Risk Adjustment ... And Quality Adjustment

• Abstract
• Introduction
• Goals of a Risk Adjustment Arrangement
• Designing Risk Adjustment Systems for the Right Sector
• Is Risk Adjustment Worthwhile?
• HMOs and PPOs
• Designing the Risk Adjustment Model Properly
• Data for Managing Risk Adjustment
• The Importance of Transition
• The Need for Quality Measurement, Reporting and Adjustment
• Moving Risk Adjustment Forward

Abstract

Individual decisions about joining a health plan or choosing a physician organisation for care can lead to substantially different disease burdens among these entities. This kind of "adverse selection" produces a concentration of high-risk and high-cost patients that, if not adjusted for through differential payments, can create significant financial inequity. The primary goal of a risk adjustment arrangement, then, is to encourage the providers of health insurance (the "health plans") and the providers of health care to compete strictly on the basis of quality and efficiency of care, not risk selection. Risk adjustment systems must be designed to suit different situations. Risk adjusting health plan premiums may be more appropriate in relation to health plans serving the elderly and disabled (Medicare), the poor (Medicaid), and the small group market, where risk variation between such plans may be more pronounced than it is in the commercial HMO (health maintenance organisations) arena. If HMOs and PPOs are in the same risk adjustment pool, the system must be carefully designed to account only for variations in risk, not underlying cost or relative efficiency. Risk adjusting payments from health plans to providers may, however, produce more significant results than developing models adjusting premiums paid to the plans because there is likely more risk variation among providers than plans.

Whether health plan premiums or provider payments are the focus, a risk adjustment system must have a significant enough effect in shifting money and equalising financial burdens to balance the effort and investment required to develop the system in the first place. The system should be based on risk factors or diagnostic information, not costs (lest it simply reward more expensive or more frequent utilisation per se), and should promote good data capture, especially the encounter data necessary in capitated systems. A careful transition period to the new risk adjustment model will promote understanding of the potential for gains (and losses) from the adjustments, as well as expose the potential for "gaming" the system. Finally, risk adjustment should be accompanied by the measurement and reporting of service quality and the adjustment of payments on this basis as well. It is necessary, but not sufficient, to adjust health plan or provider payments based on a larger proportion of high-risk individuals. It is equally important to adjust payments for how well these individuals’ health care is actually managed.

Managed care works best in a competitive environment in which health plans and provider groups succeed or fail based on public accountability for quality as well as price. Quality should include both the proactive "health maintenance" of low-risk individuals and the excellent care management of high-risk individuals. Designing a framework in which this kind of "quality competition" can thrive means designing appropriate risk adjustment mechanisms as well as better performance measurement tools.

Introduction

Most individuals have choices when it comes to selecting the organisation that will provide their health insurance (referred to in this paper as "health plan") or their actual health care. Since they will act in their own financial self-interest, they will choose a health plan with a generous drug benefit if they require medications for a chronic illness, or a plan with low co-payments for office visits if they must see their doctor regularly, or a physician known for clinical excellence in treating HIV, diabetes, or heart disease if they have one of those conditions. This collection of individual decisions can result in substantially different disease burdens among different health plans or health care providers, where this process of "adverse selection" produces a concentration of high-risk and high-cost patients. Unless this additional risk is adjusted for through differential payments, the adverse selection can create significant financial inequity.

Risk adjustment is an important tool for a growing, managed care environment, but it is not a "magic bullet". Risk adjustment systems must be designed for the right situations and must be powerful enough to provide benefits proportionate to their cost. Such systems require a significant investment of resources - time, human resources, and the capital needed for effective data gathering and analysis. Ensuring a proper return on the investment, therefore, requires a thoughtful and carefully targeted approach.

This paper reviews key aspects of risk adjustment, including the need to adjust for quality of care delivered, not just for risk factors among managed care enrollees and patients.

Goals of a Risk Adjustment Arrangement

The primary goal of a risk adjustment arrangement is to encourage the providers of health insurance (referred to here as "health plans") and the providers of health care to compete strictly on the basis of quality and efficiency of care, not risk selection. A successful risk adjustment model will, therefore, counter the negative effects of competition based on risk avoidance and the possibility that health plans and providers doing a better job than their competitors of serving and managing higher-cost patients would be penalised financially by adverse selection.

Designing Risk Adjustment Systems for the Right Sector

Risk adjustment systems must be designed to suit different situations. For example, a large employer offering no choice of health plans would have little or no interest in risk adjusting its own premiums based upon changes in the risk factors of its employees. And, at least among some groups of large employers dealing with large HMOs (health maintenance organisations), it is not clear that there is always sufficient risk variation among the health plans to make a risk adjustment system worth the effort (as the Pacific Business Group on Health ^{[ 2 ]} has found). It may be that employer/purchaser discussions about risk adjusting health plan premiums may be more appropriate in relation to health plans serving the elderly and disabled (Medicare), the poor (Medicaid), and the small group market, where risk variation between plans may be more pronounced than it is in the commercial HMO arena.

Are health plans even the most fruitful sector on which to focus? The IHA membership actually believes that developing models that risk adjust payments from health plans to providers may produce more significant results than developing models adjusting premiums paid to health plans. This is because there is likely more risk variation among providers than plans, eg there is probably more risk variation between the large teaching hospital and the small community hospital than there is between the insurance plans that might cover both of their patients. In the short term, providers - especially physician groups - may need to make very significant investments in data systems to allow for both risk and quality measurement, however, and it not clear who would provide the capital for those investments.

Is Risk Adjustment Worthwhile?

A risk adjustment system must have a significant enough effect among HMOs and/or among provider organisations to balance the effort and investment required to develop the system. California’s small group purchasing programme, begun as a State government initiative and now run privately by PBGH as PacAdvantage, risk adjusts the premiums of the competing health plans that it offers. Although it moves a small amount of total premium dollars among the participating HMOs (about 1% or less), an individual plan with higher risk factors can receive a significant premium adjustment, and so the programme works. At the provider level, the Buyers Health Care Action Group in the Minneapolis area, a local employer coalition, devised a risk adjustment system for the health care systems that competed for enrollees which moved about 5% of total premium among provider organisations in 1997 based on their relative disease burden ($6 million out $116 million). This was more than enough effect to make the system worthwhile.

HMOs and PPOs

A risk adjustment model in a purchasing pool that moves money between HMO and PPO (preferred provider organisation) plan types must be carefully structured so as to measure for true differences in risk, not cost, and to be equitable to both participating purchasers and plans. If significant payment adjustments are made from HMOs to PPOs, either HMOs or employers preferring HMOs may drop out of the pool, or there may be a ripple effect on employer contribution strategies. It is also important to underscore the point that risk adjustment is not meant to adjust for efficiency differences between different health care delivery systems and that PPOs will be more expensive than HMOs for an identical group of members. This is because HMOs will more tightly manage expenses, whereas unmanaged PPO providers will have a natural tendency to do more in order to get paid more.

Designing the Risk Adjustment Model Properly

Risk adjustment models must be based on risk factors or diagnostic information, not treatment utilisation or costs, lest they reward more expensive or more frequent utilisation per se, as opposed to adjusting for potentially higher cost patients. This requires a sophisticated risk adjustment model that uses individual health assessment survey data and/or encounter data (and/or in some models pharmacy data), as opposed to simply inpatient data. Health plans and providers in mature managed care markets, such as California, that have done a good job of reducing unnecessary hospitalisation should not be penalised in a risk adjustment system that relies only on inpatient data.

Unfortunately this is exactly what the federal government did with Medicare+Choice risk adjustment ^{[ 3 ]}.

Data for Managing Risk Adjustment

As more and more health care is delivered by provider organisations under capitated payments from health plans, the availability of timely and accurate encounter data ^{[ 4 ]} becomes critical. Risk adjustment is difficult enough using fee-for-service claims data (as in the Buyers Health Care Action Group model in Minnesota). It is extremely challenging when clean encounter data with adequate diagnosis information is not readily available from provider groups. Evidence suggests that data will get better as provider organisations mature, but clearly better incentives or new requirements are needed. IHA’s new Pay for Performance initiative in California, which covers physician organisations serving commercial HMO enrollees, does reward groups that can provide good encounter data. The federal efforts to move beyond inpatient data to ambulatory data for risk adjustment in Medicare should also help in this regard.

The Importance of Transition

The initial implementation of a risk adjustment model needs a carefully planned transition from the existing system. Health plans and/or providers need to clearly understand the goals, structure and limits of the new risk adjustment model. The potential for "gaming", whether it is upcoding data or finding ways of attracting healthy enrollees and discouraging costly ones, needs to be squarely addressed. A simulation will help the players understand the potential impact, and all parties involved must understand that risk adjustment is a zero sum game in which one organisation’s financial gain is made possible by some other organisation’s equivalent financial loss.

The Need for Quality Measurement, Reporting and Adjustment

Risk adjustment should be accompanied by the measurement and reporting of service quality and the adjustment of payments on this basis. It is necessary, but not sufficient, to adjust health plan or provider payments based on higher numbers of HIV or diabetic enrollees, for example. It is equally important to adjust payments for how well these enrollees’ health care is actually managed. Are appropriate diagnostic and treatment protocols being followed? Is care being managed proactively? Is health status optimised? These direct rewards for quality would be in addition to the indirect rewards for quality - that is, employees and private individuals switching to plans or provider organisations with better quality, with higher risk adjusted payments attached to higher risk enrollees.

Moving Risk Adjustment Forward

Managed care works best in a competitive environment in which health plans and provider groups succeed or fail based on public accountability for quality as well as price. Quality should include both the proactive "health maintenance" of low-risk individuals and the excellent care management of high-risk individuals. Designing a framework in which this kind of "quality competition" can thrive means designing appropriate risk adjustment mechanisms as well as better performance measurement tools.

As health plans consolidate and grow larger, the risk variation between them may decrease, making risk adjustment more appropriate at the provider level than the plan level. Appropriate incentives (positive and negative) must be designed to ensure the accuracy and timeliness (and confidentiality) of patient encounter data, which will be needed for effective quality adjustment as well as for any risk adjustment arrangement not based solely on health assessment survey data. In the Pay for Performance initiative noted above, physician organisations that do not meet a certain level of annual encounter data submissions to their contracting health plans are simply not eligible for the bonus programme. Those that do, however, can earn positive financial incentives for certain types of information technology investment as well as for clinical scores backed up by good data. The ability to make both risk and quality adjustments will be significantly enhanced by the movement to electronic medical records and electronic data interchange, although progress in this area is extremely slow in the US health care system.

In the interim, investment in better risk adjustment tools must proceed. There are different risk adjustment methodologies under development, and a few with "real-world" testing, each with its own advantages and disadvantages. Some are better predictors of expenditures than others, and some are more costly to implement than others. All require a significant investment of time and money to design, test, refine and implement. The difficulties inherent in risk adjustment should not be used as a reason to avoid experimentation and testing, but neither is risk adjustment a universal good. It must be approached with the same care as any investment - by picking the right time and place and maximising the returns.