Gr ow th in health care costs and insurance

Transcription

1 HMO Market Penetration And Costs Of Employer-Sponsored Health Plans Higher market penetration by managed care leads to lower employer health plan costs. b y La u r e n c e C. B ake r, Jo e l C. C a n t o r, S t e ph e n H. Lo n g, an d M. S u s a n M a rq uis Using two employer surveys, we evaluate the role of increased health maintenance organization (HMO) market share in containing costs of employer-sponsored coverage. Total costs for employer health plans are about 10 percent lower in markets in which HMOs market share is above 45 percent than they are in markets with HMO enrollments of below 25 percent. Gr ow th in health care costs and insurance premiums slowed substantially during the 1990s. 1 Many believe that the growth of health maintenance organizations (HMOs), which have captured an increasing share of the employment-based insurance market, played a large part in attenuating the previous trend. 2 Although the existing literature suggests that HMO penetration can influence spending and therefore may be able to influence premiums, no studies have examined the overall effect of HMO penetration on total premiums. 3 This paper takes on that task, focusing on employer-sponsored health coverage, which covers 74 percent of the population under age sixty-five. We use data from two surveys of employers to examine the relationship between HMO market share and premiums This is the result of lower premiums for HMOs than for non-hmo plans, as well as the competitive effect of HMOs that leads to lower non-hmo premiums for employers that continue to offer these benefits. Slower growth in premiums in areas with high HMO enrollments suggests that expanded HMO market share may also lower the long-run growth in costs. 1 for employer health plans in 1993 and These data encompass two periods that saw a wide range of HMO penetration rates. We also investigate whether premiums for employer coverage are influenced through competitive effects of HMOs on premiums for non-hmo plans. Data And Methods n Data. Our data come from the 1993 and 1997 Robert Wood Johnson Foundation Employer Health Insurance Surveys. The 1993 survey interviewed 22,347 private employers in ten states: Colorado, Florida, Minnesota, New Mexico, New York, North Dakota, Oklahoma, Oregon, Vermont, and Washington. 4 The 1997 survey interviewed 21,545 private employers throughout the nation. The sample was concentrated in the sixty communities Laurence Baker is an assistant professor of health research and policy at the Stanford University School of Medicine. Joel Cantor is professor of public policy and director, Center for State Health Policy, at Rutgers University in New Jersey. Steve Long and Susan Marquis are senior economists at RAND s Washington office. H E A L T H A F F A I R S ~ S e p t e m b e r / O c t o b e r Project HOPE ThePeople-to-PeopleHealth Downloaded from HealthAffairs.org Foundation, on March Inc. 01, 2018.

2 2 followed by the Community Tracking Study and in twelve states having significant smallgroup rating reforms. 5 These cases were supplemented by a sample from the remainder of the continental United States. 6 The sampling frame for both surveys was the Dun s Market Identifiers national census of employment establishments. Within the geographic units described above, the frame was stratified by the number of workers at the establishment. Both surveys were conducted by telephone. The interview was conducted with the person(s) in each establishment most knowledgeable about the employer s health benefit plans and workers characteristics. The response rate to the 1993 survey was 71 percent; to the 1997 survey, 60 percent. Because we use the data to measure average health insurance premiums and HMO penetration in a market area, we restricted our analysis sample for 1993 to employers in one of the metropolitan statistical areas (MSAs) in the ten states covered. For 1997 we limited our sample to employers in one of the fifty-one MSAs that are part of the Community Tracking Study; the other nine Community Tracking sites are rural areas. To measure premiums in the market area, we included only employer-sponsored plans for which a premium was reported by a respondent. We distinguish between two kinds of employersponsored plans: HMOs and non-hmos. 7 We limited our sample to the MSAs for which we had reported single premium data for a minimum of twenty-five non-hmo plans. Our resultant sample for 1993 covers fifty MSAs and includes 7,624 employers and 9,850 health insurance plans (6,806 non-hmos and 3,044 HMOs). 8 For 1997 we cover forty-three MSAs and have data from 6,083 employers and their 7,832 health insurance plans (4,323 non- HMOs and 3,509 HMOs). 9 Our two study samples include ten MSAs in common. We also conducted all of our analyses for each year on this subset of the areas to investigate whether any differences in findings between years is attributable to the different mix of markets that we studied. We observe the same relationship between HMO penetration and premiums for these ten MSAs that we see for all of the study markets. Therefore, we do not separately report the results for the ten areas. n Measures. Our goal is to examine whether premiums for employer-sponsored insurance plans vary between MSAs with different rates of HMO penetration. Health insu rance prem iums (wh ich include the amounts paid by both the employer and employee) were self-reported by respondents for each plan offered, separately for single and for family coverage. We classify plans as an HMO or a non-hmo based on the respondent s selfassessment of plan type, aided by complete definitions of plan types as needed (HMO, preferred provider organization, or traditional indemnity in both surveys and point-of-service plan in the 1997 survey). We report all premiums in 1997 dollars; 1993 premiums were adjusted to 1997 dollars using the medical component of the Consumer Price Index. In addition to comparing premium levels across market areas, we also compare the change in premiums from 1996 to 1997 to investigate whether growth in premiums varies depending on HMO penetration. The change measure is based on an establishment-level question included in the 1997 survey about change in total health insurance cost per enrollee compared with one year ago. The two surveys also obtained information about the scope of benefits of each health insurance plan, including the deductible, coinsurance rate, amount of out-of-pocket maximum, benefits included (prenatal, maternity, mental health, substance abuse, prescription drugs, and dental in 1993; vision, mental health, prescription drugs, and dental in 1997), whether the plan is medically underwritten, and the waiting period for coverage of preexisting conditions. We adjust premiums for differences among plans in these characteristics as described below. The HMO penetration rate in each MSA was measured from the survey data as the percentage of employees in the MSA enrolled in their employer s health benefit plan who were enrolled in an HMO. For 1993 we compared H E A L T H A F F A I R S ~ V o l u m e 1 9, N u m b e r 5

3 the analysis results using the survey measure of HMO penetration to results using an estimate of MSA market share developed from Group Health Association of America data on enrollment and market shares of all HMOs in the counties within the MSA. 10 Because the results using the alternative measure of HMO penetration were similar to those using the survey measure, they are not reported here. We use data from two other sources to measure and control for community characteristics that might affect premium levels: the Area Resource File to measure income per capita in the MSA, and the data files of the Health Care Financing Administration s average adjusted per capita cost (AAPCC) index to measure the county per capita payment rates as a proxy measure of price differences between areas. The measure that we use is the average payment rate for Aged Part A and Aged Part B. The MSA area measure is computed as the population-weighted average of the measure for each county in the MSA. n Methods. We compare average premiums for employer-sponsored plans in MSAs with varying degrees of HMO penetration. To control for differences among MSAs in the characteristics of employers or the health insurance benefits they offer, we estimate and present premiums that are adjusted for these differences. We use multivariate regression analysis to examine the relationships between area HMO market share and premiums. We estimate separate models for single premiums and for family premiums. We also perform separate regressions for each of the two study years. The range of HMO penetration rates observed in the data changed substantially between 1993 and Had we pooled the two years of data, differences in premiums attributable to differences in HMO penetration rates would be highly confounded with any effects of time on premiums. Therefore, we chose to fit separate models for the two years. 11 The key independent variable in our models is HMO market share in the MSA. For each year we divided MSAs into four groups defined by approximate quartiles of the penetration distribution. We include indicator variables for the groups. Since HMO market shares rose substantially between 1993 and 1997, separate categorizations are used for each year. For 1993 we divide MSAs into groups with less than 10 percent, percent, percent, and over 25 percent market share. For 1997 we divide MSAs into groups with less than 25 percent, percent, percent, and over 45 percent market share. We created categorical measures for HMO shares to avoid imposing a functional form on the relationship between increasing penetration rates and premiums. Differences among MSAs in other characteristics might confound our analysis if these differences correlate with HMO penetration. Therefore, we control for plan benefits, characteristics of the establishment, and characteristics of the market. Plan benefits, enumerated above, are included in the model to control for differences in the composition of health plans across areas and changes induced by market competition. In addition, we include an indicator for whether the plan is a self-insured or purchased plan. Establishment characteristics in the regression include the size of the firm, percentage of workers who are in a union, number of years in business, industry, and amount of turnover in the workforce in a year. We also control for risk differences in the employee population using the sex and age mix of the employee population and an indicator of whether the employer was ever denied insurance. Factors that affect premium levels also may influence HMOs decisions to establish in an area. Failure to account for this could lead to biased estimates. We minimize this problem by including area characteristics that are likely to be correlated with both premium levels and HMO penetration. These include income per capita as a measure of area demand for services and AAPCC as a measure of health care prices. The regression models are fitted to data in which the insurance plan is the unit of observation. However, the data are weighted so that each MSA receives equal weight. That is, 3 H E A L T H A F F A I R S ~ S e p t e m b e r / O c t o b e r

4 4 the unit of analysis is effectively the MSA, and the coefficients on our HMO indicator variables reflect the average premium for MSAs in the group. We treat the MSA as the unit of analysis so that each metropolitan area contributes equally to our estimate of the effect of HMO penetration, rather than giving higher weight to areas with greater population. W ithin the MSA, establishments are weighted by sampling probabilities to represent the mix of establishments in the MSA. If an employer offers multiple plans, plans are weighted by their enrollment shares. We use the regression to predict the adjusted mean premium paid by employers for plans of a given type in MSAs in each of the four penetration groups for each study year. These predicted values adjust the premiums to the mean value of the plan, establishment, and area characteristics in our data. Differences between markets with different HMO market shares are tested using two-tailed t- tests of statistical significance. We first perform this analysis including premiums from all health plans, to assess the overall effect of HMOs on premiums. We then redo the analysis once using only data from HMO plans and a second time using only data from non-hmo plans. This enables us to examine whether the effects we observe on all premiums are the result of changes in HMO or non-hmo premiums. Results There was substantial growth in the share of employees in their employer-group plans who enrolled in HMOs between 1993 and In 1993 the average rate of HMO enrollment in the fifty markets we studied was 21 percent, whereas the rate was 37 percent in the fortythree markets included in our 1997 data. This reflects real growth in HMO shares over the period and is not a result of the difference in the surveyed communities. Among the ten communities in common to both surveys, HMO shares rose from 27 percent in 1993 to 42 percent in n Premium levels. Overall, premiums for employer-sponsored plans are lower in areas with high HMO enrollment. This can be seen in Exhibit 1, which reports average adjusted premiums, as well as in Exhibits 2 and 3, which plot average adjusted single and family premiums across all plans by the mean HMO penetration rate for the four groups in both survey years. In markets with HMO shares above 25 percent, total premiums for single and family coverage were significantly lower than in other markets in both years. In 1997 premiums in markets with shares above 35 percent tended to be lower than premiums in markets with shares of percent (Exhibit 1). In part, premiums are lower overall in markets with higher HMO shares because HMO premiums are often lower than non-hmo premiums, particularly in 1997 (Exhibit 1). Thus, increases in HMO market share, and workers corresponding shift from non-hmo to HMO plans, reduce overall premiums by moving people from more expensive to less expensive plans. HMO premiums also could be lower in areas with higher HMO market shares if, for example, growth in HMO market share fosters competition among HMOs. When we examine the relationship between HMO market share and HMO premiums, we find some significant differences, but there is not a consistent relationship between HMO market share and HMO premiums. n Competitive influence. In addition, our results support the hypothesis that there is a competitive influence of HMO penetration on non-hmo premiums. These premiums are lower in markets with HMO shares above 25 percent than in markets with a smaller HMO presence. In 1993 single and family premiums for non-hmo coverage were significantly lower in markets with HMO enrollment shares above 25 percent. In 1997 non-hmo premiums for single coverage were significantly lower in MSAs with HMO penetration between 25 and 45 percent than in areas with low penetration; non-hmo premiums for family coverage were significantly lower in all areas with HMO shares above 25 percent than in the low-penetration markets. There is some evidence that non-hmo premiums in- H E A L T H A F F A I R S ~ V o l u m e 1 9, N u m b e r 5

5 EXHIBIT 1 Average Adjusted Monthly Premiums, By Health Maintenance Organization (HMO) Penetration Rate, 1993 And 1997 HMO penetration, 1993 data Less than 10 percent percent percent 25 percent or more $ b d $ b 169 b $ b HMO penetration, 1997 data Less than 25 percent percent percent 45 percent or more b 167 b c b 169 b HMO penetration, 1993 data Less than 10 percent percent percent 25 percent or more HMO penetration, 1997 data Less than 25 percent percent percent 45 percent or more $ b c 400 b 394 b d $ b $ c b 401 b 415 b SOURCE: 1993 and 1997 Robert Wood Johnson Foundation Employer Health Insurance Surveys. NOTE: MSA is metropolitan statistical area. a In 1997 dollars. b Significantly different from the premium of the lowest HMO penetration group in the survey year, p =.05, two-tailed test. c Significantly different from the premium of the lowest HMO penetration group in the survey year, p =.10, two-tailed test. d Too few plans; the average number of observed plans in the MSAs is fewer than five creased in areas with the highest penetration (over 45 percent) relative to non-hmo premiums in areas with mid-range penetration (35 45 percent). However, non-hmo premiums in markets with HMO penetration above 45 percent did not differ significantly from the premiums in areas with HMO enrollments in the range of percent. n Rates of premium growth. Markets with high HMO penetration also appear to have experienced much slower recent growth in premiums than other areas have. The average reported increase in premiums between 1996 and 1997 in the MSAs with HMO shares below 25 percent was 5.3 percent, in contrast to an increase of 1.3 percent in the markets with HMO shares of 45 percent or more (Exhibit 4). The rates of premium increase differed significantly among all of the market areas except those with HMO shares of 35 to 45 percent versus those with higher market penetration. Conclusions HMOs have captured an increasing share of the employment-based insurance market. The pace of change is very fast; the penetration rates in the ten communities we compared rose by 56 percent in the brief four-year period between our two surveys. Our findings suggest that this expansion has led to lower costs for employer-sponsored health plans. Compared to markets with HMO enrollments of less than 25 percent, in 1997 premiums were about 8 10 percent lower in mar- H E A L T H A F F A I R S ~ S e p t e m b e r / O c t o b e r

6 EX H IBIT 2 Average Adjusted Monthly Employer Premiums For Single Coverage, By Health Maintenance Organization (HMO) Penetration Rate, 1993 And 1997 Premium in 1997 dollars HMO penetration rate (percent) SOURCE: 1993 and 1997 Robert Wood Johnson Foundation Employer Health Insurance Surveys. NOTE: Average HMO penetration rates for the four market-share groups in the study are 3, 15, 23, and 41 percent for 1993 and 17, 30, 40, and 56 percent for kets with HMO enrollments of 45 percent or more. We observe similar effects of increasing HMO market share in The savings occurred both because HMO premiums were lower than premiums for non-hmo plans and because the expanding role of HMOs had spillover benefits in lower premiums for non- HMO plans. Specifically, the competitive effect of HMOs leads to lower premiums for non-hmo plans in markets with HMO enrollments above about 25 percent. The savings we observe reflect reductions in average premiums across all employees. It is sometimes argued that HMOs have achieved EX H IBIT 3 Average Adjusted Monthly Employer Premiums For Family Coverage, By Health Maintenance Organization (HMO) Penetration Rate, 1993 And 1997 Premium in 1997 dollars HMO penetration rate (percent) SOURCE: 1993 and 1997 Robert Wood Johnson Foundation Employer Health Insurance Surveys. NOTE: Average HMO penetration rates for the four market-share groups in the study are 3, 15, 23, and 41 percent for 1993 and 17, 30, 40, and 56 percent for H E A L T H A F F A I R S ~ V o l u m e 1 9, N u m b e r 5

7 EXHIBIT 4 Reported Change In Employer Premiums, By Health Maintenance Organization (HMO) Penetration Rate, Less than 25 percent percent percent 45 percent or more 5.3% 3.6 a 2.0 a 1.3 a SOURCE: 1997 Robert Wood Johnson Foundation Employer Health Insurance Survey. a Significantly different from the premium of the lowest HMO penetration group in the survey year, p =.05, two-tailed test. savings only by selecting the healthiest and least expensive patients, or by obtaining price concessions from doctors and hospitals that are then simply shifted onto other insurers. If this were true, then the reductions in HMO premiums would be accompanied by increases in premiums for non-hmo plans; the net effect on overall spending would be small. Our results suggest that this is not generally the case. In fact, we find that non-hmo premiums tend to fall with rising HMO market share, precisely the opposite of what one would observe if HMOs simply selected healthy patients and shifted costs to other insurers. We had some indications of adverse selection in non-hmo plans in areas where HMOs have captured close to half the market. Non-HMO premiums were higher in areas with market shares higher than 45 percent in 1997 than in areas with percent HMO share, although this was not statistically significant. The largest difference in the premiums between HMOs and non-hmos also tends to be in areas with the highest HMO enrollment. We note, though, that while premiums paid for those staying in non-hmo plans may be higher in these areas, overall average premiums still fell as HMO market share increased. In addition, the growth in premiums in recent years appears to be lower in areas with high HMO penetration. This suggests that in addition to a one-time lowering of prices, the cost differences between areas may reflect a more lasting phenomenon as well. However, we need to observe changes in premiums over a longer time period before drawing conclusions about long-run effects. n Caveats and future research needs. Although our results on health costs are encouraging, they are limited to two crosssections covering a short time period, and they suggest several other issues for research. Our data do not provide evidence of how the savings are obtained for example, through efficiencies in the delivery system, through provider discounts, or through slowing the adoption of new technologies. We cannot measure the exact distribution of these effects between HMOs and non-hmos. The growth of HMOs may have spurred the growth of other types of managed care plans, which may be a factor in the lower premiums in high HMO areas. We have not examined how the distribution of types of non-hmo plans varies across markets with different HMO penetration levels, or how a different mix of types of non-hmo plans affects premiums within markets with the same degree of HMO penetration. We also cannot address the implications for quality of care delivered in markets with high levels of HMO activity. Finally, it is difficult to know how these results can be extrapolated into the future. HMOs may have expanded initially in areas where there potential to reduce costs was greatest. We included some controls for this, but our results still may overestimate the effect of changing HMO penetration and to some extent reflect HMOs location decisions. If so, future expansions in other areas may not achieve the same magnitude of premium savings. Moreover, other analysts have found that in addition to the HMO market share, the level of competition among HMOs also affects premiums. 13 Industry consolidation, therefore, 7 H E A L T H A F F A I R S ~ S e p t e m b e r / O c t o b e r

8 8 may affect future savings. Similarly, consolidation among hospitals or physician groups may affect the savings from increased managed care penetration. The pace of change in markets is remarkably rapid, which further complicates forecasting from these findings. There is an ongoing need for careful monitoring of these issues, especially because of the changing environment. This research was supported by Grant no and no from the Robert Wood Johnson Foundation (RWJF). Any views expressed herein are solely those of the authors, and no endorsement by the RWJF or the authors affiliations is intended or should be inferred. NOTES 1. P.B. Ginsburg and J.D. Pickreign, Tracking Health Care Costs, Health Affairs (Fall 1996): ; D.M. Cutler and L. Sheiner, Managed Care and the Growth of Medical Expenditures, in Frontiers in Health Policy Research, ed. A.M. Garber (Cambridge, Mass.: MIT Press, 1998), ; and R. Kuttner, The American Health Care System: Employer-Sponsored Health Coverage, New England Journal of Medicine 340, no. 3 (1999): G.A. Jensen et al., The New Dominance of Managed Care: Insurance Trends in the 1990s, Health Affairs ( Jan/Feb 1997): 5 136; and J.R. Gabel, P.B. Ginsburg, and K.A. Hunt, Small Employers and Their Health Benefits, : An Awkward Adolescence, Health Affairs (Sep/Oct 1997): For studies that have looked at partial effects, see L.C. Baker, The Effect of HMOs on Fee-for- Service Health Care Expenditures: Evidence from Medicare, Journal of Health Economics 16, no. 4 (1997): ; R. Feldman, B. Dowd, and G. Gifford, The Effect of HMOs on Premiums in Employment-Based Health Plans, Health Services Research 27, no. 6 (1993): ; D. Gaskin and J. Hadley, The Impact of HMO Penetration on the Rate of Hospital Cost Inflation: , Inquiry 34, no. 3 ( 1997): ; and T.M. Wickizer and P.J. Feldstein, The Impact of HMO Competition on Private Health Insurance Premiums, , Inquiry 32, no. 3 (1995): J.C. Cantor, S.H. Long, and M.S. Marquis, Private Employment-Based Health Insurance in Ten States, Health Affairs (Summer 1995): P. Kemper et al., The Design of the Community Tracking Study: A Longitudinal Study of Health System Change and Its Effects on People, Inquiry 33, no. 2 (1996): Research Triangle Institute, 1997 Employer Health InsuranceSurvey: Final Methodology Report (Research Triangle Park, N.C.: RTI, 1998). 7. The latter comprise preferred provider organizations, point-of-service plans, and traditional indemnity plans. 8. Of the original sample of 22,347 establishments, 32 percent were excluded because they did not offer insurance. Among employers offering insurance, 32 percent were not in metropolitan areas; 16 percent were excluded because premiums were not reported, and 2 percent, because of insufficient plan observations in the MSA. Offering establishments included in the analysis were somewhat smaller than those in the MSAs that were excluded; 72 percent of those included were establishments of firms with fewer than fifty employees, compared with 58 percent of the excluded establishments. Offering establishments included in our analysis also were less likely than those excluded to be retail establishments (17 percent versus 25 percent) and more likely to be financial and professional institutions (46 percent versus 42 percent). 9. Of the,829 establishments in one of the fiftyone metropolitan tracking sites, 33 percent were excluded because they did not offer insurance. Among employers offering insurance, 26 percent of cases were excluded because they did not report premiums, and 3 percent were excluded because of insufficient plan observations in the MSA. As in 1993, employers in our sample were somewhat smaller than excluded ones; 68 percent of included cases were establishments of firms with fewer than fifty employees versus 62 percent of excluded cases. Differences by industry between included and excluded cases were smaller than in See Baker, The Effect of HMOs, for a description of the methods for developing the GHAAbased measure. 11. In addition, there were small differences between the two surveys in the control variables in the regression. The 1993 survey collected data about the joint distribution of the age and sex of workers; the 1997 survey measured the marginal distributions of each. The two surveys also differed slightly in the scope of benefits measured, as noted in the text.. We have controlled for a number of factors that may themselves be affected by HMO market penetration, such as AAPCC and the scope and breadth of benefits. Thus, our tests of effects of HMO penetration are conservative tests. 13. D. Wholey, R. Feldman, and J.B. Christianson, The Effect of Market Structure on HMO Premiums, Journal of Health Economics 14, no. 1 (1995): H E A L T H A F F A I R S ~ V o l u m e 1 9, N u m b e r 5