The associations between substance use disorders, schizophrenia-spectrum disorders, and Axis IV psychosocial problems

Transcription

1 Soc Psychiatry Psychiatr Epidemiol (2005) 40: ORIGINAL PAPER DOI /s Michael T. Compton. Paul S. Weiss. Joyce C. West. Nadine J. Kaslow The associations between substance use disorders, schizophrenia-spectrum disorders, and Axis IV psychosocial problems Accepted: 9 June 2005 / Published online: 25 October 2005 Š Abstract Background Substance abuse among individuals with schizophrenia-spectrum disorders (SSDs) is associated with a range of adverse psychosocial outcomes in the areas of occupational functioning, housing stability, economic independence, access to health care, and involvement with the legal system. The aim of this study was to estimate the effects of substance use disorders (SUDs), SSDs, and dual diagnosis with both disorders on the risk for six important Axis IV psychosocial problems. This was accomplished using a large dataset of patients who are representative of individuals in routine US psychiatric practice. Method Weighted data from the 1999 Study of Psychiatric Patients and Treatments from a practice-based research network of the American Psychiatric Institute for Research and Education were analyzed. Some 615 US M. T. Compton, MD, MPH. N. J. Kaslow, PhD Dept. of Psychiatry and Behavioral Sciences Emory University School of Medicine Atlanta (GA), USA M. T. Compton, MD, MPH (*) Dept. of Psychiatry Grady Memorial Hospital Box Jesse Hill Jr. Drive, S.E. Atlanta (GA) 30303, USA Tel.: / Fax: / mcompto@emory.edu P. S. Weiss, MS Dept. of Biostatistics Rollins School of Public Health of Emory University Atlanta (GA), USA J. C. West, PhD, MPP American Psychiatric Practice Research Network American Psychiatric Institute for Research and Education Arlington (VA), USA psychiatrists provided detailed clinical, psychosocial, and health services information on 1,843 patients, including 285 patients with one or more SUDs without an SSD, 180 patients with a diagnosis of an SSD without substance abuse comorbidity, and 68 dually diagnosed patients. Logistic regression models were used to determine effect estimates (adjusted odds ratios), and corresponding 95% confidence intervals were calculated. Results After adjusting for sociodemographic variables and for SSD diagnosis, SUD diagnosis was independently associated with increased risk for five of the Axis IV psychosocial problems of interest (occupational problems, housing problems, economic problems, problems with access to health care services, and problems related to interaction with the legal system/ crime) when compared to all other psychiatric patients (n=1,310). After adjusting for the sociodemographic variables and for SUD diagnosis, SSD diagnosis (compared to all other psychiatric diagnoses) was associated with Axis IV economic problems, but not with the other five psychosocial problems of interest. The presence of both an SUD and an SSD diagnosis (dual diagnosis) was associated with a greater risk for four of the six Axis IV psychosocial problems studied, compared to the risks associated with either diagnosis alone. Limiting the substance of abuse to alcohol resulted in similar findings. Conclusions Although SUDs are associated with increased risk for poor social adjustment, the comorbidity of SUDs and SSDs is associated with greatly compounded psychosocial burdens. These findings, from a large sample of representative US psychiatric patients, demonstrate the ongoing need for improved services and policies for those specially burdened patients with the dual diagnosis of both an SSD and substance abuse or dependence. Š Key words schizophrenia substance abuse dual diagnosis comorbidity axis IV psychosocial problems SPPE 964

2 940 Introduction The lifetime prevalence of substance use disorders (SUDs), excluding nicotine, in individuals with schizophrenia is over 40%, a rate estimated to be 4.6 times that of the general population [10]. Characteristics associated with comorbid SUDs in the context of schizophrenia-spectrum disorders (SSDs) include: male gender, younger age, higher parental social class, higher premorbid cognitive functioning, higher IQ scores, forensic history, and family history of substance abuse [4, 9, 11, 17, 20]. Substance abuse among individuals with SSDs is associated with a range of clinical and psychosocial difficulties including: poorer outcomes from psychosis, higher rates of relapse of illness, decreased medication adherence and more missed appointments, more hospital admissions, increased risk of illness and injury, increased prevalence of medical problems, loss of educational and employment opportunities, unstable housing and homelessness, financial hardship, and increased criminal prosecution and incarceration [9, 10, 20]. Dual diagnosis with an SUD and schizophrenia or other psychotic disorders is clearly associated with psychosocial problems in multiple domains. However, little research has been conducted to quantify the magnitude of the social impairment related to this comorbidity in large, representative samples of patients. The aim of this analysis was to estimate the effects of (1) SUDs, (2) SSDs, and (3) combined SUD and SSD (referred to as dual diagnosis herein) on the risk for six important Axis IV psychosocial problems (problems with primary support group, occupational problems, housing problems, economic problems, problems with access to health care services, and problems related to interaction with the legal system/ crime). This was accomplished using logistic regression models and the calculation of odds ratios (ORs), adjusted for a number of sociodemographic variables, from a dataset including 1,843 patients being treated in routine US psychiatric practice. It was predicted that compared to either diagnosis singly, the presence of a dual diagnosis would confer greatly increased risk for social adjustment problems. Furthermore, effect estimates of this elevated risk were determined. Analyses were also conducted to examine these effect estimates when controlling for several important potential confounders, including access to health care, private vs public treatment setting, and inpatient vs outpatient status. The impact of alcohol use disorders (AUDs) and comorbid AUD and SSD also was examined in analyses using restricted samples. Subjects and methods Data were obtained from the 1999 Study of Psychiatric Patients and Treatments (SPPT), a cross-sectional survey conducted with participants of the American Psychiatric Institute for Research and Education (APIRE) American Psychiatric Practice Research Network (PRN). Details of PRN studies have been published previously [12, 22]. Briefly, the PRN is a nationwide network of psychiatrists practicing in public and private settings who are American Psychiatric Association (APA) members providing at least 15 h/week of direct patient care. Participants in the study included 324 (52.7%) volunteer psychiatrists and 291 (47.3%) randomly selected psychiatrists from the APA membership. Respondents (n=615 PRN psychiatrists) and nonrespondents (n=169 PRN psychiatrists) were comparable with respect to gender, age, race, and region of practice. Each participant completed general information on 12 consecutive patients according to a randomly assigned start-time, as well as more detailed data on a subsample of three patients. Diagnoses were based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV, [1]). Participating psychiatrists were instructed to indicate all DSM-IV diagnoses and their corresponding codes for each patient, but not to record provisional or rule-out diagnoses. In these analyses, the SSD classification refers to three DSM-IV diagnoses: schizophrenia, schizophreniform disorder, and schizoaffective disorder. The other diagnoses within the DSM-IV schizophrenia and other psychotic disorders group (delusional disorder, brief psychotic disorder, shared psychotic disorder, psychotic disorder due to a general medical condition, substance-induced psychotic disorder, and psychotic disorder not otherwise specified) were not included. The SUD classification in these analyses included DSM-IV substance abuse and substance dependence of the following six drugs: alcohol, amphetamines, cannabis, cocaine, hallucinogens, and opioids. Polysubstance dependence also was included in the SUD classification herein. Because the focus of this study was on the impact of SSD/SUD dual diagnosis in general, current or past, all diagnosed SUDs meeting this specification were included, regardless of whether they were coded as currently meets criteria, in partial remission, or in full remission. It is widely held that SUDs are often chronic, relapsing, and remitting disorders, and thus, excluding those patients who currently met the criteria for partial or full remission did not seem appropriate. DSM-IV substance-induced disorders (including substance intoxication and substance withdrawal) were not included. The group referred to as dually diagnosed in this report consisted of patients who have both an SSD and an SUD, as they are defined above. The six DSM-IV psychosocial problems of interest (problems with primary support group, occupational problems, housing problems, economic problems, problems with access to health care services, and problems related to interaction with the legal system/ crime) also are consistent with DSM-IV definitions. Examples of each psychosocial problem from the DSM-IV were provided to participating psychiatrists in the glossary accompanying study materials. These examples were as follows: (1) problems with primary support group include death of a family member, health problems in family, disruption by separation/divorce/estrangement, removal from the home, remarriage of a parent, sexual or physical abuse, parental overprotection, neglect of a child, inadequate discipline, discord with siblings, or birth of a sibling; (2) occupational problems include unemployment, threat of job loss, stressful work schedule, difficult work conditions, job dissatisfaction, job change, or discord with boss or coworkers; (3) housing problems include homelessness, inadequate housing, unsafe neighborhood, or discord with neighbors or landlord; (4) economic problems include extreme poverty, inadequate finances, or insufficient welfare support; (5) problems with access to health care services include inadequate health care services, transportation to health care facilities unavailable, or inadequate health insurance; and (6) problems related to interaction with the legal system/crime include arrest, incarceration, litigation, or victim of crime. Several DSM-IV Axis IV variables were not included as outcomes in this analysis, including problems related to the environment, educational problems, and other psychosocial and environmental problems. Rather, the investigation targeted the six variables presented herein because they represent the specific core dimensions of social adjustment of greatest interest to

3 941 the research team. In addition to health services characteristics and axial diagnostic variables gathered by the SPPT, a number of sociodemographic characteristics were recorded, including age, gender, race/ethnicity, marital status, and employment status. The data were weighted with a three-stage propensity score weighting scheme to generate nationally representative estimates based on a random sample of APA members and their patients [12]. In the first stage, respondents of the 1996 National Survey of Psychiatric Practice (NSPP) were weighted to adjust for discrepancies in characteristics known about the general APA member population (including age, sex, race/ethnicity, etc.). In the second stage, PRN psychiatrists were weighted to adjust for discrepancies in characteristics known about the NSPP sample (including demographics, involvement in medical research, practice setting, etc.). The final stage adjusted for differences in the probability of any one patient being selected for the study as a function of the number of patients seen by the psychiatrist during the sampling period [12]. Findings presented herein are based on weighted estimates. Because weighted counts were not in integer values, numerical data presented in Tables 1 3 are not in whole numbers. Data were analyzed using the SUDAAN statistical software [18]. SUDAAN adjusts for the complex sampling design and the weighting procedures. In all analyses, standard errors were adjusted for clustering of patients within psychiatrists practices. Logistic regression models were used to determine ORs, and corresponding 95% confidence intervals (CIs) were calculated. Specifically, the crude association between SUD diagnosis and the six psychosocial problems were examined first. Then, a stratified analysis was conducted based on the presence or absence of a comorbid SSD diagnosis. Confounding was assessed by the databased method of comparing crude ORs to the adjusted ORs (aors) obtained from the stratified analysis. Effect modification was assessed by comparing the two stratum-specific ORs. To quantify the independent effect of SUD diagnosis and SSD diagnosis on each psychosocial problem studied, logistic regressions were conducted controlling for sociodemographic characteristics. Similarly, aors were calculated for the effect of the dual diagnosis (SSD and comorbid SUD) on the psychosocial problems, compared to those patients in the dataset without either diagnosis. Because the access to health care Axis IV problem may be a mediator of the effects of diagnosis on the other outcomes, a determination was made regarding whether or not the effect of SUD, SSD, or combined SUD and SSD was attenuated by good vs poor access to health care. That is, the access to health care variable was included as a predictor in the five logistic regression models of the other outcome variables. Additional models were run to assess for changes in effect estimates when adjusting for (1) private vs public treatment setting and (2) inpatient vs outpatient status. A final set of modeling was conducted to ascertain whether or not substance abuse that is limited to alcohol has a different effect on psychosocial problems compared to abuse that includes illicit substances. Results Among the 1,843 patients about whom detailed diagnostic and treatment information was obtained, 285 (15.5%) had one or more SUDs without an SSD and 180 (9.8%) had a diagnosis of an SSD without a comorbid SUD. Some 68 patients (3.7%) had a dual diagnosis of both an SUD and an SSD. The remaining 1,310 (71.1%) of the psychiatric patients in the dataset had diagnoses other than SSDs or SUDs. The exact diagnostic distributions for the three subgroups of interest are provided in Table 1. Basic sociodemographic data for all four subgroups (patients with one or more SUDs, patients with an SSD without an SUD, patients with both an SUD and an SSD, and patients with neither an SUD nor an SSD) are shown in Table 2. In general, significant differences between groups are in the expected directions. For example, patients with dual diagnosis were more likely to be younger and male, compared to their counterparts with SSDs without a comorbid SUD. Also as expected, patients with neither an SUD nor an SSD were more likely to be female, more educated, married, and employed, compared to those with both an SSD and an SUD. Table 3 shows the prevalence rates of each of the six Axis IV psychosocial problems in each of the three diagnostic groups of interest. Testing of differences between the SUD group vs the dually diagnosed group, Table 1 Diagnostic distributions of patients with one or more SUDs without an SSD (n=285), patients with an SSD without a comorbid SUD (n=180), and those dually diagnosed with both (n=68) SUD a SUD and SSD a n=285 (15.5%) n=68 (3.7%) Alcohol abuse or dependence (65.5%) 44.9 (65.7%) Cannabis abuse or dependence 49.7 (17.5%) 15.5 (22.7%) Cocaine abuse or dependence 33.3 (11.7%) 15.9 (23.2%) Opioid abuse or dependence 26.4 (9.3%) 0.8 (1.1%) Amphetamine abuse or dependence 8.2 (2.9%) 1.1 (1.7%) Hallucinogen abuse or dependence 0.3 (0.1%) 0 (0%) Polysubstance dependence 35.8 (12.6%) 10.0 (14.6%) SSD SUD and SSD n=180 (9.8%) n=68 (3.7%) Schizophrenia, paranoid type 81.4 (45.1%) 24.6 (36.0%) Schizophrenia, disorganized type 2.8 (1.5%) 0 (0%) Schizophrenia, catatonic type 0.8 (0.4%) 0 (0%) Schizophrenia, undifferentiated type 25.7 (14.2%) 10.2 (14.9%) Schizophrenia, residual type 5.3 (2.8%) 0.8 (1.1%) Schizophreniform disorder 4.1 (2.3%) 0.3 (0.4%) Schizoaffective disorder 60.7 (33.7%) 32.5 (47.6%) Counts may not be in integer values due to the use of weighted data a Percentages sum to greater than 100% because patients may have had more than one SUD

4 942 Table 2 Basic sociodemographic characteristics of the study subgroups Significant comparisons a SUD SSD Both SUD and SSD Neither SUD nor SSD n=285 (15.5%) n=180 (9.8%) n=68 (3.7%) n=1,310 (71.1%) Age (mean, SE) 40.9 (1.15) 45.0 (1.24) 37.4 (1.54) 40.2 (0.83) 2,3 2,4 Female gender (n, %) (41.3) 83.9 (48.1) 11.1 (16.3) (58.7) 1,3 1,4 2,3 3,4 Race/ethnicity (n, %) Black/African American 39.3 (13.8) 28.7 (15.9) 16.2 (24.0) (8.6) NS Hispanic 20.8 (7.3) 18.6 (10.3) 4.5 (6.6) 88.6 (6.8) NS White (76.0) (70.6) 43.4 (64.2) (80.7) NS Other 8.3 (2.9) 5.8 (3.2) 3.6 (5.3) 52.1 (4.0) NS Years of education (mean, SE) 12.9 (0.27) 12.1 (0.24) 11.5 (0.39) 12.9 (0.19) 1,3 3,4 Never married (n, %) 91.1 (32.4) 94.7 (53.2) 41.0 (60.3) (40.6) 1,2 1,3 3,4 Not working (n, %) (66.6) (82.4) 58.1 (84.9) (57.1) 1,2 1,3 2,4 3,4 Patient s status (n, %) Inpatient 69.7 (24.5) 49.8 (27.6) 19.2 (28.1) (10.7) 1,4 2,4 Outpatient (64.7) (65.3) 46.1 (67.4) (83.7) 1,4 2,4 Partial/intermediate 30.7 (10.8) 12.8 (7.1) 3.1 (4.5) 74.2 (5.7) NS Counts may not be in integer values due to the use of weighted data a Based on analysis of variance and chi-square tests with a significance criterion adjusted for six pairwise comparisons (p ) NS not significant (at p ) and between the SSD group vs the dually diagnosed group revealed only two statistically significant differences. Dually diagnosed patients were significantly more likely to have housing problems than were patients with an SSD only (33.4 compared to 16.6%; t= 2.08; df=609; p=0.04). Dually diagnosed patients also were significantly more likely to have problems with access to health care than were patients with an SSD only (28.3 compared to 8.6%; t=2.45; df=609; p=0.01). When assessed singly, (SUD diagnosis without an SSD, and SSD without a comorbid SUD) both diagnostic categories were independently associated with the Axis IV psychosocial problem areas in the expected direction. There were no meaningful differences between the crude (unadjusted) ORs (pertaining to the impact of the presence of an SUD on each outcome) and the aors obtained from the stratified analysis (pertaining to the impact of an SUD on each problem, adjusted for the effect of the schizophrenia-spectrum diagnostic category). Thus, in the association between SUD diagnosis and Axis IV psychosocial problems, schizophrenia-spectrum diagnosis is not a confounder. Furthermore, stratum-specific estimates did not differ significantly, indicating that there was no evidence of effect modification occurring between the two diagnostic variables (when comparing the ORs pertaining to the impact of SUD diagnosis on each outcome stratified by presence or absence of a schizophrenia-spectrum diagnosis). After reviewing initial logistic regression models of the six Axis IV psychosocial problems containing only the two diagnostic categories of interest as independent variables (data not shown), models also containing relevant sociodemographic covariates were developed. In the determination of aors and 95% CIs, each of the three groups of interest (SUD, SSD, and both SUD and SSD) was compared to the group of patients in the dataset with neither an SUD nor an SSD (n= 1,310, 71.1%). After adjusting for these sociodemographic variables (age, gender, race/ethnicity, marital status, and employment status) and for SSD diagnosis, SUD diagnosis was independently associated with five of the six Axis IV psychosocial problems (Table 4). With respect to SUD diagnosis, significant aors for Axis IV problems ranged from 2.21 for occupational problems to 3.75 for legal system/crime problems. After adjusting for the sociodemographic variables and for SUD diagnosis, Table 3 Prevalence rates of psychosocial problems in the three diagnostic categories of interest, comparing patients dually diagnosed with both an SUD and an SSD to those with either diagnosis alone Axis IV psychosocial problem SUD SSD Both SUD and SSD n=285 (15.5%) n=180 (9.8%) n=68 (3.7%) Primary support (62.1%) 92.8 (51.4%) 36.6 (53.5%) Occupational (47.9%) 66.4 (36.8%) 33.3 (48.8%) Housing 58.2 (20.5%) 29.9 (16.6%) a 22.8 (33.4%) a Economic (41.2%) 76.3 (42.3%) 36.2 (53.0%) Access to health care 45.1 (15.9%) 15.5 (8.6%) b 19.3 (28.3%) b Legal system/crime 77.9 (27.4%) 10.5 (5.8%) 12.6 (18.5%) Counts may not be in integer values due to the use of weighted data a Statistically significant difference, p=0.04 b Statistically significant difference, p=0.01

5 Table 4 Adjusted odds ratios and 95% confidence intervals calculated from logistic regression models controlling for age, gender, race/ethnicity, marital status, and employment status SUD SSD Both SUD and SSD n=285 (15.5%) n=180 (9.8%) n=68 (3.7%) Axis IV psychosocial problem aor 95% CI aor 95% CI aor 95% CI Primary support , , , 1.92 Occupational 2.21* 1.59, , * 1.71, 4.50 Housing 2.66* 1.66, , * 2.00, 8.04 Economic 2.43* 1.71, * 1.44, * 3.04, 9.28 Access to health care 2.95* , * 2.02, 9.16 Legal system/crime 3.75* 2.30, , 4.49 In the determination of aors and 95% CIs, each of the three groups of interest (SUD, SSD, and both SUD and SSD) is compared to the group of patients in the dataset with neither an SUD nor an SSD (n=1,310; 71.1%). SUD includes alcohol and other substance of abuse/dependence * Statistically significant, p< schizophrenia-spectrum diagnosis was significantly independently associated only with Axis IV economic problems (aor=2.18). When compared to patients with neither an SUD diagnosis nor an SSD, dually diagnosed patients were at much higher risk for Axis IV psychosocial problems than were patients with an SUD diagnosis alone or an SSD alone. This was true for four of the six psychosocial outcomes assessed. The aors pertaining to problems with primary support group were not significant for either SUD alone, SSD diagnosis alone, or the dual diagnosis. For problems related to interaction with the legal system/crime, the SUD diagnosis alone conferred a greater risk, with no elevation in risk associated with the dual diagnosis. Dually diagnosed patients were at much greater risk for occupational problems, housing problems, economic problems, and problems with access to health care than were patients with either diagnosis alone. For example, when controlling for the effects of sociodemographic variables, those with an SUD diagnosis had an odds of economic problems 2.4 times that of psychiatric patients without an SUD diagnosis or a schizophrenia-spectrum diagnosis; those with an SSD diagnosis had an odds of economic problems 2.2 times that of psychiatric patients without an SUD diagnosis or a schizophreniaspectrum diagnosis; and those with the dual diagnosis had an odds of economic problems 5.3 times that of psychiatric patients without an SUD diagnosis or a schizophrenia-spectrum diagnosis. To determine whether or not the effects of SUD, SSD, or combined SUD and SSD were attenuated by good vs poor access to health care, the access to health care variable was included as a predictor in the five logistic regression models of the other outcome variables. The access to health care variable was significantly associated with each of the other five Axis IV psychosocial problems of interest. Including access to health care in the models resulted in relatively minor changes in the model coefficients, without clear evidence of mediation. For example in the model examining the occupational problems outcome, when including the access to health care variable as a predictor, the aor for SUD diagnosis decreased from 2.21 (1.59, 3.08) to 2.03 (1.46, 2.83). Similarly, in the model examining the economic problems outcome, when including the access to health care variable as a predictor, the aor for schizophrenia-spectrum diagnosis decreased from 2.18 (1.44, 3.31) to 2.15 (1.42, 3.25). Thus, while the relationship between diagnosis and the outcome variables may be partly confounded by access to health care, diagnosis and access to health care both Table 5 Adjusted odds ratios and 95% confidence intervals calculated from logistic regression models controlling for age, gender, race/ethnicity, marital status, and employment status AUD SSD Both AUD and SSD n=144 (7.8%) n=180 (9.8%) n=40 (2.2%) Axis IV psychosocial problem aor 95% CI aor 95% CI aor 95% CI Primary support , , , 1.70 Occupational 2.50* 1.63, , * 1.91, 6.36 Housing 2.38* 1.21, , * 1.34, 8.56 Economic 1.95* 1.22, * * 2.08, 8.40 Access to health care 3.05* 1.56, , * 1.68, 8.87 Legal system/crime , , 2.06 In the determination of aors and 95% CIs, each of the three groups of interest (AUD, SSD, and both AUD and SSD) is compared to the group of patients in the dataset with neither an SUD nor an SSD (n=1,310; 71.1%). AUD includes alcohol abuse and dependence without illicit drug abuse or dependence * Statistically significant, p<0.05

6 944 contribute independently to the outcomes in most cases. In no instances did the change in coefficients (before and after controlling for access to health care) exceed 20%. The model coefficients of the six logistic regression models were also assessed when adjusting for two important potential confounders: private vs public treatment setting and inpatient vs outpatient status. When adjusting for these characteristics in separate models, the model coefficients changed slightly, indicating some confounding. However, statistically significant aors remained significant, without remarkable changes compared to the initial effects estimates shown in Table 4. For example, in the model examining housing problems, when controlling for private vs public practice setting, the aor for SUD diagnosis decreased from 2.66 (1.66, 4.26) to 2.50 (1.48, 4.22), and the aor for schizophrenia-spectrum diagnosis decreased from 1.50 (0.90, 2.51) to 1.19 (0.69, 2.05). Similarly, also in the model examining housing problems, when controlling for inpatient vs outpatient status, the aor for SUD diagnosis decreased to 2.34 (1.43, 3.82), and the aor for schizophrenia-spectrum diagnosis decreased to 1.41 (0.85, 2.34). The six logistic regression models were applied to a more restricted sample to determine whether or not substance abuse that is limited to alcohol has a different effect on psychosocial problems compared to abuse that includes illicit substances. In these models, the AUD group (consisting of patients with alcohol abuse or alcohol dependence without SUDs involving illicit substances, n=144) were compared to the group of patients in the dataset with neither an SUD of any type nor an SSD (n=1,310). Similarly, the small group of patients with an AUD and an SSD (n=40) also was compared to the group of patients in the dataset with neither an SUD of any type nor an SSD. As shown in Table 5, in five of the six models, the aors were very similar to those obtained in the original model that included all SUDs, and the significance of the aors did not change. However, in the model examining problems related to interaction with the legal system/crime, the aor for the SUD diagnosis was 3.75 (2.30, 6.09) compared to the nonsignificant aor of 1.75 (0.90, 3.43) associated with AUD diagnosis. Similarly, the estimate associated with the SUD/schizophrenia-spectrum dual diagnosis was 1.92 (0.82, 4.49) compared to 0.72 (0.27, 2.06) with the AUD/schizophrenia-spectrum comorbidity. Of note, the aors and 95% CIs in the middle columns of Tables 4 and 5 (comparing the SSD group with the group of patients with neither an SUD nor an SSD) are slightly different (though the statistical significance of effect estimates did not change). This is because of subtle effects on parameter estimates due to the sample restriction, which changed the underlying contingency tables. Discussion No previous research could be located from a representative sample of patients in routine US psychiatric practice that provides effect estimates of the impact of SUDs, SSDs, and combined SUD and SSD (dual diagnosis) on the risk for distinct Axis IV psychosocial problems. The results of these analyses are best summarized as six specific main findings. First, SUD diagnoses are clearly associated with five of the six Axis IV psychosocial problems, when controlling for the effects of important sociodemographic characteristics and schizophrenia-spectrum diagnosis. This is not surprising given previous documentation that substance abuse and dependence have prominent adverse effects on psychosocial outcomes. Second, interestingly, when controlling for the effects of sociodemographic variables and SUD diagnosis, SSD diagnosis had an independent effect on the economic outcome only, when compared to other psychiatric patients (see below). Third, again when controlling for the effects of important sociodemographic variables, those with a dual diagnosis (both an SUD and an SSD) were at higher risk for occupational, housing, economic, and access to health care problems than were those with either diagnosis alone. Interestingly, as discussed further below, this was not true for Axis IV problems related to the legal system/crime those with an SUD alone were at greater risk for legal system/crime problems compared to other psychiatric patients (aor=3.75; 2.30, 6.09) than were those with the dual diagnosis (aor=1.92; 0.82, 4.49). The fourth and fifth main findings relate to further analyses using these logistic regression models with the addition of potential confounders. Including access to health care as a predictor in the other five models of Axis IV problems led to minimal changes in effect estimates. Similarly, controlling for private vs public treatment setting and controlling for inpatient vs outpatient status caused minimal changes in estimates. Sixth, when the samples of SUD patients and dual-diagnosis patients were restricted to alcohol abuse and dependence (excluding illicit substance abuse/dependence), similar effect estimates were obtained. One exception, however, was that the SUD category was strongly related to legal system/crime problems, whereas the AUD category was not. This is not surprising given the criminal difficulties that often accompany illicit drug abuse and dependence. Perhaps most importantly, these analyses revealed that when compared to psychiatric patients having neither an SUD nor an SSD, dual-diagnosis patients had a much higher risk for psychosocial problems than patients with either diagnosis alone. For example, compared to psychiatric patients with neither an SUD nor a schizophrenia-spectrum diagnosis, dually diagnosed patients had nearly three times the odds of

7 945 having occupational difficulties, about four times the odds of having problems in the areas of housing and access to health care, and more than five times the odds of having economic problems. The magnitude of these effects is impressive, carrying important implications for policy and treatment services. Clearly, there are complex associations between SUDs, SSDs, and psychosocial problems. The finding that dually diagnosed patients are at greatly increased risk for poor social outcomes is consistent with previous research [5, 13, 19]. Two of our findings are of particular interest. First, although both SUD diagnosis and dual diagnosis were significantly associated with most Axis IV psychosocial problems, schizophrenia-spectrum diagnosis without comorbidity was associated only with one of six psychosocial problems (economic difficulties), when compared to other psychiatric patients. Thus, in the context of routine US psychiatric care, much of the effect that SSDs have on worsening social outcomes beyond that of other mental illnesses may be related to sociodemographic factors and comorbid substance abuse. It could also be that those patients with schizophrenia-spectrum diagnoses without comorbid substance abuse have a form of the illness characterized by both better psychosocial outcomes and lower risk of substance abuse and dependence. It should, however, be emphasized that the overall levels of psychosocial problems observed among patients with other mental illnesses were relatively high and not drastically different from the rates reported for the SSD sample in Table 3. Second, we found that compared to other psychiatric patients, those with an SUD without an SSD were at greater risk for legal system/ crime problems than were those with the dual diagnosis. This suggests that patients in psychiatric care with schizophrenia-spectrum diagnoses with a comorbid SUD may be less likely to be arrested or incarcerated, or involved in litigation or victimization compared to those with SUD diagnoses without an SSD. This could be related to associated negative symptoms, a lower likelihood of being involved in the drug culture (e.g., possession, trafficking) aside from the use of the substance, or the diversion of patients with obvious severe mental illnesses from the criminal justice system to treatment settings. However, this finding could also be the result of a selection bias if those with one or more SUDs and legal/criminal problems were more likely to be in treatment than those without legal/criminal problems and if this were not true of patients with dual diagnosis and other psychiatric diagnoses. Another interesting finding is that cocaine abuse or dependence was more prevalent among the dual-diagnosis group compared to the group of patients with one or more SUDs only (23.2 and 11.7%, respectively). Conversely, opioid abuse or dependence was much less prevalent among the dual-diagnosis group (1.1 compared to 9.3% in the SUD group). The dual-diagnosis group was also characterized by a large predominance of males (83.7%) compared to the SUD group (58.7%). The study findings should be considered in light of several methodological limitations. First, the 1999 SPPT utilized a cross-sectional, observational research design that relied exclusively on the report of treating psychiatrists. Nonetheless, the survey design does offer certain strengths for the purposes of characterizing patients being treated by American psychiatrists. For example, the naturalistic, real-world data from a survey of volunteer and random APA members offer advantages over smaller observational reports that may be influenced by local trends in care or service structures. Second, the SUD categorization used for this report did not separate those with abuse vs dependence, though distinguishing between these two types of SUD may have important implications for assessment and prognosis in dually diagnosed individuals [3]. Third, regarding reported diagnoses by psychiatrists, structured diagnostic interviews were not used. Prior PRN research that compared rates of comorbidity from the SPPT and the National Comorbidity Survey suggests that some disorders (especially anxiety disorders) may be less likely to be identified and diagnosed as DSM-IV disorders by clinicians in routine psychiatric practice compared to patients assessed through comprehensive structured diagnostic interviews [21]. However, it is unlikely that SSDs would have been undiagnosed by practicing PRN psychiatrists. Fourth, it should be recognized that the six outcome variables of interest are not independent outcomes. For example, occupational problems are clearly associated with economic problems, and economic problems are associated with housing problems. Thus, the six domains studied herein should not be viewed as completely independent aspects of psychosocial functioning. Yet, given the general dearth of research on psychosocial functioning with this level of detail, these analyses sought to make use of as much specificity in the data as possible. Despite these shortcomings, the findings that emerged have potentially important implications for clinicians and policy-makers in the area of schizophrenia and substance abuse comorbidity. Patients with psychotic disorders who have a comorbid SUD have greatly compounded risks for psychosocial problems. Specialized integrated treatment approaches may provide greater benefit for the comorbid population in comparison to routine care [2, 7, 8, 23], though many substance abuse treatment programs do not accept patients with severe psychiatric disorders. In many communities, there are major deficiencies in psychosocial services for those with SSDs and comorbid substance abuse [6, 14, 16]. Recent research has shown that the treatment of comorbid mental illness and substance abuse appears to be inadequate, despite the well-known high prevalence of the comorbidity. For example, Rothbard et al. [15] recently documented that

8 946 rates of utilization of ambulatory substance abuse treatment services were very low in the range of 1 to 2% among those with schizophrenia receiving Medicaid in two large urban counties in Pennsylvania. They suggested that this may be related to an insufficient number of trained clinicians and dually licensed programs that can deliver integrated services. Furthermore, the substantial burden on service providers of obtaining and maintaining licensing permits from separate regulatory agencies to provide both mental health and substance abuse treatment also may represent a system-wide impediment to the implementation of appropriate integrated services for dually diagnosed patients. These findings, from a large sample of representative US psychiatric patients, demonstrate the ongoing need for improved services and policies for those specially burdened patients with the dual diagnosis of both an SSD and alcohol and/or illicit drug abuse or dependence. Based on the prevalence of comorbidity and its impact on functioning in multiple domains, substance abuse among those with SSDs represents an important public health problem requiring further research and special program planning. Š Acknowledgements These analyses were supported by a Glaxo- SmithKline Health Services Research Scholarship provided by the American Psychiatric Institute for Research and Education (APIRE) to the first author during his Postdoctoral Fellowship in Community Psychiatry/Public Health. Development and support of the American Psychiatric Practice Research Network (PRN) has been generously funded by the American Psychiatric Foundation, the John D. and Catherine T. MacArthur Foundation, the Center for Mental Health Services, and the Center for Substance Abuse Treatment. References 1. American Psychiatric Association (2000) Diagnostic and statistical manual of mental disorders, Text revision, 4th edn. American Psychiatric Association, Washington, DC 2. Barrowclough C, Haddock G, Tarrier N et al (2001) Randomized controlled trial of motivational interviewing, cognitive behavior therapy, and family intervention for patients with comorbid schizophrenia and substance use disorders. Am J Psychiatry 158: Bartels SJ, Drake RE, Wallach MA (1995) Long-term course of substance use disorders among patients with severe mental illness. Psychiatr Serv 46: Cantor-Graae E, Nordström LG, McNeil TF (2001) Substance abuse in schizophrenia: a review of the literature and a study of correlates in Sweden. Schizophr Res 48: Cuffel BJ, Shumway M, Chouljian TL et al (1994) A longitudinal study of substance use and community violence in schizophrenia. J Nerv Ment Dis 182: Drake RE, Mueser KT (2000) Psychosocial approaches to dual diagnosis. Schizophr Bull 26: Drake RE, McHugo GJ, Noordsy DL (1993) Treatment of alcoholism among schizophrenia outpatients: 4-year outcomes. Am J Psychiatry 150: Hellerstein DJ, Rosenthal RN, Miner CR (2001) Integrating services for schizophrenia and substance abuse. Psychiatr Q 72: Hunt GE, Bergen J, Bashir M (2002) Medication compliance and comorbid substance abuse in schizophrenia: impact on community survival 4 years after a relapse. Schizophr Res 54: Kavanagh DJ, McGrath J, Saunders JB et al (2002) Substance misuse in patients with schizophrenia: epidemiology and management. Drugs 62: Modestin J, Gladen CJS, Christen S (2001) A comparative study on schizophrenia patients with dual-diagnosis. J Addict Dis 20: Pincus HA, Zarin DA, Tanielian TL et al (1999) Psychiatric patients and treatments in 1997: findings from the American Psychiatric Practice Research Network. Arch Gen Psychiatry 56: Räsänen P, Tiihonen J, Isohanni M et al (1998) Schizophrenia, alcohol abuse, and violent behavior: a 26-year followup study of an unselected birth cohort. Schizophr Bull 24: Ridgely MS, Goldman HH, Willenbring M (1990) Barriers to the care of persons with dual diagnoses: organizational and financing issues. Schizophr Bull 16: Rothbard AB, Kuno E, Hadley TR et al (2004) Psychiatric service utilization and cost for persons with schizophrenia in a Medicaid managed care program. J Behav Health Serv Res 31: Salloum IM, Moss HB, Daley DC (1991) Substance abuse and schizophrenia: impediments to optimal care. Am J Drug Alcohol Abuse 17: Sevy S, Robinson DG, Solloway S et al (2001) Correlates of substance misuse in patients with first-episode schizophrenia and schizoaffective disorder. Acta Psychiatr Scand 104: Shah BV, Barnwell BG, Bieler GS (2001) SUDAAN user s manual, release 8.0: software for the statistical analysis of correlated data. Research Triangle Institute, Research Triangle Park, NC 19. Smith J, Hucker S (1994) Schizophrenia and substance abuse. Br J Psychiatry 165: Swofford CD, Scheller-Gilkey G, Miller AH et al (2000) Double jeopardy: schizophrenia and substance abuse. Am J Drug Alcohol Abuse 26: Wilk JE, West JC, Narrow WE et al (2002) Comorbidity patterns in routine psychiatric practice. Poster presentation American Psychiatric Association meeting 22. Zarin DA, Pincus HA, West JC et al (1997) Practice-based research in psychiatry. Am J Psychiatry 154: Ziedonis DM (2004) Integrated treatment of co-occurring mental illness and addiction: clinical intervention, program, and system perspectives. CNS Spectr 9:892 ( ,925)