Judith D Kasper. Handbook of Social Studies in Health and Medicine. Editor: Gary L Albrecht, Ray Fitzpatrick, Susan C Scrimshaw. Sage Publications, 2000.
Attributes of both individuals and societies contribute to the health of populations. Among these, the relative contribution of health-care services provided by trained professionals and aimed at preventing or curing disease is viewed by some as inflated. Rene Dubos observed 40 years ago that:
The greatest strides in health improvement have been achieved in the field of diseases that responded to social and economic reforms after industrialization. (Dubos 1959: 139)
More recently, Evans and Stoddart have argued that ‘the factors which affect health … go well beyond health care per se.’ (1990: 1350).
While the prominence of health-care services as a feature in the broader landscape of population health is subject to debate, there is widespread acceptance of the view that individuals should be able to obtain health services when illness strikes, and should have access to certain proven health-care interventions known to prevent or reduce the risk of disease. Most developed industrialized societies have implemented systems that guarantee basic health care as a right of citizenship. This is not the case in the United States, where financial barriers and the existence of a large number of uninsured people in particular remain central facts in any discussion of health-care utilization or barriers to care. Following the failure of large-scale health system reform in the first Clinton administration, political consensus around improving opportunities for coverage among uninsured children gradually developed, resulting in the State Children’s Health Insurance Program in the Balanced Budget Act of 1997. Other such government-sponsored efforts, for example, expanding Medicare eligibility to those 55-64 years old, face major political obstacles, however. Financial barriers to care have dominated concerns about access in the United States, and set this country apart from other industrialized nations in the health care area. One consequence has been diminished awareness of other barriers that are found not only in the United States but also in countries that have removed payment as an obstacle to care.
This chapter provides an overview of the state of knowledge concerning barriers to health care, drawing on literature from health services research and the social sciences. The scope is limited to Western developed nations, primarily the United States, United Kingdom, and Canada. Barriers to health care in developing nations are frequently only one of many challenges facing the public and private infrastructures of these countries. A recent report on health inequalities in South Africa (Hirshowitz and Orkin 1995), for example, noted that in addition to scarcity of medical clinics and providers, and access problems related to costs, distance, and waiting time, the majority African population was ‘worse off’ in almost every aspect of their lives, including lack of electricity and clean drinking water, poor sanitation, and overcrowded housing. While some of the findings on barriers to care in countries with well-developed and sophisticated health systems may be applicable to Third-World countries, the form barriers take and priorities among them are apt to differ substantially. Rural health problems in the US or Canada, for example, are of a different order from those in a country like South Africa, where half of the African population is rural and one-third relies on walking as their principal means of transportation.
The term ‘health care’ will be used to encompass preventive in addition to medical services. This overview begins with a brief historical perspective followed by a discussion of current definitions of access and barriers to care. Empirical research is reviewed relating to barriers to entry to care, and barriers that arise once the care process is underway. Finally, new conceptual and methodological developments are discussed.
A Historical Perspective on Barriers to Care: Early Concerns and Conceptual Models
Early Concerns about Whether People Were Getting Needed Care
In the United States, the initiation of the Committee on the Costs of Medical Care is often depicted as the point of origin for a population perspective on health-care use, as well as concern about whether people were able to obtain medical care when needed (Anderson 1990; Committee on the Costs of Medical Care 1932; Roemer 1985). The population surveys conducted by the Committee found lower-income people, although in poorer health, received less medical care than those with higher incomes (Committee on the Costs of Medical Care 1932). The United States did not implement a national health insurance program in the wake of the Committee’s report, a pattern of inaction that persists (for comprehensive studies of the politics of health care in the United States, see Anderson 1990; Weissert and Weissert 1996). Today, only those aged 65 and older, under the Medicare program established in 1965, enjoy universal entitlement to physician and hospital care, although continued increases in beneficiary cost-sharing over the years have led some to argue these pose a financial barrier for those with low incomes (American Association of Retired Persons 1995).
In other developed countries, recognition of the need to ‘protect, promote and restore’ (Canada Health Act 1984) the health of citizens through access to health services led to government-funded national systems of health care. The goals of these programs have been to promote ‘equal treatment for equal need’ (Smaje and Le Grand (1997) on the British National Health Service), and provide ‘access to healthcare services without financial or other barriers’ (Canada Health Act 1984). The organizational approaches to these goals differ. In the UK, the establishment of the National Health Service in 1948 linked each patient to a general practitioner who could be consulted as needed at no cost to the patient, both ensuring a point of entry to the health-care system and eliminating financial barriers. The Canadian health-care system, established under the Canada Health Act of 1984, is administered at the provincial level. It guarantees coverage to all and prohibits direct charges to patients. While direct access to hospital-based physicians is restricted, as in the UK, access to community practitioners is not limited to a single practitioner.
Empirical research on access and barriers to care has been a major focus of health services research in the United States. Resources in the form of government-sponsored health survey data and federal research funding have fostered such research. Equally important, perhaps, is that access has remained in the forefront of US health policy issues because of the lack of universal coverage. Even in Great Britain and Canada, however, there continue to be evaluations of equity in access to health services (cf. Badgley 1991; Benzeval and Judge 1996; Birch and Abelson 1993; Eyles et al. 1995; O’Donnell and Propper 1991; Smaje and Le Grand 1997), with attention focused on noneconomic barriers such as social class and education.
Research Developments—Population-Based Surveys and Behavioral Models for Care-Seeking
Empirical research on access and barriers to care has been fostered by two developments: (1) conceptual models which provided a framework for understanding determinants of health behavior, and service use specifically, and (2) the availability of population-based data, usually in the form of large national surveys, which made it possible to conduct empirical analyses and produce national estimates of interest to health policymakers. Andersen was explicit about this linkage, stating that the behavioral model of health services use he and others developed ‘was intended to assist in the analysis of national survey data’ (1995: 1).
Today in the United States, there are two ongoing federally funded national population-based surveys that are routinely used to evaluate access and barriers to care, among other issues—the National Health Interview Survey sponsored by the National Center for Health Statistics (cf. Bloom et al. 1997; Cohen et al. 1997; Simpson et al. 1997), and the Medical Expenditure Panel Survey sponsored by the Agency for Health Care Policy and Research (cf. Weinick et al. 1997). In addition, there are surveys specifically designed to monitor access to care, such as those sponsored by the Robert Wood Johnson Foundation (see Aday et al. 1980, 1984; Berk et al. 1995; Freeman et al. 1987, for results). While health surveys are more numerous in the United States, similar national databases are available in many other developed countries (cf. General Social Surveys in Canada; General Household Survey in Great Britain).
Although various conceptual frameworks and models have been employed in studying access and barriers to care, two of the most widely-used are discussed here—the behavioral model of health services use (Aday and Andersen 1975; Andersen 1968, 1995) and the health belief model (Rosenstock 1966; Strecher and Rosenstock 1997). The origins and objectives of these models have been described elsewhere by their developers (Andersen 1995; Strecher and Rosenstock 1997), and critical appraisals are not lacking (see Good (1994) on the health belief model; see Pescosolido and Kronenfeld (1995) on utilization models such as Andersen’s). Briefly, the health belief model deals with the process by which individuals assess their risk from disease or poor health habits, evaluate the seriousness of the risk, weigh the benefits of action, and grapple with barriers to action such as pain, costs, and inconvenience. This model has been applied to a broad spectrum of health behavior that includes differences in willingness to change poor health habits or adopt healthy ones, compliance with medical regimens, and some types of health service use, preventive care in particular (see Janz and Becker (1984) for a review of findings). Andersen’s behavioral model of health services was from the outset intended to address use of health-care services; the bulk of studies using this model have concentrated on physician care. The basic model, although still evolving, consists of factors that predispose service use (e.g., demographic, social structure, health beliefs), enable use (e.g., personal/family/community financial resources, service availability), and indicate need for care. As Pescosolido and Kronenfeld (1995) have noted, these models have components in common but different points of emphasis. For example, health beliefs are disaggregated into several discrete elements (e.g., perceived susceptibility, perceived severity, perceived benefits) in the health belief model which has social psychology underpinnings, but constitute only one of several characteristics predisposing service use in Andersen’s model, which draws on a sociological prospective.
From the perspective of studies of access and barriers to care, the Andersen model has been more influential. Andersen’s model is more narrowly targeted on health services use, while the health belief model is applicable to a broad array of help-seeking behavior. Furthermore, Andersen and colleagues (Aday and Andersen 1975; Andersen 1968, 1995) suggested that a key application of the model was to evaluate equity in health service use. Because health services are intended to address needs for care, if characteristics other than need, such as insurance coverage (enabling) or race (predisposing or, to the extent it correlates with poverty, enabling) are predictors of use, these relationships suggest inequities in access. The influence of enabling characteristics in particular, such as insurance coverage and income, in empirical analyses of physician use have been a major focus of access studies that employ this model. Finally, as Mechanic (1979) observed, Andersen’s model when applied to data from large surveys has been more successful in accounting for variance in service utilization than models that emphasize psychosocial factors.
The role of ‘need for care’ in the two models is very different. In the health belief model need is not addressed directly, although the end result of various ‘calculations’ that a person undertakes to form a decision to change behavior or seek treatment could be characterized as a determination of need for care. Such ‘calculations’ include: Am I at risk? What are the consequences of inaction? How effective is the course of action being considered? How difficult will it be to implement? Andersen and colleagues distinguish between ‘perceived need,’ a self-evaluation of health such as overall health, symptoms, or functional difficulty, and ‘evaluative need,’ a professional judgement concerning health status and need for care such as diagnoses. Measures of need in empirical analyses using Andersen’s model, whether perceived or evaluative, generally reflect degrees of good or ill health. These have been shown repeatedly to correlate with illness-related care, but are of less relevance in understanding use of preventive services or health promoting/damaging behavior. In the health belief model, on the other hand, determination of need equates with a willingness to act, which provides a framework for understanding behavior that is not driven by illness or poor health, but by the desire to avoid these states. These different perspectives on need may explain why Andersen’s model has largely been applied to access to illness-related services, while the health belief model has found broader application in studies of access to preventive health services and screening behavior.
Current Definitions of Access and Barriers to Care
Defining Access and Barriers to Care
Access to care and barriers to care are often used interchangeably, and will be here. Reference to barriers to care seems to be gaining currency, however. According to a recent Institute of Medicine (Institute of Medicine) report, access is:
A shorthand term for a broad set of concerns that center on the degree to which individuals and groups are able to obtain needed services from the medical care system. (1993: 4)
In a reformulation of Andersen’s model, the same report recasts predisposing and enabling characteristics as various barriers to care—personal (e.g., attitudes, education, cultural), financial (e.g., poverty, insurance coverage), and structural (e.g., service availability, transportation). Another report on access issued recently by the Robert Wood Johnson (RWJ) Foundation (1993) also attributes lack of access to specific types of barriers—economic, supply and distributional, and sociocultural.
While this shift is not entirely new (see Aday 1975), a Medline literature review since 1960 shows greater use of the ‘barriers’ terminology in the last decade. Describing factors that affect service use as barriers to care may stem from several factors. First, analyses of access, using Andersen’s model in particular, identified certain characteristics, such as lack of insurance coverage or presence of a usual source of care, as influences that facilitated or interfered with access. Later studies have focused on identifying determinants or correlates of the presence or absence of these ‘barriers.’ Secondly, the Institute of Medicine report expresses the view that ‘the most important consideration’ in access to health services is:
Whether opportunities for good health outcomes are systematically denied to groups in society. (1993: 4)
A focus on barriers to care that deny such opportunities logically follows, and as already noted, the Institute of Medicine report identifies several. It is noteworthy that the Institute of Medicine statement proposes health outcomes rather than equal service use as the measure of whether access is adequate. As Birch and Abelson (1993) point out in discussing the goals of the Canadian health system, achieving equity may depend on some types of inequality, for example allocating more health resources to those with greater needs for care. While equality of health outcomes cannot be guaranteed by equitable access alone, considering barriers to care that influence health outcomes represents a shift from prior models that focused on barriers to unequal utilization.
Finally, the emphasis on barriers to care reflects the policy orientation of many researchers engaged in analyzing the health-care system. ‘Barriers’ provide targets for policy intervention. For example, stating that poverty is a barrier to utilization makes the target for action clear; noting that service use decreases with income, does not. Empirical analysis of large data sets, particularly those that provide national estimates, has also been compatible with producing policy-relevant findings that demonstrate the size and scope of access problems. The health belief and utilization models were offspring of the social sciences and informed by the desire to understand human behavior within the context of health-care use and help-seeking generally. As Gray and Phillips (1995) suggest, drawing policy implications from such perspectives presents a challenge. In general, policymakers are more interested in knowing which actions to take to address problems, than in the complexities of human behavior.
Barriers to Entry, Barriers in the Care Process, and Barriers from the Consumer’s Perspective
Much of the research on barriers to care has focused on entry to services, usually measured by contact with a physician. A focus on entry to physician care has dominated research on access, in part because physicians are key points of entry to the health-care system. Initiation of preventive care, such as cancer screening or well-child visits, are specialized services where interest in factors affecting entry to care has also generated considerable research. As research on population subgroups, such as the severely mentally ill and those with chronic physical illness, has grown, so has recognition of the need to consider other service sectors such as specialty mental health care and long-term care.
More recently, barriers at other phases of the care process, following entry to care, have emerged as important. This is particularly true in the United States, where the spread of managed care has sparked interest in the effects on access of such organizational policies as putting physicians at financial risk for patient care and restricting specialty referrals. Characteristics previously established as barriers to care entry are still of interest as possible influences on whether a patient receives appropriate care within healthcare systems. For example, in a study of health outcomes among Medicare-covered elderly people in managed care, Ware et al. (1996) found worse health outcomes for the poor elderly relative to those better off over a 4-year period.
A renewed interest in quality of care has also generated more studies that evaluate access to services considered standard, or appropriate treatment for specific diagnoses or health problems. Examples include receipt of ophthalmo-logical exams by diabetics (Weiner et al. 1995), and access to certain types and appropriate doses of medications by schizophrenics (Lehman et al. 1998). Such studies often emphasize variations among providers in delivering services, ignoring, for the most part, the patient’s role in the process.
Finally, some access studies are including measures that reflect the consumer’s perspective on problems, such as difficulty or delays in obtaining care. Interest has grown in the consumer point of view in the United States (Knickman et al. 1996) as managed care has disrupted established relationships with doctors and introduced a new layer of bureaucracy to the process of seeking health care. This line of inquiry might be viewed as a much pared-down version of the ‘perceived barriers’ component of the health belief model. Results are usually interpreted in the context of the Andersen model, however; reported difficulty or delays due to costs are seen as alternative measures of financial or other barriers. Findings are also interpreted as indicating ‘unmet need for care.’ Berk et al. (1995) reported unmet needs from the 1994 Robert Wood Johnson Foundation access survey, for example, using questions about not getting needed medical services including prescription drugs, dental care, and mental health care in addition to general physician care.
Barriers to Entry to Care
Characteristics of Individuals and their Social Environment
Poverty and Socioeconomic Status
There is extensive literature on the relationship of socioeconomic status to health and the pernicious effects of poverty on health (see Haan et al. 1989; Marmot et al. 1997; Robert and House, Chapter 1.8 in this volume; Williams and Collins 1995). Despite skepticism about the ability of health services to significantly intervene in this relationship (Evans and Stoddart 1990; Williams 1990), the extent to which poverty or low socio-economic status contributes to problems in obtaining care has been a major focus of access studies and remains a key test of health system equity. Measures of socioeconomic status vary in definition and use (see Williams and Collins (1995) for a discussion). Access studies in the United States tend to use income and/or education. Studies in other countries often focus on social class.
National data from the United States, prior to the implementation of public programs for the elderly and poor documented that 71 per cent of high-income people, but only 56 per cent of low-income people saw a doctor in a year (Andersen et al. 1976). Coverage of the elderly and many of the poor, through Medicare and Medicaid, has reduced income differences in likelihood and volume of physician contacts in the US population. However, between 1991 and 1993, while average physician contacts per person in a year were similar, or slightly higher, for poor compared with nonpoor men and women, among those reporting their health as fair or poor, average contacts for the poor were much lower. Average contacts for the poor were 13 per cent and 15 per cent for men and women, respectively, versus 17 per cent and 23 per cent, respectively, for the nonpoor.
In Canada and the UK, where the introduction of universal coverage and public financing of care has largely eliminated the relationship between income and utilization (Badgley 1991; Eyles et al. 1995), the continued influence of ‘class position’ has been noted (Badgley 1991). Based on a comparison of family physician use in Canada in 1985 and 1991, Eyles et al. (1995) suggest that although needs among poorer income groups have increased relative to richer groups, utilization has not increased correspondingly. Furthermore, while the relationship between use and income has been shown to disappear in various analyses, these authors still find relationships with other measures of socio-economic status, including education and region of residence ‘which would appear to represent nonincome based barriers to access’ (Eyles et al. 1995: 638).
Andersen et al. (1976) included one preventive service in a health survey done in the early 1960s and found:
The low income population was least likely to have had a physical within a year and most likely to report never having had an examination. (1976: 8)
Recent data on preventive services document continued disparities by income. Vaccination levels for children 19-35 months of age in the United States are higher for nonpoor compared with poor children. Recent data indicated that only 59 per cent of poor children in this age group had been immunized against preventable childhood illnesses such as polio, measles, and diphtheria, compared with 71 per cent of non-poor children (National Center for Health Statistics 1995). In 1990, among adult women, 73 per cent of poor compared with 84 per cent of nonpoor women were found to have received a pap test in the previous 3 years. Similarly, although breast cancer screening among women over 50 has increased dramatically, in 1990, only 22 per cent of poor women compared with 46 per cent of nonpoor women had received a mammogram in the previous year. For both cervical and breast cancer, which screening tests are designed to detect at an early stage, women living in low-income areas were more likely to be diagnosed after metastasis than women living in high-income areas − 14 per cent and 5 per cent, respectively, for cervical cancer, and 22 per cent versus 46 per cent, for breast cancer (Robert Wood Johnson Foundation 1993). It is tempting to attribute income differences in preventive service use solely to lack of insurance coverage, but differences persist even when these services are financed. Among the elderly on Medicare, for example, a higher percentage of the nonpoor have received the one-time immunization for pneumococcal pneumonia (Robert Wood Johnson Foundation 1993).
Despite the effectiveness of implementation of universal coverage as a remedy for socioeconomic differences in many types of utilization, evidence from other countries shows differences by social class in the use of preventive services. Benzeval et al. (1995) note findings of lower immunization rates among lower social classes in the UK and ‘an inverse class gradient in relation to attendance at health checks and other preventative services’ (1995: 166). As reported by Maclntyre (1997), the Black Report also took note of ‘inequalities in use of health services, particularly and most worryingly of the preventive services’ (1997: 727). Maclntyre observes that social class is too often simply a control variable in analyses of health-related behavior, and infrequently the subject of investigation—a charge reminiscent of criticisms of the use of race in many analyses of access of healthcare use in the United States. Benzeval et al. (1995) echo the need for more research on factors that underlie treatment-seeking behavior which may account for social class differences.
Ethnic and Racial Minorities
The use of race as an explanatory variable in health services research has been challenged recently on the grounds that it often functions merely as a proxy for other characteristics, primarily socioeconomic status (LaVeist 1994; Schulman et al. 1995). Greater efforts to differentiate the effects of socioeconomic status from race are also called for (LaVeist 1994; Nickens 1995; Schulman et al. 1995; Williams 1994). In addition, race and ethnic group membership also ‘proxy’ for culture, which can incorporate attitudes, beliefs, and preferences. This perspective on race is usually not explicit, but emerges as analysts attempt to interpret results that indicate effects of race/ethnicity independent of socioeconomic status. Despite conceptual fuzzi-ness about what race represents, Williams (1994) and Williams and Collins (1995) point to one key reason to continue to focus on racial differences—the existence of discrimination in the health-care system as in other segments of society. He suggests:
The failure of socioeconomic indicators to completely account for racial differences in health (also results from) the failure of most studies to consider the effects of racism on health. (Williams and Collins 1995: 366)
Race and ethnic differences in patterns of care have been documented repeatedly for physician use, and for many other services as well, including prenatal care (LaVeist et al. 1995), mental health care (Wells et al. 1987), nursing home use (Mui and Burnette 1994), interventions for coronary artery disease (Ford and Cooper 1995) and emergency room use (White-Means 1995).
Disentangling the effects of race/ethnicity from socioeconomic status and other factors influencing service use is difficult. Stein et al, (1991) studied mammography use among white, black, and Hispanic women and found that cost concerns exerted the largest effect on mammography use among Hispanic women. Although impoverishment was greater among the Hispanic women, the authors suggest that this was partly a result of acculturation, specifically inability to speak English. In addition, use by both black and Hispanic women was related to perceptions of the examination as painful and fears about radiation levels. It appears from this study that eliminating financial barriers might eliminate a significant portion of the difference in mammogram use by race/ethnicity, but clearly not all.
Acculturation is sometimes advanced as an explanation for the effects of ethnic group membership on access, particularly in studies that include Hispanic subgroups. Results are mixed, in part because measures of acculturation are not consistent across studies. Chesney et al. (1982) and Wells et al. (1987) found effects; Marks et al. (1987) and Markides et al. (1985) did not. Solis et al. (1990), using national survey data with large samples of three Hispanic subgroups, examined several preventive services, including a routine physical examination, a dental checkup, vision testing, blood pressure testing, pap test, and breast examination. They found lower use of services by Mexican Americans, than by Puerto Ricans or Cuban Americans, but attributed these differences to insurance coverage and having a regular source of care. Only one measure of acculturation, language, was associated with recency of screening examinations; several others were not, including country of origin, contact with homeland, and parental expectations of children.
The persistence of race differences in utilization despite removal of financial barriers is illustrated by studies of the elderly population. A recent report by the agency that administers the Medicare program showed that 22 per cent of elderly blacks received a flu shot in 1995 compared with 43 per cent of whites (Health Care Financing Administration 1996). Another study examined 32 procedures and tests, that included services for cardiac, cerebrovascular, and orthopedic procedures and mammograms, and found that whites were more likely than blacks to receive services for 23 procedures and had a particular advantage in access to higher technology or newer services (Escarce et al. 1993). The authors offer a variety of hypotheses for further investigation: differences in prevalence and severity of clinical conditions; financial barriers associated with more and better private supplemental coverage for whites; the reliance of blacks on different providers (neighborhood health centers and hospital outpatient departments rather than private physicians); the effect of race on physician and institutional decision making; patient preferences and health beliefs. These correspond to already familiar explanations for race/ethnic differences in patterns of utilization—inadequate measurement of health differences and effects of SES related to race, quality of care differences, racism, and psychosocial factors that reflect culture or education. Efforts to explore these explanations are few, however, in part because the data that are the mainstay for many analyses of access to care, national surveys and administrative claims, are inadequate to the task.
Smaje and Le Grand (1997) point out that in the UK:
Ethnicity is a dimension of possible inequity … that has received much less attention than that arising from other forms of social stratification such as social class or income. (1997: 485)
However, their study found few differences in service use among Indian, African, Pakistani, Bangladeshi, and Chinese people compared with whites. Only the Chinese displayed consistently lower utilization, which the authors attribute to possible differences in cultural views of health, as well as residential patterns.
In Chapter 2.1 of this volume, McElroy and Jezewski define culture as ‘a normative framework for decision making and behavioral strategies.’ Studies on access to care that derive conceptually from the Andersen model and make use of large surveys have demonstrated that socioeconomic status and race/ethnicity influence patterns of use, but have had little success in measuring the impact of more elusive concepts such as culture. About 20 years ago, Mechanic (1979) noted the ‘gap between [the] two literatures dealing with physician utilization’ regarding the importance of psychosocial factors—one based on more theoretical, small sample studies focusing on illness behavior, and the other using multivariate statistical techniques to analyze large samples.
Attention to culture, and psychosocial factors, such as health beliefs and attitudes that derive in part from culture, remains an underrepresented area in the literature on access and barriers to care. One striking example of the importance of this type of research comes from outside the field. A recently published book by Anne Fadiman (1998), The Spirit Catches You and You Fall Down: A Hmong Child, her American Doctors and the Collision of Two Cultures, describes the inability of Western physicians and a Cambodian family to find common ground in the treatment of a young girl with epilepsy, given their conflicting perspectives on the meaning of illness and the roles and responsibilities of patients and doctors. Two other recent studies are useful illustrations of the importance of addressing culture and health beliefs if we are to move beyond mere speculation about the effects of noneconomic variables on access to care.
Using a large national survey sample, Fiscella et al. (1998) examined skepticism toward medical care which they describe as:
The final common pathway for disparate socioeconomic and cultural forces that generate doubts about the relative ability of conventional medical care to improve one’s health. (1998: 181)
Multivariate analyses confirmed a relationship between a four-item scale measuring skepticism and several behaviors: not having health insurance, not having a regular source of care, uniformly lower health-care utilization, and unhealthy behaviors such as smoking and not using a seatbelt. The study also indicated that certain demographic characteristics were associated with skepticism—being male, younger, lower income, less educated, and white. One motivation for this study was to raise awareness among the medical and health services community of attitudinal factors in patient behavior. The authors argue that efforts to hold physicians and health plans accountable for performance by measuring various indicators of patient health or compliance with preventive care guidelines do not consider the limits of provider influence over patient attitudes and behavior. The study clearly emphasizes the importance of some of the psychosocial factors in health behavior that have received short shrift in many access and utilization studies. Interestingly, references from the body of work on health beliefs in the social science literature are not cited. The authors do note, however, that in their study, skepticism toward medical care explains only a modest amount of variation in utilization:
Consistent with previous studies (Mechanic 1979 is cited) that have examined the effect of multiple psychosocial factors (Fiscella et al. 1998: 188)
This lack of explanatory power in multivariate analyses has clearly worked against attention being given by those in health services research to psychosocial and cultural factors in studies of access and barriers to care. White-Means (1995), a health economist, notes that ‘empirical models typically exclude measures of culture and attitudes’ (1995: 219). However, in a study of emergency medical use, she shows that race may represent cultural and attitudinal differences that affect utilization of care, and should not be viewed as ‘simply correlated with a subset of economic variables,’ (1995: 210). Using as a measure of culture a tradition of use of home remedies, and as a measure of attitudes a trust in the traditional medicine system, she demonstrates that both are related to emergency room use by black elders, although neither is a factor in use by whites.
Characteristics of Health-Care Organization and Financing
The literature regarding the deleterious impact of being uninsured on access to care and utilization is extensive. Two lines of research continue to be pursued. The first is the continued monitoring of the size and scope of the uninsured population in the United States. Declines in employment-based coverage (Holahan et al. 1995), and gaps in Medicaid eligibility that leave many poor people without coverage (Davis 1997), remain central issues in policy research. Among the major objectives of many national population-based health surveys is to provide data on insurance coverage for purposes of characterizing the numbers and characteristics of those without coverage. The second and more recent research initiative is to document the consequences of being without coverage by linking perceptions of unmet needs or poor health outcomes to inadequate access among the uninsured. Davis, addressing the Association for Health Services Research, called for more studies of this type, noting that there are only a few studies that document the health consequences of being uninsured and ‘as a result, many believe that the uninsured are able to get care when they need it’ (1997: 645).
Other aspects of insurance coverage in the United States, and effects on access, have also been the focus of research, including reductions in use as insurance co-payments increase (Newhouse et al. 1981; Simon et al. 1996), lower use among Medicare beneficiaries without supplemental private coverage (Blustein 1995), and the consequences of less comprehensive private insurance benefits (Short and Banthin 1995). As long as variations in coverage and multiple types of coverage and plans characterize US health care, the extent to which these variations represent barriers and the determinants of variations in coverage—employment, family composition, income, disability—will be a focus of health policy and health services research.
Regular Source of Care
The importance of a regular source of care (usually a specific physician, but including health-care organizations such as clinics) was proposed as a key enabling characteristic affecting utilization in Andersen’s initial model. In the words of the Institute of Medicine report, having a regular source of ambulatory care ‘has traditionally been viewed as a sine qua non for access to medical care’ (1993: 159). Having a point of contact for care is seen as facilitating both illness-related and preventive care. Studies show that lacking a regular source of ambulatory care is associated with not seeing a physician (Aday et al. 1984), not receiving recommended medical care and receiving less preventive care (Hayward et al. 1991), less continuity of care (Becker et al. 1974), and delays in seeking care (Sox et al. 1998).
In the United States, the proportion of the population with no usual source of care ranges from 10 to 15 per cent. Elderly people and very young children are more likely to have a usual source of care than others, as are higher-income people, and white compared with black or Hispanic individuals (Aday et al. 1984; Kasper and Barrish 1982). Not surprisingly, a much higher percentage of people without health insurance report no usual source of care. Among poor and nonpoor children, for example, 87 per cent of those under age 4 without insurance and 74 per cent of those 5-17 years old were without a usual source of care, compared with 94 per cent and 89 per cent, respectively, of those with public coverage (Robert Wood Johnson Foundation 1993).
There also has been investigation of the effects of types of usual source of care, in particular seeing a specific physician versus obtaining care at a regular site with no specific doctor, or at sites such as hospital outpatient departments and emergency rooms (Cornelius et al. 1991; Kasper 1987). One recent study found that having a regular doctor, compared with a regular site but no regular doctor, improved the likelihood of a physician visit for people in poor health, but made no difference for mammography among women over 50 or childhood immunizations (Lambrew et al. 1996).
In the United States, reliance on hospital outpatient departments or emergency rooms as a regular source of care has been associated with low income, minority ethnic status, public insurance coverage, and lack of coverage (Aday et al. 1984). For example, in 1982, 27 per cent of poor nonwhites indicated this type of regular source of care. This pattern seems likely to change dramatically as states enroll Medicaid populations into managed care plans that restrict direct access to emergency rooms. One recent study (Shah-Canning et al. 1996) suggests, however, that altering the heavy reliance of inner-city poor families on the hospital emergency room may be difficult. While 95 per cent of families seeking pediatric ER care reported a usual source of care, most did not attempt to contact their regular source prior to visiting the ER.
Some analysts question a uniform interpretation of lack of a usual source of care as an access barrier because the majority of those with no usual source indicate they feel no need for one (Hayward et al. 1991). Lambrew et al. (1996) argue that this perception is at odds with reality, since individuals with no usual source of care consistently appear ‘at risk of receiving less timely and appropriate care’ (1996: 148). Other criticisms relate to the inability of this construct to reflect more complex patterns of provider relationships. Persons with chronic disease, for example, make use of multiple regular providers, and access to nonphysician providers may be equally critical.
Lack of a usual source of care is an access indicator that appears largely irrelevant to studies of the British or Canadian health systems. In Great Britain, individuals are explicitly linked to a general practitioner. In both countries the supply of primary care physicians is greater and the distribution more even than in the United States. Furthermore, direct access to hospital-based practitioners is quite limited (requiring a referral from a general practitioner), so reliance on hospital outpatient departments or ERs as regular sources of care is precluded. Even in the United States, as the percentage of Americans in managed care continues to climb, absence of a regular source of care may diminish in value as an access indicator at least among the insured because most plans require selection of a primary provider or clinic as the first point of contact.
Early studies of access indicated significant regional and rural/urban differences in access to care in the United States. Data from 1963, for example, showed the rural farm population was less likely to have a physical examination, and that central city dwellers and the rural farm population were least likely to see a doctor and most likely to report having no usual source of care (Andersen et al. 1976). At present, however, residence has declined in importance in considerations of access, in part because large residential differences have disappeared. In a study from the 1980s, Freeman et al. (1987) noted that:
After many years of national attention to achieving a more equitable geographic distribution of health resources, rural Americans, on average, appear to be receiving as much medical care as urban residents. (1987: 17)
Designation of certain areas as medically underserved for purposes of augmenting physician supply was introduced in the mid-1970s, although there is skepticism about how well this designation reflects access problems (Berk et al. 1983; Kleinman and Wilson 1977). Community Health Centers, subsidized by various levels of government, were also established in the 1960s to provide primary care services in inner cities and rural areas. A review by Blumenthal et al. (1995) of studies on the impact of these centers, by and large showed improvements in access to services.
Disparities in the distribution of physicians remain. For example, the distribution of providers continues to be associated with population income. In 1990, there were 75 pediatricians per 100000 children under age 19 in high-income counties in the United States, and 33 per 100000 in low-income counties. The distribution of internists followed a similar pattern (Robert Wood Johnson Foundation 1993). However, the implications of physician supply for access to care are not clear. Grumbach et al. (1997), using data from urban areas in California, found physician supply to be unrelated to access once insurance status, income, and race/ethnicity were taken into account. They argue that:
A more geographically equitable distribution of physicians in the US is unlikely to compensate for a less than egalitarian system of health insurance. (1997: 82)
They also suggest that greater attention to race and ethnicity, rather than to numbers of physicians, may go further toward addressing unmet needs, since minority physicians are more likely to care for minority patients.
In the United States there are still concerns about the distribution of providers and services. As managed care networks develop, which potentially exclude community health centers and urban hospitals, there are fears about the availability of ‘safety net’ providers for vulnerable populations, such as the inner-city poor and the seriously mentally ill. Data on geographic differences in treatment patterns, attributed to a variety of factors including physician practice patterns and supply, have also led to concerns about high, and possibly unnecessary, utilization, in addition to whether lower levels of use indicate access problems (cf. Chassin et al. 1987).
Barriers within Systems of Care or Treatment Episodes
There is growing concern in the United States about the ability to obtain needed and appropriate care, independent of concerns about gaining entry. Managed care organizations attempt to reduce utilization of high-cost, specialty, and hospital-based care, and some fear these reductions will mean poorer quality care or will fall disproportionately on those who are sickest and most vulnerable. The development of Outcomes research’ has also focused attention on the care process. Concerns about barriers to care have expanded beyond physician access to include whether specialty care is available, and whether care is provided appropriately, for example, according to recognized treatment guidelines and standards of care.
The distinction between studying quality of care and studying barriers or access is not always clear in outcomes research. Some investigators start from observed variations in practice or treatment modalities, and attempt to determine whether these affect outcomes, and if so, why; others start from a known outcome, such as hospitalizations that are considered medically unnecessary, and attempt to identify predictors or correlates of these events. In the first type of study, the role of access relative to other forces that affect outcomes can not always be clearly identified. A study by Ware et al. (1996), for example, found differences in health outcomes for elderly, and poor, chronically ill patients treated in a Health Maintenance Organization (HMO) and fee-for-service systems over a 4-year period. In interpreting their results, the authors raise several areas for future exploration that may explain these differences, including clinical factors, and variations in quality of care such as comprehensiveness, service coordination, treatment queues, and continuity. While noting that HMOs reduce utilization, the authors do not speculate on whether differences in access to services is a causal factor in their findings, although some of the process measures suggested for future research, specifically comprehensiveness and treatment queues, may reflect access differences. Another study of this type is that by Shaughnessy et al. (1994), who explored differences in outcomes for home health patients in Medicare fee-for-service and HMO care. The reduced levels of disability in fee-for-service patients were, in this instance, linked to a higher volume of home health visits.
The second type of study seems more likely to reflect directly on access issues in the process of care. Two recent articles that focus on hospitalizations which could be prevented by access to primary care illustrate this approach. The first, by Bindman et al. (1995), calculated hospitalization rates for five chronic conditions (asthma, hypertension, congestive heart failure, chronic obstructive pulmonary disease, and diabetes) for different zip code areas in urban California. Interviews were conducted with random samples of adults living in these areas who were asked about access to care, chronic medical conditions, and propensity to seek care. Controlling for various characteristics, the authors found that within zip codes, rates of preventable hospitalizations for these chronic diseases were inversely related to access to care. A second article, by Pappas et al. (1997), focused on ‘potentially avoidable hospitalizations’ using a list of twelve diagnoses for which hospitalization ‘can be avoided if ambulatory care is provided in a timely and effective manner’ (1997: 811). These authors reported higher rates of potentially avoidable hospitalizations for persons living in middle- and low-income compared with high-income areas, and for blacks compared with whites, among people under age 65. Hospitalizations for ambulatory care-sensitive conditions such as asthma, severe ear/nose/throat infections, and bacterial pneumonia have also been shown to vary between children living in low- and high-income areas (Robert Wood Johnson Foundation 1993). These studies represent efforts to demonstrate that poor access to good quality primary care leads to subsequent utilization that is costly and unnecessary, bringing access considerations directly into the stream of research concerned with quality and outcomes of care.
Where is Research on Access and ‘Barriers to Care’ Heading, and What Methods are Needed to Answer New Questions?
New Access Indicators
Interest in health outcomes and the role of access to specific types of services at appropriate stages of treatment will likely grow. Several factors will influence the development of access indicators that are meaningful in terms of health or treatment outcomes. At present there are a limited number of diseases and conditions for which access to specific services or treatments have been proposed as significant influences on health outcomes. These include access to primary care in childhood asthma to prevent emergency care and hospitalizations (Robert Wood Johnson Foundation 1993), cancer screening for early detection of breast and cervical cancer (Institute of Medicine 1993; Robert Wood Johnson Foundation 1993), and office-based blood tests and eye examinations for diabetics to prevent health declines (Weiner et al. 1995). Highly prevalent or costly diseases and conditions are the most likely candidates for further development of such indicators because there is greater interest in monitoring health conditions that affect large numbers of people or consume significant resources. For conditions that do not meet these criteria, however, research and policy interest may be more limited. Secondly, these types of indicators require knowledge about the relationship of service use to health or treatment outcomes that at present is limited to selected conditions and diseases. For example, preventable hospitalizations for angina or hypertension among chronically-ill adults have been suggested as indicators of inadequate primary care, reflecting poor access or quality (Institute of Medicine 1993). For other diseases of the elderly, such as arthritis or dementia, for example, too little is known about the relationship between service use and outcomes to suggest an indicator that would measure the consequences of good or poor access. In these instances, hospitalization cannot be interpreted as indicating either one. Finally, the development of indicators of access with consequences for health outcomes also requires, in many instances, a variety of data sources, not only surveys but utilization data with diagnostic information and sometimes other specialized data sets such as disease registries, which contain information on severity. As a result, these types of study are often complicated to design and implement, time-consuming, and expensive.
The Re-Emergence of Noneconomic Influences
Although insurance coverage and financing remain at the center of access concerns in the United States, it is clear that even when no financial barriers exist, cultural and behavioral factors influence access. There is some indication these are receiving greater attention, as reflected by discussions of what race conveys, and what lies behind differences by socioeconomic status. These attributes are chiefly viewed as attaching to individuals and influencing their behavior; however, the response of health-care providers and institutions to persons with particular characteristics also may influence access, though these have received less attention. For example, it has been suggested that racism is a factor in lower use of nursing homes among blacks in the United States (Falcone and Broyles 1994), and one study in the UK suggested class differences in the amount of information patients receive from their general practitioners (Cartwright and O’Brien 1976).
Examining access indicators that affect outcomes for individual diseases and conditions may also help questions about the role of psychosocial factors in utilization to resurface. As Andersen has suggested:
If we examine beliefs about a particular disease, measure need associated with that disease, and observe the services received to deal specifically with the disease, the relationships will probably be much stronger than if we try to relate general health beliefs to global measures of need and a summary measure of all services received. (Andersen 1995: 2)
The Impact of Managed Care
Managed care has elevated concerns about certain organizational aspects of health care as potential barriers. These include financial arrangements with participating physicians which may create incentives for providing less care, organizational policies that require prior authorization or restrict referral to some types of services, and limited recourse for patients to appeal denials of care. There is little empirical data about the prevalence of such policies or their impact on patient access to care. Some of the conceptual, measurement, and data developments needed to address the specific effects of managed care on access are beginning to be addressed (cf. Aday in this volume; Docteur et al. 1996; Gold 1998; Kasper 1998). Rapid changes in the industry and development of new organizational features, for instance point-of-service care, which eases restrictions on specialty access but at a price, make it difficult to study managed care and to generalize from findings. Attention to systematic differences in utilization and selected health outcomes for those who are especially vulnerable (e.g., chronically ill, low income) may prove useful in monitoring access, at least in the short run.
Models and Data
Reformulations of the models of Andersen and Rosenstock have been proposed (cf. Andersen 1995; Institute of Medicine 1993; Strecher and Rosenstock 1997), and the usefulness and durability of each suggests that they will continue to be relied on in access studies. Rosenstock’s model highlighted the primacy of decision making by individuals and patients, and Andersen’s model emphasized the influence of an individual’s place in the social structure. Studies of outcomes and care quality, however, often draw on Donabedian’s framework for evaluating quality of care, in which barriers become one of many aspects of the structure or process of care that may influence outcomes (Donabedian 1988). Because this framework emphasizes the behavior of medical organizations and professionals, it risks diminishing the role of patient behavior as a source of variation in health outcomes. On the other hand, applying this framework to treatment outcomes will unavoidably draw attention to some psychosocial aspects of patient behavior that are not included in most studies of barriers to care, such as compliance, attitudes toward treatment, and expectations of treatment. Explanations of these complex behaviors are likely to benefit from the social science perspective that produced the early models.
Pescosolido (1992) has made a compelling case that much can still be learned about using medical care by viewing it through the lens of social science theory, for example, as a type of ‘help-seeking strategy’ embedded within social networks and influenced by social interaction. Others (cf. Mechanic 1989) have noted the potential contributions of less-used ethnographic or qualitative methods to understanding medical care utilization. Greater interest in the effects of access on health outcomes could lead research in these directions. Analysis of secondary data from large health surveys, which has dominated medical utilization studies in part because of advantages over primary data collection in terms of time and resource constraints (Mechanic 1989), will not meet the objectives of health outcomes research. General population surveys are structured to provide data on the population at large, and sociodemographic subgroups, rather than on people with specific health conditions, who are usually the focus of health outcomes studies. Furthermore, data to address health outcomes must reflect the health-care experience of individuals over time, and draw on multiple sources, including providers, insurers, patients, and possibly family members or caregivers.
There is potential for increased use of qualitative methods in outcomes research. Qualitative data can provide a fuller and more nuanced depiction of the complexities of individual behavior and motivation than is possible from standardized questionnaires. In addition, survey questions and content for outcomes studies cannot be off-the-shelf since they must be sensitive to access issues for individuals in different organized care settings and with disease or condition-specific service needs. Ethnographic studies can provide guidance with regard to meaningful question wording and content. Such methods could be especially useful in instances where little is known about patient or provider behavior and attitudes. Many examples come to mind, such as the impact of culture or religious beliefs in selecting among established treatment alternatives over the course of chronic diseases such as AIDS or schizophrenia, or the growing attraction for both physicians and patients of nontraditional treatments and natural remedies. Because most studies of health outcomes are done within the medical, public health, and health services research communities, the potential contribution of ethnographic studies is not always recognized. At the same time, it is not clear to what extent social scientists with this type of expertise will be attracted to these issues.
Access to care has been a key measure of health system performance. Financial barriers to access, primarily being uninsured, have been a major focus of research and policy in the United States, but even in societies that have eliminated financial barriers to access, research on noneconomic barriers to access continues. A considerable body of research exists that documents barriers related to personal characteristics and organizational aspects of care. There is no question that ability to pay has a major impact on whether people choose to seek care, but research tells us that once this barrier is removed, others remain. Mechanic posed a question 20 years ago that is still relevant:
Why [do] persons with similar complaints behave so differently and why [does] the same person with comparable symptoms at various times choose to seek medical care on one occasion but not another. (Mechanic 1979: 394)
The effort to generate new access indicators that affect health outcomes in addition to utilization of care, and continued reformulation of access models, indicates that interest in these questions is not waning. Continuing to monitor barriers to access, and who is affected by them, is critical, but a simultaneous effort is needed to understand the mechanisms by which some longstanding and well-documented barriers exert influence. Access research represents the merging of social science research on one aspect of human behavior with health policy and medical concerns about providing care to those in need. The expertise of both will be needed to address the more complex questions that lie ahead.