Commentary—Assessing the Impact of Managed Care Patient Protection Laws: Problems and Pitfalls

In their article “Impacts of Managed Care Patient Protection Laws on Health Services Utilization and Patient Satisfaction with Care,” Sloan, Rattliff, and Hall use data from three rounds (1996–1997, 1998–1999, and 2000–2001) of the Community Tracking Study (CTS) to assess the impact on health utilization and patient satisfaction of “patient protection” laws. Each of the three rounds of the CTS surveyed approximately 32,000 individuals about the care received by themselves and their families (a total of approximately 60,000 persons).

Patient protection laws include those that address: the size and scope of provider networks, the range of covered benefits, the procedures essential to access covered benefits, and the financial incentives used by managed care plans to affect provider behavior. Specifically, they include gag-clause bans, direct-access to specialist laws, any-willing-provider laws, freedom-of-choice laws, mandated minimum length of stays for delivery laws, and laws limiting the size of physician withholds and bonuses. The basic purpose of these laws is to improve the care provided to enrollees of managed care plans. The proportion of American workers enrolled in managed care plans has risen from 4 percent in 1977 (Gabel 1999) to 93 percent in 2001 (Supreme Court to Rule on Patients' Rights 2003), and all but three states have enacted a comprehensive set of patient protection laws (Sloan and Hall 2002). Thirty-five states enacted patient protection laws between 1997 and 1999.

Proponents of these laws maintain that they help ensure that enrollees have a reasonable choice of providers at the time of service, do not face limits on services that may adversely affect their health, are not arbitrarily denied access to covered services, and are not treated by a physician who faces inordinately high financial risks for ordering necessary services. Opponents of these laws (who often refer to them as “anti-managed care” laws) maintain that they reduce the flexibility of plans in their dealings with physicians, prevent plans from limiting particular services in order to increase efficiency, preclude plans from implementing cost-effective utilization management techniques, and prohibit plans from employing financial incentives that encourage cost-effective behavior on the part of health care providers. Moreover, opponents maintain that these laws increase the number of people who are unable to afford health insurance.

Several studies of any willing provider laws have shown that they increase the cost of care (Hellinger 1995; Jensen and Morrisey 1999; Rogal and Stenger 2001, and there have been a number of studies of the impact of specific patient protection laws (Hellinger 1996). This study is an important addition to this literature because it is the first to examine the impact of patient protection laws in their entirety, employing multivariate regression analytic techniques to examine a rich data base to evaluate the impact of controversial laws.

Supporters of patient protection laws insist these laws protect patients and strengthen public trust while opponents insist they increase the cost of care. Yet, this study finds that patient protection laws have had little or no impact on either the utilization of health care services or on patient trust in the health care system.

There are two plausible explanations for these findings, and the two explanations are not incompatible. The first explanation is that patient protection laws have (as found by the authors) indeed had little impact. It is certainly possible that recent changes frequently attributed to patient protection laws (e.g., increased size of provider networks and reduced requirements for enrollees to access services from specialist physicians) have resulted from a confluence of market forces and not as a result of legislation.

It also is possible that these laws have little impact because they have not been actively enforced. There is evidence that the length of stay for many maternity cases was below the mandated minimum stated in Illinois' minimum length of stay law (Rogal and Stenger 2001). There is also reason to believe that any willing provider laws were not actively enforced during the time period studied by the authors because of a court order in 1997 blocking Arkansas' broad-based any willing provider law, passed in 1995 (Crowley 2003; McLean and Richards 2003; Bleed 2004a.

Arkansas' 1995 any willing provider law was blocked because both the U.S. District Court and the U.S. Court of Appeals found that it was barred under federal rules pursuant to employee benefits as specified in the Employee Retirement Income Security Act of 1974. It was not until after the April 2003 decision of the Supreme Court to uphold a similar any willing provided law that had been passed in Kentucky that the Arkansas law was enforced (Bleed 2004b).

Moreover, part of the reason why the authors found that patient protection laws had little effect may be related to their choice of dependent variables. In this study, the authors examine the impact of patient protection laws on six measures of utilization (number of overnight hospital stays, number of emergency room visits, number of outpatient surgical procedures, number of office visits to physicians and other health care professionals, whether the respondent had a mental health visit, and whether or not the most recent visit was to a specialist). The authors do not examine the impact of these laws on the cost of services. Yet, much of the opposition to these laws is based on the premise that these laws increase the cost of providing a service, and it is certainly possible that these laws increase the cost of care without affecting the utilization of care. Indeed, there is evidence from California that laws that limit the ability of health plans to selectively contract with providers (e.g., any willing provider laws) increase the price that insurers pay for health services and in turn increase health care premiums (Zwanziger, Melnick, and Bamezai 2000).

The second explanation for the findings is that the analytic design of this study is unable to detect the impact of patient protection laws on utilization or patient trust because of specification problems in the design as they relate to the magnitude of the impact (i.e., the magnitude of the impact of these laws was not sufficiently great to be identified in this study because of limitations in the design of the study).

The application of quantitative methods to policy analysis is always problematic. Nonetheless, researchers routinely apply multivariate statistical techniques that were developed to measure the impact of changes in a specific factor on a variable of interest in an experimental setting to examine instead social, political, and economic processes. These settings in contrast have few variables held constant, many variables unobserved and many others measured inexactly, frequent problems with reverse causality, and circumstances where the complex motives, attitudes, and strategies of those involved may only be inferred (Dowd and Town 2002).

Thus, a key question regarding health services research that provides information for public policy makers is whether the statistical techniques and data employed by the researcher are capable of providing “meaningful” answers about the causal relationships between the variables of interest. Meaningful in this sense denotes the capability of identifying and quantifying causal relationships among variables that are both statistically and practically significant. Unquestionably, the ability of this research to accurately estimate the impact of patient protection laws on utilization and trust is dependent on the severity of the specification problems of its analytic design.

Among the specification problems in the analytic design of this study is the omission of variables. Omission of variables may result in spurious correlation and lead to biased estimates. One variable likely to affect the utilization of health care services is the supply of health care providers. However, the authors do not include variables measuring either the supply of physicians or hospital beds in the area. Managed care penetration rate—another variable excluded from this study—has been shown to affect utilization (Baker 1999). This rate is also likely to affect patient trust. A third exclusion from this study is insurance status that is likely to affect both utilization and patient trust. For example individuals without insurance are less likely to utilize services and to have less trust in the health care system.

Errors in variables may be a serious specification problem in this study, because the package of patient protection laws passed is not uniform across states. Moreover, the absence of baseline information in the study is also a problem because most states had enacted some types of patient protection legislation before they enacted their “package” of patient protection laws.

Another factor that complicates the estimation of the impact of patient protection laws is that major changes in the way health care is provided (e.g., increases in managed care penetration, health plan and hospital mergers, and expanding provider networks for almost all health plans) occurred at the same time as the enactment of patient protection legislation. This makes it difficult to isolate the impact of patient protection laws from the impact of these other changes.

Complicating the situation further is the Balanced Budget Act (BBA) of 1997, which provided Medicaid recipients and Medicare beneficiaries the same protections provided in most patient protection laws (Marsteller and Bovbjerg 1999). Included in the BBA of 1997 is an access to service provision that permits direct access to a women's health specialist to provide routine and preventive health services, a restriction on rules that inhibit patient–physician communications (i.e., a gag rule), and a provision that requires plans to have in place a grievance system that resolves appeals within 30 days. Although this law was passed in 1997, the final regulations implementing it were not promulgated until 2001. Furthermore, the Newborns and Mothers Health Protection Act was passed by Congress in 1996 and this law prohibited “drive-through” deliveries by mandating that health plans provide for a minimum 48-hour hospital stay (Kuper 1998). These federal actions may attenuate the impact of state laws on the utilization of health care services and weaken the impact of state laws on trust to the extent that citizens fail to differentiate between federal and state actions.

Moreover, there are jurisdictional issues. State laws do not apply to persons whose care is paid for by Medicaid, Medicare, and other federal programs, or to patients whose care is paid for by a self-insured employer-sponsored health benefit plan. State laws only apply to persons whose care is paid by employer-sponsored health benefit plans that are not self-insured. Ideally, an analysis of the impact of state patient protection laws should differentiate between persons whose care is under the jurisdiction of these laws from persons whose care is not under the jurisdiction of state law.

This study may also contain a reverse causality problem because patient distrust may increase the likelihood that a law to protect patient rights is passed in the state. In such a situation the empirical analyses may understate the impact of the package of patient protection legislation because these laws would be more likely to be enacted in areas where the level of patient distrust was relatively high before enactment of the law. This would not be a serious problem if the authors employed panel data (i.e., if they had collected data from the same individuals during each of the three waves of interviews). However, the CTS is a cross-sectional data set (i.e., data were collected from different individuals each of the three times it was conducted). The authors acknowledge this limitation and state that, “Particularly judged ex post, CTS is not well-suited for analyzing effects of individual patient protection laws” (p. 655).

Clustering is another methodological limitation of this approach. The authors estimate separate equations for each member of a household. Yet, the responses from each household member are likely to be correlated thus violating the assumption that the explanatory variables were independently distributed. If this correlation is not taken into account then the standard errors for the coefficients will be underestimated, and the statistical significance of the coefficients will be overestimated. There are techniques to adjust for clustering effects including generalized estimating equations (Liang and Zeger 1986) which would help redress this concern.

Notwithstanding the aforementioned issues, I believe that the authors have made a noteworthy contribution to our understanding of the impact of patient protection legislation on the use of health care services and on the way people view their health care system. Indeed, the authors are aware of the limitations of their analyses and have provided a good discussion of several possible reasons why they did not find significant results. Future research in this area should examine the impacts of specific types of patient protection legislation utilizing a longitudinal database (e.g., examine the impact of direct access legislation for obstetrician/gynecologists using information collected from the same women both before and after implementation of the law).