Effective risk adjustment is a critical aspect of healthcare reform. Accurate and feasible methods for measuring risk variations in the healthcare market are needed to compensate health plans appropriately for the health risks their enrollees represent.

• Risk adjustment in the healthcare market is a major business and IT issue that needs to be addressed for healthcare reform so that health insurers can be compensated for adverse selection.
• There are many business and IT approaches to risk adjustment, each with its own positive and negative attributes.
• The member data that is needed to create risk adjustments is daunting, and projections require sizable amounts of data and analysis.

• Enhance communication among researchers, employers and policymakers to stay educated on the different approaches for dealing with the problems of risk segmentation in the healthcare market. Further demonstrations and comparative analyses of risk adjustment tools must be conducted.
• Understand the data challenges of gathering sufficient information to make risk adjustment work. Risk adjustment involves an enormous amount of effort to obtain data that is difficult to measure. Incorporate incentives for participation in risk assessments during the enrollment process to gather more-accurate, self-reported health information.
• Evaluate more-sophisticated tools that can adequately predict risk in settings with mixed populations, inpatient and outpatient utilization, and multiple data sources. Risk adjustment tools have been in development for decades. However, many use a fairly narrow database. All the patient's risk factors (such as age, diagnosis and comorbidities) should be considered when using risk adjustment to determine costs of care.

U.S. health insurers currently have an incentive to attract healthy enrollees and try to exclude sicker or higher-risk enrollees. They gain more by competing on the basis of risk selection practices than by competing on the basis of efficiency and quality. A potential method for solving the problems of risk selection is risk adjustment. Risk adjustment is an analysis mechanism to help reduce both the negative financial consequences for health plans that enroll high-risk users and the positive financial consequences for health plans that enroll low-risk users. The absence of mechanisms for pooling and pricing risk is a greater problem in the individual market. As a result, many health insurers do not participate in the individual market or offer policies at prohibitively high premium rates.

Risk adjustment can be a corrective tool that may help reorient the current incentive structure of the insurance market and reduce the negative consequences of enrolling high-risk users by somehow compensating insurance plans according to the health risk of the enrollees they take on. It requires some means of calculating the expected healthiness of a pool of people and the fair transfer of payment. Robust analytics and predictive modeling are just a couple of tools needed to support effective risk adjustment.

Effective risk adjustment is a crucial aspect of healthcare reform and needs to be addressed, because, as the bill for U.S. healthcare reform rolls out, health insurance coverage for the uninsured population in the United States will increase. A key question that health insurers are raising is: "How can we be protected against adverse selection?"

Healthcare reform will require that all health plans in the individual and small group markets be subject to the same system of risk adjustment. The secretary will be required to prequalify entities conducting risk adjustment. However, these entities cannot be owned or operated by insurance carriers. The secretary can define qualified risk adjustment models, which can be used by the states, or states can choose to develop their own risk adjustment model, but it must produce similar results and not increase federal costs. There will also be a reinsurance mechanism under which all insurers in a state will have to pay an amount proportional to their insurance premiums into a reinsurance fund. This fund will then pay out to insurers based on how many high-risk enrollees they have.

The Patient Protection and Affordable Care Act (PPACA) directs the establishment of the American Health Benefit Exchange. The exchange is intended to facilitate the purchase of qualified health plans by small employers and individual purchasers. These exchanges are buying centers that will allow health insurers to compete directly on price and benefits in a comparable market. They will be a game changer for health insurers and will turn health insurance choices into commodities, with buying decisions based largely on price rather than value-based decision making.

The Achilles' heel for exchanges will be adverse selection. Current exchanges tend to attract (or to have dumped upon them) sicker and more costly enrollees. For healthcare reform exchanges to work, they will need to establish a risk adjustment mechanism for all health plans in a given geographical area so that insurers are protected from adverse selection. The exchanges will require insurers to accept all applicants without consideration of the applicant's health, and it will further prohibit or significantly limit premium variation related to health status. Current rating and underwriting practices will need to change, because exchanges will eliminate risk-based underwriting for the small group and individual markets. Exchanges will have the authority to set insurance prices and impose mandatory minimum medical-loss ratio requirements.

We can learn from the initiative in Massachusetts, where legislation was enacted in 2006 that would provide nearly universal coverage to all state residents. The bipartisan legislation combined individual responsibility, through a mandate to purchase insurance, with government subsidies to ensure affordability. The Commonwealth Health Insurance Connector Authority was formed to manage Massachusetts' initiative for universal healthcare coverage for the Commonwealth of Massachusetts. The Connector Authority is an independent public entity that was established to facilitate the availability, choice and adoption of private health insurance for eligible individuals and employer groups, as regulated by the Massachusetts Health Care Reform Act of 2006.

The Connector Authority has publicly shared goals for the program to:

• Secure fair and reasonable (not excessive) rates
• Mitigate risk selection and bidding gamesmanship
• Protect members from large premium differentials
• Align health plan payment with actual health risk and care management goals
• Increase transparency and simplicity

Risk adjustment was the basis for the program to help offset risk selection issues and establish a fairer payment mechanism. The project was based on the use of predictive modeling-based risk adjustment to allocate program rates across managed care organizations that will be underwriting the uninsured who are now covered under the program. By introducing predictive modeling that would better align payment to population acuity, the Connector Authority hoped to level the playing field among health plans, allowing for competitive bidding among a greater number of plans.

Verisk Health was chosen to work on the project and assist with risk-adjustment model selection. Verisk Health solutions apply business analytics to disparate data sources to filter out the members with the greatest chance of incurring high costs based on their illness burden, lifestyle indicators and omissions in care. The tools combine medical, pharmacy, eligibility, and other care management information into one comprehensive database and then apply clinical algorithms and predictive models that transform data into meaningful information.

The predictive modeling tool will be used to align the monthly capitation payment with the risk of the enrollees in each managed care organization. From the same pool of money, payments will be aligned to individual plans, taking dollars away from plans with better-than-average risk and providing dollars to other plans with worse-than-average risk. The predictive modeling tool will not be used to help set the rates, but to allocate dollars across the plans. The predictive modeling program will be set up so that payment will be better aligned with risk in the short term. In the long term, the software may be used in creative ways to better understand the acuity of the population and provide some insight into ways that specific disease management programs can be structured. In addition to the Connector Authority using the model, the participating health plans also use the same models, thus creating a consistent view of risk for both parties. This is similar to the approach that Medicare Advantage plans use — Medicare (models created by DxCG for CMS) and the plans use the same risk adjustment models to ensure a consistent methodology.

The challenge in developing risk adjustment mechanisms lies in finding ways of measuring or assessing the risk to project accurately who falls into the category of higher risk and how much their expenses will exceed those of the average enrollee. No single tool will be enough to solve all the problems risk brings to the system, but in conjunction with other tools and reforms, well-designed risk adjustment mechanisms can help make health plans less cautious about the risk involved in covering high-cost enrollees. The solutions will likely vary depending on the population served, but done well, risk adjustment can be used to motivate health plans to seek out high-risk enrollees.

The first step in implementing a satisfactory risk adjustment system is developing an adequate risk assessment tool. Risk assessment is a method that health insurers can use to evaluate the predicted overall healthcare claim dollars for each member relative to the average members in a given patient population. There are two types of risk assessment models:

1. Prospective or predictive models, which use data on a member from a previous year to estimate the member's future expenses and set health insurance premium rates for a given patient population
2. Concurrent models, which draw on member data collected in the current year to explain expenses in the same period (Typically, health insurers use concurrent models to profile physicians and other healthcare professionals because these models capture more of the costs of actual utilization during a year, while prospective models only make predictions of future utilization.)

The risk assessment process entails feeding claims data and data from pharmacies, laboratories and member-reported information into risk-modeling programs. The methodology that these models use to predict risk and/or determine costs of care varies, but all modeling software produces a relative risk score for each member in a population when data of that population is run through the software. The relative risk score demonstrates what the population's predicted risk or predicted cost of care will be to the health insurer.

Risk adjustment can now be used to adjust payments to health insurers or physicians in order to reflect the differences in risk, as measured by the risk assessment process. Predictive modeling systems can analyze data from one or more sources to predict which members will incur the highest costs during the next 12 months.

There are notable differences between risk assessment models and risk adjustment models. Predictive risk assessment models are used to predict claim dollars, whereas risk adjustment models are designed to predict health status. The models also have different data requirements. Predictive risk assessment models do not have any restrictions on the type of information fed them, while risk adjustment models require detailed diagnosis and pharmacy data and generally exclude utilization, claim dollars, and procedure code information.

There are vendors in the U.S. healthcare market that have developed a variety of models. A few examples include:

• Diagnostic Cost Groups (DCGs) from Verisk Health. DCGs use age, sex and diagnoses generated from patient encounters with the entire medical delivery system to infer which medical problems are present and what their likely effect on healthcare costs for a given population will be. They also allow the ability to determine the distribution of each problem to the overall risk/cost. (This is what makes them applicable in both the financial and medical management settings.)
• Episode Risk Groups (ERGs) from Ingenix. ERGs are episode-based, and the methodology used is a basic illness classification system that uses a series of clinical and statistical algorithms to combine related services.
• Medstat Medical Episode Grouper (MEG) from Thomson Reuters. MEG classifies disease-specific episodes of care based on organ system, etiology and severity of illness.
• Adjusted Clinical Groups (ACGs) from Johns Hopkins University. ACGs cluster health plan members, having similar comorbidities, into groups that have similar resource requirements and clinical characteristics.

Enhance communication among researchers, employers and policymakers to stay educated on the different approaches for dealing with the problems of risk segmentation in the healthcare market. Further demonstrations and comparative analyses of risk adjustment tools must be conducted.

Understand the data challenges of gathering sufficient information to make risk adjustment work. Risk adjustment involves an enormous amount of effort to obtain data that is difficult to measure. Incorporate incentives for participation in risk assessments during the enrollment process to gather more-accurate, self-reported health information.

Evaluate more-sophisticated tools that can adequately predict risk in settings with mixed populations, inpatient and outpatient utilization, and multiple data sources. Risk adjustment tools have been in development for decades. However, many use a fairly narrow database. All the patient's risk factors (such as age, diagnosis and comorbidities) should be considered when using risk adjustment to determine costs of care.