Age differences in the influence of race, SES, and psychiatric morbidity on healthcare utilization and expenditures.

ABSTRACT

We examined how race, socio-economic status (SES), and psychiatric morbidity impact healthcare service utilization and how the impact of these predictors varies for respondents 65 years of age and over compared to those under age 65. We analyze the Household Component of the 1996 Medical Expenditure Panel Survey (MEPS), focusing on 12,303 African American and white adults in the sample. Rates of service utilization and type of service utilization are compared by race, gender, and mental disorder status. Race and psychiatric morbidity differences in costs by type of service utilization are investigated. Negative binomial and OLS regression models are examined by age and gender to discern the impact of race and psychiatric morbidity net of SES. Among respondents under the age of 65, African Americans and respondents with low SES have particular patterns of service utilization-avoiding outpatient care and physician visits. This pattern of utilization yields low costs early in life, however low SES and race are predictive of higher healthcare costs among respondents over age 65. Psychiatric impairment had a strong impact on patterns of service utilization, and its impact on healthcare costs increased with age, particularly for men. Our findings suggest that patterns of health service utilization early in the life course may have a substantial impact on future health care costs, and that alleviation of mental health problems could substantially decrease health expenditures.

**********

Race differences in health service utilization have been shown to influence health care costs, and recent research on the Medicare elderly has demonstrated that race disparities are exacerbated by psychological comorbidity (Husaini et al., 2002, 2003a, 2003b). Psychiatric disorders may strongly influence preventive health behaviors as well as the management of illness, and especially the control of chronic illnesses such as diabetes, hypertension, and cardiovascular diseases. Only a few studies have investigated how race may influence the impact of psychiatric morbidity on service utilization and costs, (Husaini et al., 2000, 2002, 2003a, 2003b; Druss, Rohrbaugh & Rosenheck, 1999; Zhang, Rost & Fortney, 1999). Research on younger adults may find contradictory results, since younger populations are healthier and more able to forego health service utilization--and the costs associated with avoiding preventive health measures and medical treatment will not be evident until later in life. Further, elderly citizens in the United States enjoy a form of universal coverage through Medicare, and health care coverage has a strong influence on health service utilization and associated costs (Carrasquillo, Lantigua & Shea, 2001). Thus, age-specific race differences in patterns of service utilization could have a profound impact on health outcomes and the costs of care. If African Americans avoid regular health service utilization early in life, this could complicate the identification and management of chronic illnesses--thus leading to heavy service utilization and costs when African Americans are elderly.

Our study examined data from the 1996 Medical Expenditure Panel Survey (MEPS) to assess race variations in psychological morbidity, health service utilization, and healthcare costs. First, differences in utilization and medical costs were presented for whites and African Americans with and without mental health disorders. Second, while taking into account the complex sampling design of the MEPS, we estimated negative binomial regression models of health service utilization. Models of health service utilization were estimated separately across age groups (respondents age 1864 are differentiated from those over 65) and by gender, and estimates were compared across models to examine variations in effects by age. Finally, we analyzed ordinary least squares (OLS) regression models for complex sampling designs for healthcare costs by gender and age group--comparing the effects of race, mental disorder, and socioeconomic factors.

PSYCHIATRIC MORBIDITY, RACE, SES, AND HEALTHCARE COSTS

Mental illness was not only a medical problem requiring treatment, it also influences other health related behaviors in ways that can contribute to other illnesses, magnify their impact, and prevent their mitigation. Research has generated divergent findings regarding race differences in the prevalence of mental illness (Williams et al. 1997; Kessler et al., 1994), however research consistently finds higher rates of psychopathology and more serious forms of psychopathology among those with lower socio-economic status (SES) (Dohrenwend, 2000; Miech & Shanahan, 2000; Turner & Lloyd, 1999). Psychiatric morbidity may be especially influential for health outcomes among African Americans and among those with few economic and educational resources. Further, the influence of psychiatric morbidity on health outcomes may be age dependent, with mental illness having a greater impact on physical health later in life.

Mental illness can exacerbate the consequences of physical ailments by hindering adherence to physicians' directives, including failing to schedule or show up for follow up appointments; avoiding routine outpatient treatment; not taking prescribed medications regularly; and ignoring advice regarding diet and lifestyle (Haug & Ory, 1987; Panzarino, 1998). Psychiatric morbidity was often linked with substance abuse, which can lead to hypertension, liver damage, and other physiological problems (McAlpine & Mechanic, 2000; Muntaner, Eaton, Diala & Dohrenwend, 1998; Sharp & Getz 1998; Brennan & Moos, 1996). Depression, in particular, was linked to obesity and overeating, which can contribute to hypertension and cardiovascular disease (Carpenter, Hasin, Allison & Faith, 2000). Several studies have reported that psychiatric morbidity significantly increases rates of utilization of emergency and inpatient services, and that total healthcare costs are nearly twice as high for those with a psychiatric diagnosis (Husaini et al., 2002, 2003a, 2003b; Druss, Rohrbaugh & Rosenheck, 1999).

Race, Social Class, and Service Utilization

Race and social class play a prominent role in determining patterns of health service utilization which influence overall healthcare costs. Previous research suggests that minorities have relatively low health service utilization rates (Hough et al., 1987; Black, Rabins, Geman, Mcguire & Roca, 1997), and notably these studies deal with the general population, rather than focusing on the elderly. Race and social class are seen as barriers to regular treatment and the early detection of disease (Kington & Smith 1997; Phillips, Mayer & Aday 2000; Weinick & Krauss 2000). Minorities and people of low SES are more likely to lack insurance coverage, and are more likely to require public assistance for health care (Monheit & Primoff-Vistnes, 2000). Options for medical care are restricted for those on public assistance, and barriers to service utilization may force people to forgo regular doctors visits and to seek treatment only when illness was severe.

Race and SES have substantial effects on the frequency and type of services utilized. Indeed, Gornick (2000, 2001) provides a convincing demonstration of black-white differences in preventive care, regular physician visits, and early treatment procedures performed on Medicare beneficiaries. Whites have substantially higher rates for procedures like cataract removal, colonoscopy, and angioplasty--procedures that improve health and prevent more serious complications (Gomick, 2001). In contrast, Gornick (2001) found that procedures associated with the failed management of chronic illness are much more often performed on African American Medicare beneficiaries, e.g., lower limb amputation, arteriovenostomy, and excisional debridement of necrotic tissue. Additionally, studies have found that African Americans are more likely to utilize emergency services, suggesting that health problems are not addressed until they are more severe (Bazargan, Bazargan & Baker, 1998; Husaini et al., 2002; Maynard, Ehreth, Cox, Peterson & McGann, 1997).

Race, SES, and Healthcare Costs

Most analyses of costs have focused on important issues of the influence of type of coverage (Carrasquillo et al., 2001; Yelin, Herndorf, Trupin & Sonneborn, 2001), on out of pocket expenses (Rasell & Bernstein, 1995), and on the concentration of expenditures in the population (Berk & Monheit, 2001). Studies examining race differences in overall expenditures are rare, and have been limited to studies of Medicare recipients (Husaini et al., 2000, 2002, 2003a, 2003b), where controls for SES are not possible using HCFA reimbursement data. Further, studies of the elderly cannot probe into the age-specific patterns of health behaviors that might help explain higher healthcare costs among the African American elderly, since they only focus on a particular period of the life course and individuals who survive to that age.. The existing research suggests two important connections relevant for the impact of race and SES on health service utilization and costs: (1) the prevalence of mental illness may vary by race and SES--with minorities and the disadvantaged having higher rates of psychiatric morbidity; and, (2) race, SES, and psychiatric morbidity may be associated with particular patterns of health service utilization--with minorities, the poor, and those with mental disorders having lower rates of utilization of preventive services, and higher rates of utilization of inpatient and emergency services.

These distinctive patterns of service utilization should yield a particular age-specific relationship between race, SES, and healthcare costs. Particularly, we should expect: (1) that race and SES differences in service utilization patterns will be more extreme among younger adults, when illness is less likely to be severe; (2) costs will be lower for younger African Americans and younger respondents who are socioeconomically disadvantaged, because the disadvantaged avoid healthcare services when they are young. Conversely, service utilization and costs for African Americans and the economically disadvantaged will be higher at older ages when undetected and untreated illnesses become more severe. In our examination we will test whether or not race differences in service utilization and costs hold when controls for SES and psychiatric morbidity are employed, and how these effects vary with age.

Data and Methods

We analyzed data from the 1996 Medical Expenditure Panel Survey (MEPS), a nationally representative survey of 10,500 households taken from a subsample of the 1995 National Health Interview Survey (NHIS), and funded by the Agency for Healthcare Research and Quality. We used the Household Component (HC), which provided data on health coverage, service utilization, expenditures, and individual characteristics of respondents. Face-to-face household interviews were conducted in five rounds over 2.5 years. We restricted our analyses to the 12,303 African American and white respondents over the age of 18, because of the relatively small number of other minorities and for clarity regarding comparisons. In the 1996 MEPS, African Americans were oversampled, and comprised 16% of the total sample when other minorities were excluded. Our estimates used the person weights for the MEPS 1996 panel to account for these sampling issues, and we used robust estimation techniques that corrected for complex sampling design and maintained representativeness for the subsample of respondents we examined (Cohen, 1997). The MEPS obtained a remarkably high overall completion rate, with 70 percent of respondents retained including both panel attrition and non-response from the original NHIS.

Measures Healthcare costs

Our ultimate dependent variable was total expenditures for all inpatient, outpatient, emergency, and physician visits. We examined expenditures rather than charges since these reflect the ultimate costs of care that must be paid by providers and patients. We also examined costs associated with five categories of service utilization using the expenditure data: (1) inpatient expenditures; (2) outpatient expenditures; (3)physician expenditures; (4) emergency expenditures; and (5) other expenditures. For brevity, we confined our multivariate analyses to the measure of total expenditures, and this measure also generated a more accurate accounting of the relationships between costs and the varied types of service utilization. Unfortunately, MEPS data did not allow us to separate costs by morbidity type. Thus, we cannot separate mental health and non-mental health expenditures, and this inflated costs for those with psychiatric disorders. Husaini and colleagues (2002) estimated that psychiatric costs were between 6 and 8 percent of total costs for elderly respondents.

Healthcare Service Utilization

We examined five measures of healthcare service utilization. (1) Total number of inpatient nights; (2) number of hospitalizations; (3) number of outpatient visits; (4) number of physician visits; and (5) number of emergency visits.

Psychiatric morbidity

Participants in the MEPS were asked to report if they had been diagnosed with a number of specific psychiatric disorders (e.g., affective psychosis, organic psychosis, depression, paranoid states, etc.). In addition, self-reported psychiatric symptoms were classified into specific diagnostic categories by the MEPS staff using a standardized classification scheme. Specific diagnoses were collapsed into a dummy indicator of mental disorder (1 = mental disorder, 0 = no disorder); 11.5% of respondents reported having been diagnosed with some form of mental illness. This proportion of psychiatric morbidity was similar to the level reported for other populations in previous studies (Kessler et al., 1994).

Socioeconomic Measures

The MEPS collected data on total household income, which was measured in dollars. In our regression results we multiplied unstandardized coefficients for income by 10,000. We also examined the independent effects of educational attainment on health service utilization and costs. We employed an indicator of highest degree completed, which had values: (1) no degree; (2) GED; (3) high school diploma; (4) bachelor's degree; (5) master's degree; and (6) doctoral degree.

Health Coverage

We examined the type of health coverage using four classifications: (1) No insurance; (2) Public Only (e.g. Medicare, Medicaid); (3) Public and Private (e.g. Medicare and supplemental private insurance); and (4) private only. Dummy variables indicated the type of coverage. In the total sample and in the age 18-64 analyses we used no insurance as the comparison category. Since virtually all elderly respondents had Medicare, we used public-only as the comparison category for the analyses of respondents over the age of 65.

Demographics 

We controlled for a variety of demographic factors that might influence health service utilization and costs. First, we compared African Americans with whites using a dummy indicator (1 = African American, 0 = White). Because race and gender interact to influence patterns of health service utilization and costs (Husaini et al., 2002), we also employed dummy indicators for race and gender, comparing White males, African American females, and African American males to white females (the omitted category) in the multivariate models. We examined the impact of age in years using a measure that was truncated at the upper end at age 90 for confidentiality purposes (all respondents 90 and over are coded 90 by MEPS). Rural residence was indicated by a dummy measure identifying those living in non-metropolitan areas. Region was examined, and we used a dummy indicator for Southern in the multivariate models.

Statistical Analyses

We take into account the complex multistage sampling design of the MEPS. Estimates were generated using the STATA statistical analysis program with the PWEIGHT and Robust Cluster options that allowed us to take into account the complex sampling design of the MEPS. This strategy greatly expands range of statistical analyses that can be estimated, while correctly estimating the standard errors and allowing data to be analyzed as subsets of the entire sample (Cohen, 1997). First, we compared the mean levels of service utilization and healthcare costs for African Americans and whites by mental disorder status. Second, we estimated multivariate models of health service utilization using a negative binomial generalization of the Poisson regression model. We chose this model because service utilization was examined as a frequency count, and ancillary analyses detected substantial overdispersion in Poisson models for each type of service utilization, thus making the negative binomial model appropriate (Cameron & Trevedi, 1998). The negative binomial regression models predicted the frequency of service utilization, with the inverse of the natural log of the estimate interpreted as the expected percentage increase or decrease in the frequency of utilization given a one unit increase in the independent variable, while other factors were held constant (Cameron & Trivedi, 1998). We estimated models for the total sample, and for women and men separately. The gender specific models employed a dummy indicator for African Americans, while the total sample models examined differences across race-gender groupings. Third, we used ordinary least squares (OLS) regression models adjusted for complex sampling design for total healthcare expenditures by age group and gender. Finally, using a slope comparison test (Marascuilo & Levin, 1983), we tested the significance of the difference in the slopes across the two age group subsamples to show how the effects of race, SES, and mental disorder vary with age.

Results

Race, Mental Disorder, and Service Utilization

In Exhibit 1, we found that for both African Americans and whites, having a mental disorder dramatically increased all types of service utilization. When compared to those who did not report psychiatric morbidity, respondents with a mental disorder had more nights in the hospital, a higher number of hospitalizations, more outpatient visits, a greater number of physician visits, and nearly double the number of visits to the emergency room (more than double for African Americans). One key race difference in service utilization was also evident in Exhibit 1. African Americans made significantly fewer visits to the physician when compared with whites, and African Americans with a mental disorder had significantly fewer physician visits than whites with a mental disorder.

Psychiatric Morbidity and Service Utilization: Multivariate Models, Age 18-64

As can be seen in Exhibit 2, having a mental disorder had a significant positive impact on each indicator of health service utilization, controlling for other factors. Indeed, the magnitude of the impact of having a mental disorder was among the strongest predictors of all forms of service utilization. Further, our gender specific models showed that the effect of mental disorder on service utilization was substantially stronger among men. The importance of this finding was amplified when one considers that about 11 percent of the sample reported having a psychiatric diagnosis. Having a mental illness increased the expected number of nights spent in the hospital by 260 percent (exp.(1.28) = 3.60) when other factors were held constant, and for men the impact was even stronger, generating a 390 percent increase (exp.(1.59) = 4.9). Controlling for other influences, having a mental disorder roughly doubled the number of visits to the hospital (exp.(0.672 = 1.96), outpatient services (exp.(0.676) = 1.97), and physician visits (exp.(0.749) = 2.11). The impact of mental illness was weaker among women, and stronger among men for inpatient nights, hospitalizations, and physician visits. Having a mental disorder also significantly increased the number of visits to the emergency room, producing a 61 percent increase in emergency service utilization, and we did not find substantial gender differences in this effect.

Race, Gender & Service Utilization:

As can be seen in Exhibit 2, in the younger age group African American men had significantly lower rates of health service utilization when compared to white women for all forms of services with the exception of outpatient services and emergency visits. When SES, health coverage, psychiatric morbidity and other covariates were taken into account, African American men had around half as many inpatient nights (exp.(-0.591 = 0.554), and less than half the number of hospital visits (exp.(-0.938) = 0.391) and physician visits (exp.(-805 = 0.448) when compared to white women. Further, compared to white women, African American women made significantly fewer physician visits, but spent more nights in the hospital. White men had significantly lower rates of service utilization than did white women on two indicators: number of hospitalizations, and physician visits.

SES, Health Coverage, & Service Utilization

As can be seen in Exhibit 2, in the younger age group, income had a negative effect on the number of nights spent in the hospital and the number of hospitalizations. Education had a significant positive impact on the number of visits to a physician, and a significant negative impact on the utilization of emergency services. Gender specific models showed that education had a stronger negative impact on hospitalizations for women. Education had a significant positive impact on outpatient visits for women, which may indicate a pattern of utilization of preventive services. For women, income also had a stronger negative impact on the number of visits to the emergency room than it does for men (exp.(b) of 0.930 for women vs. 0.995 for men). While income may provide healthier environments that allow one to avoid hospitalization, education seemed to motivate forms of service utilization that helped maintain health--regular doctors visits--and avoid costly emergency services.

Compared to those with no medical coverage, respondents with private coverage had significantly higher rates of utilization, on most measures with the exception of inpatient nights and emergency visits where the effect of insurance was negative (Exhibit 2). Compared to those with no coverage, those with private insurance made significantly more visits to a physician, had a significantly higher number of hospitalizations, and made significantly fewer trips to the emergency room, controlling for socioeconomic and other factors. Increased levels of service utilization were found among those with some form of public assistance. Having either public only, or a mix of public and private medical coverage substantially and significantly increased every form of health service utilization. Indeed, when SES and other factors were taken into account, respondents with both private and public coverage were expected to have had more than twelve times the number of nights in the hospital as respondents with no coverage (exp.(2.52) = 12.48). This dramatic difference was partly a function of the tendency for individuals with severe illnesses to have publicly funded disability coverage, along with supplemental private insurance. Further, this effect was driven in part by a dramatic gender difference in the impact of dual public/private coverage on hospitalizations. The exp(b) for public and private versus no coverage for inpatient nights was 27.9 for men, while only 3.8 for women. Still, younger MEPS respondents with only public coverage also had dramatically higher levels of service utilization when compared to those with no coverage, or those with only private insurance. Additionally, across most forms of service utilization, the impact of coverage was larger for men--men's health service utilization was more elastic and dependent on coverage.

Service Utilization: Multivariate Models, Age 65 and Over Psychiatric Morbidity & Service Utilization

Consistent with the results for respondents age 18-64, our analyses showed that rates of service utilization were significantly higher for respondents with a psychiatric disorder for all types of services with the exception of outpatient visits (Exhibit 3). Elderly MEPS respondents with a psychiatric disorder spend more than twice as many nights in the hospital as those with no mental illness (exp.(0.821) = 2.18). Older respondents with a mental disorder had 67% more hospitalization (exp.(0.511) = 1.67), 43 percent more physician visits (exp.(0.361) = 1.43), and 51 percent more visits to emergency facilities (exp.(0.414) = 1.51). Men with a mental disorder had even higher rates of service utilization compared to men with no disorder. This was particularly true for inpatient nights and hospitalizations, where the effect of mental disorder for men was substantially higher than the impact among women.

Race, Gender & Service Utilization:

Service utilization rates were lower for African Americans and especially African American men, in the 18-64 subsample. However, among the older MEPS respondents service utilization rates were higher for African Americans--with the exception of lower numbers of physician visits for African American men (Exhibit 3). Indeed, the gender specific models showed that higher rates of service utilization for African Americans were limited to the female subsample. Exhibit 3 shows that controlling for SES and other factors, African American men spent significantly more nights in the hospital, had more hospitalization, but made significantly fewer more visits to the doctor than white women. African American women used significantly more outpatient services than did their white counterparts. African American women had more outpatient visits than white women. We also found that African American men had significantly fewer doctor visits than white men. These race-gender effects are significantly different when compared to the results for the younger respondents presented in Exhibit 2. Comparing the difference between slopes between Exhibit 2 and Exhibit 3 demonstrated that the difference in the number of inpatient nights between African American men and white women increased significantly with age (t = 2.44). Similarly, the gap in the number of hospitalizations between African American men and white women widened significantly with age, with a t-ratio of 5.15. Comparing the difference between African American women and white women on outpatient service utilization, we also found significant age differences (t = 3.26) with the African American women differing more substantially from white women after age 65. Service utilization for white men also increased significantly in the older age group, particularly for inpatient nights and number of hospitalizations.

SES, Health Coverage, & Service Utilization

Among older MEPS respondents, income continued to have a negative impact on health service utilization (Exhibit 3). Across every type of service, with the exception of outpatient visits, income significantly reduced service utilization. The impact of income on hospitalizations and inpatient nights was stronger among women. The influence of education on outpatient visits became significant and positive in the older group. Income also had a stronger negative influence on outpatient visits for women. In the older subsample, the prevalence of coverage through Medicare made it necessary to use public assisted coverage as the omitted category in the analyses. Some MEPS respondents claimed to have private insurance only, and these were separated from others using a dummy indicator for comparison with respondents who had public coverage only. Once SES and other factors were taken into account, respondents with only Medicare coverage were indistinguishable from respondents who claimed private insurance coverage in their rates of service utilization. Net of other factors, respondents with both Medicare and private insurance had significantly more hospitalizations and made significantly more visits to physicians. The impact of dual coverage was stronger among women for hospitalizations, but stronger among men for physician visits.

Race, Mental Disorder & Healthcare Costs

In Exhibit 4, we examined healthcare costs by race and psychiatric disorder. Respondents with a mental disorder had substantially and significantly higher costs when compared with respondents who did not present psychiatric morbidity. Large standard errors for the relatively small number of African Americans with a mental disorder, combined with skewed costs (Berk & Monheit, 2001), left some of these differences statistically insignificant. However, the patterns were in the expected direction. Inpatient costs were nearly double for respondents with a mental disorder, and total costs revealed a similar difference in magnitude. Whites without a mental disorder averaged $2,346 in healthcare costs, while whites with psychiatric morbidity had total expenditures of $4,283. African Americans without psychiatric morbidity had $1,638 in total expenditures, while those with a mental disorder had average expenditures totaling $3,090.

There were several race related differences in health care costs. First, total costs were significantly lower for African Americans when compared to their white counterparts with or without a mental disorder. A large portion of the race-related differences in total costs was a function of significantly lower expenditures on outpatient and physician services for African Americans. We investigated the relationship between race and healthcare costs further by controlling for SES and other factors, adding the interaction between race and gender, and by dividing the sample by age groups.

HEALTH CARE COSTS: MULTIVARIATE MODELS Psychiatric Morbidity and Healthcare Costs

Ordinary Least Squares (OLS) regression models for total health care expenditures were estimated and test statistics for the differences in slopes across the two age-group subsamples were calculated (Exhibit 5). The contrasts in the service utilization findings were striking when viewed across the age groups. It was clear from the results in Exhibit 5 that psychiatric morbidity had a profound impact on healthcare costs. Expenditures for those who reported a psychiatric disorder were significantly and substantially higher, even after controls for SES, medical coverage, and a host of other factors. On average, MEPS respondents age 18-64 with a psychiatric disorder had $1,466 higher medical expenditures than did respondents who reported no mental illness. Results from respondents age 65 and over revealed that the effect of psychiatric morbidity increased with age. Controlling for other factors, older respondents who reported a mental disorder had health care costs that were $2,868 higher than those who do not present psychiatric illness. The difference in the effect of mental illness on healthcare costs across the two subsamples was significant (t = 2.18), evidencing increasing costs of mental illness over the life course.

Race, Gender, & Healthcare Costs:

The race-gender findings in Exhibit 5 were also revealing. Low levels of service utilization early in the life course determined that healthcare costs for African Americans were significantly lower than costs for white women. However, in the 65-and-over subsample, we saw a reversal in the relationship. Healthcare expenditures for African American women and white men were substantially higher than for white women. The small number of older African Americans prevented their higher costs from being statistically significant. Importantly, our results showed that the effect of race and gender on costs reversed across these age groups for African American men, as indicated by the t-tests comparing coefficients. Here, our findings demonstrated that costs for African American men compared to white women were significantly higher in the over-65 age group when compared to the 18-64 age group. Gender specific models indicated that this finding was driven primarily by differences among women (who were most numerous in the subsample). Among women, the race difference in costs went from negative and significant in the younger age group to positive and significant in the older age group. For men, the finding was less dramatic, with negative cost effects for African Americans in the younger age group and less negative race differences in the older age group.

SES, Health Coverage, & Healthcare Costs

The effect of socioeconomic factors on health expenditures also varied over the life course (Exhibit 5). The negative impact of income on health care costs became much greater in the older subsample. Income was significantly more predictive of health costs among the elderly. Interestingly, net of other factors, the effect of education diminished in the older subsample, though the difference was not significant. Indeed, gender specific models showed that the protective effect of education on health care costs was limited to women in the older age group. The negative impact of education on health expenditures gained magnitude among women, but actually reversed with men in the older age group. Over age 65, education increased health costs for men, but decreased them for women.

We are unable to compare type of coverage effects across the subsamples, since the 65 and over group effectively had universal coverage through Medicare. However, in the full sample and for respondents age 18-64 expenditures were lowest for those who had no insurance coverage (Exhibit 5). Respondents with private insurance, a mix of public and private coverage, and public medical assistance had significantly higher expenditures than did those with no coverage. MEPS respondents with private insurance had the lowest expenditures among those with some form of coverage, and their health care costs were significantly lower than those for respondents with public assistance. Health costs were highest for respondents who had both private and public assistance.

DISCUSSION AND CONCLUSIONS

Our research investigated age differences in the effects of race and SES on health service utilization and costs. While this current research has not uncovered the mechanisms through which race, SES, and psychiatric morbidity influence overall health and ultimately healthcare costs over the life course, it does provide an important starting point for future investigations. Particularly, our findings point to the potential usefulness of using a life course perspective when analyzing the relationship between race, psychiatric morbidity, and health service utilization and costs.

Our study had several important limitations. First, we were unable to track respondents over the life course to determine how patterns of health service utilization early in life influence might influence health outcomes later in the life course. Second, our study could not discern the specific reasons for health service utilization, and thus could not distinguish between preventive health service utilization and care that was received for acute or chronic morbidity. Third, we could not separate service utilization and costs associated with psychiatric morbidity from services and costs for physical ailments.

Importantly, we found that race, SES, and psychiatric morbidity had significant, independent effects on health service utilization and healthcare costs. Psychiatric morbidity should be a key component of any health cost model. Without accounting for the mental condition of patients, prediction of service utilization and costs will be much less accurate. Mental health impacts patterns of service utilization and healthcare costs in similar magnitude to the effect of type of health coverage. Since psychiatric morbidity afflicts more than 10 percent of the population, this produces enormous additional burdens on healthcare services, and potentially exorbitant costs for providers. Importantly, these additional costs are not simply a function of mental health care costs, which are typically only a small fraction of overall healthcare costs (Husaini et al., 2002a). For all adults, those with a mental disorder had higher rates of service utilization, and higher healthcare expenditures. Further, our findings demonstrated that the impact of mental illness on healthcare costs increased significantly with age.

Our results have a number of important policy implications. First, effective treatments for a variety of mental disorders have been developed over the last few decades. Yet, insurance plans, HMO's, Medicare, and Medicaid have all treated mental health issues as an extravagance. The expectation has been that mental health care should be paid for primarily by the beneficiary, thus encouraging a rationing of service utilization from the demand side. Indeed, underpayment for psychiatric treatment has even led physicians to deliberately misdiagnose mental illnesses in order to make sure that patients are able to receive treatment they could otherwise not afford (Rost, Smith, Matthews & Guise, 1994). Yet, the strategy of skimping on payment for mental health problems appears to be backfiring. While new treatments for manic depression, major depressive disorders, schizophrenia, and other mental illnesses are expensive, treating mental illness may well forestall the development of physical illnesses, and increase the effectiveness and efficiency of the management of physical illnesses. Our findings may provide insights relevant to the somewhat contested literature on the cost-offset thesis for the treatment of mental illness (Zhang et al., 1999). While we have not been able to fully test this argument, our results provide a basis of support and a starting point for future research on a life course approach to the cost-offset hypothesis.

Our research has demonstrated that race differences in the patterns and levels of service utilization and race differences in healthcare expenditures are not explained away by socioeconomic factors or differences in the type of healthcare coverage. We found that race matters because it influences age-specific patterns of care. Barriers to healthcare--whatever these may be, above and beyond socioeconomic factors and type of coverage--caused African Americans to significantly underutilize basic health services in adulthood, from age 18-64. Race differences were particularly striking for seeing a physician--which may be the most important way to mitigate the impact of chronic illnesses through early detection and effective management of disease. Because of their underutilization of services, African Americans had lower costs when they are under age 65. However, the race pattern in service utilization reversed after age 65, with African Americans having higher rates of some forms of health service utilization, and growing healthcare costs in late life as a multiplicity of chronic and acute conditions accumulate and require care.

Decreasing barriers to care and increasing service utilization may seem inefficient--perhaps particularly for insurance companies and HMOs who are not responsible for high health care costs later in life when the burden of payment shifts to Medicare. In fact, our study suggests that cost savings when people are young lead to high expenditures later in the life--and this shifts costs firmly onto the shoulders of Medicare. Racial inequities in health service utilization over the life course seem to be increasing the strain on our only stop-gap system of healthcare for the elderly. Indeed, given that most African Americans have private coverage insurance in the early life course (Carrasquillo et al., 2001), this shifts the burden from for-profit providers to the state.

Future studies should focus on qualitative differences in service utilization over the life course. Studies should investigate how race and psychiatric morbidity influence the likelihood of obtaining particular types of preventive services such as mammograms, blood pressure screening, and prostate cancer screening. Further, these patterns of early detection and intervention may be linked to future expenditures on diseases that can be treated effectively and efficiently when detected early. Additionally, studies should also examine the impact of mental illness on the life-long management of chronic diseases like diabetes and cardiovascular disease. Accomplishing this will require tracking large numbers of respondents in panel studies over long periods of time, and systematically assessing their patterns of service utilization and health expenditures. Of course, a focus on minority populations will also require considerably larger sample sizes, in order to garner the statistical power needed to assess differences across types of service utilization and particular forms of morbidity (Ferraro & Wilmoth, 2000). Tracking respondents over a longer period of the life course will also allow a better assessment of the effects of changes in insurance and federal policies on specific utilization and health outcomes. Further, by examining life histories of care and disease, researchers could more accurately assess the offsetting costs of the treatment of mental illness and regular access to healthcare.

Future research along these lines could help healthcare administrators and policymakers predict the demand for health services and the costs of care more accurately. Further, by discerning points of need and targeting resources into particular communities and age groups, administrators and policymakers may be able to reduce expenditures while simultaneously improving public health. Healthcare administrators and policymakers should work in tandem with healthcare providers to help reduce long term costs associated with the underutilization of preventive health services. Further, health administrators should be aware that while frequent physician visits might increase the short term costs of management of chronic illnesses, long term savings are likely.

EXHIBIT 1

MEAN SERVICE UTILIZATION BY RACE AND MENTAL DISORDER
FULL SAMPLE 1996 MEPS

                                WHITE             AFRICAN AMERICAN

                                   Mental
Service Utilization     No         Disorder       No
                        Disorder                  Disorder

Inpatient nights        0.462      1.323 (a)(a)   0.450 (b)(b)
No. Hospitalizations    0.102      0.202 (a)(a)   0.090
Outpatient Visits       0.597      1.003 (a)(a)   0.524 (b)
Physician Visits        3.800      7.744 (a)(a)   2.875 (a)(a)(b)(b)
Emergency Visits        0.160      0.266 (a)(a)   0.159 (b)(b)

Service Utilization     Mental
                        Disorder

Inpatient nights        1.500 (a)(a)(c)(c)
No. Hospitalizations    0.141
Outpatient Visits       0.902
Physician Visits        5.936 (a)(a)(b)(b)(c)(c)
Emergency Visits        0.350 (a)(a)

Notes:
a = difference from white with no disorder, significant at 0.05
level, two tailed

as = difference from white with no disorder, significant at
the 0.01 level, two tailed

b = difference from white with mental disorder, significant
at 0.05 level, two tailed

bb = difference from white with mental disorder, significant
at the 0.01 level, two tailed

c = difference from African American with no disorder, significant
at 0.05 level, two tailed

cc = difference from African American with no disorder, significant
at the 0.01 level, two tailed

EXHIBIT 2

NEGATIVE BINOMIAL REGRESSION MODELS OF HEALTH SERVICE UTILIZATION
AGES 18-64: 1996
(N = 9,660: 5,069 FEMALE, & 4,591 MALE)

                 Inpatient Nights         No.
                                          Hospitalizations

                 B          Exp (B)       B          Exp (B)

Income            -0.100     0.904 (b)     -0.077     0.926 (c)
  (female)       (-0.086)   (0.918)       (-0.096)   (0.908) (c)
  [male]         [-0.132]   (0.876) (a)   [-0.067]   [0.935] (a)
Education         -0.059     0.942         -0.059     0.943
  (female)        -0.003    (1.00)        (-0.078)   (0.924) (c)
  [male)         [-0.110]   [0.896]       [-0.063]   [0.938]
Age                0.035     1.03 (c)       0.012     1.01 (b)
  (female)        -0.012    (1.01)        (-0.006)   -0.994
  [male]          [0.077]   [1.08] (c)     [0.050]   [1.05] (c)
Private Only       0.098     1.10           0.209     1.23 (a)
  (female)        -0.082    (1.09)         (0.195)   (1.22)
  [male]          [0.146]   [1.16]         [0.334]   [1.40]
Public/Private     2.52      12.48 (c)      1.08      2.95 (c)
 (female)         (1.29)     (3.63) (c)    (0.973)   (2.65) (c)
  [male]          [3.33]    [27.91 (c)     [1.21]    [3.35] (c)
Public Only        1.31       3.71 (c)      1.18      3.25 (c)
  (female)        (1.27)     (3.56) (c)    (1.03)    (2.80) (c)
  [male]          [1.57]     [4.81] (c)    [1.32]    [3.74]

Rural             -0.034      0.967         0.009      1.01
  (female)        (0.055)    (0.946)      (-0.148)    (0.862)
  [male]         [-0.092]    [0.912]       [0.208]    [1.23] (b)
South              0.024      1.02          0.094      1.10
  (female)        (0.183)    (1.20)        (0.124)    (1.13)
  [male]         [-0.304]    [0.737]      [-0.013]    [0.987]
Black Male        -0.591      0.554 (a)    -0.938      0.391 (c)
Black Female       0.439      1.55 (a)     -0.010      0.990
White Male        -0.072      0.931        -0.318      0.727 (c)
Black
  (female)        -0.504     (1.66) (a)    (-0.013)   (0.987)
  [male)         [-0.579]    [0.560]       [-0.619]   [0.538] (a)
Mental             1.28       3.60 (c)       0.672     1.961 (c)
Disorder          (0.936)    (2.55) (c)     (0.441)   (1.55) (c)
  (female)        [1.59]     [4.90] (c)     [1.10]    [3.03] (c)
  [male]

                 Outpatient Visits         Physician Visits

                 B          Exp (B)         B          Exp (B)

Income            -0.023      0.977         -0.009     0.990
  (female)       (-0.017)    (0.983)        (0.001)   (1.00)
  [male]         [-0.029]    [0.971]       [-0.018]   [0.982]
Education          0.006      1.01           0.019     1.02 (a)
  (female)        (0.072)    (1.07) (a)     (0.012)   (1.01)
  [male)          [0.099]    [0.906] (b)    [0.013]   [1.01]
Age                0.041      1.04 (c)       0.021     1.02 (c)
  (female)        (0.033)    (1.03) (c)     (0.013)   (1.01) (c)
  [male]          [0.056]    [1.05] (c)     [0.032]   [1.03] (c)
Private Only      0.290       1.34           0.346     1.41 (c)
  (female)       (0.294)     (1.34)         (0.347)   (1.41) (c)
  [male]         [0.333]     [1.40]         [0.367]   [1.44] (c)
Public/Private    0.755       2.13 (c)       0.813     2.05 (c)
 (female)        (0.897)     (2.45) (c)     (0.923)   (2.51) (c)
  [male]         [0.594]     [1.81] (c)     [0.733]   [2.08] (c)
Public Only       1.17        3.22 (c)       0.849     2.34 (C)
  (female)       (1.11)      (3.03) (c)     (0.720)   (2.05) (c)
  [male]         [1.39]      [4.01] (c)     [0.9901   [2.69] (c)

Rural             0.065       1.07          -0.157     0.855 (c)
  (female)       (0.261)     (1.29) (a)    (-0.128)   (0.880) (b)
  [male]        [-0.204]     [0.815]       [-0.187]   [0.829] (c)
South            -0.466       0.628 (c)      0.020     1.02
  (female)      (-0.425)     (0.653) (c)   (-0.009)   (0.991)
  [male]        [-0.547]     [0.579] (b)    [0.056]   [1.05]
Black Male       -0.430       0.651         -0.805     0.448 (c)
Black Female     -0.285       0.753         -0.300     0.741 (c)
White Male       -0.272       0.762         -0.526     0.591 (c)
Black
  (female)      (-0.235)     (0.790)       (-0.277)   (0.758) (b)
  [male)        [-0.217]     [0.804]       [-0.313]   [0.731] (c)
Mental            0.676       1.97 (c)       0.749     2.11 (c)
Disorder         (0.627)     (1.87) (c)     (0.599)   (1.82) (c)
  (female)       [0.702]     [2.02] (c)     [0.992]   [2.70] (c)
  [male]

                 Emergency Visits

                 B          Exp (B)

Income            -0.024     0.976
  (female)       (-0.073)   (0.930) (a)
  [male]         [-0.005]   [0.995]
Education          -.0101    0.904 (c)
  (female)       (-0.097)   (0.908)
  [male)         [-0.101]   [0.904] (a)
Age               -0.011     0.989 (c)
  (female)       (-0.010)   (0.990) (c)
  [male]         [-0.013]   [0.987] (c)
Private Only      -0.232     0.793 (c)
  (female)       (-0.222)   (0.801) (c)
  [male]         [-0.224]   [0.799] (c)
Public/Private     0.420     1.52 (c)
 (female)         (0.601)   (1.82) (c)
  [male]          [0.105]   [1.11]
Public Only        0.688     1.99 (c)
  (female)        (0.701)   (2.02) (c)
  [male]          [0.626]   [1.87] (c)

Rural              0.214     1.24 (c)
  (female)        (0.206)   (1.22) (c)
  [male]          [0.241]   [1.27] (c)
South             -0.091     0.918
  (female)        -0.048    (1.05)
  [male]         [-0.268]   [0.764] (b)
Black Male        -0.268     0.765
Black Female      -0.076     0.927
White Male        -0.044     0.957
Black
  (female)       (-0.171)   (0.842) (a)
  [male)         [-0.110]   [0.896]
Mental             0.477     1.61 (c)
Disorder          (0.474)   (1.61) (c)
  (female)        [0.493]   [1.64] (c)
  [male]

Notes:

(a) = significant at the 0.05 level, two tailed
(b) = significant at the 0.01 level, two tailed
(c = significant at the 0.001 level, two tailed

EXHIBIT 3
NEGATIVE BINOMIAL REGRESSION MODELS OF HEALTH SERVICE UTILIZATION
AGES 65+: 1996 MEPS
(N=2,189 TOTAL, 1285 FEMALE, 904 MALE)

                        Inpatient Nights        # Hospitatizations
                        B         Exp (B)         B         Exp (B)

Income               -0.18        0.835 (c)      -0.100      0.904 (b)
  (female)          (-0.206)     (0.813) (a)    (-0.181)    (0.834) (a)
  [male]            [-0.138]     [0.871]        [-0.057]    (0.945)

Education             0.020        1.02        -0.022        0.978
  (female)          (-0.045)      (1.05)      (-0.010)      (0.990)
  (male)            [-0.050]      [0.951]     [-0.023]      [0.977]

Age                   0.030       1.04 (b)      0.022        1.02 (a)
  (female)          (-0.042)     (1.04) (a)    (0.032)      (1.03) (a)
  (male)             [0.041]     [1.04] (c)    [0.011]      [1.01]

Private Only          0.071        1.19        -0.011        0.989
  (female)          (-0.901)      (2.46)       (0.302)      (1.35)
  (male)            [-0.154]     [0.214] b    [-0.459]      [0.632]

Public/Private        0.088        1.09         0.254       1.29 (a)
  (female)          (-0.289)      (1.34)       (0.522)     (1.69) (c)
  (male)            [-0.139]      [.870]       [0.026]      [1.03]

Rural                 0.205        1.23         0.282       1.33 (a)
  (female)          (-0.081)      (1.08)       (0.212)     (1.24)
  (male)             [0.462]      [1.59]       [0.344]     [1.41] (a)

South                -0.351       0.704 (a)    -0.239       0.787 (b)
  (female)          (-0.269)     (0.764)      (-0.290)     [(0.748) (a)
  (male)            [-0.566]     [0.568] (c)  [-0.189]     [0.827] (a)

Black Male            0.566       1.76 (a)     0.453        1.57 (a)
Black Female          0.344       1.41        -0.018        0.982
White Male            0.576       1.77 (b)     0.508        1.66 (c)

Black
  (female)           (0.397)      (1.49)       (0.046)      (1.05)
  (male)             [0.093]      [1.10]      [-0.136]      [0.873]

Mental                0.821       2.18 (c)      0.511       1.67 (c)
Disorder             (0.481)     (1.62) (a)    (0.330)     (1.39) (b)
  (female)           [1.42]      [4.14] (c)    [0.738]     [2.09] (c)
  (male)

                       Out patient Visits        Physician Visits
                        B         Exp (B)         B         Exp (B)

Income               -0.080        0.923       -0.035       0.965 (c)
  (female)          (-0.109)      (0.897)     (-0.023)     (0.977) (a)
  [male]             [-.035]      [0.966]     [-0.044]     [0.957] (c)

Education             0.140       1.15 (b)      0.026        1.03
  (female)           (0.197)     (1.22) (b)    (0.019)      (1.02)
  (male)             [0.030]      [1.03]       [0.035]      [1.04]

Age                  -0.021       0.979 (b)     0.009       1.01 (b)
  (female)          (-0.022)      (0.978)      (0.004)     (1.00)
  (male)            [-0.0161]     [0.984]      [0.017]     [1.02] (c)

Private Only          0.046        1.05         0.156        1.17
  (female)           (0.347)     (1.41) (a)    (0.197)     (1.22) (b)
  (male)            [-0.417]     [0.659] (c)   [0.099]      [11.10]

Public/Private        0.094        1.10         0.174       1.19 (c)
  (female)           (0.119)      (1.13)       (0.144)     (1.15) (a)
  (male)             [0.001]      [1.00]       [0.217]     [1.24] (c)

Rural                 0.592       1.81 (a)        0.037        1.04
  (female)           (0.699)     (2.01) (a)      (0.053)      (1.05)
  (male)             [0.307]     [1.36]          [0.039]      [1.04]

South                -0.242       0.785 (a)    -0.051        0.950
  (female)          (-0.156)      (0.856)     (-0.062)      (0.940)
  (male)            [-0.336]     [0.715] (a)  [-0.042]      [0.959]

Black Male            0.148        1.16        -0.383       0.681 (c)
Black Female          0.683        1.98 (b)     0.113        1.11
White Male            0.204        1.23         0.035        1.04

Black
  (female)           (0.668)     (1.95) (c)    (0.094)      (1.10)
  (male)            [-0.152]      10.859]     [-0.385]     [0.680] (c)

Mental                0.016        1.02         0.361       1.43 (c)
Disorder             (0.023)      (1.02)       (0.322)     (1.38) (c)
  (female)           [0.089]      [1.09]       [0.449]     [1.57] (c)
  (male)

                       Emergency Visits
                        B         Exp (B)

Income               -0.172       0.842 (c)
  (female)          (-0.132)     (0.876) (b)
  [male]             [0.204)     [1.23] (c)

Education            -0.003        0.970
  (female)           (0.024)      (1.02)
  (male)            [-0.054]      [0.947]

Age                   0.009       1.01 (b)
  (female)           (0.030)      (1.03)
  (male)             [0.030]     [1.03] (c)

Private Only         -0.169        0.844
  (female)           (0.177)      (1.19)
  (male)             [-1.15]      [0.316]

Public/Private        0.124        1.13
  (female)           (0.093)      (1.10)
  (male)             [0.166)      [1.18]

Rural                 0.070        1.07
  (female)           (0.112)      (1.12)
  (male)             [0.018]      [1.02]

South                -0.256        0.774
  (female)          (-0.224)      (0.799)
  (male)            [-0.319]      10.727]

Black Male            0.153        1.17
Black Female         -0.246        0.781
White Male            0.012        1.01

Black
  (female)          (-0.214)      (0.807)
  (male)             [0.083]      [1.09]

Mental                0.414       1.51 (b)
Disorder             (0.384)     (1.47) (c)
  (female)           [0.516]      11.68]
  (male)

Notes:

a = significant at the 0.05 level, two tailed

b = significant at the 0.01 level, two tailed

c = significant at the 0.001 level, two tailed

EXHIBIT 4 MEAN HEALTHCARE COSTS BY RACE AND MENTAL DISORDER
FULL SAMPLE 1996 MEPS

                                   White
                              No            Mental
Healthcare Costs           Disorder         Disorder

Inpatient Expenditures     826.98           1,500.89 aa
OutpatientExpenditures     343.68           457.71 a
Physician Expenditures     547.97           926.13 aa
EmergencyExpend tures      63.62            102.94 aa
Other Expenditures         563.43           1295.00 aa
Total Expenditures         2,346            4,282.67 aa

                              African American
                              No            Mental
Healthcare Costs           Disorder         Disorder

Inpatient Expenditures     620.65 abb       951.43
OutpatientExpenditures     175.51 abb       272.59
Physician Expenditures     341.25 aabb      544.26 bb
EmergencyExpend tures      52.08 aabb       53.24 b
Other Expenditures         448.69 bb        1,268.91 aacc
Total Expenditures         1,638.18 aabb    3,090.43 aabcc

Notes:

a = difference from white with no disorder, significant at
0.05 level, two tailed

aa - difference from white with no disorder, significant at
the 0.01 level, two tailed

b = difference from white with mental disorder, significant
at 0.05 level, two tailed

bb - difference from white with mental disorder, significant
at the 0.01 level, two tailed

c = difference from African American with no disorder,
significant at 0.05 level, two tailed

c = difference from African American with no disorder, significant
at the 0.01 level, two tailed

EXHIBIT 5
OLS REGRESSION MODELS FOR TOTAL HEALTHCARE EXPENDITURES

                             Ages 18-64
                          (Adult)
                            B               [beta]

Income                   -53.76            -0.015 (a)
  (female)              (-48.71)          (-0.014)
  [male]                 [50.46]          [-0.013]
Education               -188.89            -0.032 (c)
  (female)              (-92.76)          (-0.019)
  [male]               [-280.97]          [-0.041] (a)
Age                       38.09             0.054 a
  (female)               (32.97)           (0.057) (c)
  [male]                 [44.79]           [0.054] (a)
Private Ins. Only        868.99             0.045 (c)
  (female)              (961.77)           (0.060) (c)
  [male]                [758.75]           [0.034] (c)
Public/Private          6045.62             0.113 (c)
  (female)            (2,559.85)           (0.060) (c)
  [male]             [10,057.97]           [0.159] (a)
Public Only            2,351.19             0.070 (c)
  (female)            (1,976.30)           (0.077) (c)
  [male]              [2,944.66]           [0.069] (c)
Rural                   -299.49            -0.014
  (female)             (-264.69)          (-0.015)
  [male]               [-413.51]          [-0.017]
South                   -160.64            -0.009
  (female)              (-56.59)          (-0.004)
  [male]               [-233.93]          [-0.011]
Black Male            -1,100.64            -0.031 (c)
Black Female            -661.43            -0.020 (c)
White Male              -168.80            -0.010
Black
  (female)             (-497.35)          (-0.025) (a)
  [male]             [-1,091.59]          [-0.037] (c)
Mental Disorder         1465.81             0.056 (c)
  (female)             (1482.48)           (0.076) (c)
  [male]               [1474.96]           [0.043] (c)
[R.sup.2]                  0.025
  (female)                (0.019)
  [male]                   [.036]

                                                            Adult vs.
                        Ages 65 +                           Elderly
                        (Elderly)                           t-test
                           B               [beta]

Income                  -313.54            -0.047 (c)       3.582
  (female)             (-323.71)          (-0.051) (c)
  [male]                [307.62]          [-0.044] (a)
Education                -63.05            -0.009           n.s.
  (female)             (-127.19)          (-0.022)
  [male]                 [38.05]           [0.001]
Age                      122.68             0.077 c         2.611
  (female)              (124.58)           (0.095) (a)
  [male]                [119.63]           [0.060] (c)
Private Ins. Only      1,231.20             0.023           n.a.
  (female)            (1,969.06)           (0.043)
  [male]                [199.58]           [0.003]
Public/Private         1,112.06             0.051 (a)       n.a.
  (female)              (865.34)           (0.048) (c)
  [male]              [1,538.40]           [0.058] (a)
Public Only                                                 n.a.
  (female)                 --                --
  [male]
Rural                  1,107.09             0.044 (c)       4.569
  (female)            (1,188.76)           (0.056) (c)
  [male]              [1,063.68]           [0.036] (c)
South                 -1,386.58            -0.062 (c)       3.101
  (female)           (-1,566.46)          (-0.083) (c)
  [male]             [-1,178.18]          [-0.045]
Black Male               787.41             0.013           2.832
Black Female           1,106.67             0.023           n.s.
White Male             1,603.46             0.074 (c)       3.034
Black
  (female)            (1,091.11)           -0.035
  [male]               [-655.80]          [-0.014]
Mental Disorder        2,867.90             0.082 (c)       2.184
  (female)            (2,332.70)           (0.081) (c)
  [male]              [3,861.36]           [0.0901 (a)
[R.sup.2]                  0.033
  (female)                (0.044)
  [male]                  [0.027]

Notes:
a = significant at the .05 level, two tailed
b = significant at the. 01 level, two tailed
c = significant at the .001 level, two tailed

REFERENCES

Bazargan, M., Bazargan, S., & Baker, R. (1998). Emergency department utilization, hospital admissions, and physician visits among elderly African American persons. The Gerontologist, 38(1): 25-36.

Berk, M.L., & Monheit, A.C. (2001). The concentration of health care expenditures, revisited. Health Affairs, 20:9-18.

Black, B.S., Rabins, P.V., Geman, P., McGuire, M., & Roca, R. (1997). Need and unmet need for mental healthcare among elderly public housing residents. The Gerontologist, 37:717-28.

Brennan, P.L., & Moos, R.H. (1996). Late life problem drinking: personal and environmental risk factors for 4 year functioning outcomes and treatment seeking. Journal of Substance Abuse, 8:167-80.

Cameron, A.C., & Trivedi, P.K. (1998). Regression Analysis of Count Data. Cambridge: Cambridge University Press.

Carpenter, K.M., Hasin, D.S., Allison, D.B., & Faith, M.S. (2000). Relationships between obesity and DSMIV depressive disorder, suicide ideation, and suicide attempts: results from a general population survey. American Journal of Public Health, 980:251-57.

Carrasquillo, O., Lantigua, R.A., & Shea, S. (2001). Preventive services among Medicare beneficiaries with supplemental coverage versus HMO enrollees, Medicaid recipients, and elders with no additional coverage. Medical Care, 39:616-26.

Cohen, S.B. (1997). An evaluation of alternative PC-based software packages developed for the analysis of complex survey data. The American Statistician, 51: 285-92.

Dohrenwend, B.P. (2000). The role of adversity and stress in psychopathology: some evidence and its implications for theory and research. Journal of Health and Social Behavior, 41: 1-19.

Druss, B., Rohrbaugh R.M., & Rosenheck, R.A. (1999). Depressive symptoms and health costs in older medical patients. American Journal of Psychiatry, 156: 477-79.

Eisenberg, J.M. (1999). Quality research for quality healthcare: the data connection. Health Services Research, 35: xii-xvii.

Ferraro, K.F., & Wilmoth, J.M. (2000). Measuring morbidity: disease counts, binary variables, and statistical power. Journal of Gerontology: Social Sciences, 55:173-89.

Gornick, M.E. (2001). Disparities in Medicare services: potential causes, plausible explanations, and recommendations. Minority Health Today, 2:17-30.

Gornick, M.E. (2000). The association of race/SES and use of Medicare services. A little-known failure in access to care. Annals of the New York Academy of Science, 896: 497-500.

Haug, M.R., & Ory, M.G. (1987). Issues in elderly patient-provider interactions. Research on Aging, 9: 344.

Hough, R.L., Landsverk, J.A., Karno, M., Burnam, A., Timbers, D.M., Escobar, I., & Regier, D.A. (1987). Utilization of health and mental health services by Los Angeles Mexican Americans and non-Hispanic whites. Archives of General Psychiatry, 44: 702-08.

Husaini, B.A., Levine, R., Summerfelt, T., Holzer, C.E., Newbrough, R., Bragg, R., Cain, V.A., & Pitts, D. (2000). Prevalence and cost of treating mental disorders among elderly recipients of Medicare services. Psychiatric Services, 51:1245-47.

Husaini, B.A., Sherkat, D.E., Levine, R., Bragg, R., Holzer, C.E., Anderson, K., Cain, V.A., & Moten, C. (2002). Race, gender, and health care service utilization and costs among Medicare elderly with psychiatric diagnoses. Journal of Aging & Health, 14: 79-95.

Husaini, B.A., Sherkat, D.E., Moonis, M., Levine, R., Holzer, C.E., & Cain, V.A. (2003a). Race and gender differences in the impact of dementia on health service utilization and costs. Psychiatric Services, 54: 92-6.

Husaini, B.A., Sherkat, D.E., Levine, R., Holzer, C.E., Cain, V.A., Craun, C., & Hull, C. (2003b) The influence of race, gender, and mental disorder on healthcare service utilization and costs among the Medicare elderly. Research in Healthcare Financial Management, 8:19-30.

Kessler, R.C., McGonagle, K.A., Zhao, S., Nelson, C.B., Hughes, M., Eshleman, S., Witchen, H.U., & Kendler, K.S. (1994). Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States. Archives of General Psychiatry, 51: 8-19.

Kington, R.S., & Smith. J.P. (1997). SES and racial and ethnic differences in functional status associated with chronic diseases. American Journal of Public Health, 87: 805-10.

Marascuilo, L.A., & Levin, J.R. (1983). Multivariate statistics in the social sciences: a researcher's guide. Monterey, CA: Brooks/Cole.

Maynard, C., Ehreth, J., Cox, G.B., Peterson, P.D., & McGann, M.E. (1997). Racial differences in the utilization of public mental health services in Washington state. Administrative and Policy in Mental Health, 24:411-24.

McAlpine, D.D., & Mechanic, D. (2000). Utilization of specialty mental health care among persons with severe mental illness: the roles of demographics, need, insurance, and risk. Health Services Research, 35: 277-92.

Miech, R.A., & Shanahan, M.J. (2000). SES and depression over the life course. Journal of Health and Social Behavior, 41 : 437-50.

Monheit, A.C., & Vistnes, J.P. (2000). Race/ethnicity and health insurance status: 1987 and 1996. Medical Care Research Review, 57(Supplement 1): 11-35.

Muntaner, C., Eaton, W.W., Diala, C., Kessler, R.C., & Sorlie, P.D. (1998). Social class, assets, organizational control, and the prevalence of common groups of psychiatric disorders. Social Science and Medicine, 47, 2043-53.

Panzarino, P.J. (1998). The costs of depression: direct and indirect; treatment versus nontreatment. Journal of Clinical Psychiatry, 59:11-14.

Phillips, K.A., Mayer, M.L., & Aday, L.A. (2000). Barriers to care among racial/ethnic groups under managed care. Health Affairs, 19: 65-75.

Rasell, E., & Bernstein, J. (1995). An assessment of health care expenditures within and across racial and ethnic groups. American Economic Review, 63: 127-31.

Rost, K., Smith, R., Matthews, D.B., & Guise, B. (1994). The deliberate misdiagnosis of major depression in primary care. Archives of Family Medicine, 3: 333-37.

Sharp, M.J., & Getz, J.G. (1998). Self process in comorbid mental illness and drug abuse. American Journal of Orthopsychiatry, 68: 639-44.

Turner, R.J., & Lloyd, D.A. (1999). The stress process and the social distribution of depression. Journal of Health and Social Behavior, 40: 374-404.

Weinick, R.M., & Krauss, N.A. (2000). Racial/ethnic differences in children's access to care. American Journal of Public Health, 90:1771-74.

Williams, D.R., Yu, Y., Jackson, J.S., & Anderson, N.B. (1997). Racial differences in physical and mental health. Journal of Health Psychology, 2: 335-51.

Yelin, E., Herrndorf, A., Trupin, L., & Sonneborn, D. (2001). A national study of medical care expenditures for musculoskeletal conditions: the impact of health insurance and managed care. Arthritis Rheum, 1160-69.

Zhang, M., Rost, K.M., & Fortney, J.C. (1999). Depression treatment and cost- offset for rural community residents with depression. Journal of Social Service Research, 25:99-110.

This paper was completed with partial funding from the Agency for Healthcare Research and Quality; R305007, R24HS11640, and NIMH R24MH59748 to Tennessee State University, Baqar A. Husaini, Principal Investigator.

Baqar A. Husaini Tennessee State University

Darren E. Sherkat Southern Illinois University 

Robert S. Levine Meharry Medical College 

Clinton Craun Tennessee State University

Pamela C. Hull Tennessee State University

Van A. Cain Tennessee State University

Barbara S. Kilbourne Tennessee State University