A Conceptual Model of the Effects of Health Care Organizations on the Quality of Medical Care

AN IMPORTANT recent change in the health care system has been the rapidly increasing enrollment of patients in various types of managed care.¹ One of the consequences of this shift is an increasing focus on the ways that the financing and organization of medical services affect the quality of care. A related development is a focus on the assessment and reporting of the quality of care provided by health plans. Measuring plan-level quality of care and determining how plans affect quality requires that one specify what constitutes a health plan (ie, the unit of analysis) and describe the characteristics of plans in a standardized way.

Health maintenance organizations (HMOs) have traditionally been classified into 4 types or models according to features of their structure: staff, independent practice association (IPA), group, and network. This taxonomy used to be effective for research, reporting, and contracting purposes because it described relatively homogeneous and mutually exclusive organizational types. Two recent developments in the HMO market greatly reduce the usefulness and validity of this classification scheme. First, "mixed-model" HMOs frequently incorporate characteristics of more than 1 of these formerly distinct types.^{2 - 3} Second, a variety of provider organizations, ranging from large groups to fully integrated delivery systems, have evolved, leading to tremendous heterogeneity within individual plans and plan types in the factors that might affect the quality of care. Thus, it is increasingly difficult to determine which aspects of health care organizations affect quality of care.

These developments pose problems for clinicians, managers, health services researchers, and policymakers who need to understand variations in, and the determinants of, quality.⁴ Measures of quality that are aggregated and classified using traditional definitions of plans and plan types do not permit one to assess accurately the organizational characteristics that affect the quality of care. New conceptual models are needed to foster an understanding of the characteristics of health care organizations that either facilitate or impose barriers to quality.

In this article, we first describe the current health care environment, focusing on structural elements, and describe why older classifications of health plans are problematic. Second, we present a framework for describing how managed care and other health care organizations influence physician behavior and quality of care. Finally, we propose a strategy for assessing the impact of organizations' characteristics on physician behavior and provide several examples of how such a strategy could be implemented. A similar framework could be useful for assessing the cost of providing medical care and other aspects of organizational performance. Furthermore, although we focus on physician behavior, it is also possible to analyze other levels of organizational performance. We discuss some of these possible extensions of our framework.

While managed care has a long history in the United States, the adoption of the federal Health Maintenance Organization Act in 1973 (as well as the term health maintenance organization) marked the beginning of an era of accelerating growth for the managed care industry.^{5 - 6} In the 1970s and 1980s, HMOs typically were 1 of 4 distinct types. Staff-model HMOs hired and supervised salaried physicians. Group-model HMOs contracted with a single group of physicians without directly hiring them. Independent practice association models contracted with individual physicians or small groups in private practice and often arose when groups of physicians affiliated into IPAs for the purposes of securing contracts. In contrast, the fourth type, network HMOs, contracted with 2 or more groups and were largely centrally organized by HMOs as a means of securing discounted contracts with providers.

Initially, this classification described organizational types with relatively distinct features and might have represented a reasonable typology for research and quality assessment. For instance, these types of organizations generally compensated physicians differently. Group- and staff-model HMOs tended to have salaried physicians, or paid physicians a single capitated payment for services for the entire year. Independent practice association and network models tended to reimburse physicians on a discounted fee-for-service (FFS) basis.² Even these reimbursement mechanisms, however, often took different forms in different regions of the country.

While the above typology continues to be commonly used, it fails to describe accurately new forms of managed care. Many new plans now have components of both IPAs and groups and share many of the same physician groups and intermediary organizations (eg, IPAs). For instance, a network-model plan may simultaneously contract with multiple individual physicians, large group practices, and 1 or more IPAs. Also, some IPAs or groups bear risk for the costs of care while others pass risk through to individual physicians. In 1995, almost 75% of enrollees in managed care plans were in IPAs or such mixed-model plans. Furthermore, most managed care organizations (MCOs) now have multiple options, or "products," for patients to choose from.⁷ Thus, the evolution of health care organizations has surpassed the current taxonomy, and even experienced observers of the health care system cannot adequately distinguish different health plans from one another.⁸ In a recent survey of 108 managed care plans, Gold and colleagues² found that plan type (classified as either group/staff or IPA/network) explained little of the variation found in the contractual arrangements between HMOs and physicians.

These developments have been further complicated by a steady shift from solo practice toward complex group practice arrangements.⁹ Furthermore, individuals and practices have often joined either IPAs with varying levels of centralized control and exclusivity or medical management companies with thousands of member physicians, all of which can contract with MCOs. Any of these different types of physician organizations can accept responsibility for varying degrees of financial risk and may be responsible for medical management functions traditionally performed by HMOs. Moreover, several metropolitan areas have seen greater integration of health care delivery organizations as physician groups and hospitals have merged to form integrated delivery systems capable of providing all aspects of care.^{10 - 12} According to a recent survey, 23% of hospitals nationwide and over 40% of hospitals with more than 15% of their revenue derived from managed care now participate in physician-hospital arrangements such as physician-hospital organizations or medical service organizations.¹³

Within an MCO, contracts can be with any combination of the above provider groups (Figure 1). Physician groups, in turn, can contract with multiple organizations. Revenues from FFS payments and/or capitation flow from the MCO to any of the above entities. These entities, however, may then act as a filter that changes the incentives to physicians. Furthermore, management control mechanisms or normative influences emanate from any or all of the above-mentioned entities.¹⁴ Consequently, even within a particular plan or type of organization, there is tremendous heterogeneity in the types of settings of physicians' practices that is not captured by simply describing the type of plan.

To understand better the influence of organizations on the delivery of medical care, we think it is necessary to describe organizations in terms of their constituent characteristics and study the effects of those characteristics individually and/or in combination. In addition, it is important to consider different potential units of analysis within organizations, including physicians, physician groups, IPAs, and other contracting entities before attempting to understand how organizational factors influence practice patterns and the delivery of care. In the following discussion, we describe the ways health care organizations can influence care, focusing on activities that are mediated by physicians. We then address the ways in which multiple levels within health care delivery systems can influence practice and suggest a framework that can be used to assess these issues.

Health care organizations influence the quality of health care in 4 general ways. First, organizations directly determine the nature and capabilities of their providers, including the amount of resources devoted to each type of provider. For instance, health plans can selectively contract with physicians and hospitals and define referral resources within their networks of providers or physician groups. Physician groups can consist of several practitioners in a single specialty or in multiple specialties. Second, organizations can also influence care by direct contact with enrollees, such as patient education about flu shots or screening mammograms, and copayments.^{15 - 16} Third, organizations can adopt a broader population-focused mission characterized by public health and educational programs aimed at the larger community. Finally, health care organizations can directly influence physician behavior. We believe that the most important influences of managed care will be those mediated by physicians and focus on this area. Below, we present a framework for describing the ways organizations influence physician behavior.

Organizations influence physician behavior in many ways, but direct influences can be grouped in 4 basic dimensions or domains. The first group of influences consists of financial incentives. The second group consists of administrative or management strategies such as utilization review, referral requirements, and profiling systems linked to administrative sanctions. The third group includes the structural characteristics of practices. These are the structural components of care such as the physical composition and location of the practice site, availability of services, staffing patterns, and governance. The last group of influences is information or normative influences, such as the professional culture of the organization and the nature of professional interactions that help to determine institutional practice styles or norms, including formal or informal education and feedback.

Financial Incentives. The theoretical impact of financial incentives on physicians has been well described.^{2 ,17 - 18} The basic methods of reimbursement as well as additional payments or penalties in the form of bonuses, surplus payments, or withholds all carry incentives to provide more or fewer services depending on whether income or substitution effects dominate. For instance, when the fee is greater than the costs of providing services, FFS payment is an inducement to provide additional services as physicians are reimbursed more for each additional unit of service. Alternatively, some types of capitation provide incentives to deliver fewer services. Hybrid systems can use withholds or bonuses ("risk-related payments") that are tied to utilization or other measures that can mitigate the more extreme effects of either basic payment system. For instance, some managed care plans and physician groups have adopted payment schemes that explicitly reward improved performance on selected measures of quality such as immunization rates or measures of patient satisfaction.¹⁹ The influence of such rewards is related to the proportion of payment that is received in the form of risk-related payments and the extent to which these payments are related to the individual physician's performance as opposed to being spread out over a larger pool of physicians.²⁰

At the plan level, the use of payment systems designed to decrease total expenditures is growing. In a recent survey for the Physician Payment Review Commission, Gold et al² found that 56% of IPA and network HMOs compensate primary care physicians via capitation, whereas only 12% use FFS without the addition of any withholds or bonuses. Plan-level payment policies can be important, especially when patients from a particular plan constitute a significant proportion of a physician's patient panel. For instance, a salaried physician at an HMO might receive 100% of his or her pay in the form of salary from that particular health plan. On the other hand, an IPA physician who directly receives 5% of his or her income from a particular plan will likely not be strongly influenced by that plan. In the middle are physicians who receive a significant proportion of their compensation from a particular plan or in a form (such as pure capitation) that does not substantially differ among several plans. Nevertheless, we know little about how these effects are transmitted to practicing physicians nor we do know the effects of multiple contractual arrangements on practicing physicians.

We know that physicians often deal with multiple MCOs that have radically different methods of contracting. More important, many physicians do not actually experience financial incentives directly. Physician organizations can adopt payment policies that reflect the aggregate incentives faced by the organization or that are designed to mitigate the effects of incentives on the provision of care. In conjunction with the American Association of Health Plans, the Health Care Financing Administration has recently proposed collecting data from MCOs on the arrangement that their "middle tiers" have with physicians.²¹

We know little about how physician contracting organizations have responded to prepaid care. Full capitation gives physician organizations the flexibility to experiment with innovative care delivery systems that might incorporate different resources such as nurse practitioners and outreach services that were not always reimbursed under the FFS system.

Thus, we suggest that there are 3 main components related to the use of financial incentives that bear examination when determining the influence of such incentives on the provision of care: (1) the incentives attached to a particular patient and the reimbursement mechanisms of his or her plan; (2) the overall risk experienced by the physician or physician group for all patients (this overall risk might be reflected in the nonfinancial plans and policies that groups adopt to influence care [eg, use of case managers], but are likely most related to determining the formula for compensating physicians within the group); and (3) the exact method of compensating physicians. In some cases this will be the same as the risk experienced by the group (eg, when the "group" is actually a solo practice), while in other cases, a physician's compensation will not have a direct link with the risk experienced by the group. An example of the latter might be groups that use pure salary when the overall group compensation includes risk-related payments in the form of bonuses or withholds. In such situations, it is crucial to characterize the compensation schemes of individual physicians.

Management Strategies. Management or administrative strategies include case management practices, critical pathways or guidelines, formularies, electronic reminders, and various approaches thought to directly influence physician behavior.^{22 - 24} Hillman²⁵ described these controls as "direct instructions on how a physician should or should not act in specific clinical circumstances."

The dominant administrative control mechanism has been utilization review. Traditionally, utilization review has been used by plans to control spending and promote standard protocols of treatment. Physicians often felt that such controls were burdensome and intrusive and generally resented the increased administrative effort. In the era of capitation, however, many physician organizations have adopted the same practices to control costs. A recent survey of capitated medical groups within a large network-model HMO in California found that 100% of these physician groups had instituted some form of utilization review such as preauthorization.¹⁴

The collection and dissemination of comparative data to physicians (profiling) is increasing, although the effect of these strategies is poorly understood.^{26 - 29} Most MCOs analyze utilization data for individual physicians. Seventy-nine percent of physician groups in the above-mentioned survey retrospectively profiled physician utilization.¹⁴ The use of guidelines and protocols is also increasing. Numerous health plans, IPAs, medical groups, and hospitals have promulgated guidelines and care paths to facilitate and guide medical decision making. In addition, physicians may be more receptive to guidelines than in the past. Most guidelines, however, have not been successful in changing physician behavior.^{30 - 32} Nonetheless, many organizations are now using computerized medical records and other automated systems that will likely result in improved adherence with guidelines and could improve both the quality and efficiency of delivered care.^{33 - 34}

Structural Characteristics. The structure and organization of practice, including the physical surroundings, availability of services, and staffing mix, are poorly understood influences on the quality and costs of care.^{35 - 38} Some structural components, such as the availability and convenience of colleagues for consultations and facilities for diagnostic testing and more urgent patient evaluation, can have a profound impact on the practice of medicine. For instance, a physician with accessible x-ray facilities may be able to defer an emergency unit evaluation for a patient with suspected pneumonia. Additionally, the availability of a colleague for a "curbside" consultation may expedite a diagnostic workup and decrease the number of referrals. An on-site laboratory may increase the utilization of outpatient testing. The use of nonphysician personnel such as nurse practitioners, physician assistants, or nursing assistants also can increase the efficiency of practice as well as patient and physician satisfaction.^{39 - 41}

Research on firms suggests that governance structure influences performance.^{42 - 44} Research on medical groups also suggests that the governance structure of group practices (eg, presence of a full-time or part-time medical director) is associated with the adoption of management strategies designed to improve the processes of care, although the link to performance is not well established.^{45 - 46} Are physicians involved in the top management of the organization? Do those physicians who are involved continue to practice medicine? How is practice policy determined?

Information or Normative Influences. The interactions and feedback from professional colleagues are potentially important determinants of physician behavior. Such interactions can range from chance encounters to formal outpatient management or morbidity and mortality conferences. The types and degree of specialization in the practice as well as the formalization of practice responsibilities all may have an effect on the information sharing, development of group norms, and adaptation of common practice styles.

Large variations between regions in the use of medical procedures support the hypothesis that physicians in a particular institution or community adopt similar practice styles.⁴⁷ These variations are not fully explained by patient characteristics or the background and supply of physicians. The emphasis of particular organizations on preventive care, evidence-based medicine, caring for patients with chronic diseases, or research are other examples of the importance of the organizational culture on the provision of care.^{48 - 49} While difficult to measure, these influences should be considered in any model of organizational influences.

For each individual patient, one can define a system of care, often consisting of multiple organizations, all of which influence care. The constituent variables that influence physician behavior can arise from any or all of the tiers or levels that are present in a particular patient's system of care (Figure 2). The first and most basic level is the physician practice site, which can range from a solo practitioner's office to a large multispecialty group. The practice site can have a simple relationship with a larger entity, such as a practice site in a staff-model HMO, or can be related to multiple entities such as often occurs within IPA or network-model HMOs.

At the highest level of the system is the parent MCO, which consists of 1 of a set of mutually exclusive health care delivery organizations. These organizations include very large health plans such as Harvard Pilgrim Health Care and Kaiser Permanente, as well as integrated delivery systems that contract directly with employers. Between these 2 extremes there can be several other layers that include physician organizations such as IPAs and physician-hospital organizations, as well as divisions of health plans that can be differentiated along product lines (preferred provider organizations vs HMOs) or regional or operational lines

Not all organizational layers of a system of care have an equal impact on physicians. In fact, the particular attribute of the system to be studied or the specific parameter of interest may well arise from just a single organizational layer. The appropriate unit for measurement will thus depend on 2 major factors: (1) the organizational characteristics relevant to the hypothesis being tested, and (2) the variability of those characteristics across different levels of the delivery system.

This framework has several important implications for the way we describe and analyze the effects of health care organizations. First, and perhaps most important, we think that simply using different MCOs as the unit of analysis will not adequately capture the within-organization heterogeneity in the quality of care provided by physicians working with those organizations, nor will it allow us to understand better the ways in which organizational characteristics influence physicians. Second, it is critical to characterize carefully the practice within which physicians provide care (Figure 2). Physician practices serve as a nexus through which many of the organizational influences described above operate. For example, all physicians in a given practice typically have similar structural, administrative, and financial arrangements and have common normative influences.

Third, when the outcomes we are interested in are defined by patient experiences (eg, quality of care), it is important to specify units of analysis that have similar relationships to patients. Thus, rather than adopt self-described organizations as the only unit of analysis when analyzing quality of care or organizational influences, it is important to specify logical groupings of physicians that have the same organizational influences in relation to specific patients. For example, in our framework, if one wanted to study the influence of different financial arrangements, one would need to characterize the incentives experienced by physicians, rather than the incentives used by a particular organization (Figure 3). One possibility is that it would be necessary to use physician practices or groups as the unit of analysis because different groups implement financial incentives in different ways. Alternatively, it might be that all physicians in a particular IPA were subject to similar management strategies. In short, the unit of analysis should be defined by answering the following question: "For the variables of interest (eg, a particular management strategy), what groupings of physicians have similar experience of those organizational influences?"

Theoretically, one could imagine different levels of aggregation for analyzing the influence of different organizational characteristics (Figure 1). In reality, however, we think it is likely that there will always be heterogeneity within levels of aggregation higher than the physician practice group. This does not mean that it is impossible to study higher-level organizations as units of analysis and, in fact, one must do so to study organizational influences that are not mediated by physicians. What it does mean is that rather than using simple descriptive labels of organizations, it is important to characterize the ways in which organizations at different levels arrange and transmit the influences described above and to relate these to the experiences of specific patients.

This approach acknowledges the complexities of contemporary health care organizations, yet also suggests a simplified approach to their study. When designing a study, the unit of analysis will depend on the question being asked. For organizational effects that are not mediated by physician behavior, one must aggregate up to MCOs or levels within MCOs. For effects mediated by physicians, one must also examine the level of the physician.

For example, assume that the hypothesis to be tested is that patients enrolled in health plans that pay physicians using capitation undergo fewer tests than those enrolled in health plans that pay physicians using FFS. To address this question, one would describe for all patients the nature of the following: plan-level payments (to the group); the overall payments experienced by the group from all health plan contracts; and the specific method of reimbursement for practicing physicians, whether or not related to plan payment policies. Thus, at the practice level, it would be important to describe both the aggregate financial incentives affecting the practice as well as those passed through to physicians. In addition, the practice policies put in place to augment these goals should be assessed. In practice, some physicians might receive a salary from only 1 employer and not directly feel the effects of incentives from managed care plans. These same physicians, however, might have to follow guidelines for referrals within the practice. Other physicians might directly experience incentive payments from a managed care plan as well as additional payments from 1 or more intermediary organizations. Thus, the sum effect of incentives from multiple sources and other programs or policies would have to be integrated to determine the net effect at the level of the physician.

For a second example, some economists hypothesize that physicians treat patients according to the modal insurance coverage (ie, in terms of deductibles, copayments, and the like) in a given area (the "norms" hypothesis).⁵⁰ A supply-side version of this hypothesis suggests that physicians have a single style of practice that is a reflection of the aggregate organizational influences in addition to physician-specific determinants such as education and training.⁵¹ Using our approach, one could test this hypothesis at the physician level by studying financial incentives and disincentives for specialist referral use. For each patient, one could determine the incentives to refer experienced by the physician from the system of care defined by a particular patient. Then one could determine the aggregate incentives experienced by each physician across all of his or her patients and test to see if referral use was related to the aggregate incentives or to the incentives defined by the individual patient.

Analysis of the effects of managed care has largely focused on the use of financial incentives and rules or regulations within specific health plans. We suggest that we need to consider a broader range of factors affecting physician behavior and should change our frame of reference from characterizing the behavior of large MCOs to studying the constituent characteristics of such organizations and their subunits.

We propose 4 domains of organizational attributes that might directly influence physician practice patterns: financial incentives, management controls, structural capabilities, and the normative environment. Only by examining all 4 of these influences will we adequately capture the specific programs and policies of health care organizations. We think that, at a minimum, the physician practice should be a primary focus, as this is the organizational layer that serves as the final common pathway to physicians. The challenge now is to characterize and measure the effects of these characteristics and define best practices that could influence positively the quality and efficiency of care delivered.

To simplify the explication of our framework we have focused on 1 aspect of organizational performance (ie, quality of medical care) and stressed the role that physicians play in influencing quality. There are other important aspects of organizational performance, such as costs and efficiency, and organizations can affect the health of populations directly (eg, through health education programs or by facilitating access to efficacious services). We think the framework described here offers a useful first step in conceptualizing the analysis of these important features of organizational performance.