Abstract
Objectives. We compared the influence of substance abuse with that of other comorbidities (e.g., anxiety, HIV) among people with mood disorder (N = 129 524) to explore risk factors for psychiatric hospitalization and early readmission within 3 months of discharge.
Methods. After linking Medicaid claims data in 5 states (California, Florida, New Jersey, New York, and Texas) to community-level information, we used logistic and Cox regression to examine hospitalization risk factors.
Results. Twenty-four percent of beneficiaries with mood disorder were hospitalized. Of these, 24% were rehospitalized after discharge. Those with comorbid substance abuse accounted for 36% of all baseline hospitalizations and half of all readmissions.
Conclusions. Results highlight the need for increased and sustained funding for the treatment of comorbid substance abuse and mood disorder, and for enhanced partnership between mental health and substance abuse professionals.
Mood disorders are currently the leading cause of psychiatric hospitalization.1,2 About 46% of all psychiatric inpatients in Maryland state general hospitals,1 for example, are admitted for major depressive disorder or bipolar disorder, and within a year of discharge, between 20% and 50% are likely to be readmitted.3–5 Many researchers have concluded, however, that few clinical characteristics can reliably predict either hospitalization or rehospitalization.3,4 Several studies suggest that comorbidity with substance abuse is associated with the first lifetime hospitalization for mood disorder,2 increased risk of psychiatric hospitalization more generally,6 and readmission after discharge.7 Most American studies on the topic, however, have used small to midsize local samples, and findings from much larger investigations using Danish case registers,7 for example, or information on US military populations8 may not generalize to the US general population or to low-income people such as those receiving Medicaid.
There do not appear to be any large-scale American studies comparing the role of substance abuse and other factors in either hospitalization or early readmission among Medicaid beneficiaries with mood disorder, even though Medicaid is the largest payer (covering 44% of all costs) of public mental health services.9 It is the only insurance plan available to many low-income persons, who rely on it exclusively for coverage of outpatient mental health or substance abuse treatment.
By linking 1999 to 2000 Medicaid claims data in 5 US states (California, Florida, New Jersey, New York, and Texas) to data on neighborhood characteristics (from the US Census at the zip code level and from Area Resource Files10 at the county level), we examined hospitalization for mood disorder and readmission within 3 months of discharge. After adjusting for the effects of gender, ethnicity, age, location, community characteristics, and dual eligibility with Medicare, we (1) investigated the influence of substance abuse on hospitalization for mood disorder, (2) compared the influence of substance abuse with that of other co-occurring conditions (anxiety, personality, or major medical disorder), and (3) assessed whether the influence of substance abuse was greater with major depressive disorder or bipolar disorder.
Research on comorbidity in mood disorder has yielded 3 major conclusions.2–4,11–24 First, comorbid substance abuse can have harmful negative consequences, including exacerbation of mood disorders, nonadherence to psychiatric medications, poor response to prescription drugs, and physical illness (e.g., liver disease). Second, anxiety and personality disorders contribute to poor response to treatment and increased risk of psychiatric hospitalization, and substance abuse can co-occur with them. Third, for 25% to 70% of persons with mood disorder who also have 1 or more of these comorbid conditions, diagnosis and treatment is much more complex in both inpatient and outpatient settings.
Apart from comorbidity, sociodemographic characteristics can increase the likelihood of hospitalization or readmission for mood disorder; their influence must be examined in studies predicting the use of hospital care from substance abuse or other risk factors. The evidence, however, is mixed. Some researchers have found that people initially hospitalized at a younger age are more likely to be rehospitalized later,25 but others have failed to find substantial differences.3 Similarly, Herrell et al.8 found that Blacks were less likely than were Whites to be hospitalized for mood disorder, but other studies suggest that psychiatric hospital admission in general is more common among Blacks than Whites.26,27
Similar to sociodemographic characteristics and comorbidity with substance abuse or other psychiatric disorders, medical comorbidity can affect hospitalization and readmission outcomes for mood disorder. People with mood disorders are more likely than is the general population to have health problems.28–31 The side effects of medications, difficulty accessing high-quality medical care, and unhealthful behaviors such as poor diet, lack of exercise, substance abuse, and smoking all increase the risk for cardiovascular disorders, cancer, hypertension, diabetes, and hyperlipidemia. Better outpatient treatment of these and other physical illnesses is needed to prevent hospitalization, prolong life, and contribute to overall well-being.
In short, there have been a variety of studies on hospitalization or readmission for mood disorder, but in the United States they are relatively small in scale, are typically cross-sectional and descriptive rather than longitudinal and predictive in nature, and they fail to focus on large numbers of low-income people who rely on Medicaid-funded services. Our study appears to be the first to examine the role of substance abuse and other risk factors in psychiatric hospitalization and readmission using a multistate sample of Medicaid beneficiaries with major depressive disorder or bipolar disorder. It also appears to be among the first to compare the impacts of substance abuse and of psychiatric or medical comorbidities after control for such local influences as median income and unemployment rate, as well as a wide variety of other factors.
On the basis of published research, we hypothesized that substance abuse increases risk for both hospital admission and readmission, even after inclusion of control variables. In addition, we tested the null hypothesis that substance abuse is equally influential in admission and readmission for both major depressive disorder and bipolar disorder, as well as the alternative hypothesis that drug or alcohol use is more influential in one or the other mood disorder—for example, because of the greater biological basis of bipolar disorder.32 Finally, we tested the null hypothesis that, after adjustment for other factors, substance abuse comorbidity is no more influential than other types of comorbidity. Because there was no strong evidence or theoretical basis for hypothesizing that substance abuse had a greater impact than did other risk factors on hospitalization for mood disorder, we chose an open-ended exploration of the various factors’ impacts.
METHODS
With the exception of context variables, data were taken from the Medicaid Analytic eXtract (MAX)33 for all Medicaid beneficiaries in 5 states (California, Florida, New Jersey, New York, and Texas) over a 2-year observation period (1999–2000; N = 5 979 180). MAX contains person-level information on demographic characteristics, dual eligibility with Medicare, primary and secondary diagnoses, and service use. The data set is maintained by the Centers for Medicare and Medicaid Services and extracted from the Medicaid Statistical Information System. Our analyses were limited to persons receiving care on a fee-for-service basis, because data on medical encounters financed through managed-care companies are possibly incomplete. To focus on nonelderly, community-living adults with uninterrupted public health insurance, we selected only persons aged 18 to 64 years without residence in long-term care and with full-year Medicaid coverage. Contextual variables were obtained by linking MAX with Area Resource Files and Census Public Use Summary Files.34
Measures
Major mood disorder.
Diagnostic categorization of individuals was based on diagnoses assigned by service providers and recorded on claims using diagnostic codes from the International Classification of Diseases, Ninth Revision, Clinical Modification35 (ICD-9-CM; see Table 1 for codes identifying psychiatric and medical conditions). Analysis focused on persons with primary or secondary diagnoses of either major depressive disorder or bipolar disorder (including manic disorders). Only beneficiaries with at least 1 inpatient or 2 outpatient diagnoses of major depressive disorder (n = 85 861) or bipolar disorder (n = 34 882) were selected in order to help eliminate from analyses beneficiaries who may have been diagnosed erroneously with mood disorder (e.g., hastily in a single brief encounter) by only 1 outpatient mental health service provider. In preliminary analysis, people identified in this way had an average of 12 inpatient or outpatient diagnoses of major depressive disorder and 20 of bipolar disorder.
TABLE 1.
International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) Codes of Psychiatric and Medical Conditions Used: Medicaid Analytic eXtract, 1999–2000
| ICD-9-CM Codes | |
| Psychiatric conditions | |
| Major depressive disorders | 296.2–296.3 |
| Bipolar disorders | 296.0–296.1, 296.4–296.8, 301.13 |
| Alcohol-related disorders | 291, 303 |
| Drug-related disorders | 292, 304, 305 |
| Anxiety-related disorders | 300 |
| Personality disorders | 301.0, 301.20, 301.22, 301.4, 301.50, 301.6–301.7, 301.81–301.83, 301.9 |
| Medical conditions | |
| Diabetes | 250 |
| Cancer | 140–172, 174–198, 199.1, 200–208 |
| Congestive heart failure | 428 |
| Dyslipidemia | 272.0–272.4 |
| HIV/AIDS | 042–044, V08 |
| Hypertension | 401 |
Comorbidity.
The same selection procedure of 1 inpatient or 2 outpatient diagnoses was used to identify beneficiaries with co-occurring substance abuse, anxiety disorders, and the most commonly occurring personality disorders. Some of the more frequently occurring comorbid medical conditions were also chosen for analysis on the basis of prior studies linking them to mood disorder.28–31 In addition to diagnostic information, data on personal characteristics, including age, ethnicity, gender, state, and dual eligibility with Medicare, were also taken from MAX.
Contextual variables.
Urbanicity was operationalized with rural–urban continuum codes36 linked to MAX data by county of hospitalization. It was defined on a 9-point continuum collapsed into 3 categories of metropolitan (urban) areas, nonmetropolitan areas adjacent to an urban location, and nonmetropolitan areas not adjacent to an urban location. Data on community context (at the zip code level) from the US Census were divided into quartiles, or 4 categories (lowest, highest, and 2 midrange categories) of area-specific median incomes, unemployment rates, and percentage of Blacks and resident non-US citizens within the zip code. These community characteristics were chosen because of their association with mood disorder or other serious mental illness in prior studies.37
Inpatient service use.
Psychiatric hospitalization and readmission for mood disorder within 3 months of discharge was identified by primary or secondary diagnoses as recorded in MAX inpatient claims. Hospital use was defined as a binary variable indicating use versus nonuse.
Analysis
Logistic regression was used to predict mood disorder hospitalization in the 2-year observation period from drug abuse, alcohol abuse, and all other independent variables. After fitting separate equations for each of these variables, we fit an adjusted model with all of them to estimate the influence of substance abuse in relation to other covariates.
For those with a hospital admission, rehospitalization within 3 months of discharge was modeled by Cox proportional hazards regression to take account of both whether readmission was observed and, if so, time to rehospitalization. Independent variables in rehospitalization analyses included the same factors used to predict baseline hospitalization. In admission and readmission analyses, individuals were clustered within counties.
For both admission and readmission, we compared the influence of comorbid substance abuse with that of other risk factors (e.g., psychiatric or medical comorbidity) by observing effect sizes in regression models (e.g., odds of hospitalization or readmission hazard associated with drug abuse vs anxiety disorder). After adjustment for all of the covariates, predictions of admission or readmission in the entire sample did not differ substantially from predictions of admission or readmission for major depressive disorder or for bipolar disorder; therefore, only regression results for the entire sample of persons with mood disorder are presented.
For bivariate comparisons of people with and without drug or alcohol abuse, we plotted Kaplan–Meier survival curves and used log-rank tests to compare differences in time to hospital readmission. In addition, we examined interdiagnostic differences in readmission risk by fitting separate proportional hazard models for the influence of drug or alcohol abuse on readmission for major depressive disorder vs bipolar disorder and then performing the t test on unstandardized parameter estimates and their standard errors to see whether the association was greater in one diagnostic group than the other. Results relating to comparisons of major depressive disorder and bipolar disorder are presented in the text rather than in the tables. All tests of statistical significance were 2-tailed.
RESULTS
About one quarter of persons with mood disorder were hospitalized at least once during the 2-year observation period (30 888 of 129 524, or 24%). Among those hospitalized, about one quarter were readmitted within 3 months of discharge (7457 of 30 888, or 24%). Rates differed (although determinants were similar) between major depressive disorder and bipolar disorder in relation to both hospitalization (major depressive disorder: 15 594 of 85 861, or 18%; bipolar disorder: 10 283 of 34 882, or 30%; P < .001) and rehospitalization (major depressive disorder: 3559 of 15 594, or 23%; bipolar disorder: 2083 of 10 283, or 20%; P < .001). To highlight interdiagnostic differences, these figures were derived after exclusion from analyses of 8781 people (5011 of whom were part of the sample of 30 888 people hospitalized at baseline) with 1 inpatient or 2 outpatient diagnoses of both major depressive disorder and bipolar disorder during the 2-year observation period.
Among people with mood disorder (major depressive disorder, bipolar disorder, or both), 15% (19 747 of 129 524) had comorbid drug or alcohol abuse, but these individuals accounted for 36% (10 995 of 30 888) of all baseline hospitalizations and half (3726 of 7457) of all rehospitalizations within 3 months of discharge. A description of the sample, the percentage hospitalized during the 2-year follow-up, unadjusted and adjusted odds ratios of hospitalization, the percentage rehospitalized within 3 months of discharge, and adjusted rehospitalization hazard ratios are shown in Table 2.
TABLE 2.
Rates of Psychiatric Hospitalization and Rehospitalization Among Medicaid Beneficiaries in 5 States With 1 Inpatient or 2 Outpatient Diagnoses of Mood Disorder: Medicaid Analytic eXtract, 1999–2000
| No.a (%) | Hospitalized, % | Unadjusted ORb (95% CI) | Adjusted ORc (95% CI) | No. Rehospitalized Within 3 moa (%) | Adjusted Hazard Ratiod (95% CI) | |
| Demographic variables | ||||||
| Total sample | 129 524 (100) | 24 | 7457 (24) | |||
| Gender | ||||||
| Women (Ref) | 86 795 (67) | 23 | 1.00 | 1.00 | 4457 (23) | 1.00 |
| Men | 42 729 (33) | 26 | 1.23*** (1.19, 1.26) | 1.11*** (1.05, 1.17) | 3000 (27) | 1.08* (1.02, 1.15) |
| Race/ethnicity | ||||||
| White (Ref) | 63 014 (49) | 26 | 1.00 | 1.00 | 3645 (23) | 1.00 |
| Black | 17 537 (14) | 29 | 1.16*** (1.18, 1.20) | 1.05 (0.95, 1.16) | 1377 (28) | 1.01 (0.94, 1.09) |
| Hispanic | 15 841 (12) | 20 | 0.74*** (0.71, 0.77) | 0.87* (0.76, 0.99) | 792 (25) | 1.04 (0.96, 1.12) |
| Other | 33 132 (26) | 20 | 0.72*** (0.70, 0.75) | 0.92 (0.81, 1.05) | 1643 (25) | 0.94 (0.86, 1.02) |
| Age, y | ||||||
| 18–29 (Ref) | 14 730 (11) | 28 | 1.00 | 1.00 | 1170 (28) | 1.00 |
| 30–39 | 28 547 (22) | 29 | 1.06* (1.01, 1.11) | 0.91* (0.85, 0.98) | 2083 (25) | 0.97 (0.90, 1.06) |
| 40–49 | 39 442 (31) | 26 | 0.87*** (0.84, 0.91) | 0.80*** (0.70, 0.90) | 2394 (24) | 0.92 (0.85, 1.01) |
| 50–59 | 33 490 (26) | 19 | 0.59*** (0.57, 0.62) | 0.63*** (0.53, 0.76) | 1408 (22) | 0.82*** (0.75, 0.90) |
| 60–64 | 13 315 (10) | 14 | 0.41*** (0.39, 0.44) | 0.50*** (0.38, 0.65) | 402 (22) | 0.90 (0.77, 1.06) |
| Location | ||||||
| Urban (Ref) | 88 413 (72) | 24 | 1.00 | 1.00 | 5319 (25) | 1.00 |
| Semi-urban | 23 667 (19) | 24 | 1.04* (1.00, 1.07) | 0.90 (0.76, 1.07) | 1243 (22) | 0.98 (0.88, 1.10) |
| Rural | 11 208 (9) | 23 | 0.97 (0.93, 1.02) | 0.86 (0.70, 1.07) | 494 (19) | 0.86 (0.73, 1.01) |
| State | ||||||
| California (Ref) | 42 948 (33) | 18 | 1.00 | 1.00 | 2159 (27) | 1.00 |
| Florida | 16 474 (13) | 32 | 2.08*** (2.00, 2.17) | 2.27*** (1.76, 2.93) | 1199 (23) | 0.80*** (0.70, 0.91) |
| New Jersey | 8 209 (6) | 31 | 2.02*** (1.91, 2.12) | 1.45** (1.11, 1.89) | 523 (21) | 0.63*** (0.57, 0.70) |
| New York | 44 022 (34) | 25 | 1.50*** (1.45, 1.55) | 1.29* (1.03, 1.60) | 2624 (24) | 0.74*** (0.68, 0.80) |
| Texas | 17 871 (14) | 23 | 1.34*** (1.28, 1.39) | 1.41* (1.07, 1.85) | 952 (23) | 0.85** (0.76, 0.95) |
| Contextual variables | ||||||
| Median income, $ | ||||||
| ≤ 27 535 (Ref) | 31 102 (25) | 23 | 1.00 | 1.00 | 1918 (27) | 1.00 |
| 27 536–34 344 | 30 982 (25) | 22 | 0.96* (0.92, 1.0) | 1.00 (0.92, 1.09) | 1628 (24) | 0.98 (0.88, 1.09) |
| 34 345–44 158 | 31 039 (25) | 24 | 1.06** (1.02, 1.10) | 1.07 (0.93, 1.22) | 1672 (23) | 0.90* (0.82, 1.00) |
| ≥ 44 159 | 31 030 (25) | 25 | 1.12*** (1.08, 1.16) | 1.09 (0.90, 1.31) | 1788 (23) | 0.88** (0.81, 0.96) |
| Unemployment, % | ||||||
| ≤ 1.80 (Ref) | 29 969 (25) | 25 | 1.00 | 1.00 | 1628 (23) | 1.00 |
| 1.81–2.69 | 29 909 (25) | 23 | 0.92*** (0.88, 0.96) | 1.02 (0.94, 1.11) | 1526 (23) | 0.99 (0.93, 1.06) |
| 2.70–3.71 | 29 742 (25) | 23 | 0.94** (0.91, 0.98) | 1.06 (0.93, 1.19) | 1703 (25) | 1.04 (0.96, 1.12) |
| ≥ 3.72 | 29 942 (25) | 22 | 0.89*** (0.85, 0.92) | 0.97 (0.86, 1.09) | 1613 (25) | 0.98 (0.89, 1.07) |
| Black, % in neighborhood | ||||||
| ≤ 2.36 (Ref) | 30 645 (25) | 22 | 1.00 | 1.00 | 1491 (22) | 1.00 |
| 2.37–6.97 | 30 565 (25) | 23 | 1.04* (1.00, 1.08) | 1.05 (0.93, 1.17) | 1657 (24) | 1.06 (0.99, 1.13) |
| 6.98–22.49 | 30 686 (25) | 24 | 1.11*** (1.07, 1.15) | 1.03 (0.91, 1.16) | 1725 (24) | 0.98 (0.91, 1.05) |
| ≥ 22.50 | 30 560 (25) | 26 | 1.20*** (1.16, 1.25) | 0.99 (0.85, 1.16) | 2062 (27) | 1.01 (0.90, 1.13) |
| Non-US citizens, % in neighborhood | ||||||
| ≤ 4.69 (Ref) | 30 880 (25) | 28 | 1.00 | 1.00 | 1741 (20) | 1.00 |
| 4.70–12.61 | 30 825 (25) | 25 | 0.87*** (0.84, 0.90) | 0.96 (0.87, 1.06) | 1815 (23) | 1.09 (1.00, 1.18) |
| 12.62–21.99 | 30 771 (25) | 22 | 0.72*** (0.69, 0.75) | 0.92 (0.82, 1.03) | 1742 (26) | 1.17*** (1.08, 1.28) |
| ≥ 22.00 | 30 847 (25) | 19 | 0.62*** (0.60, 0.64) | 0.85 (0.71, 1.02) | 1659 (28) | 1.20** (1.06, 1.37) |
| Type of mood disorder and insurance | ||||||
| Type of mood disorder | ||||||
| MDD (Ref) | 85 861 (71) | 18 | 1.00 | 1.00 | 3559 (23) | 1.00 |
| BPD | 34 882 (29) | 30 | 1.88*** (1.83, 1.94) | 1.93*** (1.58, 2.36) | 2083 (20) | 0.92** (0.87, 0.97) |
| Dual eligibility with Medicare | ||||||
| Yes (Ref) | 38 241 (30) | 26 | 1.00 | 1.00 | 1138 (11) | 1.00 |
| No | 91 283 (71) | 23 | 0.82*** (0.80, 0.85) | 0.69*** (0.56, 0.84) | 6319 (30) | 2.53*** (2.29, 2.79) |
| Medical comorbidity | ||||||
| Cancer | ||||||
| Yes | 3 246 (3) | 26 | 1.14** (1.05, 1.23) | 1.26*** (1.10, 1.46) | 289 (34) | 1.46*** (1.25, 1.70) |
| No (Ref) | 126 278 (98) | 24 | 1.00 | 1.00 | 7168 (24) | 1.00 |
| Congestive heart failure | ||||||
| Yes | 3 317 (3) | 35 | 1.77*** (1.64, 1.90) | 1.67** (1.16, 2.41) | 454 (39) | 1.76*** (1.57, 1.97) |
| No (Ref) | 126 207 (97) | 24 | 1.00 | 1.00 | 7003 (23) | 1.00 |
| Diabetes | ||||||
| Yes | 16 822 (13) | 26 | 1.16*** (1.12, 1.21) | 1.29*** (1.16, 1.45) | 1275 (29) | 1.26*** (1.15, 1.39) |
| No (Ref) | 112 702 (87) | 24 | 1.00 | 1.00 | 6182 (23) | 1.00 |
| Dyslipidemia | ||||||
| Yes | 11 968 (9) | 23 | 0.96 (0.92, 1.01) | 1.08 (0.93, 1.26) | 808 (29) | 1.10 (0.97, 1.25) |
| No (Ref) | 117 556 (91) | 24 | 1.00 | 1.00 | 6649 (24) | 1.00 |
| Hypertension | ||||||
| Yes | 28 184 (22) | 27 | 1.23 (1.19, 1.26) | 1.41*** (1.26, 1.58) | 2329 (31) | 1.28*** (1.17, 1.39) |
| No (Ref) | 101 340 (78) | 23 | 1.00 | 1.00 | 5128 (22) | 1.00 |
| HIV/AIDS | ||||||
| Yes | 4 091 (3) | 40 | 2.21*** (2.07, 2.35) | 1.37*** (1.26, 1.50) | 572 (35) | 1.25 (0.99, 1.67) |
| No (Ref) | 125 433 (97) | 23 | 1.00 | 1.00 | 6885 (22) | 1.00 |
| Psychiatric comorbidity | ||||||
| Anxiety | ||||||
| Yes | 24 385 (19) | 39 | 2.43*** (2.36, 2.51) | 2.03*** (1.58, 2.59) | 3170 (34) | 1.38*** (1.21, 1.57) |
| No (Ref) | 105 139 (81) | 21 | 1.00 | 1.00 | 4287 (20) | 1.00 |
| Personality disorder | ||||||
| Yes | 5 426 (4) | 67 | 7.32*** (6.91, 7.76) | 4.25*** (3.21, 5.62) | 1329 (36) | 1.53*** (1.45, 1.71) |
| No (Ref) | 124 098 (96) | 22 | 1.00 | 1.00 | 6128 (23) | 1.00 |
| Substance abuse | ||||||
| Drug use | ||||||
| Yes | 17 244 (13) | 57 | 5.72*** (5.53, 5.91) | 3.35*** (3.10, 3.63) | 3464 (35) | 1.58*** (1.47, 1.69) |
| No (Ref) | 112 280 (87) | 19 | 1.00 | 1.00 | 3993 (19) | 1.00 |
| Current alcohol use | ||||||
| Yes | 7 484 (6) | 65 | 6.72*** (6.40, 7.06) | 3.14*** (2.82, 3.49) | 1834 (38) | 1.46*** (1.28, 1.66) |
| No (Ref) | 122 040 (94) | 21 | 1.00 | 1.00 | 5623 (22) | 1.00 |
Note. OR = odds ratio; CI = confidence interval; MDD = major depressive disorder; BPD = bipolar disorder.
Totals vary because of missing data.
Unadjusted ORs reflect the probability of psychiatric hospitalization from sociodemographic characteristics, comorbidity, and context.
Adjusted ORs for hospitalization were used to control for all of the independent variables (first column of the table).
Adjusted hazard ratios for time to early rehospitalization were used to control for all of the independent variables (first column of the table). Persons who were not readmitted within 3 months were not included.
*P < .05; **P < .01; ***P < .001.
Substance Abuse and Baseline Hospitalization
People with drug or alcohol abuse had baseline hospitalization odds that were more than 3 times greater than those of their nonabusing counterparts after adjustment for all other independent variables (Table 2). These odds were comparable to the adjusted odds of admission relating to comorbid anxiety or personality disorder. By contrast, the adjusted odds of admission associated with co-occurring substance abuse were much greater than were those for medical comorbidity. People with co-occurring health problems had an adjusted risk between 9% and 67% greater than that of their counterparts without those problems; however, except for dyslipidemia, the smaller effect size was significant.
Both drug and alcohol abuse was more influential when co-occurring with hospitalization for major depressive disorder than for bipolar disorder (for drug abuse, b = 1.76 ±0.02 with major depressive disorder, b = 1.48 ±0.03 with bipolar disorder, t = 7.77, P < .001; for alcohol abuse, b = 1.99 ±0.03 with major depressive disorder, b = 1.59 ±0.05 with bipolar disorder, t = 6.86, P < .001).
Substance Abuse and Early Readmission
The relationship between substance abuse and time to rehospitalization is shown in Figure 1. People with drug or alcohol abuse had an adjusted readmission hazard rate that was substantially greater (58% and 46%, respectively) than that of their counterparts without drug or alcohol abuse (Table 2). This adjusted effect size was almost identical to the risk conferred by comorbid anxiety disorder, personality disorder, cancer, or congestive heart failure, and substantially greater in magnitude than that conferred by comorbid diabetes, hypertension, HIV/AIDS, or dyslipidemia.
FIGURE 1.
Percentage of Medicaid beneficiaries with mood disorder who were not readmitted to the hospital within 90 days of discharge, by presence (n = 10 995) versus absence (n = 19 893) of substance abuse: Medicaid Analytic extract, 1999–2000.
Note. There was a statistically significant difference between the 2 groups (χ21 = 923.57; P < .001).
Interdiagnostic differences between major depressive disorder and bipolar disorder were not found in the adjusted readmission hazards of people with versus without drug abuse, or of people with versus without alcohol abuse.
Other Hospitalization Determinants
Apart from substance abuse, 7 other factors were statistically significant determinants of hospitalization after adjustment for all the covariates. First, among those with mood disorder, men were 11% more likely to be admitted than were women. Second, Hispanics were 13% less likely to be hospitalized than were Whites. Third, relative to the youngest age group (18–29 years), older people were less likely to be admitted. Fourth, compared with Californians, people in other states were between 29% (New York) and 127% (Florida) more likely to be hospitalized. Fifth, Medicaid beneficiaries without dual eligibility with Medicare were 31% less likely to be admitted than were their counterparts with just Medicaid.
Other Rehospitalization Determinants
Similar to what was found for baseline hospitalization, the adjusted hazard for rehospitalization was 8% greater for men than for women, and the hazard decreased by 3% to 18% with ascending age group relative to the youngest age group (only the 18% decrease in the 50- to 59-year age group was statistically significant). However, other findings for readmission exhibited a pattern that contrasted with risk of initial admission. Adjusted rehospitalization hazards for residents of California were 15% to 37% higher than for the other 4 states. Also unlike what was found for hospitalization, the adjusted odds of readmission were 12% lower in communities with the highest median incomes than in those with the lowest median incomes and 20% higher in communities with the highest percentage of non-US citizens than in those with the lowest percentage. In addition, the adjusted hazard for readmission was more than 2.5 times greater for persons without dual eligibility with Medicare than for those with dual eligibility.
DISCUSSION
Almost one quarter of Medicaid beneficiaries with mood disorder in 5 states were hospitalized in 1999 and 2000, and within 3 months, almost one quarter of discharged inpatients were readmitted. People with drug or alcohol abuse comorbidity accounted for more hospitalizations than has previously been recognized: 36% of all baseline hospitalizations and 50% of all readmissions. These results suggest the need for better services for this population as well as adequate funding for such services. They also suggest a need to consider new strategies for providing these services to Medicaid beneficiaries.
A key issue is the need for greater integration among service delivery systems for substance abuse, mental health, and medical care. Many studies have noted the adverse impacts of a lack of coordination among categorically funded and separately delivered services; these include problems relating to lack of cross-specialty knowledge and training, difficulties (such as lack of transportation, time-consuming applications, or referral procedures) in accessing 1 type of service when another is used, unnecessary use of 2 separate agencies, and lack of communication between agencies about progress and setbacks in care.38 Concerted efforts at better service coordination might help prevent hospitalizations and readmissions of Medicaid beneficiaries with both mood disorder and substance abuse—a population that is only a minority (15%) of beneficiaries with mood disorder.
Service integration initiatives have not always achieved their aims, however, and careful evaluation is needed.39 Less formal initiatives that cross lines of discrete service systems (e.g., service delivery teams, joint task forces, cross-trainings) can help prevent hospitalizations for people with comorbidities. These collaborations could also include mutual assistance programs, which have been found to improve outcomes for both mental health and substance abuse treatment.40
The strong influence of substance abuse on hospitalization for mood disorder was roughly equivalent to the influence of comorbid personality or anxiety disorder. People with personality or anxiety disorders accounted for 37% of hospitalizations (11 351 of 30 888) and 33% of readmissions (3750 of 7457); the former figure is almost identical to that for substance abuse and the latter is only moderately less. This suggests that interventions aimed at preventing hospitalization for mood disorder could beneficially target all people with co-occurring substance abuse, personality, or anxiety disorders. People with any of the 3 major types of comorbidity could benefit from services that have been shown to prevent hospitalization, such as family involvement in care41 and a combination of psychotropic medication and psychotherapy,42,43 especially after these approaches are altered to address specific needs relating to substance abuse, anxiety, or personality disorder in persons with mood disturbance.
Compared with the association of comorbid substance abuse, anxiety, or personality disorders with hospitalization, that of medical comorbidity is more modest but still merits some attention. Providers charged with identifying at-risk people with dual disorders could be trained to assess for specific co-occurring medical conditions, including cancer, congestive heart failure, diabetes, hypertension, HIV/AIDS, and other health problems that increase the risk of hospitalization or rehospitalization.
Several other findings merit attention even though they are largely unrelated to our hypotheses (except as control variables). First, although it is difficult to interpret study results on community characteristics without more in-depth exploration, the significant influence of local median incomes and non-US citizenship on risk for rehospitalization underscores the need for more information on these and other understudied macrolevel or socioeconomic determinants of hospital recidivism. Second, the reasons why Hispanics were less likely to be hospitalized at baseline need clarification. Finally, the low hospitalization and high readmission rates in California highlight the need for further research on state variations (e.g., Medicaid policies, hospital admission criteria) in patterns of psychiatric hospitalization.
Other conclusions drawn from our study must also be considered speculative or exploratory because of research limitations, especially those inherent in the use of administrative data.44 Clinical history or symptom-level description of people with mood disturbance, substance abuse, or other disorders was unavailable, as were data on severity of symptom or stage and history of illness. In addition, diagnoses could have been miscoded or made erroneously by practitioners, and no other source of clinical information was available for triangulation. Finally, we were unable to examine admission or readmission among people who use private funds for hospitalization or rely on non-Medicaid insurance coverage.
Regarding hospitalization for mood disorder, future studies could examine (1) processes, stages, or trajectories of substance abuse influence over time, or across repeated hospitalizations; (2) the influence of specific drugs, drug and alcohol combinations, or combinations of drugs and prescribed medications; (3) the benefit over time of psychiatric medication, psychotherapy, or other outpatient services, and the impact of intervention nonadherence; (4) the benefits of referral and use of services after psychiatric hospital discharge; and (5) the influence of various combinations of comorbid conditions (e.g., substance abuse and anxiety or HIV/AIDS). Finally, all of these suggestions for research could be applied to the study of hospitalization for schizophrenia, the elderly and adolescents, and privately and publicly insured persons.
Acknowledgments
Research was supported by the National Institute of Mental Health (grant 2 R01 MH60831-04) and by the Agency for Healthcare Research and Quality through a cooperative agreement with the Center for Research and Education on Mental Health Therapeutics (1 U18 HS016097-01).
Human Participant Protection
This study was approved by the institutional review board of Rutgers University.
References
- 1.Brown SL. Variations in utilization and cost of inpatient psychiatric services among adults in Maryland. Psychiatr Serv 2001;52:841–843 [DOI] [PubMed] [Google Scholar]
- 2.Minnai GP, Tondo L, Salis P, et al. Secular trends in first hospitalizations for major mood disorders with comorbid substance use. Int J Neuropsychopharmacol 2006;9:319–326 [DOI] [PubMed] [Google Scholar]
- 3.Bridge JA, Barbe RP. Reducing hospital readmission in depression and schizophrenia: current evidence. Curr Opin Psychiatry 2004;17:505–511 [Google Scholar]
- 4.Ezquiaga E, Garcia A, Bravo F, Pallares T. Factors associated with outcome in major depression: a 6-month prospective study. Soc Psychiatry Psychiatr Epidem 1998;33:552–557 [DOI] [PubMed] [Google Scholar]
- 5.Scott J, Pope M. Self-reported adherence to treatment with mood stabilizers, plasma levels, and psychiatric hospitalizations. Am J Psychiatry 2002;159:1927–1929 [DOI] [PubMed] [Google Scholar]
- 6.Hendryx M, Russo JE, Stegner B, Dyck DG, Ries RK, Roy-Byrne P. Predicting rehospitalization and outpatient services from administration and clinical databases. J Behav Health Serv Res 2003;30:342–351 [DOI] [PubMed] [Google Scholar]
- 7.Kessing LV. The effect of comorbid alcoholism on recurrence in affective disorder: a case register study. J Affect Disord 1998;53:49–55 [DOI] [PubMed] [Google Scholar]
- 8.Herrell R, Henter ID, Mojtabai R, et al. First psychiatric hospitalizations in the US military: the National Collaborative Study of Early Psychosis and Suicide. Psychol Med 2006;36:1405–1415 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Mark TL, Buck JA. Components of spending for Medicaid mental health services, 2001. Psychiatr Serv 2005;56:648. [DOI] [PubMed] [Google Scholar]
- 10.Health Resources and Services Administration Area Resource File (ARF). Available at: http://www.arfsys.com. Accessed June 18, 2008
- 11.Bauer MS, Altshuler L, Evans DR, Beresford T, Willford W, Hauger R. Prevalence and distinct correlates of anxiety, substance, and combined comorbidity in a multi-site public sector sample with bipolar disorder. J Affect Disord 2005;85:301–315 [DOI] [PubMed] [Google Scholar]
- 12.Boylan KR, Bieling PJ, Marriott M, Begin H, Young LT, Macqueen GM. Impact of comorbid anxiety disorders on outcome in a cohort of inpatients with bipolar disorder. J Clin Psychiatry 2004;65:1106–1113 [DOI] [PubMed] [Google Scholar]
- 13.Brieger P, Ehrt U, Marneros A. Frequency of comorbid personality disorders in bipolar and unipolar affective disorders. Compr Psychiatry 2003;44:28–34 [DOI] [PubMed] [Google Scholar]
- 14.Cassidy F, Ahearn EP, Carroll BJ. Substance abuse in bipolar disorder. Bipolar Disord 2001;3:181–188 [PubMed] [Google Scholar]
- 15.George EL, Miklowitz DJ, Richards JA, Simoneau TL, Taylor DO. The comorbidity of bipolar disorder and axis II personality disorders: prevalence and clinical correlates. Bipolar Disord 2003;5:115–122 [DOI] [PubMed] [Google Scholar]
- 16.Goodwin RD, Stayner DA, Chinman MJ, Wu P, Tebes JK, Davidson L. The relationship between anxiety and substance use disorders among individuals with severe affective disorders. Compr Psychiatry 2002;43:245–252 [DOI] [PubMed] [Google Scholar]
- 17.Freeman MP, Freeman SA, McElroy SL. The comorbidity of bipolar and anxiety disorders: prevalence, psychobiology, and treatment issues. J Affect Disord 2002;68:1–23 [DOI] [PubMed] [Google Scholar]
- 18.Haywood TW, Kravitz HM, Grossman LS, Cavanaugh JL, Jr, Davis JM, Lewis DA. Predicting the “revolving door” phenomenon among patients with schizophrenic, schizoaffective, and affective disorders. Am J Psychiatry 1995;152:856–861 [DOI] [PubMed] [Google Scholar]
- 19.Hirschfeld RMA, Vornik LA. Bipolar disorder—costs and comorbidity. Am J Managed Care 2005;11:s85–s90 [PubMed] [Google Scholar]
- 20.McElroy SL, Altshuler LL, Suppes T, et al. Axis I psychiatric comorbidity and its relationship to historical illness variables in 288 patients with bipolar disorder. Am J Psychiatry 2001;158:420–426 [DOI] [PubMed] [Google Scholar]
- 21.Nolen WA, Luckenbaugh DA, Altshuler LL, et al. Correlates of 1-year prospective outcome in bipolar disorder: results from the Stanley Foundation Bipolar Network. Am J Psychiatry 2004;161:1447–1454 [DOI] [PubMed] [Google Scholar]
- 22.Regier DA, Farmer ME, Rae DS, et al. Comorbidity of mental disorders with alcohol and other drug use. Results from the Epidemiological Catchment Area (ECA) Study. J Am Med Assoc 1990;264:2511–2518 [PubMed] [Google Scholar]
- 23.Sorvaniemi M, Hintikka J. Recorded psychiatric comorbidity with bipolar disorder—a Finnish hospital discharge register study. Nord J Psychiatry 2005;59:531–533 [DOI] [PubMed] [Google Scholar]
- 24.Young LT, Cook RG, Robb JC, Levitt AJ, Joffe RT. Anxious and non-anxious bipolar disorder. J Affect Disord 1992;29:49–52 [DOI] [PubMed] [Google Scholar]
- 25.Kessing LV, Andersen EW, Andersen PK. Predictors of recurrence in affective disorder—analyses accounting for individual heterogeneity. J Affect Disord 2000;57:139–145 [DOI] [PubMed] [Google Scholar]
- 26.Cooper-Patrick L, Gallo J, Powe N, Steinwachs DM, Eaton WW, Ford DE. Mental health service utilization by African Americans and whites: The Baltimore Epidemiological Catchment Area Follow-Up. Med Care 1999;37:1034–1045 [DOI] [PubMed] [Google Scholar]
- 27.Snowden LR, Holschuh J. Ethnic differences in emergency psychiatric care and hospitalization in a program for the severely mentally ill. Community Ment Health J 1992;28:281–191 [DOI] [PubMed] [Google Scholar]
- 28.Carney CP, Jones LE. Medical comorbidity in women and men with bipolar disorders: a population-based controlled study. Psychosom Med 2006;68:684–691 [DOI] [PubMed] [Google Scholar]
- 29.Carney CP, Woolson RF, Jones L, Noyes R, Jr, Doebbeling BN. Occurrence of cancer among people with mental health claims in an insured population. Psychosom Med 2004;66:735–743 [DOI] [PubMed] [Google Scholar]
- 30.Kilbourne AM. The burden of general medical conditions in patients with bipolar disorder. Curr Psychiatry Rep 2005;7:471–417 [DOI] [PubMed] [Google Scholar]
- 31.McElroy SL, Frye MA, Suppes T, et al. Correlates of overweight and obesity in 644 patients with bipolar disorder. J Clin Psychiatry 2002;63:208–213 [DOI] [PubMed] [Google Scholar]
- 32.Smoller JW.FinnFamily, twin CT, and adoption studies of bipolar disorder. Am J Med Genet 2003;123:48–58 [DOI] [PubMed] [Google Scholar]
- 33.US Department of Health and Human Services Medicaid Analytic eXtract (MAX) general information. Available at: http://www.cms.hhs.gov/MedicaidDataSourcesGenInfo/07_MAXGeneralInformation.asp. Accessed June 18, 2008
- 34.US Census Bureau Census Public Use Summary Files. Available at: http://www.census.gov/main/www/cen2000.html. Accessed June 18, 2008
- 35.International Classification of Diseases, Ninth Revision, Clinical Modification. Hyattsville, MD: National Center for Health Statistics; 1980. DHHS publication PHS 80-1260 [Google Scholar]
- 36.US Department of Agriculture Measuring rurality: rural–urban continuum codes. 2004. Available at: http://www.ers.usda.gov/Briefing/Rurality/RuralUrbCon. Accessed December 21, 2007
- 37.Fortney J, Rushton G, Wood S, et al. Community-level risk factors for depression hospitalizations. Adm Policy Ment Health 2007;34:343–352 [DOI] [PubMed] [Google Scholar]
- 38.Prince JD, Austin MJ. Inter-agency collaboration between child welfare and child mental health systems. Soc Work Ment Health 2005;4:1–19 [Google Scholar]
- 39.Bickman L. Implications for evaluators from the Fort Bragg Evaluation 1. Am J Eval 1996;17:57–74 [Google Scholar]
- 40.Moos R, Schaefer J, Andrassy I, Moos B. Outpatient mental health care, self-help groups, and patients’ one year treatment outcomes. J Clin Psychology 2001;3:273–287 [DOI] [PubMed] [Google Scholar]
- 41.McFarlane WR. Multifamily Groups in the Treatment of Severe Psychiatric Disorders. New York, NY: Guilford Press; 2002 [Google Scholar]
- 42.Hegerl U, Plattner A, Moller HJ. Should combined pharmaco- and psychotherapy be offered to depressed patients? A qualitative review of randomized clinical trials from the 1990s. Eur Arch Psychiatry Clin Neurosci 2004;254:99–107 [DOI] [PubMed] [Google Scholar]
- 43.Miklowitz DJ, George EL, Richards JA, Simoneau TL, Suddath RL. A randomized study of family-focused psychoeducation and pharmacotherapy in the outpatient management of bipolar disorder. Arch Gen Psychiatry, 2003;60:904–912 [DOI] [PubMed] [Google Scholar]
- 44.Max MB, Lynn JL, Using administrative data to study hospice care. Bethesda, MD: National Institute of Health; 2003. [Google Scholar]