Inpatient mental health facility utilization.
Virnig B, Huang Z, Lurie N, Musgrave D, McBean AM, Dowd B. Does Medicare Managed Care Provide Equal Treatment for Mental IllnessAcross Races?. Arch Gen Psychiatry. 2004;61(2):201–205. doi:10.1001/archpsyc.61.2.201
While disparities in access to care are well documented, little is known
about the quality of mental health care received by racial and ethnic minorities.
We examined the quality of mental health care received by elderly enrollees
in Medicare + Choice plans.
An observational study was performed using individual-level Health Plan
Employer Data and Information Set data. From 4182 to 5 016 028 individuals
65 years or older and enrolled in Medicare + Choice plans in 1999 were involved
in different measures. Rates of mental health inpatient discharges, average
length of stay, percentage of members receiving mental health services, rates
of follow-up after hospitalization for mental illness, optimal practitioner
contacts for antidepressant medication management, and effective acute- and
continuation-phase treatment were assessed.
Compared with whites, minorities received substantially less follow-up
after hospitalization for mental illness. The 30-day follow-up rates for whites,
African Americans, Asians, and Hispanics were 60.2%, 42.4%, 54.1%, and 52.6%,
respectively. Minorities also had lower rates of antidepressant medication
management for newly diagnosed episodes of depression. The rates of optimal
practitioner contacts for whites, African Americans, Asians, and Hispanics
were 12.5%, 12.0%, 11.1%, and 10.6%; the rates of effective acute-phase treatment
were 60.1%, 48.5%, 40.7%, and 57.6%; and the rates of effective continuation-phase
treatment were 46.7%, 32.7%, 31.9%, and 39.6%, respectively. The statistically
significant disparities persisted after adjusting for effects of age, sex,
income, plan model, profit status, and region of the country.
The overall quality of mental health care for people enrolled in Medicare
+ Choice managed care plans is far from optimal. There are large and persistent
racial differences that merit further attention to better understand their
underlying causes and solutions.
Mental illness is a common, debilitating problem for the elderly. Itis estimated that about 20% of the adult population 55 years and older experiencesome mental disorder, including depression, anxiety, and age-related cognitiveimpairment.1 Treating mental illness in theelderly is complicated because of stigma, financial barriers, underrecognitionof disease, and challenges inherent in distinguishing mental illness fromother disease processes. Effective treatment is complicated by comorbidities,age-related physiological changes, and risk of polypharmacy.
The recent Surgeon General's supplemental report Mental Health: Culture, Race, and Ethnicity2 reportedthat " . . . racial and ethnic minorities bear a greater burden from unmetmental health needs and thus suffer a greater loss to their overall healthand productivity."2(p3) This reportfurther documented how little is known about quality of care received by racialand ethnic minorities with mental health needs. Subsequent research has documenteddisparities in treatment for patients with schizophrenia and for Medicaidpatients with depression.3- 6 Treatingminority elderly persons with mental illnesses poses the combined challengesassociated with race and age.7 However, todate, very little research has focused on this population.
In 1999, approximately 6.2 million elderly persons (or 16% of all beneficiaries)were enrolled in Medicare managed care plans, also known as Medicare + Choice(M + C) plans.8 However, the quality of carereceived by mentally ill M + C beneficiaries is largely unstudied. The factorsthat affect differences in quality of care for minorities, in general, arepoorly understood. Likewise, it is not fully understood whether minoritiesexperience a different quality of mental health care than whites and, if theydo, why such differences exist. Several studies have reported significantproblems with utilization and quality of mental health care for racial andethnic minorities9,10 in the generalpopulation. A recent Institute of Medicine report comprehensively cataloguedstudies documenting differences in treatment between whites and minorities.7 However, by and large, the Institute of Medicine reportaddressed differences in treatment, in contrast to quality of care. Threerecent studies on racial variation in quality of care provided in M + C plansreported large racial differences in the rate of follow-up after hospitalizationsfor mental illness between African Americans and whites, as well as differencesin quality of care for the general M + C population, including differencesin the likelihood of seeing a primary care provider or a specialist, and differencesin experience with care.7,11,12 Suchstudies naturally raise the question of whether the racial differences inM + C plans are greater or lesser than the racial disparities in fee-for-serviceMedicare, especially in view of the hypothesis that quality improvement programsin managed care organizations are likely to reduce disparities in care. Whilethis study cannot address that question, it does explore the disparities inM + C plans in greater detail.
The Health Plan Employer Data and Information Set (HEDIS) offers a wayto study health services provided to Medicare beneficiaries enrolled in managedcare plans. These standardized performance measures are selected to assessthe adequacy and effectiveness of care.13 TheM + C plans are required to submit annual HEDIS reports that include plan-level(often referred to as summary) and individual-level data. This analysis extendsprevious work11,12 to more completelyexamine racial differences in the use of mental health services and more completelydescribe the effectiveness of mental health care provided to enrollees inM + C plans.
This analysis links individual-level HEDIS data with the beneficiary'sdemographic information, allowing the impact of age and race on receipt ofspecific mental health services to be examined.
Individual-level HEDIS data for reporting year 2000 (based on 1999 experience)were merged with demographic data obtained from the Centers for Medicare andMedicaid Services. The 301 M + C plans that submitted individual-level HEDISdata included information on 7 498 496 persons. The average numberof records per contract was 27 875 (range, 1189 to 484 738 personsper contract).
Individual records were identified via the Health Insurance Claim (HIC)number, a unique identification number used by Medicare. The HICs were mergedwith the 1999 Medicare Denominator file to obtain information on the age,race, sex, and state and county of residence. The HICs were also merged withthe Group Health Plan master file to confirm that each submitted record showedcorresponding plan enrollment during the contract year of question.
Individuals were excluded from this analysis if they did not have avalid HIC, if their race was classified as unknown or other, if there wasno evidence of managed care enrollment in either the denominator or GroupHealth Plan master file, or if they were younger than 65 years in 1999. Entirecontracts were excluded from this analysis if their submitted records failedto achieve at least 95% match on HIC.
The analysis focused on the HEDIS 2000 measures related to mental healthservices,14 listed below.
Mental health inpatient discharges: Numberof inpatient discharges from a hospital or treatment facility with any mentalhealth diagnoses except alcohol- and drug-related diagnoses or mental retardation.
Average length of stay for mental health inpatientstay: Number of days in a hospital or treatment facility with any mentalhealth diagnoses except alcohol- and drug-related diagnoses or mental retardation.
Percentage of members receiving mental healthservices: Percentage of members receiving any inpatient, day/night(ie, partial hospitalization), or ambulatory mental health services duringthe measurement year.
Seven-day and 30-day follow-up after hospitalizationfor mental illness: The percentage of members hospitalized for treatmentof depression, schizophrenia, attention-deficit disorder, or personality disorderswho were seen on an ambulatory basis or were in day/night treatment with amental health provider within 7 days or 30 days of hospital discharge.
Antidepressant medication management optimalpractitioner contacts: The percentage of members who were diagnosedas having a new episode of depression, treated with antidepressant medication,and who had at least 3 follow-up contacts with a primary care practitioneror mental health practitioner.
Antidepressant medication management effectiveacute- or continuation-phase treatment: The percentage of members whowere diagnosed as having a new episode of depression, were treated with antidepressantmedication, and continued taking an antidepressant drug during the entire12-week acute-treatment phase or for at least 6 months of continuation-treatmentphase.
The number of plans reporting mental health utilization varied frommeasure to measure (Table 1),in part because of rules that require reporting only for measures with a totaleligible population of 30 or more. Measures such as utilization rates requirethe denominator to be the entire covered population and, as a result, allplans report the measure. In contrast, when the denominator is defined asusers of services (eg, 7-day follow-up rate), measures are not reported byplans with fewer than 30 users of inpatient mental health care. This reportingrule generally affects small plans rather than large plans. The final columnin Table 1 indicates that thepercentage of submitted records that are included in the analysis ranges from79% to 93%.
Race was obtained directly from the 1999 Medicare denominator file.The categories included in this analysis were white, African American, Asian,and Hispanic. Persons whose race was listed as Native American were not includedin this analysis because of small numbers of M + C enrollees. Every plan hadat least some minority representation, ranging from 1.6% to 68.0% overall.All but one plan had at least some African American members, with a medianof 5.5% African American and a maximum of 68%. Similarly, 97.8% of plans hadat least some Asian members, with a median of 0.5% and a maximum of 40%, and97.8% of plans had at least some Hispanic members, with a median of 1.8% anda maximum of 29.6%.
We imputed indirectly on the basis of the median disposable householdincome by ZIP code for households with persons 65 years and older with theuse of figures from the 1990 US Census Bureau, because data from the 2000census were not yet available.15,16
Medicare plan structure and profit status were obtained from HealthCare Financing Administration's Monthly Report on Medicare Managed Care Plansfor December 1999,17 which reflects informationthat specific plans provided to the Health Care Financing Administration.Plan structure was defined as group, staff, and independent practice association.Profit status was defined as for-profit and not-for-profit. Region of thecountry was defined by means of a 4-level census designation: Northeast, South,Midwest, and West.
All analyses were conducted with the SAS system (SAS Institute Inc,Cary, NC). All measures were age and sex adjusted by direct standardizationmethods.18 Logistic regression models wereused to estimate adjusted odds ratios and to test for the presence of interactionsbetween race and other model elements. Bonferroni corrections were used toadjust for multiple comparisons.
In 1999, 2.4% of M + C enrollees received at least some mental healthcare. Overall, 0.35% of persons received inpatient care and 2.06% receivedtheir mental health care solely in ambulatory settings. The discharge ratefrom inpatient facilities was 5.8 per 1000 enrollees per year. The averagelength of stay was 7.9 days. Day/night care was the least frequently usedform of mental health services, with approximately 0.11% of beneficiariesreceiving care in that setting. The percentage of M + C enrollees receivingmental health care showed considerable racial variation and ranged from 11.7per 1000 members for Asians to approximately 24 per 1000 for whites and Hispanics(Table 2). Because most personsreceiving mental health care did so in an ambulatory setting, racial variationin ambulatory mental health care followed similar patterns. The percentageof members receiving inpatient and day/night care was substantially lowerthan rates of ambulatory care.
Although inpatient utilization increased with age, African Americansand whites had similar inpatient mental health utilization across age groups(Figure 1). Asians, in contrast,used less inpatient mental health care than African Americans and whites forall age groups. Hispanics also had less inpatient mental health care, andit did not increase with age to the degree that it did for African Americansand whites.
The 7-day and 30-day follow-up rates for individuals after inpatientmental health admissions were 35.5% and 58.2%, respectively, and there wasstrong variation across racial groups in follow-up rates (Table 3). African Americans and, to a lesser degree, Hispanics andAsians had lower follow-up rates than did whites.
Multivariate regression analysis confirmed significant differences betweenAfrican Americans and whites after adjusting for other factors such as planprofit status, health maintenance organization (HMO) model type, and regionof the country. The odds ratio associated with a 30-day follow-up visit forAfrican Americans compared with whites was 0.5 (95% confidence interval, 0.4-0.6),controlling for age, sex, income, number of admissions, average length ofstay, plan profit status, HMO model type, and region of the country. The coefficientsfor interaction terms were not statistically significant for race and age,plan profit status, HMO model, or region of the country. These associationswere maintained when plans were stratified by percentage of minority enrollment.
For persons with only one inpatient stay (96.6%), aggregate inpatientdays (which may be summed across multiple stays) equals length of stay forthe single hospitalization. This allows for an examination of the relationshipbetween length of stay and posthospitalization follow-up. Overall, lengthof stay was positively related to follow-up rate. The 30-day follow-up ratefor persons with a 4-day length of stay was 47.6%, compared with 55.2% fora 7-day stay and 59.2% for a stay longer than 14 days (P<.01). The association between length of stay and 30-day follow-updid not vary by race. In addition, racial variation in 7-day and 30-day follow-uprates was not explained by differences in length of stay.
For individuals diagnosed as having a new episode of depression, therates of optimal practitioner contacts, effective acute-phase treatment, andeffective continuation-phase treatment were low, at 11.7%, 58.6%, and 43.1%,respectively. Racial variation in the 2 effective acute- and continuation-phasetreatments showed a pattern consistent with global rates of ambulatory mentalhealth care; whites were more likely to receive effective acute-phase andcontinuation-phase treatment for antidepressant use. Despite these strongpatterns associated with age, racial differences persisted after age and sexadjustment. The coefficients for interaction terms between race and otherpredictors were not statistically significant (Table 4).
Multivariate regression analyses confirmed these findings. For effectiveacute-phase and continuation-phase antidepressant medication management, thedifferences between African Americans and whites and between Asians and whitesremained highly significant after controlling for age, sex, income, numberof inpatient admissions, average length of inpatient stay, plan profit status,HMO model type, and region of the country. The odds ratios for effective acute-and continuation-phase treatment for African Americans compared with whiteswere 0.64 (P<.01) and 0.54 (P<.01), respectively; for Asians compared with whites, 0.41 (P<.01) and 0.45 (P<.01),respectively; and for Hispanics compared with whites, 0.87 (P = .12) and 0.72 (P = .04). The coefficientsfor interaction terms between race and age, plan profit status, HMO model,and region of the country were not statistically significant. These associationswere maintained when plans were stratified by percentage of minority enrollment.
Consistent with 2 recent Surgeon General's reports,1,2 thesedata show that the mental health care for the elderly is suboptimal and thatthe quality of care is worse for racial and ethnic minorities than it is forwhite elderly persons. Particularly telling are the different patterns forinpatient hospitalizations and their follow-up. Rates of inpatient hospitalizationwere comparable across racial groups, but rates of follow-up were significantlydifferent across races. Once outpatient care had been accessed, the proportionswith an optimal number of visits were similar. However, the use of antidepressantsassociated with those visits varied dramatically across racial groups. Thissuggests that the quality of ambulatory care is particularly problematic forracial and ethnic minorities. Like follow-up rates, antidepressant medicationmanagement varied considerably across racial groups. While whites had ratesof effective management that were of concern, they were, nonetheless, significantlyhigher than the rates for African Americans, Hispanics, and Asians. The suboptimalmental health care experienced by elderly patients in M + C plans ought tobe a source of concern because undertreated mental health care needs are associatedwith increased rates of long-term institutionalization, increased levels ofhealth care use in all areas, and increased mortality rates.7
This study based its assessment on HEDIS measures. These measures arethe result of a consensus process but may not represent complete agreementwithin the mental health community. The low overall rates of antidepressantmanagement may be attributed to this sort of professional disagreement. Lessclear, however, is whether low follow-up rates after inpatient stays or differencesin antidepressant use across racial groups can be dismissed as professionaldisagreement.
There are some limitations to this analysis. First, our study was limitedto Medicare enrollees in M + C plans. We have no information on the qualityof care received by beneficiaries in fee-for-service Medicare. The M + C enrolleesare healthier than fee-for-service enrollees,19 aremore likely to live in urban areas,20 and maydiffer in other, unmeasured ways. Second, there is no information availableabout reasons for failure to receive adequate follow-up care. It is likelythat several factors related to access to and quality of care play a role,including receipt of a recommendation for follow-up and adherence to physician-recommendedfollow-up visits.7 Likewise, it is possiblethat generalists are providing this mental health care and that follow-upambulatory care visits are listed with some other diagnosis or are somehowpart of a bundling arrangement and are not recorded by the plan's administrativedata. Similarly, out-of-plan use will not be reflected in HEDIS measures.Third, HEDIS data report on rates of services received, not necessarily servicesneeded. The lack of variation in inpatient hospitalization rates may not meanthat there was no variation in the underlying level of need for inpatientcare. Similarly, the variation in rates of ambulatory care does not implythat appropriate care was received. Measures of antidepressant managementwere limited to people taking prescription antidepressants and, while theremay be legitimate professional disagreement about appropriate use of antidepressantsin the elderly, there is less disagreement about management for persons takingantidepressant medications. These quality measures do not examine whetherall persons who might benefit from antidepressant therapy are receiving it;they examine the patterns of care for persons taking prescription antidepressantmedications. There is always concern about the coding of race by Medicaresources. However, studies of the accuracy of Medicare's race code suggestthat the primary error is mistakenly identifying some persons as white whoshould be classified in another racial group.21 Theeffect of this misclassification, if any, is likely to be bias toward findingfewer racial differences. However, the true impact of this potential misclassificationon this study is not known.
This study documents strong racial differences in HEDIS indicators ofquality of mental health care for persons receiving antidepressant medicationtherapy and inpatient hospital care. The results of the analysis are disturbingin terms of both the racial disparities and the overall low quality of mentalhealth care. Recent Surgeon General's reports highlight the need to take thequality of mental health care of the elderly seriously and emphasize the importanceof measuring and addressing racial disparities in care. This study providesevidence of how much improvement is still needed.
Corresponding author and reprints: Beth Virnig, PhD, MPH, Divisionof Health Services Research and Policy, University of Minnesota, 420 DelawareSt SE, MMC 729, Minneapolis, MN 55455.
Submitted for publication July 2, 2002; final revision received July21, 2003; accepted July 22, 2003.
This study was supported by a contract from the Centers for Medicareand Medicaid Services, Baltimore, Md.
This study was presented in part as a poster at the annual meeting ofthe Academy for Health Services Research; June 23-25, 2002; Washington, DC.