Risk Adjustment

Risk adjustment is any mechanism for reflecting differences in health status across individuals or groups of individuals. When risk adjustment is used in studies of costs or other outcomes of hospitalizations, episodes, or specific procedures, it is often called case mix adjustment. The diagnosis-related groups (DRGs) classification system is used to make risk-adjusted payments for hospital admissions. DRGs are also used to identify groups of hospitalizations that are sufficiently similar that it makes sense to try to identify systematic factors associated with variations in their content, costs, or health care outcomes. A fairly comprehensive discussion of risk adjustment for measuring cost and other health care outcomes is given by Iezzoni (1997). Risk adjustment is also used to describe a process for measuring or predicting health care expenditures (or other outcomes) of individuals or groups of individuals during a specified period of time, such as a year. A distinction is sometimes made between the process of measuring differences in medical risk (risk assessment) and the methods used to modify payments to providers, based on these assessed differences (risk adjustment). Risk adjustment is being used by public insurers, employers, health plans, hospital and provider networks, and even by individual providers. Glazer and McGuire (2001) define “formal risk adjustment” as the use of formulas that incorporate demographic, diagnostic, or other health information to calculate individual-level health premiums. They document that formal risk adjustment is widely used by public payers such as Medicare, Medicaid, and state governments, but used less often by private employers, who may have other strategies for offsetting risk selection incentives. Risk adjustment is also increasingly used within health plans to evaluate provider efficiency, and to identify, manage, and evaluate the management of individual high-cost cases.

Many kinds of information may be used for risk adjustment. The most widely used risk adjusters are age and gender, which explain 10-fold differences in expected health care costs across individuals. Employment status and whether or not a person is in a nursing home are examples of other demographic variables that have been used. The use of such demographic information alone leaves considerable opportunity for risk selection by health plans, because people with similar demographics may differ widely in their health care needs. In recent years health plans and payers have explored other types of information for risk adjustment, including survey-based measures, particularly self-reported health status data, and prior utilization measures. Some groups are beginning to use pharmacy information for risk adjustment; however, concerns about incentives (that derive from paying more for patients when they receive additional drugs) may limit the use of such data for adjusting payments. A clear preference has emerged for using diagnostic information for risk adjustment, with many employers and public programs currently basing payment to health plans on diagnostic information, age, and sex.

[Page 491]The U.S. Medicare program is a leading proponent of using risk adjustment to correct problems of risk selection. Prior to January 2000, capitation payments to health plans with Medicare risk contracts were risk adjusted using only demographics, and risk selection (that was not offset by risk-adjusted payments) was convincingly shown to be a serious problem (Brown et al., 1993). Since 2000, Medicare has been calculating part of its risk-adjusted premium payments to plans using inpatient diagnoses based on a payment formula called the Principal Inpatient Diagnostic Cost Group (PIP-DCG) model. The program has announced that from 2004 onward, it is expanding its risk adjustment formula to recognize multiple diagnoses from both inpatient and outpatient encounters.

...