faculty podcasts and more. Get Details online at Subscribers receive a 50% Discount off the $295 Registration Fee! Call 209.577.4888 to register Catching Up With….Rong Yi..continued from page 12 Rong Yi PhD continued: I was looking for work, and my husband’s job was in Boston. I’ve been with the company ever since. Within DxCG, I started as a research associate, then became a senior research associate, then a senior scientist and then the senior scientist and then principal in the analytical pharmacy group. I kept some of my scientist responsibilities as I started to take on consulting work, and in February of this year I formalized my consulting role with my current title. PMN: How did predictive modeling become one of your responsibilities? When did you first hear about it, and when did you first start deploying it? How has it changed over time? RY: “Predictive modeling” wasn’t a catchphrase when I started, although by late 2004, it had started to be. It’s different from when DxCG was first founded on risk adjustment, which is based on a set of very rigorous principles. That focus was on how to design a system that accurately reflects the disease burden of an underlying population and provides the right incentives. It has to be fair. Sicker patients cost more money, and plans that are responsible for more risk should be correctly rewarded. DxCG was founded on those principles, which actually represent more of a public health kind of approach. Predictive modeling has taken us one step further. Risk assessment laid out the foundation for estimating healthcare cost, utilization and risk in the future, but it’s based on the payer’s view. There are things that could improve predictive accuracy that we don’t use because of incentive concerns -- things like procedure information and pharmacy information traditionally have not been used in the risk adjustment world, like they have in predictive modeling, where they’ve added predictive accuracy to understanding future costs. Here’s an example: Risk-adjusted payment is based on diagnoses. Predictive modeling’s goal is to predict, as accurately as possible, healthcare outcomes, quality, readmissions and the like. That’s a very big difference. Early on, when I started at DxCG -- which, by the way, because it had come out of a public payment and health economics approach, for a time stayed largely within the risk-adjustment domain -- we were not using procedural or pharmacy information. As the industry moved forward, though, we also moved forward, and we’ve stayed ahead of the curve, becoming and remaining a leader in predictive modeling. Now we’re using all kinds of information – procedure data, prior cost data, prior utilization data and more -- to get better and better predictive accuracy. For me personally, it’s very exciting, personally experiencing the growth of DxCG and of the predictive modeling industry. PMN: What occupies a typical day or week for you? What functions, activities and workload are you typically engaged in? RY: There’s quite a bit of variability, especially with my new consulting role. I’m more client-based than before, so I work very closely with our sales group and our marketing group. When a lead comes in, I write up a proposal, basically designing in my head what the study should look like and what we should be doing. Consulting is an interesting world for me personally. I’ve done research, cranking the models, making sure everything works, for a number of years. Now I also work closely with clients to convey the potential of the work we can perform. My consulting activities have created a role for me to work with our sales group to identify opportunities for our software to be adapted to meet our clients’ business needs. That allows me to be more supportive. I helped develop all the older models, so I know them intimately. Now I work more on the customization side. PMN: Is there anything else you'd like to share about yourself with the predictive modeling community? RY: I really have learned a lot from my colleagues and from our competitors. Healthcare is my passion. As a health economist, what interests me particularly is the complexity of the system and all the different players. And I’m delighted by the prestige of being associated with an organization that’s viewed as a gold standard in the industry. © 2009, Health Policy Publishing, LLC. All rights reserved. No reproduction or electronic forwarding without permission. page 11
  12. 12. 12 Catching Up With….Rong Yi PhD An industry leader expands the reach of predictive modeling across the globe Rong Yi PhD, vice president of consulting at Verisk Health Inc., Boston, has probably done as much as anyone in the business to expand the reach of predictive modeling in healthcare beyond the borders of the United States and onto the shores of foreign countries throughout the world. She’s leading a large project with the Health Authority of Abu Dhabi, for example, to develop health risk assessment tools and predictive models that use lifestyle and medical claims to predict disease progression and life expectancy. She already led adoption of US-devised risk adjustment models for the German market, including its migration to ICD-10 coding, and helped train German health services researchers to use the newly developed models. She also led research projects for the Kassenärtzliche BundesVereinigung (Physicians Union of Germany) to predict physician activity levels, risk-adjustment payments and payments by specialty. And she’s working with care management companies, private insurers and consultants in Germany and the United Kingdom to expand the use of DxCG models in care management, physician payment, revenue forecasting and other areas. Yi has also presented in China, opening new markets for DxCG there and in Taiwan, and will return to her native country soon to speak on risk adjustment and predictive modeling at the International Health Economics Association’s 7th Annual Congress in Beijing in July. In her current post at Verisk, she leads the firm’s analytic consulting practice and works with domestic and international clients to solve business problems that are not typically addressed within standard product offerings. Yi’s been responsible for the research and development of Verisk/DxCG’s risk adjustment and predictive models for healthcare, worker’s compensation and disability, and, clearly, for its international healthcare markets. Before that, she served as a senior scientist at Verisk and as the firm’s analytic services principal. Prior to joining Verisk in 2001, she was a research assistant and a research fellow at Boston University, where she was also part of the adjunct faculty at the School of Management and the College of Arts and Sciences. Rong Yi PhD • PhD in economics, Boston University, 2001 • BS in economics, Peking University, Beijing, 1995 • In the commercial health insurance space: --Led the patient-centered medical home pilot project for a physician-owned health plan --Led three biannual releases of DCG/HCC and RxGroup models --Responsible for developing the majority of the predictive models in DxCG’s current software releases --Designed and executed a risk-adjustment methodology for Episode Treatment Groups and Medical Episode Groups --Developed PM for a large health plan and its contracting physician groups for pay-for-performance programs --Designed a risk-adjustment study for a healthcare purchaser consortium to compare hospital efficiency • In the Medicare, Medicaid and uninsured space: --Led risk-adjustment model selection & implementation project for Massachusetts Health Insurance Connector Authority --Working on predictive analytics for the Medicare Advantage and Part D markets --Led and designed research to incorporate DxCG’s relative risk scores into the Medical Expenditure Panel Surveys data --Led and executed the research to estimate disease burden and healthcare needs for the uninsured using PM • And in the workers’ compensation and disability management space: --Developed predictive models and case identification tools for a leading workers’ compensation insurer to prospectively identify and manage slowly emerging high-cost claims based on initial medical claims following injury • Recent conference presentations include: --“Risk Adjustment and Predictive Modeling,” at the 5th Annual World Health Care Congress – Europe, in Brussels --“Risk Adjustment and Predictive Modeling for Medicaid,” at the 2008 National Predictive Modeling Summit --“Predictive Modeling Research Methodologies and Applications for Workers’ Compensation and Bodily Injuries,” at a 2008 Insurance Services Office User Conference --“Understanding the Healthcare Needs for the Under- and Uninsured Population,” at the 2007 National Congress of the Under and Uninsured --“Understanding Geographic Variation in the Privately Insured Population in the United States,” at the International Health Economic Association 2005 Research Meeting, in Barcelona Predictive Modeling News: What path did you take to your present position, starting right out of college? Was it the career path you envisioned when you started? Rong Yi PhD: While studying for my PhD in economics at Boston University, I had to find a way to finance my own education, so I taught various courses, including finance, financial markets and macro-economics. One of my advisors was DxCG co-founder Randy Ellis, one of the best-known health economists in the world, who led me to that company, which became my first full-time job continued on page 11 Published by Health Policy Publishing, LLC ● 209-577-4888 ●