Medicaid has become one of the most expensive programs in most state budgets (1,2). Many states, faced with budget shortfalls, have sought to curtail Medicaid spending by decreasing reimbursement for providers, restricting eligibility and benefits, and increasing copayments for beneficiaries (3). Unfortunately, such strategies limit access to care and increase the number of uninsured persons (4). As increased attention is paid to Medicaid spending, it is important that policy makers understand how the characteristics of the covered population and the different components of the Medicaid benefit drive costs. In this regard, little information is available about the general medical costs of Medicaid beneficiaries with psychiatric disorders.
To control health care costs, many states have adopted Medicaid managed care programs. For general medical services, Medicaid managed care includes health maintenance organizations (HMOs), primary care case management programs, and disease management initiatives. For behavioral health, many states have chosen to carve out the behavioral health benefit and shift the risk for management of that benefit to a behavioral health organization (5,6,7,8,9,10). In 2004 a total of 21 states, including the District of Columbia, had managed behavioral health programs, 14 of which included behavioral health carve-outs (11). To control pharmacy costs, states often shift the risk for drug coverage, including psychotropic medications, to HMOs, and states are increasingly looking at pharmacy benefit management contracts and formulary limitations. This multiplicity of management arrangements makes it difficult to develop coherent systems that integrate the delivery of medical, behavioral health, and pharmaceutical care across funding streams.
In Colorado, behavioral health organizations have established at-risk contracts to provide all mental health services for the Medicaid population except for the treatment of substance use disorders and organic mental disorders. The state's contracts with the medical HMOs specifically exclude mental health services as a covered benefit. This benefit design means that visits to primary care physicians' offices that are billed with a psychiatric diagnosis are not a covered benefit. The benefit design also limits access to substance abuse treatment and means that mental health services provided in general medical settings are reimbursed only if they are provided by a specialty mental health provider who is contracted through the carve-out behavioral health organization.
Colorado Access, a nonprofit Medicaid health plan, is Colorado's largest Medicaid HMO and also holds the Medicaid behavioral health carve-out contract in Denver. As such, Colorado Access is uniquely poised to examine the costs of care for Medicaid beneficiaries with psychiatric disorders within the medical plan. In this retrospective study, we used Colorado Access claims and encounter data for patients covered under both its HMO and its behavioral health organization to determine the prevalence of identified psychiatric disorders, whereabouts in the system those patients received care, and the various contributors to the costs of care for those patients. This information can inform the design of delivery systems to more effectively address both the behavioral health needs and the medical needs of Medicaid beneficiaries in HMOs.
Patients included in the analyses were members, age 18 years and older, of the Medicaid HMO and behavioral health carve-out of Colorado Access between January and December 2000. To be included, members had to be eligible for services on both the physical and the behavioral health sides of the plan at any point during the period studied. Colorado Access claims and encounter data were used for all segments of this analysis. A subset of the population with psychiatric disorders was composed of Denver Medicaid recipients with severe and persistent mental illness (N=300). These patients are managed outside of the Medicaid behavioral health capitation program as a result of a class action lawsuit and were excluded from this analysis because Colorado Access does not manage their behavioral health care.
Sociodemographic data were obtained from the health plan's administrative database and included age, gender, Medicaid aid category, number of months enrolled in the plan, and Disability Payment System (DPS) score. The DPS is a case-mix adjustment model developed for Medicaid beneficiaries (12). Information about other variables, such as ethnicity and socioeconomic status, was unavailable for this study.
For prevalence rates and comparisons, the cohort was categorized into several psychiatric diagnostic clusters: any psychiatric diagnosis (ICD-9 codes 295.0-319.0), depressive disorders (296.20-296.36, 300.4, 311), bipolar disorder (300.00-300.02, 300.21-300.3), psychotic disorders (295.00-295.99, 298.9), anxiety disorders (300.0-300.09, 300.2-300.29, 300.3), and substance abuse or dependence (303.90-305.90, excluding 305.10) (13). Patients were assigned to a psychiatric diagnostic cluster if their claims history contained any of the aforementioned diagnoses, regardless of the type of claim submitted. Patients with more than one psychiatric diagnosis were classified in all appropriate categories. For each comparison of outcomes that targeted a particular diagnosis, patients were assigned to one of three mutually exclusive groups if they had the psychiatric diagnosis of interest, any psychiatric diagnosis except the target diagnosis, or no psychiatric diagnosis.
Prevalence rates were calculated on the basis of the presence of at least one submitted claim in which a psychiatric diagnosis was listed. Health care costs are based on the amount paid by Colorado Access through each line of business for the care of service recipients. For subcapitated clinics and providers, the Medicaid Fee Schedule for Colorado was used to calculate the costs of encounters. Because of the highly skewed distribution of the cost data, tests of statistical significance for cost comparisons were based on log-transformed costs.
Statistical analysis was performed by using SPSS Version 11.5 for a personal computer. Prevalence estimates were calculated by using the number of patients with each diagnosis as the numerator and the total number of adults who obtained services during the study period as the denominator. To compare differences in overall health care costs among participants, analysis of covariance (ANCOVA) was used. The procedure was performed for each diagnosis relative to the other two groups. Covariates in the ANCOVA models included gender, aid category, age, months enrolled in the plan, and DPS score. For each analysis, when the result of the main-effects F test for global differences among the groups was significant, Tukey pairwise comparisons were carried out. Because of the large number of analyses reported with results in the same direction as hypothesized, a Bonferroni correction for multiple statistical comparisons was believed to be too conservative, so the actual p values for each analysis are reported (14). Results of statistical tests were considered significant at p<.05, and all confidence intervals were set at 95 percent.
During the study period 61,256 members were enrolled in the Medicaid HMO and 58,186 members were enrolled in the behavioral health carve-out. Of these, 8,027 adults were eligible to receive services in both the medical plan and the behavioral health plan. Of these dually eligible members, 5,166 (64 percent) received services only through the medical plan, 105 (1.3 percent) received services only through the behavioral health plan, and 1,229 (15.3 percent) received services through both. Thus the overall penetration rate was 79.6 percent for the HMO and 16.6 percent for the behavioral health organization. These 6,500 unique adult members who received services constitute the study population.
The mean±SD age of the 6,500 members was 42.7±18 years, with a range of 18 to 98 years. A total of 4,739 members (72.9 percent) were women. The aid category distribution was as follows: Aid to the Needy and Disabled and Aid to the Blind, 2,379 members (36.6 percent); Temporary Assistance for Needy Families, 2,262 members (34.8 percent); Old Age Pension, 1,359 members (20.9 percent); Baby Care-Kids Care, 358 members (5.5 percent); and other, 85 members (1.3 percent). The mean number of months of eligibility during the study period was 9.5±3.9. Continuous enrollment criteria were not applied because the goal was to study a representative sample of enrollees that would reflect the population that a Medicaid health plan might actually serve in a given year.
Of the 6,500 patients, 2,597 (39.9 percent) were identified as having a psychiatric diagnosis. t1 presents the sample sizes, mean age, gender distribution, and prevalence rates for the diagnostic clusters.
The presence of any psychiatric diagnosis increased a patient's overall health care costs by a factor of 2.24 ($6,995 compared with $3,121 for patients with no psychiatric diagnosis). The overall health care costs include medical and pharmaceutical costs. Compared with costs for patients who did not have a psychiatric diagnosis ($3,121), patients with bipolar disorder had costs that were 2.74 times greater ($8,567), those with a psychotic disorder had costs that were 2.62 times greater ($8,201), those with an anxiety disorder had costs that were 2.42 times greater ($7,575), those with a substance use disorder had costs that were 2.51 times greater ($7,847), and those with a depressive disorder had costs that were 2.33 times greater ($7,284). Further details are provided in t2. F1 and F2 present the distribution of costs between the HMO and the behavioral health organization.
Costs for general medical care
The presence of any psychiatric diagnosis increased a patient's medical costs exclusive of pharmaceuticals by a factor of 1.77 ($4,690 compared with $2,649 for patients with no psychiatric diagnosis). In addition, for those with a substance use disorder costs were 2.34 times greater ($6,225), for those with an anxiety disorder costs were 2.12 times greater ($5,620), for those with a depressive disorder costs were 1.79 times greater ($4,759), for those with bipolar disorder costs were 1.10 times greater ($2,930), and for those with a psychotic disorder costs were 1.09 times greater ($2,892). Additional information is presented in t3.
Costs were higher in all categories of service for patients who had a psychiatric disorder compared with those who did not have such a disorder. Pharmaceutical costs were 167.7 percent higher ($1,264 compared with $472 for those without a psychiatric disorder). Costs for ancillary services, such as laboratory tests and radiology, were 119.0 percent higher ($124 compared with $57). Emergency department costs were 101.9 percent higher ($219 compared with $109). Costs for outpatient hospital services were 99.1 percent higher ($504 compared with $253). Inpatient costs were 91.1 percent higher ($2,779 compared with $1,454). Primary care costs were 31.5 percent higher ($883 compared with $672).
Rates of mental health specialty care
Only 852 of the patients with a psychiatric disorder (32.8 percent) saw a mental health professional during the study period, and only 696 (26.8 percent) had more than one visit with a specialty mental health provider. Members with bipolar disorder (1,917 patients, or 73.8 percent) and those with a psychotic disorder (1,766 patients, or 68.0 percent) were more likely to have seen specialty mental health providers at least once, whereas members with depression (940 patients, or 36.2 percent) had less specialty mental health contact, as did those with anxiety disorders (976 patients, or 37.6 percent) and substance use disorders (525 patients, or 20.2 percent).
Little information is available for Medicaid populations on the prevalence and costs of mental disorders. A 2003 review of data from the National Comorbidity Survey and the National Household Survey on Drug Abuse reported that mental and substance use disorders are almost twice as common in Medicaid populations as in privately insured populations (15). The 12-month prevalence rate of the 14 mental and substance use disorders surveyed in the National Comorbidity Survey was 48.3 percent in the Medicaid population, compared with 40.5 percent in the uninsured population and 28.0 percent in the privately insured population.
The study reported here, which used a different method, supports the finding that Medicaid beneficiaries have unusually high rates of mental disorders. In the study, 39.9 percent of adult Medicaid beneficiaries had a psychiatric diagnosis on a claims encounter, even after exclusion of a subset of 300 Colorado Access adult members with serious and persistent mental illness whose care was not managed by the behavioral health plan but who represented 4.6 percent of the psychiatric population covered by the health plan during the study period. This finding suggests that the 39.9 percent prevalence rate is most likely a significant underestimate of the true frequency of mental disorders in this adult Medicaid HMO population.
The finding that adult Medicaid beneficiaries have very high rates of mental disorders is not surprising given that this population has a number of risk factors, including poverty, unemployment, psychosocial stressors, and chronic medical illnesses (16,17).
The higher prevalence of mental disorders in Medicaid populations is associated with greater spending on direct mental health services compared with private insurance. A 2003 review by Mark and colleagues (18) estimated that direct spending on mental health and substance abuse services accounted for 9.3 to 13 percent of Medicaid spending, compared with 3.1 to 5.6 percent of spending in private insurance plans. Consistent with these findings, in the study reported here 10.5 percent of health care spending was for direct mental health billings. However, studies that look only at direct mental health spending and exclude pharmacy and medical costs underestimate the true costs of caring for patients with mental disorders (19).
A number of studies have documented that medical patients with comorbid psychiatric illnesses have significantly higher medical costs, but few studies have focused specifically on Medicaid populations (19,20,21,22,23,24). Using a state Medicaid fee-for-service data set, Garis and Farmer (25) examined nine chronic conditions (psychosis, depression, acid peptic illness, diabetes, congestive heart failure, respiratory illness or asthma, hypertension, and anxiety) and found that of all these conditions psychosis and depression accounted for the highest mean overall yearly costs—$6,964 and $5,505, respectively. These authors also looked at different pairings of chronic comorbid conditions and found that psychosis was a factor in five of the seven highest-cost pairs. Another analysis of a ten-state Medicaid database found that mental health and substance abuse service users accounted for 11 percent of all Medicaid enrollees but made up fully one-third of "high cost members" (defined as being in the top ten percent of enrollees by costs); most of their costs resulted from the use of non-mental health services (26).
In this study, overall health care costs were 2.24 times higher for patients who had a diagnosis of a mental disorder than for those who did not have such a diagnosis. After pharmacy costs and direct mental health service expenses were excluded, costs to the medical plan were higher by an average factor of 1.77, ranging from 1.09 for patients with psychotic disorders to 2.34 for patients with substance use disorders. As shown in F2, the relative contribution of mental health, pharmacy, and general medical services to the overall cost increases differed significantly depending on psychiatric diagnosis. Patients with bipolar and psychotic disorder were more likely to be treated with specialty mental health services and had higher specialty mental health and pharmaceutical costs. Patients with depression, anxiety, and substance use diagnoses were more likely to receive a majority of their treatment in general medical settings and had higher general medical costs.
A higher prevalence of severe and persistent mental illness in Medicaid populations compared with commercially insured populations is suggested by the finding that 5.1 percent of adults had a psychotic disorder and 4.0 percent had bipolar disorder. From a cost perspective, patients with bipolar disorder and schizophrenia had the highest overall costs of care, with factor increases of 2.74 and 2.62, respectively. However, when pharmacy and specialty mental health costs were removed, general medical costs for patients with these diagnoses were relatively similar to those for the average adult Medicaid beneficiary, with factor increases of 1.10 and 1.09, respectively.
From a Medicaid policy perspective, the high cost of pharmaceutical agents to treat these diagnoses raises the question of how financial risk for some psychotropic medications should be handled. In some states, such as Colorado, the medical plans bear financial risk for all drugs, including specialty psychotropics, such as atypical antipsychotics and mood stabilizers, even though these agents are most frequently prescribed by psychiatrists who are contracted through the behavioral health carve-out (27). This separation of prescribing patterns from accountability for costs misaligns incentives between the medical plan and the mental health providers in regard to controlling costs for these agents.
Patients with substance abuse, anxiety, and depression diagnoses also had significantly higher overall health care costs than those without a psychiatric diagnosis, with factor increases of 2.51, 2.42, and 2.33, respectively. These factor increases are larger than those reported for commercially insured patients with these diagnoses (19,20,23,28,29,30).
This study had several important limitations. It used claims data to provide information about the diagnoses of interest to clinicians and about costs of care across the entire adult membership of one Medicaid health plan. This claims-based method most likely underestimated the prevalence of mental disorders and may not provide an accurate diagnosis for any given patient (31,32,33,34,35,36). These limitations have been discussed in studies of commercially insured populations that use the same methods (18,19,22). In future, researchers could conduct diagnostic interviews of representative samples of each Medicaid enrollment category (Aid to the Needy and Disabled, Aid to the Blind, Temporary Assistance for Needy Families, Old Age Pension, and Baby Care-Kids Care), including interviews with nonusers to more accurately assess the prevalence of mental disorders in Medicaid-eligible subpopulations.
Another limitation of this study was that only data for billed mental health services were captured. With the implementation of statewide Medicaid mental health carve-outs, many primary care physicians have given up on billing for mental health-related visits. Because only mental health services that are contracted through the behavioral health carve-out are billable, mental health services provided in primary care settings were not represented in this study unless provided by a contracted behavioral health organization.
The ability to generalize these findings to other Medicaid populations is also limited because of heterogeneity among states in Medicaid benefit design, eligibility thresholds, and managed care structures (5,37,38). In addition, as has been noted for commercial plans (39), the HMO population studied probably has less mental illness than a Medicaid fee-for-service population, because disabled enrollees in Colorado and many states are allowed to opt out of managed care. Also, to enable us to report on all adult members of the health plan, this study included members who were dually eligible for Medicaid and Medicare. We were unable to capture the cost of services billed to Medicare, so we have underestimated the true cost of care for those patients. Future studies might attempt to address this deficit and examine per-member-per-month costs to better control for differences in duration of enrollment. The direction of the bias caused by these limitations would suggest that the results of this study underestimate the prevalence and costs of mental disorders in Medicaid populations.
Despite the study's limitations, the findings of significantly higher prevalence and costs of mental disorders in Medicaid populations have policy implications for the design of funding structures and clinical systems. Currently, most carved-out Medicaid behavioral health programs are designed to support community mental health systems of specialty care that prioritize serving patients with serious and persistent mental disorders. These systems have increased the number of people who gain access to mental health services, decreased hospital use, and provided flexible funding for a continuum of nontraditional community-based services (40,41). Although Medicaid behavioral health carve-outs may improve some aspects of specialty mental health care within the community mental health system, they may have exacerbated the long-standing problem of how to provide for the mental health needs of general medical patients and the primary care needs of persons with serious and persistent mental illness.
Many patients with depression, anxiety, and somatization disorders present and expect to be cared for in general medical settings (42). Consistent with other studies (23,43), this study found that only 32.8 percent of patients with a diagnosis of a mental disorder had seen a mental health professional during the study period and only 26.8 percent had more than one visit. Of all beneficiaries who received any services in this study, only 1.3 percent received specialty mental health services without some contact with the medical system. In addition to showing that patients with psychiatric diagnoses had frequent contact with the medical system, this study documented that patients with psychiatric diagnoses accounted for significantly higher costs to the medical plan. These differences in costs remained statistically significant even after Kronick scores (12) were used to control for severity of medical comorbidity. Mental disorders, such as anxiety, depression, and substance abuse, may be associated with higher general medical costs for several reasons (19,23,30,39,44,45), but that discussion is beyond the scope of this paper.
From a funding perspective, when states carve out mental health services, policy makers frequently expect that the behavioral health carve-out will provide all needed mental health services for the covered population. However, carve-out behavioral health organizations are not usually set up to support the delivery of mental health services in general medical settings and currently have few financial incentives to provide mental health services in those settings. From a provider perspective, few clinics can afford to support co-located mental health specialists, and carve-out behavioral health organizations seldom pay primary care and other non-mental health providers to deliver services. As a result, existing primary care practice and reimbursement structures do not encourage primary care practitioners to take on and effectively deal with the time-consuming and ubiquitous mental health issues in their practices (46). On the other hand, when primary care patients are referred to off-site specialty mental health services, fewer than half follow through with even one appointment (42).
It is unclear whether more effective treatment of mental health issues can have an impact on the medical costs of Medicaid patients. The higher prevalence and cost of mental disorders may make it relatively easier to demonstrate the cost-effectiveness of mental health interventions in Medicaid populations than in commercial populations, but such cost-effectiveness remains to be demonstrated. From a quality perspective, care management models have been developed that improve the treatment of depression and anxiety in primary care (47,48,49,50). Patients who have thought or mood disorders comorbid with chronic medical conditions are especially likely to incur high costs and to be difficult to treat (25). Closer psychiatric follow-up, better coordination of care, and greater attention to psychosocial issues may help improve clinical outcomes for medical conditions as well as for psychiatric conditions among Medicaid beneficiaries with psychiatric disorders (21,44,51).
This study found that behavioral health conditions were especially common and costly in an adult Medicaid population. As a result, it is especially important that Medicaid delivery systems develop better ways to address the integration of behavioral health and physical health care for Medicaid beneficiaries. To do so Medicaid programs need to create financial incentives that encourage the physical and behavioral health systems to work together more effectively to address the behavioral health needs of Medicaid beneficiaries.
Dr. Thomas and Dr. Waxmonsky are affiliated with the department of psychiatry and Dr. Rost with the department of family medicine at the University of Colorado at Denver and Health Sciences, 4455 E. 12th Avenue, A011-99, Denver, Colorado 80220 (e-mail, email@example.com). Dr. Thomas and Dr. Waxmonsky are also with Colorado Access in Denver, with which Ms. Flanders-McGinnis is affiliated. Dr. Gabow is affiliated with Denver Health Medical Center. Dr. Socherman is with the Portland Department of Veterans Affairs Medical Center in Portland, Oregon.
Per-person costs of care for 6,500 Medicaid beneficiaries eligible for care from a Medicaid HMO and a behavioral health carve-out, by presence or absence of a psychiatric diagnosis
Per-person costs of care for 6,500 Medicaid beneficiaries eligible for care from a Medicaid HMO and a behavioral health carve-out, by indicated psychiatric diagnosis
Psychiatric diagnoses, gender, and age in a population of 6,500 Medicaid beneficiaries eligible for care from a Medicaid HMO and a behavioral health carve-out
Annual costs for physical and mental health care provided to 6,500 Medicaid beneficiaries eligible for care from a Medicaid HMO and a behavioral health carve-outa
aCosts are included for physical and mental health care and all pharmaceuticals. Because of the highly skewed distribution of the cost data, tests of statistical significance for cost comparisons were based on log-transformed costs.
Annual costs for physical health care provided to 6,500 Medicaid beneficiaries eligible for care from a Medicaid HMO and a behavioral health carve-out