In the face of the growing problem of uninsurance, U.S. policymakers

DataWatch Patterns Of Individual Health Insurance Coverage, 1996 2000 Understanding the dynamics of this volatile market will improve the chances that future reform efforts will succeed. by Erika C. Ziller, Andrew F. Coburn, Timothy D. McBride, and Courtney Andrews ABSTRACT: Information about patterns of individual health insurance coverage is limited. Knowledge gaps include the extent to which individual insurance provides transitional versus long-term coverage, and participants insurance status before and after being covered by an individual plan. In this study we use data from the 1996 2000 Survey of Income and Program Participation (SIPP) to examine how long the individually insured maintain their coverage; sources of coverage before and after enrolling in an individual health plan; and characteristics of those who rely on individual insurance coverage. Understanding the dynamics of this market will better inform federal and state insurance reform efforts. In the face of the growing problem of uninsurance, U.S. policymakers have proposed using the individual insurance market to increase coverage. 1 However, policy discussions are hampered by limitations in empirical knowledge about the individual market and its participants. 2 A limitation of prior research has been that the most readily available data sources, such as the Current Population Survey (CPS), are cross-sectional. This means that previous studies have been unable to explore patterns of individual insurance coverage over time. Little is known about the extent to which individual insurance provides transitional or longer-term coverage. We also do not know what types of insurance coverage people hold before purchasing or after leaving an individual plan. Experts observe that before making policy prescriptions, we need a clearer understanding of how people move between individual and other sources of coverage. 3 Additionally, analysts believe that volatile enrollment and disenrollment patterns have undermined the effects of state and federal individual insurance reforms. 4 These patterns have been inferred from researchers observations that many people hold both employer and individual insurance over the course of a year. 5 This study uses data from the 1996 2000 Survey of Income and Program Partic- Erika Ziller (eziller@usm.maine.edu) is a research associate at the Institute for Health Policy, Muskie School of Public Service, University of Southern Maine in Portland. Andrew Coburn is a professor and director of that institute. Timothy McBride is a professor of health policy and management, School of Public Health, St. Louis University, where Courtney Andrews is a senior research assistant. 210 November/December 2004 DOI 10.1377/hlthaff.23.6.210 2004 Project HOPE The People-to-People Health Foundation, Inc.

Individual Insurance ipation (SIPP) to describe individual insurance coverage patterns. We examine how long the individually insured maintain their coverage, sources of coverage before and after individual plan enrollment, and the characteristics of the individually insured. Understanding the dynamics of this market will better inform future federal and state insurance reform efforts. Data Sources And Methods Data. The 1996 2000 SIPP is a nationally representative, longitudinal survey conducted by the U.S. Census Bureau. 6 Through in-person and telephone interviews, SIPP collects detailed sociodemographic data and information on monthly changes in household and individual income, health insurance status, labor-force status, and participation in government-sponsored programs. The first interview of the 1996 2000 SIPP panel occurred in April 1996. Households were interviewed every four months though March 2000, for thirteen total interviews over forty-eight months. At each interview SIPP collected information for the entire sampled household for the four preceding months. The initial sample for the 1996 2000 panel comprised 40,188 households (about 95,000 people), including 55,556 adults who were ages 18 64 at the first interview. We identified respondents as having individual health insurance if they reported having health insurance that was privately purchased. If they indicated that their coverage was through a spouse, we attached spouses insurance data. Although our definition may result in some misclassification (such as coverage through an association, state-sponsored public plan, or discount medical plan), it is the best definition available through SIPP. Other studies using federal data sets have encountered similar challenges defining cases, which means that that this and other studies may overestimate individual insurance enrollment. 7 Analyses. For our descriptive analyses, we used chi-square tests of independence to determine what characteristics are associated with having individual insurance. We then counted respondents discrete individual insurance spells over the survey period and identified their source of coverage before and after being individually insured. When people exited an individual-coverage spell, we identified what characteristics were associated with whether they had employer-based coverage or public coverage (including Medicaid, Medicare, and TRICARE) or were uninsured. For those with an individual-coverage spell, we used survival analysis techniques to estimate the duration of spells beginning during the 1996 2000 panel. Statistical techniques for producing descriptive estimates of spell durations are well developed if the beginning of each spell is observed. 8 We used nonparametric techniques (the Kaplan-Meier estimator) to compute estimates of spell durations, including median spell lengths. Finally, we conducted multivariate survival analyses to isolate the independent relationships between respondents characteristics and duration of individualcoverage spells. 9 We included covariates related to health insurance coverage, in- HEALTH AFFAIRS ~ Volume 23, Number 6 211

DataWatch cluding sociodemographic and economic characteristics, employment status and job characteristics, marital status, and health status. Because individual insurance regulation varies by state, we included several regulatory measures as control variables. 10 Data on state insurance regulation come from the Assessing the New Federalism State Database Archive, produced by the Urban Institute. 11 Finally, to control for the potential impacts of the Health Insurance Portability and Accountability Act (HIPAA) of 1996, we included time as a dichotomous covariate indicating whether the spell began before or after January 1998. Study Results Characteristics of the individually insured. In the first month of the survey, 5.9 percent of adults ages 18 64 reported having privately purchased health insurance. 12 Nearly three-fourths of the individually insured were employed, although people outside the labor force had a higher rate of individual coverage (Exhibit 1). 13 Many of the characteristics associated with lower employer-based coverage are related to higher rates of individual coverage. For example, employees working part time were almost twice as likely to have individual insurance as those working forty hours a week or more. Nearly half of individually insured workers were selfemployed, while nearly three-fourths worked for a business with fewer than twenty-five employees (including the self-employed). Nearly half of the individually insured were ages 45 64, and people over age 55 were nearly three times as likely as those under age 35 to have individual coverage. Almost two-thirds were married, although widowed people had higher individual coverage rates. Very few of the individually insured had less than a high school education. While the majority of those covered were white non-hispanic, Asian Americans had the highest rate of individual insurance participation (8 percent). Rates of individual coverage were slightly higher for those with incomes at EXHIBIT 1 Characteristics Of The Individually Insured, 1996 2000 Characteristic Percent with individual insurance (N = 55,556) Percent of individually insured (n = 3,296) Total 5.9 100.0 Employment status*** Employed Unemployed Out of labor force Hours worked a **** 1 39 40+ Self-employment b **** Self-employed Not self-employed 5.6 5.1 7.7 8.1 4.7 22.3 3.2 73.7 3.6 22.7 37.8 62.2 49.0 51.0 212 November/December 2004

Individual Insurance EXHIBIT 1 Characteristics Of The Individually Insured, 1996 2000 (cont.) Characteristic Firm size b **** Fewer than 25 employees 25 99 employees More than 100 employees Income as percent of poverty level*** 0 99% 100 199% 200 299% 300 399% 400% or more Educational attainment**** Elementary High school Some college or higher Age (years)**** 18 24 25 34 35 44 45 54 55 64 Race c **** White, non-hispanic Black, non-hispanic Hispanic/Native American Asian Marital status**** Married Widowed Divorced or separated Never married Health status d Excellent/very good Good Fair/poor Percent with individual insurance (N = 55,556) 13.8 3.9 1.8 6.1 6.8 5.6 5.2 5.7 3.9 5.6 6.7 4.3 4.4 5.5 6.0 11.3 6.9 2.9 2.4 8.4 6.0 10.5 5.1 5.9 6.0 5.9 5.8 Percent of individually insured (n = 3,296) 72.0 8.4 19.6 15.3 22.1 17.3 13.4 32.0 9.1 31.6 59.3 9.1 19.0 25.6 21.0 25.2 84.1 6.1 4.8 5.0 61.0 4.2 12.3 22.5 64.2 24.1 11.8 SOURCE: 1996-2000 Survey of Income and Program Participation (SIPP). NOTE: Frequencies are based on data from the first month of the survey. a Excludes the self-employed. b Among those who indicate that they are currently employed. c Because of sample size, Hispanic and Native American (including Eskimo) respondents are collapsed into one category. d Health status data are self-reported and collected in month 13. ***p.01 ****p.001 100 200 percent of the federal poverty level. Thus, 37 percent of the individually insured fell below 200 percent of poverty. Despite expectations that adverse selection affects the individual insurance market, the likelihood of having such coverage does not vary by health status. Nearly two-thirds of the individually insured HEALTH AFFAIRS ~ Volume 23, Number 6 213

DataWatch were in very good or excellent health; only 12 percent were in fair or poor health. Spells of individual coverage. Thirteen percent of nonelderly adults had at least one spell of individual insurance coverage during 1996 2000. Most of these had only one spell of individual coverage (58.4 percent). One-third of the individually insured (31.8 percent) had two spells, and 9.8 percent had three or more spells during the four years of the study. Thus, while some of the individually insured cycled into and out of individual coverage repeatedly, the majority did not. Entering and exiting individual insurance. Exhibit 2 shows respondents coverage status immediately before and after their spells of individual insurance. Roughly two-thirds of spells began or ended with employer-sponsored coverage. Eighty-five percent of those with such coverage before a spell of individual coverage returned to an employer-sponsored plan. Thus, more than half of all individualcoverage spells (58 percent) bridged periods of employer-based insurance. One-tenth of all individual-coverage spells bridged periods of public insurance, including Medicaid, Medicare, and TRICARE. More than three-fourths (77 percent) who entered their individual-coverage spell from a public plan returned to some type of public coverage in most cases, Medicaid. Because of the definition of individual insurance cases, some of these individually insured people may have moved from Medicaid to a State Children s Health Insurance Program (SCHIP) plan covering adults or to another state-sponsored program, and back to Medicaid. Roughly one-sixth of the individually insured began or ended an individualcoverage spell uninsured. It is encouraging that 49 percent of those who started their individual-coverage spell after being uninsured ultimately gained employerbased coverage. However, 15 percent of all individually insured people ended up EXHIBIT 2 Reason Why Individual-Coverage Spells Ended And Began, 1996 2000 Reason spell ended Reason spell began Total Gained employer coverage Gained public coverage Uninsured Total Lost employer coverage Lost public coverage b Uninsured 100.0 a 68.2 14.1 17.8 69.8 a 58.1 3.1 8.6 15.2 a 3.3 10.2 1.6 15.0 a 6.8 0.7 7.5 Percent by reason for spell beginning Lost employer insurance b Lost public coverage Uninsured 100.0 100.0 100.0 85.2 22.3 48.6 4.9 72.6 9.1 10.0 5.1 42.3 SOURCE: 1996 2000 Survey of Income and Program Participation (SIPP). NOTES: Totals may not equal 100 percent because of rounding. Among new individual-coverage spells with observed beginnings and endings (n = 4,011). a Percentage of individually insured. b Includes TRICARE (formerly CHAMPUS). 214 November/December 2004

Individual Insurance without coverage when they left their individual plan. Many characteristics associated with being uninsured after an individualcoverage spell are comparable to those of the uninsured, in general (Exhibit 3). People were more likely to be uninsured if they were unemployed, worked less than full time or for a small business, or had lower family incomes. Racial and ethnic minorities were about twice as likely to become uninsured. Only 9 percent of the self-employed became uninsured after their individualcoverage spell, compared with 16.5 percent of those working for someone else; this suggests that the self-employed were less willing to drop individual insurance without an alternative source. This coverage may come from a spouse, since nearly 80 percent of married people left individual insurance for an employer-sponsored plan, compared with only 56 percent of those who never married. Younger and healthier adults were more likely to end up uninsured after leaving an individual plan than their older and sicker counterparts (Exhibit 3). Nearly 40 percent of people ages 18 24 became uninsured, compared with 5.6 percent of those ages 55 65. Seventeen percent of those in excellent or very good health EXHIBIT 3 Health Insurance Coverage After Individual-Coverage Spell, By Characteristic, 1996 2000 Type of coverage (n = 4,011) Characteristic Employer (%) Public (%) Uninsured (%) Total 69.8 15.2 15.0 Employment status*** Employed Unemployed Out of labor force 76.7 51.0 57.6 7.9 20.6 28.8 15.4 28.4 13.6 Hours worked a **** 1 39 40+ 70.5 80.7 10.9 6.0 18.6 13.3 Self-employment b **** Self-employed Not self-employed 84.9 65.9 5.7 17.6 9.4 16.5 Firm size b **** Fewer than 25 employees 25 99 employees More than 100 employees 71.6 79.2 78.9 10.9 7.6 6.3 17.5 13.2 14.8 Income as percent of poverty level*** 0 99% 100 199% 200 299% 300 399% 400% or more 56.7 61.6 70.0 75.6 77.4 20.3 20.6 16.3 11.1 10.9 23.0 17.8 13.7 13.4 11.7 HEALTH AFFAIRS ~ Volume 23, Number 6 215

DataWatch EXHIBIT 3 Health Insurance Coverage After Individual-Coverage Spell, By Characteristic, 1996 2000 (cont.) Type of coverage (n = 4,011) Characteristic Employer (%) Public (%) Uninsured (%) Educational attainment**** Elementary High school Some college or higher 49.4 68.3 75.4 30.8 16.7 10.7 19.8 15.0 13.9 Age (years)**** 18 24 25 34 35 44 45 54 55 64 54.2 74.7 76.7 75.5 69.3 7.9 6.6 10.7 13.8 25.1 37.8 18.8 12.6 10.7 5.6 Race c **** White, non-hispanic Black, non-hispanic Hispanic/Native American Asian 73.4 52.5 55.9 67.5 14.3 20.8 18.4 11.8 12.3 26.8 25.8 20.7 Marital status**** Married Widowed Divorced or separated Never married Health status d Excellent/very good Good Fair/poor 78.2 57.5 62.5 56.0 14.4 35.2 18.4 11.4 7.4 7.3 19.2 32.6 75.2 66.8 52.6 8.2 19.9 37.9 16.6 13.3 9.4 SOURCE: 1996 2000 Survey of Income and Program Participation (SIPP). NOTE: Frequencies are based on data from the first month of the survey. a Excludes the self-employed. b Among those who indicate that they are currently employed. c Because of sample size, Hispanic and Native American (including Eskimo) respondents are collapsed into one category. d Health status data are self-reported. **p.05 ***p.01 ****p.001 ended up uninsured, compared with 9 percent of those in fair or poor health. Older and sicker adults were much more likely to gain public coverage: one-fourth of people ages 55 65 and 38 percent of those in fair or poor health. Length of individual insurance spells. The median length of new individualcoverage spells was eight months (Exhibit 4). Nearly half of all spells lasted less than six months, and two-thirds lasted one year or less. Although this suggests that the majority of individual-coverage spells were short, a sizable minority of those covered had long spells. Roughly one-sixth (17 percent) of those with individual insurance coverage retained it for more than two years. 14 Exhibit 5 shows a multivariate duration analysis of the predictors of individual- 216 November/December 2004

Individual Insurance EXHIBIT 4 Distribution Of New Individual-Coverage Spell Lengths, 1996 2000 Distribution Spell length Less than 6 months 6 12 months 13 18 months 19 24 months More than 24 months Total Median spell length (months) Total number of spells Spells with observed ending Number of right-censored spells Percent of spells right-censored 48.2% 16.3 13.7 4.8 17.0 100.0 8.0 6,134 4,046 2,088 34.0% SOURCE: 1996 2000 Survey of Income and Program Participation (SIPP). coverage spell length. We found that being out of the labor force or self-employed or working for a small business each increased the length of individual-insurance spells. For example, the self-employed had spell lengths that were 20 percent longer than those who worked for someone else. People who were single, divorced, or separated remained individually insured 7 percent longer than married people. Non-Hispanic whites had individual-coverage spells that lasted 27 percent longer, and Asian Americans spells that lasted 20 percent longer, than those of other racial or ethnic groups. The spells of people in fair or poor health were 9 percent shorter than those of people in very good or excellent health. Those with prior public coverage had shorter spells, and those entering from employer coverage had longer spells, compared with those entering uninsured. Income had no independent effect on the length of individual-coverage spells, although we tested different specifications in our model, including categorical and nonlinear relationships. One explanation is that income may be positively associated with access to employer-based coverage and negatively associated with demand for individual coverage. If this is correct, then the effects of gaining employer coverage for those with higher incomes, and difficulty affording premiums by those with lower incomes, may cancel each other out. As noted above, we included measures of state insurance regulation as control variables: presence of a high-risk pool, community rating in the small-group market, guaranteed issue for small groups, and guaranteed renewal. We included time as a covariate to control for the implementation of HIPAA. Only time and smallgroup community rating were significant variables in the model, each predicting 9 percent shorter spell lengths. However, it is important not to overstate any conclusions about the impact of regulation on individual spell lengths, because these variables were included for control, not explanatory, purposes. 15 HEALTH AFFAIRS ~ Volume 23, Number 6 217

DataWatch EXHIBIT 5 Multivariate Survival Analysis Of New Individual-Coverage Spell Lengths, 1996 2000 Characteristic (referent) Coefficient Standard error Survival time ratio a Employment status (employed) Unemployed Out of labor force** Employment characteristics Works full time Self-employed****.1104.0812.0379.1847.078.038 1.08.033.039 1.20 Firm size (100+ employees) 25 99 employees Fewer than 25 employees.14016.0081.038.056 1.15 Age Age in years**** Age squared****.0573.0007.006.000 Race (black, Hispanic, Native American) White, non-hispanic**** Asian***.2405.1860.034.068 1.27 1.20 Marital status (married) Widowed Single, divorced, or separated**.0060.0660.061.030 1.07 Health status (excellent/very good) Good Fair/poor** Other characteristics Income-poverty ratio Less than high school diploma Female***.0434.0966.0028.0657.0808.028.044 0.91.003.039.027 1.08 Prior insurance coverage (uninsured) Public**** Employer-based coverage****.2115.1889.060.032 0.81 1.20 State insurance regulation High-risk pool Community rating small group*** Guaranteed issue small group Guaranteed renewal individual Spell began post-hipaa (1998)****.0355.0974.0409.0271.0922.025.030.031.030.027 0.91 0.91 Log likelihood Total N Percent censored 6859.72 6,134 34 SOURCE: 1996 2000 Survey of Income and Program Participation (SIPP). NOTE: HIPAA is Health Insurance Portability and Accountability Act of 1996. a Survival time ratios can be interpreted like logit odds ratios. Ratios below 1.0 indicate proportionally shorter spells for that characteristic, and ratios above 1.0 indicate proportionally longer spells. **p.05 ***p.01 ****p.001 218 November/December 2004

Individual Insurance Discussion Our findings suggest that participants in the individual insurance market are heterogeneous. Most individual-coverage spells begin when people enter from and exit to employer coverage. This implies that a primary function of individual insurance is to bridge gaps in employer coverage. However, an important minority of the individually insured maintain coverage for more than two years, with smallbusiness employees and the self-employed having the longest spells. As noted above, analysts have speculated that individual-coverage enrollment and disenrollment are volatile. Our findings support this hypothesis (median spell length: eight months); this lends support to insurance-industry claims that marketing and administrative costs are higher than for group coverage. Given that most of this volatility involves moving from and to employer coverage, it is unclear what policymakers can or should do to stabilize individual coverage for this group. Health status does not affect the likelihood that someone will be individually insured at a point in time, versus being uninsured or having employer-sponsored or public coverage. Given that those in poorer health are likely to have a higher demand for insurance, this suggests that insurance companies techniques for avoiding adverse selection might be effective at limiting enrollment among sicker adults. Also, many of the sickest people may have public coverage. Health status relates to duration of individual coverage, with those in fair or poor health having the shortest spells. About 40 percent left individual coverage for public insurance, perhaps because of age, disability, or financial status. Reflecting a potential failing of the individual market, nearly one-sixth of those exiting an individual insurance plan do so without another source of health insurance to take its place. We found that healthier and younger people were much more likely than their sicker or older counterparts to end up uninsured, which supports prior theory and research on adverse selection spirals in the individual market. Although younger people generally face lower premiums, costs appear to be high enough (and self-perceived health risks low enough) to make them unable or unwilling to continue buying individual coverage. Although it is unclear what effect tax credits will have on covering the uninsured generally, if sufficiently generous, they might help keep younger and healthier participants in the market. People with family incomes of 100 200 percent of poverty have a slightly elevated rate of individual coverage (37 percent of the individually insured in our sample had incomes below 200 percent of poverty). Given the financial constraints of this income group and the high cost of many individual plans, many may be purchasing plans with high deductibles or limited benefits or both (although identifying benefit structure was beyond the scope of this study). Future research on individual coverage should investigate access to health care services for those in lower income brackets. HIPAA is the principal federal reform targeting the individual insurance market HEALTH AFFAIRS ~ Volume 23, Number 6 219

DataWatch People who obtain individual coverage after being uninsured or covered by a public program would not be protected by HIPAA. over the past decade. However, given the nature of the individually insured and their patterns of coverage, it is unclear what impact HIPAA or similar state-level reforms have had on individual coverage. To be eligible for guaranteed issue of individual coverage or other HIPAA provisions, a person must have had eighteen months of continuous health insurance coverage (with at least the last day being in an employer-based plan) and have exhausted available Consolidated Omnibus Budget Reconciliation Act (COBRA) or state continuation coverage. Although the bulk of those who exit an employer-based plan may be covered by HIPAA reforms, the nearly one-third who obtain individual coverage after being uninsured or covered by a public program would not be protected by HIPAA provisions. The patterns of individual insurance coverage are complex and vary for different subgroups holding individual plans. For most, individual health insurance bridges periods of employer-based coverage, meaning that targeting policy interventions to this group (as HIPAA does) may be appropriate. However, sizable minority subgroups in the individual market are overlooked by this type of reform. Thus, policymakers should be mindful of the variability of individual insurance spells as they consider reforming or expanding access to individual health insurance coverage. The authors acknowledge the Robert Wood Johnson Foundation s Changes in Health Care Financing and Organization (HCFO) Initiative for funding this study. The conclusions and opinions expressed in the paper are the authors, and no endorsement by the University of Southern Maine, St. Louis University, or the funding source is intended or should be inferred. 220 November/December 2004

Individual Insurance NOTES 1. R. Cunningham, Joint Custody: Bipartisan Interest Expands Scope of Tax-Credit Proposals, Health Affairs, 18 September 2002, content.healthaffairs.org/cgi/content/abstract/hlthaff.w2.290 (2 August 2004). 2. M.V. Pauly and L.M. Nichols, The Nongroup Health Insurance Market: Short on Facts, Long on Opinions and Policy Disputes, Health Affairs, 23 October 2002, content.healthaffairs.org/cgi/content/abstract/hlthaff.w2.325 (2 August 2004). 3. D.L. Rogal and A.K. Gauthier, Introduction: The Evolution of the Individual Insurance Market, Journal of Health Politics, Policy and Law 25, no. 1 (2000): 3 8. 4. D.J. Chollet, Consumers, Insurers, and Market Behavior, Journal of Health Politics, Policy and Law 25, no. 1 (2000): 27 44. 5. Ibid. 6. For complete information on SIPP, see www.census.sipp.gov (14 June 2004). 7. Chollet, Consumers, Insurers, and Market Behavior ; and U.S. Government Accountability Office, Private Health Insurance: Millions Relying on Individual Market Face Cost and Coverage Trade-offs (Washington: GAO, 26 November 1996). 8. P.D. Allison, Survival Analysis using the SAS System: A Practical Guide (Cary, N.C.: SAS Institute, December 1995). 9. We estimated a parametric survival model using a Weibull distribution and PROC LIFEREG in SAS (see ibid.). 10. Chollet, Consumers, Insurers, and Market Behavior. 11. For more information on the Assessing the New Federalism (ANF) project and the state database, see Urban Institute, Assessing the New Federalism, 2001, www.urban.org/content/research/newfederalism/ AboutANF/AboutANF.htm (14 June 2004). 12. This estimate is close to those produced for other studies. For example, Deborah Chollet and Adele Kirk found that 7.7 percent of the nonelderly had individual insurance in 1996, based on CPS data. D. Chollet and A. Kirk, Understanding Individual Health Insurance Markets: Structure, Practices, and Products in Ten States (Washington: AcademyHealth, 1998). The difference may be attributable to the fact that CPS respondents are asked to identify coverage over the entire previous year, whereas SIPP collects data for each of the prior four months. 13. Many findings in this section confirm prior research on individual insurance. For example, Chollet found that rates of individual coverage are higher among adults ages 55 64, the self-employed, part-time and part-year workers, and nonworkers. Chollet also found little variation in individual coverage by income, although her analysis found slightly lower rates among those living below poverty. Chollet, Consumers, Insurers, and Market Behavior. 14. Because SIPP asks questions every four months, data often cluster at the four-month intervals (the seam effect ). This may explain why the median spell length is eight months, although 48 percent of spells last less than six months. 15. Prior research on the impact of regulation on insurance enrollment has proved complicated and yielded contradictory results. To draw strong conclusions from our analyses would have required more sophisticated diagnosis of multicollinearity in our individual and small-group reform variables, as states inclined to regulate these markets often pass multiple reform measures. HEALTH AFFAIRS ~ Volume 23, Number 6 221