|
MARKETWATCHThe Impact Of Blue Cross Conversions On Health Spending And The Uninsured
Using statewide data on health spending and uninsurance rates, we investigate the impact of Blue Cross conversions on health care costs and coverage. We find mixed results, with some conversion states improving their performance on either or both measures relative to the national average and others experiencing a decline. A multivariate analysis suggests that overall, the impact of Blue Cross conversion may be to reduce hospital and total spending, but whether this effect endures depends in part on how "conversion" is defined. State policymakers and regulators might find these results useful in considering future Blue Cross conversions.
The recent denial of Blue Cross conversion bids by regulators in Kansas (February 2002, sustained by a state Supreme Court ruling in August 2003), Maryland (May 2003), and Washington (July 2004), coupled with the voluntary withdrawal of conversion proposals by Blue Cross plans in North Carolina (July 2003) and New Jersey (August 2003), has considerably slowed the momentum toward the conversion by Blue Cross/Blue Shield plans to for-profit status.1 Nevertheless, it is considered likely that conversions and consolidations among Blues plans will continue.2 For example, early in 2004, New Jersey state officials quietly reopened discussions with the Blues plan about the possibility of conversion, and the recently approved merger of Anthem and WellPoint Health Networks will culminate in the largest U.S. health insurer.3 Such conversions have the potential to remake the corporate landscape of health care finance.4 Between 1993 and 2003, Blue Cross plans in fifteen states became for-profit, and most of these merged into two large holding companies—WellPoint and Anthem—which themselves are merging. As a result, by 31 March 2003 more than one-quarter of Blue Cross subscribers nationwide belonged to a for-profit plan. The sixty million Blues subscribers remaining in nonprofit Blues plans represent 31 percent of the commercially insured market.5 Blue Cross conversion proposals have been subjected to increasing regulatory scrutiny. Although by no means the only factor of importance, a critical consideration for regulators is whether conversion might have any adverse effects on either affordability or availability of coverage or services. Indeed, regulators’ concerns about premium increases were at least part of the motivation for the failure of conversion efforts in Kansas, Maryland, North Carolina, and Washington.6 But in all of these conversions, regulators looked at a wide array of factors when assessing whether conversion was in the public interest, including potential effects on premiums, access to coverage, and access to care. Premiums. On theoretical grounds, it is conceivable that conversions could result in lower health insurance premiums if (1) equity is a more efficient source of capital than debt or retained earnings; (2) consumers have less trust in for-profit plans and need an economic inducement to buy from them; or (3) Blues plans have unexploited market power over health care providers and face effective competition against other insurers in setting their premiums.7 Since Blue Cross controls at least half the individual market in thirty-three states and more than one-third of the group market in twenty-nine states, there also are theoretical grounds for believing that conversion could result in higher premiums in states where Blue Cross has potential market power that has not yet been fully exploited by the nonprofit plan.8 The general concern of most regulators is that stronger profit motives might result in higher premiums in health insurance product markets (such as individual or small-group markets). Coverage and access. A related concern of regulators is that higher premiums, tighter underwriting, or abandonment of historical commitments (for example, a role as insurer of last resort or a commitment to provide affordable coverage in the individual insurance market) might result in higher numbers of uninsured people. Alternatively, if Blue Cross plans have unexploited purchaser power in the market for medical care services, conversion might prompt more-aggressive bargaining with providers over payment rates, resulting in lower health care costs (but not necessarily less costly insurance coverage) and lower hospital profits that could jeopardize either the long-term viability of selected hospitals or their ability to cross-subsidize public goods such as uncompensated care, teaching, or research. Thus, a variety of theoretical reasons exist for being concerned over the potential impact of conversions; there also are countervailing theoretical reasons why these may be of no practical import. Ultimately, only empirical evidence can resolve these conflicting hypotheses. Admittedly, many other factors are at stake in a typical conversion; we have summarized the evidence regarding many of these factors elsewhere, using qualitative methods.9 The analysis that follows is intended to help inform policymakers about the potential impacts that can be measured by examining the aggregate effects of conversion on health spending, uninsurance rates, and hospital profitability. We do not pretend that these are the sole bases on which to judge the merits of a conversion, but they are important factors, and our analysis therefore provides at least some insights for regulators who are faced with judging the merits of such conversions.
Data sources and type. Information about the date and nature of Blue Cross conversions was obtained from multiple sources, including the Centers for Medicare and Medicaid Services (CMS), the Bureau of the Census, and the American Hospital Association (AHA) Annual Survey of Hospitals. Health spending. Regrettably, there are no readily available state-level time series of data on health insurance premiums; such sources would have allowed us to focus on the factor of greatest interest to regulators. However, personal health spending, by type (total, hospital, and physician), along with resident population, Medicare spending, and Medicare enrollees, are reported for all states by the CMS.10 We well recognize that health spending is only one factor affecting premium levels. In particular, reductions in the rate of increase in health spending may not necessarily translate into lower premiums if a plan commands sufficient market power to capture these savings in the form of higher profits. Nevertheless, health spending is worth examining for two fundamental reasons. First, it is at least theoretically plausible to imagine even for-profit plans passing along savings in the form of lower premiums to increase market share. Second, cutting in the other direction, lower health care spending could be a sign of possible reductions in quality or access. Hospital profitability. Hospital profitability—the ratio of total revenue to total expense—was calculated from the AHA data. This measure holds potential interest for health policy because it reflects possible constraints on hospitals’ ability to cross-subsidize essential but unprofitable services. The uninsured. Finally, state-level estimates of the percentage of people without health insurance coverage are reported by the Census Bureau based on estimates from the March Current Population Survey (CPS).11 Other data categories. In our multivariate analysis, we supplement these data with several other data sources to account for regulation of hospital services, hospital reimbursement, competition (health maintenance organization, or HMO, share), insurance coverage, and other area characteristics reflecting supply of health services and demand factors.12 Methods. We used two alternative definitions of conversion. The first dates conversion from the year in which a Blue Cross plan became a stock company either on its own or through acquisition. The second dates conversion from the year in which a plan either became a stock company, created a major for-profit subsidiary, or was acquired by a multistate mutual that subsequently converted. (Although a mutual company is technically not a for-profit investor-owned company, it is not subject to the same operational constraints faced by conventional nonprofits.)13 This change in definition predominantly affects Anthem plans, many of which were acquired during the 1990s (hence converted according to definition 2) but did not become stock companies until Anthem did so in 2001 (that is, a definition 1 conversion). Key variables. Our key variables of interest relate to per capita health spending, hospital profitability, and the percentage of the population without health insurance coverage. Statewide per capita spending (by place of service) was calculated by the authors and indexed to equivalently calculated figures for the United States (U.S. average = 100). Similar index figures were calculated from reported percentages without coverage. Technically, the March CPS asks about coverage during the preceding year, so the Census Bureau reports estimates of coverage (including uninsured rates) from the March 2003 CPS as being year 2002 estimates. However, ample evidence suggests that respondents answer these questions as if they are reporting their coverage on the day of the survey; we therefore treated March CPS figures as point-in-time estimates for the year the survey was conducted.14 Multivariate approach. We explored whether variations in health spending and the uninsured were related to Blue Cross conversions using a multivariate approach. Our multivariate models for health care spending and hospital profitability controlled for factors expected to affect health spending, including (1) certificate-of-need (CON) regulation (Section 1122, whether the state lifted acute care CON regulations); (2) hospital rate setting (Medicare prospective payment system, mandatory rate setting); (3) hospital reimbursement (Medicare and Medicaid payments as a share of hospital spending); (4) competition (HMO share); (5) insurance coverage (Medicaid, Medicare, uninsured); and (6) area characteristics. In all models, we also included state-and time-fixed effects—that is, binary variables representing each state and each year. As best as possible, these models have controlled for "endogeneity"—that is, the possibility that states whose Blue Cross plans converted were somehow different from those that did not, in ways that also affect our measured outcomes. In essence, each state serves as its own control, and the net impacts we measured can be interpreted as what happened to spending in those states relative to what might have happened absent conversion. State "fixed effects" do not completely eliminate endogeneity issues, but they provide more credible results than if we had compared states over time without taking the unmeasured differences included in the fixed effects into account. Measuring conversion effects. Because of the possibility that the effects of conversion may occur prior to conversion (as an organization changes its behavior in anticipation of conversion or to maximize the value of an initial public offering after conversion), we separately examined the effects in each of the three years prior to conversion as well as year by year following conversion up through year 3 (using a single variable to capture any effects in post-conversion year 4 and beyond). This formulation allowed us to detect whether conversion effects were short lived or did not become apparent until several years after conversion. Including three preconversion years in our analysis also allowed us to see whether there were any anticipatory effects of conversion reflecting changes in behavior to better position a Blues plan as it began to consider conversion. We also tried alternative models to see whether varying our estimated time trends by individual state made any difference. Spending categories. We examined total spending, as well as spending on hospitals and physicians separately. Except for removing our three insurance coverage variables as independent variables, we used the same covariates to predict the uninsurance index that we used for spending and hospital profits.15 Because of limitations on the availability of data, our analysis covers the period 1980–1998 for spending and 1980–2001 for uninsurance.
Exhibit 1  shows per capita health spending and uninsurance rates statewide, relative to the national average (100), before and after Blue Cross conversions (by comparing the average index value in the five years preceding conversion to the average value during available years postconversion). Unfortunately, although we can report uninsurance data through March 2004, the latest state-level spending data end in 1998, which precludes an examination of the most recent conversions and truncates the postconversion period of observation for others. Additional complications are created by the fact that each conversion is unique. At best, these figures are only broadly suggestive of the possible marketwide effect of conversion, especially since this analysis fails to take account of many other factors that might also account for these trends.
Overall results in the states. This rough cut of the raw data shows mixed results: The overall situation in some states improved following conversion (post/pre ratio less than 1) but worsened in others. Taking California as an example, the state’s per capita health spending was 102.4 percent of the national average in the five years prior to its 1993 conversion, but this dropped to an average of 93.0 percent of the U.S. average in the years for which we have data following that conversion. The ratio of these two figures, 0.908, says that health costs went down by 9.2 percent relative to the national average between the two periods. Consistent with that general picture, California also was able to improve by about three percentage points in terms of its uninsurance rate relative to that of the United States, following the Blue Cross conversion there. (Although not shown, these changes are similar if we date the California Blue Cross conversion from 1996, when it was completed, rather than 1993, when it began.) Spending and insurance. For all conversions overall, the impact on spending worsened slightly, whereas the impact on uninsurance improved, but the gross impacts (whether positive or negative) were relatively modest in size (for example, even the 4.7 percent relative reduction in uninsurance risk translated into an absolute reduction in the chances of being uninsured for an average person of less than 1 in 100; nevertheless, for those gaining coverage, this clearly was a beneficial impact). (Using definition 1, which dates conversion from the time a plan becomes a stock company [not shown], we found a slight reduction in all spending measures, but only four states contributed to this result. The uninsurance index dropped less than 1 percent). Clearly, one cannot attribute all of these changes to Blue Cross conversions, since many relevant factors were not controlled for, such as changes in per capita income or in eligibility for government programs such as Medicaid. However, statewide spending levels decreased about as often as they increased, and more often than not the uninsured rate improved, following Blue Cross conversions. Thus, as best we can infer from these admittedly crude comparisons, statewide affordability and accessibility do not inexorably deteriorate following a Blue Cross conversion. Effects over time. We tested a variety of models, ultimately selecting the one that was theoretically most defensible and which required the least restrictive set of assumptions.16 The final model shows how the effects of conversion play out over time, and it also accounts for the possibility that health spending trends (not just levels) appear to be different across states for reasons not accounted for by our other independent variables.
Analysis under definition 1.
Using the first conversion definition again, we found a significant change in overall inflation-adjusted health spending in the first two years following conversion, but these spending reductions were no longer significant after that period (Exhibit 2
Medicare spending per enrollee mirrored to some extent the pattern for total spending, with declines in spending that ranged from 3.8 percent to 6.9 percent in the first three years following conversion, but were no longer significant at conventional levels after that. Given that the effects of conversion ought to be observed more strongly in the private sector than the public sector, this may be puzzling, but several possibilities come to mind. The first is that conversion resulted in changes for Medicare supplemental coverage, in the form of higher premiums, tighter underwriting, or outright abandonment of products. It is known that Medicare supplemental coverage results in higher Medicare spending because of "moral hazard."17 But even accounting for this effect, the extent of possible reductions in supplemental coverage would have had to be rather sizable to produce the results in Medicare spending we observed. Neither media accounts nor our own key informant interviews are consistent with this story. Second, there could be some sort of spill-over effect of more aggressive bargaining with hospitals, but the problem is that our results are in the opposite direction of what might be expected. That is, if tougher bargaining lowered private patient payments to hospitals, one would expect them to respond by finding a way to increase their Medicare revenues. If converting plans were to become more aggressive in pursuing managed care, this could have beneficial spillover effects on Medicare, as it might encourage greater numbers to join Medicare HMOs. But it is odd that the effect would be more pronounced in the Medicare population than in the general population. What should be clear is that absent other supporting data, any of these possibilities is speculative.
Analysis under definition 2.
Our results for the more expansive definition of conversion (conversion is dated from the year a plan becomes a stock company, creates a major for-profit subsidiary, or is acquired by a multistate mutual that subsequently converts) were even stronger than those just described for the narrower definition (Exhibit 3
Our data provide no way of demonstrating whether the reduction in hospital spending resulted from lower hospital use or lower payments to hospitals, but these findings are consistent with our key-informant findings that conversion may result in more aggressive bargaining behavior with providers.18 If so, however, this finding does not provide a basis for determining whether lower provider payments are captured by Blue Cross plans as increased profits or instead are passed on to subscribers in the form of lower insurance rates, since our measure of health care spending focuses only on gross amounts paid to providers and not on insurance premiums charged to subscribers. Previous work on nonprofit Blue Cross plans suggests that at least some of the savings captured through lower provider payments resulting from market power are in fact shared with consumers via lower premiums, but for-profit plans may not necessarily behave the same way.19 Possible impacts on insurance rates. Possible impacts on insurance rates are suggested, however, by observing any changes in uninsurance.20 For the strictest definition of conversion, we found no significant impact of conversion on uninsurance rates, although the coefficients postconversion all were in the negative direction.21 In contrast, using the more expansive definition of conversion, we found a reduction in uninsurance risk that appears to persist in the long term even though it is not consistently significant in the short term. Caveat. An important caveat is that our findings are somewhat sensitive to how we defined conversion. When we broadened the definition to include acquisitions by multistate mutual insurers (such as Anthem prior to 2001) and the creation by nonprofit parent companies of large for-profit subsidiaries that issue stock, the effects appeared to be more enduring, and they also were associated with lower hospital profits.
This study shows that Blue Cross conversions appear to have very modest and possibly negligible impacts on overall health spending or the likelihood of being uninsured across states. Using definition 1. Using the most restrictive definition of conversion—creation of a stock company accountable to shareholders—our descriptive analysis shows that converting states overall appeared to produce gains for the general population in terms of slightly lower spending and uninsurance rates relative to what they had experienced prior to conversion. At the level of individual states, conversion was associated with approximately the same number of winners as losers.
Our multivariate results provide a more solid picture of some potentially important effects, with temporary reductions in spending for the general population and those on Medicare. These appear to be driven by reductions in hospital spending that persisted for four years or more. However, given the paucity of conversions that have been in effect this long, only a handful of states contributed to this result. Indeed, for our spending measures (which only run through 1998), Missouri (which converted in 1994) is the only state contributing to the three- and four-year postconversion results in our definition 1 analysis (Exhibit 2 The California and HMO factors. We were concerned that our results could be dominated by California, so we ran our total spending models (for both definitions) excluding California. Except for small changes in the size of the coefficients, the basic pattern and significance of the results we have reported here persisted. Likewise, we wondered whether the impact of conversion might be very different in states with high HMO penetration because they might have more competition among insurers; when we added HMO interaction terms for both the pre- and postconversion year variables, none of the coefficients was significant, so we are satisfied that our results are not masking a more nuanced story. Possible adverse relationship. It is important to emphasize that the available data cannot rule out any possibility of an adverse relationship between conversion and spending or uninsurance. In fact, in our descriptive data we saw both of these relationships in some states, such as Georgia. However, statistically, as a weighted average effect, a simple conversion is more likely to produce a beneficial (or at least benign) effect on either measure of impact. Using definition 2. This general story strengthens when we use our broader definition of conversion. In that case, our descriptive analysis suggests that conversion may be associated with a worsening of the spending picture but even greater improvements in the uninsurance picture than we observed under the narrower definition. But our multivariate results instead show that spending reductions that persist for four years or more are driven by lower hospital spending and concomitant reductions in uninsured risk. So again we are left with the overall impression that although any given state may experience a conversion differently, on average, a conversion is more likely to be associated with beneficial consequences on health spending, and perhaps on uninsurance risk (depending on the type of conversion), although possibly at the expense of reductions in hospital profits. We in fact observed sizable reductions in hospital profits. Given hospital total margins of 3.4 percent in 2000, a long-term reduction in margins of 2.7 percentage points is rather dramatic.22 In 2000, one-third of hospitals had negative margins, so in states where the average margin is cut by two-thirds, we can imagine many more hospitals facing potential financial difficulties. But we really do not know for certain, because margins were at least 7.0 percent for the top quartile of hospitals that year and 12.0 percent for those in the top decile.23 Thus, if the 2.7 percent reduction were achieved by cutting margins in the top-performing hospitals down to 3.4 percent, we would view this result much differently than if every hospital saw its margin cut by 2.7 percentage points. We cannot tell with the data we have available. Consistency with interviews. This failure to find many adverse consequences of conversion is consistent with our own key-informant interviews in four states with Blue Cross conversions as well as a systematic survey of the foundations created by these conversions.24 In this survey, foundation officials "indicated little evidence that [the conversions in their state] resulted in any major adverse impacts on the relevant populations." The study’s authors concluded that "at a macro level, previous BCBS Plan conversions do not appear to have caused massive disruptions in their respective state’s healthcare...delivery systems."25 Health plans’ concerns. What is interesting in light of our findings is that none of the plans offering conversion proposals justified them on grounds that it would result in lower premiums or reduce the number of uninsured people; instead, regulators’ concerns about premium increases at least partially accounted for the failure of conversion efforts in Kansas, Maryland, North Carolina, and Washington. Accounting for other factors. That said, the absence of definitive proof of major harms does not mean that conversions are necessarily neutral or beneficial. First, several other factors besides the ones we measure here should inform a judgment about the overall public policy impact of a Blue Cross conversion, including impact on insurance premiums, changes in underwriting and marketing, and benefits from conversion foundations. Second, as we have tried to stress throughout this paper, much uncertainty exists about the actual effects of previous conversions on costs and coverage, because of both limitations in available data and the limited number of available observations. Also, each state is unique, so even if the historical record were clear elsewhere, it is difficult to predict with great certainty what the actual effects will be in another state. The conversion of an institution that typically is a dominant force in its health insurance market and in its health policy community is a sobering step, one that almost certainly will not be undone once it is taken. Therefore, it should not be taken lightly, without careful consideration of all available evidence of the full range of competing health policy implications.
Christopher Conover (conoverc{at}hpolicy.duke.edu) is an assistant professor of public policy studies at the Center for Health Policy, Law, and Management, Terry Sanford Institute of Public Policy, Duke University, in Durham, North Carolina. Mark Hall is a professor of law and public health at Wake Forest University Medical School in Winston-Salem, North Carolina. Jan Ostermann is a research associate at Duke’s Center for Health Policy, Law, and Management. This analysis builds on work done at the request of the North Carolina Department of Insurance, under funding from Blue Cross and Blue Shield of North Carolina. The authors received helpful comments and advice from Frank Sloan and anonymous reviewers on another paper using similar data and econometrics to explore a different issue.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Study Data And Methods
Study Findings
