Emergency Department Utilization, Hospital ...

Copyright 1998 by The Cerontological Society of America The Cerontologist Vol. 38, No. 1, 25-36

This study uses a theoretical model of health services utilization to examine (a) emergency department utilization, (b) hospital admissions, and (c) office-based physician visits among a sample of 998 low-income elderly African American persons. Poisson Regression analysis was used to estimate the parameters specified in the Andersen behavioral model. Some of the more interesting results include the following: (a) a greater frequency of emergency room visits among respondents with a lower level of accessibility to physician services, (b) a lack of a significant relationship between some chronic illnesses such as diabetes and heart problems and the frequency of office-based physician visits, (c) a greater number of hospital admissions among insured persons, and (d) a significant impact of the health locus of control indexes on all three types of health care utilization. The results of this study challenge the assumption that hospital and emergency use are the results of nondiscretionary behavior. Key Words: African American, Elderly, Health care utilization

Mohsen Bazargan, PhD,2 Shahrzad Bazargan, PhD,3 and Richard S. Baker, MD' derstanding of factors that account for health care utilization should provide health care planners and providers with more appropriate strategies to facilitate both more effective utilization of health services and promotion of health status of elderly

Findings of empirical studies document the existence of substantial unmet medical needs among African Americans (Blendon, Aike, Freeman, & Corey, 1989). Previous studies indicate that elderly African Americans tend to underutilize services that could enhance their health status and quality of life (Mui & Burnette, 1994). The physical quality of life for elderly African American persons is poor, and this poverty of physical well-being is related directly to the inferior quality of health care received by this segment of our population (Jones, 1986). Although, relative to their needs, elderly African Americans are identified both qualitatively and quantitatively as underutilizers of medical care (Kart, 1991; Mui & Burnette, 1994; Reed, 1990; Wolinsky et al., 1990), reasons for persistent underutilization are poorly understood (Kulys, 1990). More information on factors that predict the use of medical services among the elderly minority is needed (Anderson & Cohen, 1989). In particular, what still remains to be answered is the differential effect of determinants on health services utilization by the noninstitutionalized elderly living in poor urban neighborhoods (Petchers & Milligan, 1988; Wan, 1982). A better un-

African Americans (Edmonds, 1990). A recent National Hospital Ambulatory Medical Care Survey indicates that compared to their White counterparts, elderly African Americans display a higher rate of visits to emergency departments and outpatient hospital clinics and a lower rate of officebased physician visits (McCaig, 1994a, 1994b; Schappert, 1993). Research on emergency-room utilization has focused primarily on the White population and those age 65 and younger (White-Means, 1995). However, some preliminary evidence on emergency department utilization among elderly African Americans (based on a small sample of patients) indicates that attitude about formal medical services and levels of access to quality medical services are significant factors in the emergency-room utilization (White-Means, 1995). In addition, there are some indications that elderly African American patients resort to emergency rooms as a source of preventive health care, chronic-condition monitoring, and prescription medicine (Craig, 1991). Recent data indicate that over 40% of all emergency department utilization by elderly African American persons has been identified as nonurgent (McCaig, 1994a).

1 The research reported in this article was supported by National Institutes of Health Grant G12 RRO 3026-09, by the National Center for Research Resources, NIH/NCRR/RCMI. We would like to thank Dr. Joel Swartz for his invaluable advice and help on statistical analysis. 'Address correspondence to Mohsen Bazargan, PhD, Charles Drew University of Medicine and Science, Research Centers in Minority Institutions and Department of Family Medicine, 1621 East 120th Street, Los Angeles, CA 90059. E-mail: mobazarg.cdrewu.edu 'Department of Psychiatry, Charles R. Drew University of Medicine and Science, Los Angeles. 'Research Centers in Minority Institutions, Program Director, Charles R. Drew University of Medicine and Science, Los Angeles.

Vol. 38, No. 1,1998

Hospitalization has been generally conceptualized as a nondiscretionary service, implying that the hospitalization is most likely to be up to the judgment of the physician (Mutran & Ferraro, 1988; Wolinsky et al., 1990). However, compared to their White coun25

Downloaded from http://gerontologist.oxfordjournals.org/ by guest on January 9, 2017

Emergency Department Utilization, Hospital Admissions, and Physician Visits Among Elderly African American Persons1

Methods

Data The entire study sample was composed of 1,114 elderly African Americans, aged 62 years and older, who resided in New Orleans, Louisiana, and were living independently. They were randomly selected from 23 senior citizen centers in the city. The sample was drawn from the service lists provided by the senior citizen center managers or directors. The participants were notified by a personal letter describing the nature of the study and selection procedures for inclusion of participants. To encourage voluntary participation, we offered each potential participant remuneration. A few days after personal letters were mailed, selected participants were contacted by telephone and solicited for participation. For those selected participants whose telephones were not listed (n = 70; 6.28%), door-to-door solicitation was utilized. During these solicitation contacts, once again, the study was briefly described and all questions regarding the research were explained. Appointments for survey interviews were made with all participants who agreed to participate in this study. Face-to-face interviews were subsequently conducted in the home of all consenting participants by trained interviewers at designated appointed times. The data were collected between February 1992 and July 1993. The refusal rate for both telephone and door-to-door solicitation was almost the same, and only 11.5% of selected participants did not agree to participate in this study. A total of 998 (out of 1,114) interviews were completed successfully. All interviews were conducted by trained African American interviewers.

Model The health services utilization framework developed by Andersen and his colleagues (Aday & Andersen, 1974; Aday, Fleming, & Andersen, 1984; Andersen, 1995; Andersen & Newman, 1973) has been widely used for investigating health services utilization by the elderly (Eve, 1988; Mui & Burnette, 1994; Nelson, 1993; Wolinsky, 1994; Wolinsky & Coe, 1984; Wolinsky, Cutler, Callahan, & Johnson, 1994; Wolinsky & Johnson, 1991; Wolinsky, Stumps, & Johnson, 1995). This model conceptualizes health care utilization as the end product of a complex pattern of interactions between predisposing, enabling, and need-for-care characteristics. (For a detailed, historical review and its application to the special case of older adults, see Wolinsky, 1990.) Predisposing characteristics exist prior to the onset of illness and include those characteristics that describe the propensity of individuals to use health care services. These propensities are characterized by demographic, social structural, and health belief variables. The second component of the behavioral framework, the enabling characteristics, refers to the individual's ability to use health care services should the need arise. The enabling component contains familial and community resources that make health services available to the individual for use. Having private health insurance, a Medicaid and Medicare card, tangible support, availability and accessibility

Measurement Table 1 displays the means, percentages, and standard deviations of the variables used in this

26

The Gerontologist


of physicians, and residential stability are some examples of enabling factors (Wolinsky & Johnson, 1991). Although the predisposing and enabling components are necessary conditions for use of health services, they are not sufficient. To use health care services, the individual must perceive some illness. Need-for-care characteristics are the most immediate cause of the use of health services. It involves both perceived and evaluated health status. Measures of perceived illness may include symptoms individuals experience, a general self-report health status, and functional disability. Evaluated health measures are actual health problems that the individual is experiencing and those that have been clinically identified or judged by health practitioners (Andersen & Newman, 1973). Andersen (1995) argued that logical expectations of the model are that perceived need will better help us to understand care-seeking behavior, whereas evaluated need will be more closely related to the type and amount of treatment that will be provided after a patient has been presented to a medical care provider.

terparts, elderly African Americans are less likely to be hospitalized (Mutran & Ferraro, 1988; Wan, 1987). Moreover, when they are admitted to hospitals, they manifest significantly more severe symptoms of illness (Buckle, Horn, Oates, & Abbey, 1992). Once hospitalized, elderly African Americans have longer average stays (Johnson, Gibson, & Luckey, 1990) and have higher hospital fatality rates than their White counterparts (Edmonds, 1990). Despite the significant evidence of the importance of ethnicity on hospital utilization, less attention has been given to its predictors within elderly minority populations. A need for more gerontological research focusing on health status, health care needs, and health service utilization among elderly African American persons has been realized (Anderson & Cohen, 1989; Edmonds, 1990; Richardson, 1992; Wolinsky et al., 1990; Wykle & Kaskel, 1991). Whereas a few researchers suggest more systematic investigations of correlates of health care utilization among minority populations (Mui & Burnette, 1994), others argue that indepth community studies of high-risk populations such as older frail minorities can generate a greater understanding of the elements that facilitate or inhibit services utilization (Petchers & Milligan, 1988). This study uses a theoretical model of health services utilization to examine the emergency department utilization, hospital admissions, and officebased physician visits among a sample of lowincome, urban elderly African American persons. A description of the model, its major components, and its major assumptions follows.

Variable Predisposing Characteristics Age Gender (Male) Education Living arrangement (Alone) Internal health locus of control Chance health locus of control Powerful others externality health locus of control

M (%)

SD

72.41 (23.9) 8.71 (47.7) 27.91 24.97

7.277

— 3.436

— 6.379 7.171

24.71

7.999

Enabling Characteristics Perceived tangible support Availability/accessibility of medical doctors Private insurance (Yes) Medicare (Yes) Medicaid (Yes) Residential stability (Yes)

18.27 2.24 (44.8) (87.7) (31.0) (77.7)

5.558 .995

Need Characteristics Perceived health status Heart conditions Hypertension Diabetes Arthritis Breathing problem Kidney problems Stomach problems Teeth problems Eye problems Ear problems Blood circulation problems Stroke Cancer Limitation of daily activities

2.07 (33.1) (61.7) (23.7) (65.8) (16.0) (6.6) (15.4) (10.2) (46.7) (11.4) (23.5) (10.5) (5.7) 47.33

.862

Health Services Utilization (in past 6 months) Number of hospitalizations None 1 2 or more Number of physician visits None 1 2 3-4 5-6 6 or more Number of Emergency Room Visits None 1 2 3 or more

.257 (81.2) (14.6) (4.2) 1.31 (62.5) (12.5) (7.8) (7.2) (6.9) (3.1)

.39 (77.5) (14.1) (4.5) (3.9)

— — — -

— — — — — — — — — — — — _ 4.840

Enabling Characteristics.—There are six measures of enabling characteristics. Three of them address the issue of health insurance. The first variable is a dichotomous indicator of whether the participant has private (excluding Medicare and Medicaid) insurance. The next two measures are also dichotomous indicators and reflect whether the respondent has a valid Medicaid and Medicare card, respectively. The fourth measure taps residential stability. Snider (1980) argued that geographically stable persons are more likely to be aware of the health services available in their community and to have established relationships with health care providers. Those individuals who have lived at the same address for 5 or more years are considered to be geographically stable (Wolinsky & Johnson, 1991). The next measure, instrumental support, is measured with a modified version of the inventory of social support behaviors (Barrera, Sandier, & Ramsay, 1981; Krause, 1986). Instrumental support consists of nine items and is the result of instrumental behavior that helps a person directly. The helping person intervenes by providing transportation, making a financial contribution, helping with work obligations, or providing some other form of material aid or direct relief (Krause, 1986). Scores on this scale could range from 9 to 36 (M = 18.27; SD = 5.56). A higher score on this indicator reflects greater instrumental support. The alpha reliability of this composite is .75.

.637

— — — 2.773

— — — — — — .935

— _ — -

study to evaluate the predictors of emergency department utilization, hospital admissions, and physician visits among our sample of elderly African American persons. Predisposing Characteristics.—There are seven measures of predisposing characteristics: age, sex, education, living arrangement, and three measures of health locus of control. Sex and living arrangement are coded dichotomously, with "male" and "living alone" coded as 1 and "female" and "living Vol. 38, No. 1,1998

27


with others" coded as 0. Education and age are measured in actual number of years. Multidimensional Health Locus of Control Scales (Wallston, Wallston, & DeVellis, 1978) were used to assess the health locus of control beliefs. These scales consist of three subscales, six items each, all pertaining to the maintenance of health. The three subscales are internal health locus of control, chance health locus of control, and powerful others externality health locus of control. The scales are presented in a 6-point Likert format ranging from 1 (strongly disagree) to 6 (strongly agree), and scores on these subscales could range from 6 to 36. The internal health locus of control subscale consists of questions that assess the extent to which individuals believe they are responsible for their health, the extent to which individuals can avoid behavior that increases the risk of disease, and the extent to which they should play an active role in coping with their illness. The reliability of this indicator was marginal: Cronbach's alpha = .683. The chance health locus of control subscale items focus on the person's views regarding his or her health and sickness as a function of external forces such as luck, accident, good fortune, and so on. The reliability estimate for this indicator was again borderline: alpha = .688. Finally, the powerful others health locus of control subscale consists of questions that assess the extent to which individuals believe that powerful others, particularly physicians, nurses, and other health professionals, are responsible for their health and illness. The alpha reliability of this composite was .776.

Table 1 . Summary of Health Care-Utilization Distribution and Other Sample Characteristics (/V = 998)

variables are shown in Table 1. The dependent variables are count measures and are extremely skewed. A Poisson regression analysis was used to estimate the parameters specified in the behavioral model. A Poisson model was chosen because it is a natural probability model for count variables, in contrast to the continuous, unbounded dependent variable assumed by standard linear regression (Hamilton, 1990). The hierarchical approach in which predisposing, enabling, and need characteristics are sequentially entered into the model used to examine the number of emergency room visits, hospital admissions, and doctor visits.

The last measure of enabling characteristics, perceived availability and accessibility of physician services is measured by a 4-item index: (1) "Would you say you have to go through too much trouble in order to see a doctor?" (2) "When you go to see a doctor, would you say you usually have to wait too long to see her/him?" (3) "Do you think it is easy for you to go to a doctor during the hours doctors are in their offices?" And (4) "If you need medical help at night, do you think it is easy to get a doctor to come to your home?" The respondent's score on this 4-item composite index could range from 0 to 4 with a higher score indicating a higher level of perceived availability and accessibility of physician services (M = 2.24; SD = 0.995). The apparent reliability of this 4-item composite is low: alpha = 0.546. This study followed the Wolinsky & Johnson (1991) study and did not include the variable of income in Table 1 as a measure of enabling characteristics. As with most surveys, the problem with income data is the large item nonresponses. Wolinsky and Johnson (1991) documented that omitting respondents without income data produced biased estimates of the other parameters, and the most appropriate strategy is to delete income from analysis.

Results

Need-For-Care Characteristic—There are 15 measures of the need for health services. The first variable, self-reported health status, is assessed by asking respondents to rate their general health condition, with response categories of 1 (excellent), 2 (good), 3 (fair), and 4 (poor). The next 13 measures are all dichotomous variables indicating whether or not the respondents have been diagnosed by a physician as having heart disease, hypertension, diabetes, arthritis, respiratory diseases, kidney, teeth, eye, ear, and blood circulation problems, stroke, or cancer. The last measure of the need for health services (limitation of daily activities) is derived from 13 questions asking respondents to report whether any given above-mentioned chronic illnesses limit their daily activities. With response categories of 1 (a great deal), 2 (some), 3 (very little), and 4 (not at

all). This scale is created by computing the unweighted sum of responses for all 13 questions. It is important to note that the limitation of daily activity scores are derived from the chronic condition questions, not ADL/IADL (activities of daily living/instrumental activities of daily living) questions. This would probably underreport limitations in activities where the individual has limitations but does not attribute them to any of the 13 listed conditions. Health Services Utilization.—There are three measures of health services utilization. All of them tap formal services: number of emergency room utilizations, hospital admissions, and physician visits within a 6-month period prior to the interviews. Analysis. — Summary distributions of the dependent variables (emergency room utilizations, hospital admissions, and physician visits within a 6- month period prior to the interviews) and independent 28

The Gerontologist


Characteristics of the Sample The mean age of the sample was 72.4 years; the range was 62 to 99 years. The majority was female (76%). Only 20.9% of the sample were currently married, 54.5% were widowed, and 24.6% were divorced, separated, or never married. The mean education was 8.7 years, with less than 48.9% of the participants indicating that they had no formal education beyond 8th grade. Slightly more than 23.7% reported completing high school. Seventy percent of the sample had a monthly income of $750 or less ($9,000 per year). The average income was $680 per month ($8,160 per year). Over 87.7% and 31.8% of the sample reported being enrolled in Medicare and Medicaid, respectively. Just over 44.8% had private heath insurance. Recent census information has indicated that 61% of the total population of elderly African American, 62 years and older in New Orleans, were female and, similar to our sample of the older elderly African American population, 51% completed less than a 9th grade education, and only 24% completed high school or had post-high school education. Sixty-one percent of the elderly African Americans in New Orleans had a yearly income of $9,999 or less (U.S. Bureau of the Census, 1990). In regard to health status, only 8.4% of the sample rated their present health as excellent. Thirty-two percent described their health as good, and 41.8% and 17.6% rated their health as fair and poor, respectively. Remarkably, similar to our study, recent reports from the Center for Disease Control and Prevention (U.S. Department of Health and Human Services, 1995) indicated that 9.3% and 17.6% of African Americans aged 65 years and older, participating in a 1992 National Health Interview Survey, reported their health status as excellent and poor, respectively (N = 2,622). In all, 7.4% reported no chronic illness. The number of chronic illnesses reported by the remainder of the sample ranged from 1 to 11 with the average number of conditions reported just over 3 (M = 3.3). The most frequently cited chronic illnesses were: arthritis (65.8%), hypertension (61.7%), eye problems (46.7%), heart trouble (33.1%), diabetes (23.7%), and

circulation problems (23.5%). Over 37.7% of participants reported that at least one condition limited their daily activities a great deal. More than 25% of respondents mentioned no limitation in their daily activities at all. Similar to our study, the 1986 Functional Limitations Supplement to the National Health Interview Survey indicated that 71.3% of elderly African Americans reported no difficulty in their daily activities (Blesch & Furner, 1993). Despite the large proportion of the sample with Medicare, 36.6% of the sample reported that "not enough medical care" was a very serious (23.4%) or somewhat serious problem (13.2%) for them personally.

associated with individuals who differ by one unit on the independent variables, and a pseudo-/?2 coefficient (Nagelkerke, 1991). In these hierarchical models, the predisposing characteristics were entered on the first step of the analysis; the enabling and need-for-care characteristics were added on the second and third step, respectively. Emergency Department Utilization.—Table 2 shows that, with respect to emergency department utilization, the predisposing factors accounted for 16.5% of the variance (R2 = .165). The enabling characteristics added another 3.5% to the explained variance (R2 = .20). The introduction of the need-for-care variables, however, brought the explained variance up to 33.6%. Table 2 also illustrates that with all predisposing, enabling, and need-for-care characteristics in the model, only eight registered significant independent impact

Overview Tables 2-4 contain the results from the hierarchical Poisson regressions. Included are the regression coefficients, the ratio of expected number of visits

Model 1 Independent Variable Predisposing Characteristics Age Male Education Lives alone Internal health locus of control Chance health locus of control Powerful other externality health locus of control Enabling Characteristics Perceived tangible support Availability/accessibility of medical doctors Private insurance Medicare Medicaid Residentially stable Need Characteristics Perceived health status Heart conditions Hypertension Diabetes Arthritis Breathing problems Kidney problems Stomach problems Teeth problems Eye problems Ear problems Blood circulation problems Stroke Cancer Limitation of daily activities R2

Model 2

Model 3

Beta

Ratio"

Beta

Ratio

Beta

Ratio3

-0.021 -0.106 0.002 0.171 -0.059 0.036 0.071

0.742** 0.915 1.016 1.187 0.474*** 1.669*** 3.019***

-0.014 0.096 0.019 0.162 -0.053 0.035 0.066

0.814 1.101 1.136 1.176 0.507*** 1.645*** 2.856***

-0.009 0.170 0.030 0.114 -0.031 0.029 0.046

0.877 1.154 1.227 1.121 0.674** 1.520** 2.071***

0.037 -0.117 -0.060 -0.036 0.364 -0.104

1.505*** 0.702* 0.942 0.964 1.439 0.902

0.031 -0.091 -0.029 -0.167 0.165 0.012

1.405** 0.768* 0.972 0.896 1.167 1.010

0.324 0.523 -0.028 -0.023 -0.075 0.029 0.229 -0.187 0.028 0.315 -0.213 0.211 0.099 0.288 -0.026

1.747*** 1.637*** 0.973 0.981 0.931 1.022 1.121 0.874 1.169 1.369** 0.873 1.196 1.062 1.143 0.781 .336

.165

3

.203

Notes: "The expected ratio for the categorical variables is the ratio of the expected number of events for participants with the characteristic to the expected number of events for participants without that characteristic. For example, those individuals with heart conditions, on average, had used the emergency rooms 1.637 times as often as persons without the heart problems. For the interval variables, the expected number of events is calculated for two hypothetical respondents such that one individual is assigned a lower score on a given variable and the other individual is assigned a higher score on the same variable (two standard deviations apart). For example, for the perceived tangible support index (with M = 18.27 and SD = 5.56), the expected number of emergency room utilizations for participants with the value of 12.71 (18.27-5.56; M - 1SD) is 1.405 times the expected number of emergency room utilizations for individuals with a value of 23.83 (18.27 + 5.56) on this index.

*p < .05; **p < .01; ***p < .001.

Vol. 38, No. 1,1998

29


Table 2. Poisson Regression Coefficients and Ratio of Expected Number of Emergency Room Utilizations Obtained From the Hierarchical Analysis of Elderly African American Health Care Utilization (N = 998)

Table 3. Poisson Regression Coefficients and Ratio of Expected Number of Hospital Admissions Obtained From the Hierarchical Analysis of Elderly African American Health Care Utilization (N = 998)

Model 2

Model 1 Independent Variable

Ratio

Beta

Ratio

Beta

Ratio

-0.022 0.024 -0.046 0.010 -0.068 0.027 0.057

0.728* 1.021 0.731* 1.103 0.418*** 1.462* 2.473***

-0.026 0.154 -0.045 0.021 -0.069 0.029 0.054

0.685** 1.167 0.734* 1.022 0.416*** 1.516* 2.387***

-0.015 0.188 -0.035 -0.023 -0.044 0.025 0.028

0.803 1.172 0.785 0.978 0.572*** 1.331 1.570*

0.023 -0.121 0.367 0.242 0.397 -0.192

1.287* 0.788 1.443* 1.274 1.487** 0.826

0.011 0.073 0.354 0.064 0.183 -0.046

1.133 1.155 1.423* 1.043 1.186 0.962

0.170 0.540 0.019 -0.096 -0.086 -0.062 0.192 0.151 0.118 0.024 -0.421 -0.056 0.272 0.622 -0.061

1.340 1.662*** 1.018 0.922 0.921 0.956 1.100 1.115 1.074 1.241 0.865 0.954 1.181 1.465** 0.555** .263

.107

.131

*p