versus those remaining with conventional fee-for-service (FFS) health plans ("stayers"). It is less commonly identified when the comparisons of "leavers" and "stayers" involve measures of health status (e.g., patient perceptions, reported chronic conditions, and functional status) (Wilensky and Rossiter, 1986; Hellinger, 1987; Robinson et al., 1991; but see Lichtenstein et al., 1991). However, when existing enrollees in HMOs and FFS plans are compared (which is preferable to studying just "leavers'' and "stayers"), the health status measures tend to be more favorable for the HMO members. This finding could also be interpreted as demonstrating the impact of better preventive and curative care, although this has not been studied explicitly (Luft and Miller, 1988).
Second, the age of a health plan or its tenure in a particular health benefit program may affect its risk profile (Neipp and Zeckhauser, 1985; Grazier et al., 1986). New health plans, particularly network-based plans, are more likely to attract individuals who are younger and less likely to have an established relationship with a physician, although specifics of their benefit design can counter this selection advantage (Sorensen et al., 1980).
Third, enrollment, continuation, and disenrollment decisions all contribute to selection dynamics. Risk pools can be dramatically affected by who joins, who stays, and who leaves during various periods (Luft et al., 1989). Those who depart one plan voluntarily for another may differ from those whose departure from a plan is involuntary.
Fourth, classifications such as "low user of care" and "high user of care" are not permanent categories (Welch, 1985). Much utilization in any given period flows from acute events for which individuals do not usually require ongoing care in subsequent years. Both high and low users of care in one year are likely to "regress" toward the mean level of use over time, with most of the effect occurring in the second year (Newhouse et al., 1989). If health plan participants remain in their original pools, the benefit from attracting an individual who has used little care is likely to diminish somewhat.
In the small-group insurance market, the response of many underwriters to this phenomenon is "durational" rating. This practice sets low initial premiums for low-risk groups and then sharply increases rates at renewal time (Hall, 1992). (The literature does not mention an equivalent use of durational rating to lower rates for high-risk groups.) In contrast, in the large-group market in which employers offer multiple health plans, the effects of regression toward the mean may be offset by continued risk selection at each year's open enrollment.
Fifth, "stay or move" decisions by very high risk individuals may be particularly important. The impact on health plan costs of members who require very high levels of care has been repeatedly documented. Between 1 and 10 percent of a group will typically account for 30 to 70 percent of its