|Home | About | Journals | Submit | Contact Us | Français|
The objective of this study was to identify long term factors associated with substance use problem among individuals affected by severe mental illness. Prospective data come from the 1994, 1998, and 2000 waves of the Maryland Mental Health Outcomes Survey conducted among a sub-cohort of adult Medicaid recipients affected by serious mental illness. We estimated factors associated with alcohol and drug problem, as well as a hierarchy of substance use problem severity constructed from the alcohol and drug problem outcomes. Drug problem was the strongest factor associated with alcohol problem, and vice versa. Conceptualizing alcohol and drug problem separately, and as a hierarchy of severity, revealed distinct profiles of significant factors. Further research is warranted to explore the utility of modeling substance use problem in terms of a hierarchy of severity.
Severe mental illness refers to a diagnosable mental, behavioral, or emotional disorder, with functional impairment that substantially interferes with one’s major life activities (Dickey, Normand, Weiss, Drake, & Azeni, 2002; Jones et al., 2004; Sokal et al., 2004). The most prevalent co-occurring condition is substance use as evidenced by estimates from the Epidemiologic Catchment Area (ECA) study indicating that up to 50% of individuals with serious mental illness have alcohol and/or drug problem (Regier et al., 1990). Routine screening for and assessment of substance use among persons with mental illness has become the accepted standard of care (RachBeisel, Scott, Dixon, 1999). Factors associated with comorbid mental disorders and substance use problem among treatment samples can be found in the literature; however, less is known about these correlates in community-based populations (see Kessler, 2004; Stinson et al., 2005). Further research is needed to understand the long-term factors associated with substance use problem among individuals affected with serious mental illness. This research is developed based on the multiple-risk-factor model within the general framework of the “secondary substance use models”, which posits that higher rates of comorbidity are the consequences of primary psychiatric illnesses leading to substance use problem (e.g.; Degenhardt, Hall, & Lynskey, 2003; Thomas, Randall, & Carrigan., 2003). Indeed, the literature indicates that serious mental illness is associated with factors such as anxiety- and depression-related symptoms, loss of functional abilities, support, financial well being and social stigma that increase vulnerability to substance use problem (Phillips and Johnson, 2001; Mueser et al., 2000). A variety of other risk factors among individuals with serious mental illness might contribute to vulnerability to substance use disorders, including poor interpersonal skills, lack of meaningful daily activities, and quality of neighborhoods (Compton, Weiss, West, & Kaslow, 2005; Mueser et al., 2000). The literature also indicates that individuals with co-occurring substance use problem were more likely to commit criminal offenses and have a history of increased rates of arrest and homelessness as compared to individuals affected with serious mental illness with no substance use problem (Grossman, Haywood, Cavanaugh, Davis, & Lewis, 1995; Swanson et al., 2006; Cuffel, Schumway, Chouljian, & McDonald, 1994; Wallace, Mullen, and Burgess, 2004;Drake, McLaughlin, Pepper, & Minkoff, 1991; Scott, 1993). Several adverse health consequences are associated with comorbid mental disorders and substance use problem, including exacerbation of psychiatric symptoms (Drake, Osher, & Wallach,1989), recurrent hospitalizations (Oflson et al., 1999; Swofford, Scheller-Gilkey, Miller, Woolwine, & Mance, 2000; Havassy, Alvidrez, & Owen, 2004), increased use of emergency services (Bartels et al., 1993; Curran et al., 2003), and premature mortality (Allebeck, 1989; Dembling, chen, & Vachon, 1999).
The present study makes several contributions to the literature. First, previous studies are generally based on treatment samples and thus are probe to selection bias, which limits their applicability to the community as a whole. Treatment-based studies also rely on diagnosis to identify a cluster of conditions that characterize serious mental illness and might not take into account associated disability. The definition of serious mental illness adopted in this study is used by the National Institute of Mental Health (NIMH, 1989) and represents the heterogeneity of individuals affected by this illness in the community in terms of variety of functional impairments (Schinnar, Rothbard, Kanter, & Jung, 1990; Ruggeri, Leese, Thornicroft, Bisoffi, & Tansella, 2000). Moreover, very few studies that examine long-term factors associated with substance use problem among individuals affected with serious mental illness (McLellan, Lewis, O’Brien, & Kleber, 2000; Galai, et al., 2003). Given the chronic aspects of these diseases, the present study further contributes to the literature by identifying factors associated with substance use problem over the course of seven years among a cohort of individuals with serious mental illness enrolled in the Maryland Medicaid program.
We used two methodological approaches to reach the aims of the study. First, using a more traditional approach to studying substance use problem, we modeled alcohol problem and drug problem as separate outcomes of interest. Any alcohol problem was compared to no alcohol problem controlling for drug problem; and any drug problem was compared to no drug problem controlling for alcohol problem. Second, using the framework adopted by the Substance Abuse and Mental Health Services Administration (SAMHSA, 2003), we constructed a hierarchy of substance use problem, comparing respondents with no substance use problem to those who had (1) alcohol problem only, (2) drug problem only, and (3) a combination of both alcohol and drug problem. This modeling approach allowed us to investigate the combined effects of alcohol and drug problem while teasing apart factors that may be specific to alcohol problem or drug problem.
Data originate from the Maryland Mental Health Patterns and Outcomes Study (MMPOS), a prospective, longitudinal study of adult Medicaid enrollees with serious mental illnesses. Eligible study participants were identified using four criteria: (1) resided in Baltimore or the Eastern Shore of Maryland during fiscal year 1993 (FY93) from July 1992 to June 1993); (2) continuously enrolled in Maryland Medicaid during FY92-94 (3) aged 21–62 in FY93; and (4) affected with MI in FY93. Individuals were classified as affected with serious mental illness if they: (a) were diagnosed with schizophrenia; bipolar disorder or major depression; or other mental health disorders; (b) had documented use of any mental health specialty visit; (c) had a documented inpatient diagnosis of bipolar disorder or major depression or other mental health disorders. A random sample of 15% (n=957) of MMPOS participants was selected to complete a self-reported questionnaire, administered by trained interviewers in FY94, FY98, and FY00. The core of each interview consisted of the same set of widely-ranging questions, including substance use, socio-demographic characteristics, functional and health status, peer and family support. The Johns Hopkins University institutional review board approved this study.
The original sample of 957 was reduced to 915 participants because the original cohort experienced 12 deaths prior to baseline and 30 subjects moved out of the sample geographic area. Cohort members could be interviewed at any of the three time points such that the composition of study participants varied throughout the study. At each time point, participants who were alive and enrolled in Medicaid were eligible to be interviewed. For the baseline survey in FY94, 681 participants completed the survey (74% response rate). At the first follow-up in FY98, 604 participants were interviewed (81% response rate). A total of 424 participants were interviewed during the second follow-up in FY00 (60% response rate). Finally, we created a long form data file with a total of 812 unique study participants that contributed at least one data point for the analyses.
As noted earlier we investigated three outcomes: alcohol problem, drug problem, and a hierarchy of substance use problem. The CAGE screener was used to identify study participants with alcohol problem during the past year. The CAGE questionnaire is an internationally used assessment instrument for identifying problem with alcohol (Ewing, 1984; Shields & Caruso, 2004). A modified version of the CAGE, referred to here as the CAGE-DRUG, was used to assess drug problem. Illegal drugs included heroin, other opiates, diverted methadone, sedatives, cocaine/crack, amphetamines, hallucinogens, marijuana, and inhalants. The following questions were posed to all study participants who reported past year use of alcohol and/or illegal drugs: (1) “have you ever felt that you should cut down on your [drinking/drug use]?” 2) “have people annoyed you by criticizing your [drinking/drug use]?” 3) “have you felt bad or guilty about your [drinking/drug use]?” and 4) “have you had a drink first thing in the morning to steady your nerves or to get rid of a hangover (eye-opener)” or “have you taken drugs to steady your nerves or to stop symptoms of withdrawal?” The CAGE has been validated in several studies. The CAGE’s sensitivity in various populations ranges from 61 to 100 percent, and its specificity ranges from 77 to 96 percent (Cherpitel 1998). A meta-analysis of the reliability and validity of the CAGE revealed a median internal consistency of 0.74 (Shields & Caruso, 2004).
A score of one or greater in the CAGE and CAGE-DRUG indicates problem with alcohol and drug, respectively. If the CAGE score was missing or zero and the study participant reported past year formal or informal alcohol treatment, he/she was classified as having alcohol problem. Alcohol treatment included attending Alcoholics Anonymous meetings, taking Antabuse, talking to a therapist about alcohol problem, or participating in group therapy programs for alcohol problem. Employing the same convention, study participants with drug problem also included those who received formal or informal drug treatment during the past year. Past year drug treatment included attending Narcotics Anonymous meetings; participating in drug detox or other recovery program; talking to a therapist about drug problem; and participating in group therapy programs for drug problem. Finally, a hierarchy of substance use problem was constructed for the study participants, all affected with serious mental illness, as follows: (1) no alcohol problem or drug problem; (2) with alcohol problem only; (3) with drug problem only; and (4) with both alcohol and drug problem.
Several socio-demographic characteristics were controlled for in the analysis, including gender (male vs. female); race (white vs. non-white); age (<30, 30–45, 46+); educational attainment (< high school (HS) diploma, HS diploma, >HS diploma); marital status (married, widowed/separated/divorced, and never married), and residence (urban vs. rural). Past year arrest for any crimes other than a traffic violation was coded as a categorical variable (no arrest, 1 arrest, 2+ arrests). Medicare status (yes/no), lifetime diagnosis of schizophrenia (yes/no), major depression (yes/no), and bipolar disorder (yes/no) were operationalized as binary variables.
Four measures of health service use were included as covariates: having a usual source of care for physical or mental/emotional health (yes/no), having an outpatient case manager during the past year to help plan and obtain services (yes/no), past year mental health visit (yes/no), and past year mental health overnight stay (yes/no). Three subjective quality of life domains were considered: overall life satisfaction, family relationships, and social relations. These items were specifically developed to assess different aspects of quality of life among individuals with severe mental illness (Lehman, 1983). Likert response patterns to these items correspond to the Delighted-Terrible scale (1=Terrible, 7=Delighted) (Andrews & Withy, 1976). Overall life satisfaction was using the following question: “how do you feel about your life as a whole?” (alpha = 0.84). Two questions were used to measure quality of life related to family relationships: (1) “how do you feel about the way you and your family act toward each other? and (2) “how do you feel about the way things are in general between you and your family?” (alpha= 0.85). The social relations domain of quality of life was quantified using three questions: (1) “how do you feel about the things you do with other people?” (2) “How do you feel about the amount of time you spend with other people?” and (3) “how do you feel about the people you see socially?” (alpha = 0.82). Quality of life constructs were operationalized as continuous variables.
Frequencies and bivariate logistic regression models were used to examine the association between the covariates and the two main outcomes: alcohol problem and drug problem. Variables that were significant at 5% level for either of the outcomes were included in the multivariable logistic models for that particular outcome. We used random intercept models with a logistic link function to identify covariates associated with alcohol problem and drug problem, respectively. The random effects approach was chosen to account for the variance attributed to individual level factors, most notably the proclivity for alcohol and/or drug problem. For the hierarchy of substance use problem we used a multinomial logistic model that included all of the covariates. This multivariate approach estimates a series of binary logistic regressions comparing each level of the dependent variable to the reference category (respondents with no substance use problem). The resulting relative risk ratios should be interpreted as the odds of having alcohol problem, the odds of having drug problem, and the odds of having both alcohol and drug problem; relative to individuals with serious mental illness with no alcohol or drug problem, respectively.
Time was included in all of the regression models to control for changes in substance use problem among respondents during the study period. We used robust standard errors to account for intra-observation correlation. We used probability weights that were developed in the data to adjust for non-response and sampling design, but inclusion of weights did not significantly alter inferences. Unweighted results were thus reported. All of the analyses were conducted using STATA 9.0 (StataCorp, 2005).
The unadjusted associations between covariates and alcohol problem are presented in Table 1. Overall, about 19% of the sample had alcohol problem. Individuals with co-occurring drug problem had almost ten times the odds of having alcohol problem as compared those with no drug problem (OR=9.34; 95% CI: 6.75, 12.93). Being female (OR=0.36; 95% CI: 0.27, 0.48) and living in a rural as compared to an urban environment (OR=0.61; 95% CI: 0.43, 0.86) were negatively associated with having alcohol problem. Compared to whites, Non-Whites had higher odds of being classified as having alcohol problem (OR=1.63; 95% CI: 1.24, 2.16). Being hospitalized for a mental or emotional impairment increased the likelihood of alcohol problem (OR=2.53; 95% CI: 1.81, 3.53). Having a lifetime diagnosis of schizophrenia (OR=2.31; 95% CI: 1.75, 3.05) or bipolar disorder (OR=1.46; 95% CI: 1.10, 1.94) increased the likelihood of alcohol problem. A slight dose-response was detected for past year arrest history (ref: no past year arrest) with the odds of alcohol problem being 3.53 for individuals arrested once during the previous calendar year (OR=3.53; 95% CI: 2.03, 6.14) and 4.52 for those arrested two or more times during the year (OR=4.52; 95% CI: 2.59, 7.92). Family support was negative associated with having alcohol problem (OR=0.90; 95% CI: 0.83, 0.97).
As expected, alcohol problem was the most important factor associated with drug problem, increasing the odds of drug problem by almost ten-fold (OR=9.34; 95% CI: 6.75; 12.93). While the magnitude of the estimates varied, the factors associated with alcohol problem were also related to drug problem (see Table 2). Being female (OR=0.55; 95% CI: 0.41, 0.73) and living in a rural (OR= 0.37; 95% CI: 0.24, 0.56) were negatively associated with drug problem. Non-Whites had higher odds of drug problem (OR=1.40; 95% CI: 1.05, 1.88). Being hospitalized for a mental or emotional problem (OR=2.40; 95% CI: 1.31, 3.53), and having a lifetime diagnosis of schizophrenia (OR=1.48; 95% CI: 1.11, 1.99), or bipolar disorder (OR=1.86; 95% CI: 1.38, 2.51) increased the odds of drug problem. A slight dose-response was also detected for past year arrest history with the odds of drug problem slightly higher for history of two or more arrests (OR=6.53; 95% CI: 3.72, 11.47) than for history of one arrest (OR=5.47; 95% CI: 3.15, 9.50). Several additional covariates were associated with drug problem. Compared to those under the age of 30, individuals over the age of 46 had a decreased likelihood of drug problem (OR=0.41; 95% CI: 0.22, 0.79). Individuals with educational attainment beyond high school had almost twice the odds of having drug problem relative to those who did not complete high school (OR=1.98; 95% CI: 1.36, 2.89). Ever being diagnosed with major depression was associated with drug problem (OR=1.76; 95% CI: 1.29, 2.41). The odds of having drug problem were lower for individuals with a case manager as compared to those without a cage manager (OR=0.68; 95% CI: 0.49, 0.94). All of the quality of life constructs were negatively associated with drug problem, naming overall life satisfaction (OR=0.87; 95% CI: 0.79, 0.96), family support (OR=0.82; 95% CI: 0.75, 0.89), and peer support (OR=0.83; 95% CI: 0.74, 0.93).
Table 3 presents the adjusted odds for alcohol problem among the study sample. No time trend for the propensity of alcohol problem over the study period was detected. The strongest factor associated with alcohol problem stayed co-occurring drug problem (OR=12.33; 95% CI: 6.53, 23.30). Female sex was negatively associated with alcohol problem (OR=0.22; 95% CI: 0.12, 0.39). Non-Whites had twice the odds of alcohol problem OR=2.04; 95% CI: 1.18, 3.54) as compared to Whites. A lifetime diagnosis of schizophrenia increased the odds of alcohol problem (OR=1.80; 95% CI: 1.03, 3.15). The association between past year mental health hospitalization and alcohol problem was significant in the bivariate analyses but attenuated in the multivariable model (OR=1.74; 95% CI: 1.01, 3.02). Residing in a rural area, past year arrest, lifetime diagnosis of depression or bipolar disorder, and perceived family support were no longer associated with alcohol problem in the adjusted multivariable model.
Table 4 presents the adjusted odds for drug problem among the study sample. No time trend was detected for a change in drug problem over the study period. Respondents with co-occurring alcohol problem had over ten times the odds of drug problem compared to their counterparts with no alcohol problem (OR=10.93; 95% CI: 5.69, 21.00). Living in a rural as compared to an urban environment was negatively associated with developing drug problem (OR=0.33; 95% CI: 0.15, 0.74). Relative to no past arrest history, individuals with two or more arrests had significantly higher odds of drug problem (OR=10.12; 95% CI: 3.22, 31.81). Perceived family support continued to have lower odds in the adjusted model (OR=0.78; 95% CI: 0.65, 0.94). Sex, race/ethnicity, age, educational attainment, psychiatric diagnoses, having a case manager and/or a usual source of care, being hospitalized for a mental or emotional condition, and the additional quality of life measures lost their significant association with drug problem in the adjusted multivariable model.
Results from the multinomial logistic regression model are presented in Table 5. No substance use problem (i.e. serious mental illness only) is the reference category. A statistically significant linear decrease in all forms of substance use problem was detected over the study period. The results indicate that being female was negatively associated with having alcohol problem by itself (OR=0.33; 95% CI: 0.20, 0.55) or in combination with drugs (OR=0.30; 95% CI: 0.17, 0.54). No sex differences were detected between respondents with drug problem only relative to those with no substance use problem. The relative risk of having alcohol problem and drugs over no substance use problem was twice for nonwhites relative to whites (OR=2.05; 95% CI: 1.14, 3.71). Age did not differentiate respondents with either alcohol or drug problem from no substance use problem; but being over the age of 46 reduced the odds of co-occurring alcohol and drug problem for the respondents (OR=0.05; 95% CI: 0.01, 0.35). Educational attainment was negatively associated with alcohol problem. The odds for having alcohol problem itself were lower for respondents with a high school degree (OR=0.48; 95% CI: 0.27, 0.84) or beyond (OR=0.40; 95% CI: 0.19, 0.87) relative to less than high school. Having completed high school was also negatively associated with developing alcohol and drug problem (OR=0.41; 95% CI: 0.21, 0.83).
In addition to demographic variables, several other factors were significant in the multinomial logistic model of hierarchy of substance use problem. Having a past year mental health hospitalization increased the odds of alcohol problem (OR=2.36; 95% CI: 1.33, 4.18), drug problem (OR=1.95; 95% CI: 1.01, 3.78), or both alcohol and drug problem (OR=4.31; 95% CI: 2.22, 8.38), relative no substance use problem. Being arrested two or more times relative to no arrest increased the odds for alcohol problem (OR=4.23; 95% CI: 1.10, 16.23), the odds for drug problem (OR= 15.68; 95% CI: 3.69, 66.67); as well as the odds of both alcohol and drug problem (OR=18.03; 95% CI: 4.48, 72.51). Family support decreased the odds of drug problem by itself (OR=0.83; 95% CI: 0.69, 0.99) or in combination with alcohol problem (OR=0.77; 95% CI: 0.63, 0.93) in comparison to individuals without substance use problem.
The present study has several noteworthy findings. More than one in seven individuals (>15%) affected with serious mental illnesses in Maryland Medicaid had alcohol and/or drug problem between 1994 and 2000. This prevalence is consistent with several findings in the literature (see Drake, Osher, & Wallach, 1989; Bartels et al., 1993; Currie et al., 2005) but our finding is higher than the 3.8% prevalence among individuals in the general US population who met criteria for past year substance use problem (Kessler, Chiu, Demler, Merikangas, & Walters, 2005). The bivariate results indicate considerable overlap among demographic factors associated with alcohol and drug problem; however, adjusting for confounders provided insight into distinct profiles of factors. Females had lower odds of alcohol problem while no sex difference was detected for the odds of drug problem. The negative association between being female and alcohol problem is consistent with previous findings in the literature (Drake et al., 1989; Moore et al., 2005), but research is mixed regarding the association between gender and drug problem. A higher prevalence of males with drug problem is typically found among treatment samples (Alexander, 1996; Dixon et al., 1998; Mueser et al., 2000) but population-based studies often do not detect this association (Van Etten, Neumark, & Anthony, 1999; Hernandez-Avila, Rounsaville, & Kranzler, 2004).
Bivariate estimates also suggest that lifetime psychiatric diagnoses were important factors associated with substance use problem among individuals affected with serious mental illness. However, in the adjusted multivariable models, the strength of the association of ever being diagnosed with schizophrenia, major depression, and/or bipolar disorder was minimized. Only a lifetime diagnosis of schizophrenia maintained its significant association with alcohol problem. Such proclivity of individuals with schizophrenia to alcohol problem has been well-documented among treatment samples (Drake et al., 1989; Mueser et al., 2000; Swofford et al., 2000; Gandhi, Bogrov, Osher, & Myers, 2003; Margolese, Carlos, Tempier, & Gill, 2006). The present study expands the investigation of the association of psychiatric diagnosis with substance use problem to include affective disorders in addition to schizophrenia. In contrast to conclusions drawn by Regier et al. (1990), we did not find significant associations between affective disorders and substance use problem. These null findings may be attributed to the homogeneity of our sample in terms of disorder severity. Indeed, by limiting our study population to those meeting the more stringent criteria of the NIMH definition of serious mental illness, we were constrained to those individuals with disabling mental illness.
Past year legal involvement did not differentiate individuals affected with serious mental illness on the basis of alcohol problem. In contrast, being arrested two or more times during the past year significantly increased the odds of drug problem. Previous studies have considered alcohol and drug problem separately and found significant associations between legal involvement and alcohol problem as well as drug problem (Grossman et al., 1995; Swanson et al., 2006). Combining alcohol and drug problem may mask the individual associations between type of substance use problem and arrest history. Research is needed to uncover the types of crimes committed by individuals affected with co-occurring serious mental illness and substance use problem. It is worth noting that the strongest factor associated with alcohol and drug problem, respectively, was the abuse of the other substance type. The magnitude of this relationship provides further evidence for the need to operationalize substance use problem in terms of a hierarchy of severity among individuals with serious mental illness.
Modeling substance use problem as a hierarchy of severity strengthens the study findings and provides a clearer understanding of the long term factors associated with substance use problem, prevalent among populations affected with serious mental illness. Those with alcohol and drug problem tended to be younger and African-American as compared to individuals with no substance use problem. Race/ethnicity did not distinguish additional patterns of substance use problem. Female sex and educational attainment were negatively associated with alcohol problem by itself as well as in combination with drug problem. Past year history of two or more arrests were associated with higher odds of any for of substance use problem. Having been hospitalized for a mental or emotional condition was associated with increased substance use problem, which might indicate missed opportunities for treatment (Leon et al., 1998). Of the quality of life domains under study, only perceived family support was found to be negatively associated with drug problem alone or in combination with alcohol problem. This conclusion is substantiated by research suggesting that while social support is important for individuals with serious mental illness, this support more often arises from family than peers (Katschnig, Freeman, & Sartorius, 2005).
Finally, it is to note that the data are self-reported (except death) with possible recall bias and social desirability bias although research has shown self-report substance use problem data to be valid (Del Boca & Noll, 2000). While acknowledging the limitations, the present study has offsetting strengths. Of utmost importance is the ability to operationalize serious mental illness in terms of diagnosis and associated disability in accordance with the NIMH definition. Moreover, alcohol and drug problem were considered separately, bearing out distinct profiles of factors depending on the type of substance use problem. The analysis of the hierarchy of substance use problem severity supported this dichotomy of substance use problem and suggested that an additional subgroup of comorbid alcohol and drug problem be recognized as distinct. Recognizing these differences can facilitate the development of more tailored prevention and intervention programs for Medicaid enrollees affected with serious mental illness.
This work was supported by grants from the National Institute on Drug Abuse (T32 DA07292, PI: Steinwachs; DA020630, PI: Alexandre) and the National Institute of Mental Health (R01 MH49250, PI: Steinwachs). This research would not have been possible without the cooperation and support from the Maryland Medicaid Program.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.