VOL. 16, NO. 1, 1990
Diagnosis of Alcohol Use Disorders in Schizophrenia
by Robert E. Drake, Fred C Osher, Douglas L. Noordsy, Stephanie C Hurlbut, Gregory B. league, and Malcolm S. Beaudett
Abstract
Reprint requests should be sent to Dr. R.E. Drake, West Central Services, 2 Whipple Place, Lebanon, NH 03766.
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
Alcohol use disorders are common comorbid conditions in schizophrenia, and their presence is associated with poor adjustment and poor treatment response. Standard alcohol assessment instruments have not been' validated for use with schizophrenic patients, and several authors have questioned the validity of these patients' selfreports. A reliable and valid screening procedure for assessing alcohol use is needed. The present study used the following three methods to evaluate a rural sample of 75 outpatients with DSM-IU-R schizophrenia or schizoaffective disorder: (1) clinical records; (2) research interviews using standard alcohol assessment instruments; and (3) case managers' ratings. In addition, consensus diagnoses, determined by combining information from all three methods with intensive case reviews, were used to determine the sensitivity and specificity of the other approaches. As expected, clinical evaluations frequently missed alcohol problems. Research interviews and case managers' ratings differentiated between alcoholic and nonalcoholic schizophrenic patients and were highly correlated. Case managers' ratings, which incorporated longitudinal observations of behavior and collateral reports as well as interview data, were more sensitive measures of current alcohol use disorders thanresearchinterviews. Subjects frequently manifested alcoholrelated problems that interfered with community adjustment without the full dependence syndrome, suggesting that schizophrenic patients may be particularly vulnerable to negative effects of alcohol.
The course of schizophrenia is frequently complicated by alcohol use. Freed (1975) reviewed studies published before 1975 and found that the reported prevalence of alcohol abuse in schizophrenia ranged from 3 to 63 percent. More recent U.S. studies have determined that between 14 and 47 percent of treated schizophrenic patients have alcohol use disorders (McLellan and Druley 1977; Alterman et al. 1981; OTarrell et al. 1983; Barbee et al. 1989; Drake et al. 1989) or are frequent, heavy users of alcohol (Test et al. 1985, 1990). The Epidemiologic Catchment Area (ECA) studies have documented the high comorbidity of alcohol use disorders and schizophrenia in community samples (Boyd et al. 1984). Aggregated data from the five ECA sites showed that schizophrenia occurred four times as frequently in alcoholics as in nonalcoholics (Helzer and Pryzbeck 1988). Though some recent studies highlight the tendency of schizophrenic patients to abuse psychotomimetic drugs (Schneier and Siris 1987), alcohol is probably the dmg they most frequently abuse (Barbee et al. 1989; Drake et al. 1989; Test et al. 1989). Alcohol use disorders in schizophrenia have been associated with several aspects of poor adjustment and poor outcome including delusions (Barbee et al. 1989), hallucinations (Noordsy et al., submitted for publication), depressive symptoms (Drake et al. 1989), disruptive behaviors (Alterman et al. 1980; Drake et al. 1989), assaultiveness (Yesavage and Zarcone 1983), poor self-care (Alterman et al. 1980; Drake et al. 1989), housing instability and
SCHIZOPHRENIA BULLETIN
58
longitudinal behavioral observations and collateral information to assess substance abuse in psychiatric patients. Confidence in assessing alcohol use is generally heightened by using several approaches concurrently and establishing concordance among them (Skinner 1984), a procedure that has thus far been neglected in studies of schizophrenia. More specifically, studies have not yet compared standard alcohol assessment instruments and case managers' ratings. Diagnosis is widely acknowledged to be difficult when both substance abuse and psychiatric symptoms are present (Lehman et al. 1989; Weiss and Mirin 1989). Several authors have attempted to validate alcohol use assessments in schizophrenic patients. Alterman et al. (1984) examined patients with clinically determined dual diagnoses of schizophrenia plus alcoholism in relation to research diagnostic interviews. Of patients with clinically determined dual diagnoses, 81 percent met lifetime diagnostic criteria for alcoholism and 69 percent met criteria for both disorders. Using a 5-point scale based on the severity of alcohol-related problems, Drake et al. (1989) found that trained clinicians were able to rate the alcohol use of schizophrenic patients with high interrater reliability and that their ratings were correlated with previous hospital diagnoses of alcoholism. Test et al. (1989) compared schizophrenic patients' selfreports of amount and frequency of alcohol use, and their judgments of whether their drinking was problematic, with assessments made by their case managers. Patients were less likely to report frequent drinking than their case managers were (27 vs. 46 percent), and heavy drinkers among them were less
likely than case managers to label their drinking as problematic (50 vs. 93 percent). For the most part, standard alcohol assessment instruments have been validated with primary alcoholics and not with dually diagnosed psychiatric patients. One exception, the Michigan Alcoholism Screening Test (MAST; Selzer 1971), has been used extensively with psychiatric patients. Although generally quite reliable and valid as a screening tool for primary alcoholics, it has been found to have good sensitivity (88-98 percent) but poor specificity (36-89 percent) with psychiatric patients (Hedlund and Vieweg 1984). Its overall accuracy may therefore be unacceptably low whenever the ratio of nonalcoholics to alcoholics is high. Recently, Toland and Moss (1989) found that the MAST failed to differentiate between alcoholic and nonalcoholic schizophrenic patients. Although their alcoholic group scored higher on the MAST (mean ± SD - 16.7 ±12.8 for alcoholic schizophrenic patients vs. 8.1 ± 9.1 for nonalcoholic schizophrenic patients), the difference was not statistically significant in their small sample (20 in each group). Several false positives were attributed to confusion on specific items, such as whether hallucinations and hospitalizations were due to heavy drinking or schizophrenia, and to the instrument's failure to differentiate between recent and remote alcohol problems. The validity of clinical diagnosis of alcoholism as the criterion variable was not addressed. As part of our clinical research program on dual diagnosis (Teague et al. 1989), we compared several methods for assessing alcohol use in schizophrenia: clinical evaluations during hospitalizations, direct
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
homelessness (Drake et al. 1989; Osher et al., in preparation), treatment noncompliance (Alterman et al. 1980; Drake et al. 1989), increased rates of other drug abuse (Barbee et al. 1989; DTake et al. 1989; Test et al. 1989), and increased rates of rehospitalization (Drake et al. 1989). The cross-sectional nature of existing studies and the strong associations of other drug abuse and medication noncompliance with alcohol abuse preclude causal inferences. Comparing studies is difficult because of the widely differing methodologies used. One reason for the variability in methods and findings is the lack of standardization in assessing alcohol use among schizophrenic patients. A reliable and valid assessment procedure is needed for clinical and research purposes (Ridgely et al. 1987; Toland and Moss 1989). In the absence of such a procedure, many studies have relied on clinicians' ratings, nonstandard instruments, or self-reported amounts of drinking, while others have used standard instruments without validation. Clinical diagnoses are generally insensitive measures of alcohol use disorders when compared with structured research interviews (McLellan and Druley 1977; Ananth et al. 1989). However, neuropsychological deficits decrease the accuracy of interview reports (Skinner and Sheu 1982), and schizophrenic patients may not be competent reporters of problems occasioned by alcohol. Two studies have found that structured interviews of schizophrenic patients failed to detect problematic drinking that was observed longitudinally by clinicians (Alterman et al. 1984; Test et al. 1989). Some authors (Safer 1987; Drake and Wallach 1989) have advocated using
59
VOL. 16, NO. 1, 1990
Methods Sample. The subjects for this study were schizophrenic outpatients treated by West Central Services, a comprehensive community mental health center in rural New Hampshire that is affiliated with Dartmouth Medical School. All active outpatient cases during 1987 with a clinical diagnosis of schizophrenia or schizoaffective disorder were considered for the study. Two research psychiatrists (R.E.D. and F.C.O.) screened hospital and mental health center records and interviewed clinically diagnosed schizophrenic patients to verify their diagnoses by DSM-III-R (American Psychiatric Association 1987) criteria. To increase confidence in the diagnoses of schizophrenia, we excluded patients for whom psychotic symptoms appeared only in association with substance abuse. Fifteen patients with inaccurate or questionable
diagnoses were eliminated. The remaining 79 patients with definite DSM-III-R schizophrenia or schizoaffective disorder were asked to participate in an interview about the role of alcohol and other drugs in their lives. Four patients, two of whom were considered alcohol abusers by their case managers, declined to participate. The final study sample of 75 patients had a mean age (± SD) of 43.6 (± 14.3) years. The sample was 48.0 percent male and 52.0 percent female. Only 14.7 percent of the sample were married; 60.0 percent were single, and 25.3 percent were separated, divorced, or widowed. Only 12.0 percent were competitively employed. Past hospitalization history, measured as total time spent in any mental hospital before the interview, ranged from a few months to 40 years (mean ± SD - 34.0 ± 77.6 months). The distribution of DSM-III-R diagnoses was as follows: schizophrenia, 89.3 percent (n = 67); and schizoaffective disorder, 10.7 percent (n = 8). Procedures and Measures. Alcohol use was independently assessed from three parallel perspectives: (1} clinical records, (2) research interviews using standard assessment instruments, and (3) ratings by clinical case managers. Information from the three perspectives was combined with intensive case review to produce consensus diagnoses. Clinical records. All available hospital discharge summaries were examined to determine whether alcohol-related problems had ever been diagnosed or detected during a hospital admission. Previous hospital discharge summaries were used as a measure of clinical practice
instead of mental health center records because hospital records are relatively independent of the case manager's perspective. Since hospitalizations occurred throughout the course of illness, clinical records were considered a lifetime measure. Five patients without discharge summaries were omitted from the analyses using this variable. The other 70 patients had a mean (± SD) number of discharge summaries available of 4.2 (± 3.1). We used a 3-point scale to indicate the rate of identification of alcohol use as a problem: never, once, and more than once. From previous work (Helzer et al. 1978; Robins et al. 1982; Ananth et al. 1989), we expected clinical records to underestimate secondary diagnoses of alcohol use disorders. Ginical records were therefore used to estimate the relative sensitivity of more structured approaches. Research interviews. Each subject was interviewed for approximately 1 hour by one of the authors who was without knowledge of other assessments. The completed research interviews, recorded verbatim, were reviewed independently by two raters to assign DSM-III-R diagnoses: no disorder, alcohol abuse, or alcohol dependence. Interrater agreement between the two independent raters on a subset of 15 randomly selected interviews was 100 percent for current and lifetime diagnoses. Research interviews included the following structured sections: 1. Alcohol Dependence Scale (ADS):
This 47-point scale consists of 25 multiple-choice questions and emphasizes the dependence syndrome, including compulsive drinking, loss of control, use of external supports to stop drinking, and frequent
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
patient interviews using standard alcohol assessment instruments, and case managers' ratings. Our sample of rural schizophrenic patients used alcohol much more frequently than other drugs (Osher et al., in preparation) and thus afforded an opportunity to examine alcohol use uncomplicated by the polydrug abuse that frequently occurs among urban schizophrenic patients (Breakey et al. 1974; Drake et al. 1989). Another advantage was that the sample was followed closely by clinical case managers who had low caseloads, training in substance abuse assessment, opportunity to make frequent behavioral observations in the community, and access to collateral information from families, community contacts, and other caregivers.
SCHIZOPHRENIA BULLETIN
60
al., submitted for publication). We also asked patients if they considered themselves alcoholic and if they thought that alcohol caused them problems. Ratings by clinical case managers. A 5-point case manager rating scale was developed as a research instrument for clinical case managers to assess the extent of alcohol-related problems among severely ill psychiatric patients in the community (Drake et al. 1989; Drake and Wallach 1989). An earlier version showed high interrater reliability and concurrent validity (Drake et al. 1989). Ratings of none (- 1), mild (= 2), moderate (- 3), severe (= 4), and extremely severe (=5) are anchored descriptively on the basis of severity of alcohol-related problems (see Appendix A). Mild indicates nonproblematic drinking; moderate indicates problematic drinking and corresponds to DSMHl-R alcohol abuse; and severe and extremely severe, which correspond to DSM-IH-R alcohol dependence, denote more serious symptoms and consequences. Case managers made current and lifetime ratings on the basis of knowledge of their clients' alcohol use during the previous 12 months and over their lifetimes, respectively. The k coefficients of interrater reliability, established by comparing ratings by clinical case managers and a team psychiatrist on a subset of 15 patients, were 055 (current) and 0.72 (lifetime). Consensus diagnosis. To establish a reference standard, we explicitly examined diagnostic disagreements between the case managers' ratings and direct interview ratings. Our research team determined the sources of disagreement and reached consensus diagnoses by first reviewing all clinical and research records, and then reviewing the cases with the entire treatment team. This procedure was used to assign current
and lifetime diagnoses according to DSM-U1-R. Data Analyses. Because several of our measures of alcohol use consist of rank-ordered scales with small numbers of rating categories, we used Kendall's T-C to measure the strength of association between scales. Kendall's T-C provides a conservative but valid measure of association for these types of data (Norusis 1987). The distinction between alcohol abuse and dependence has been controversial (Schuckit et al. 1985), and definitional criteria for abuse and dependence have been changed substantially from DSM-HI to DSMIII-R (Rounsaville and Kranzler 1989). The current distinction has not been validated in schizophrenia. Our data (Osher et al., in preparation) show few differences in adjustment between schizophrenic patients with alcohol abuse and those with alcohol dependence. Furthermore, for the purpose of planning treatment, identifying patients with alcohol use disorders (abuse or dependence) is probably the critical task (Osher and Kofoed 1989). We therefore focused our analyses on the detection of alcohol use disorder, defined as a DSM-III-R diagnosis of abuse or dependence.
Results Identification of Problem Drinkers. Table 1 summarizes rates of identification of problem drinkers (not all scales yield diagnoses) from different perspectives and using different scales. Depending on the approach, between 7 and 53 percent of the sample were identified as having a drinking problem. Rates of identification vary less when separated into measures of current and lifetime alcohol use.
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
withdrawal symptoms (Horn et al. 1984). The ADS refers to the previous 12 months and does not yield diagnoses. Instead, scores on the ADS are interpreted using the following cutoff points: no evidence of dependence (0), low level of dependence (1-13), moderate level of dependence (14-21), substantial level of dependence (22-30), and severe level of dependence (31-47). 2. CAGE: This 4-point scale consists of four questions that define the mnemonic: Have you felt you ought to Cut down on your drinking? Have people Annoyed you by criticizing your drinking? Have you ever felt bad or Guilty about your drinking? Have you had a drink first thing in the morning to steady your nerves or to get rid of a hangover (Eye-opener)? (Mayfield et al. 1974). Subjects who answer positively to two or more items are highly likely to have a lifetime alcohol problem (King 1986). 3. Michigan Alcoholism Screening Test (MAST): The MAST is a 24-item, 53-point scale developed as a brief screening instrument (Selzer 1971). As with the CAGE, the MAST does not refer to a specific time period and therefore gathers lifetime information. A score of 5 or more on the MAST indicates alcoholism; a score of 4 is suggestive of alcoholism; and a score of less than 4 indicates nonproblematic drinking. 4. Other: The interview included questions about the use of other drugs (marijuana, cocaine, speed, barbiturates, d-lysergic acid diethylamide, phencyclidine, and narcotics) in the past 12 months, use and misuse of medications prescribed by physicians, and current and lifetime amounts of consumption of alcohol and other drugs. We asked in detail about how nonpsychotic and psychotic symptoms were affected by alcohol and other drugs (Noordsy et
61
VOL. 16, NO. 1, 1990
Table 1. Frequencies of alcohol problems Cases above and below threshold No alcohol problems n (%)
Current alcohol use ADS Interview DSM-III-R rating Case manager rating Consensus DSM-III-R rating
5 14 18 19
6.7 18.7 24.0 25.3
70 61 57 56
93.3 81.3 76.0 74.7
20 29 40 33 32 38
28.6 38.7 53.3 44.0 42.7 50.7
50 46 35 42 43 37
71.4 61.3 46.7 56.0 57.3 49.3
Lifetime alcohol use Hospital identification CAGE MAST Interview DSM-III-R rating Case manager rating Consensus DSM-III-R rating Total n
75
75
Note.— ADS - Alcohol Dependence Scale. MAST - Michigan Alcoholism Screening Test. The following cutoffs were used: ADS > 14, DSM-III-R ratings > 2, case manager rating > 3, hospital identification > 2, CAGE > 2, and MAST > 5.
According to consensus diagnoses, 19 patients (25.3 percent) had a current DSM-III-R diagnosis of alcohol use disorder; 10 (13.3 percent) of these met criteria for dependence and 9 (12.0 percent) for abuse. Few subjects were identified as having moderate dependence by the ADS. The complete interview identified nearly three times as many problem drinkers as the ADS, and case managers identified an even larger proportion of subjects as problem drinkers. According to consensus lifetime diagnoses, 38 patients (50.7 percent) met DSM-III-R criteria for alcohol use disorders; 27 (36 percent) had been dependent, and 11 (14.7 percent) had been abusers. For subjects with lifetime consensus diagnoses, only half were identified as having an alcohol problem by even one
hospital discharge summary. The CAGE, case managers' ratings, and complete interviews identified more problem drinkers and yielded similar rates. The MAST identified the highest rate of alcohol use disorders. Agreement Between Measures. Table 2 shows that our various measures of alcohol use were highly interrelated. As expected, measures of current use were strongly correlated with each other (Kendall's T-C correlations between 0.44 and 0.60), and measures of lifetime use were highly correlated (correlations between 0.41 and 0.80), as compared with lower but still significant correlations between measures of current and lifetime use (correlations between 0.22 and 0.49). Thus, despite the use of a conservative
Consensus Diagnoses. Our investigations to determine the sources of disagreements and to establish consensus diagnoses revealed different results for current and lifetime assessments. For current alcohol use disorder, there were six disagreements on the distinction between no disorder and abuse, and all were resolved in consensus diagnoses of abuse. Only one patient's drinking was underestimated by the case manager but identified as abuse by interview; the case manager for this case was new, and both case records and more experienced team members quickly identified the patient as an abusive drinker. Five patients were identified as alcohol abusers by case managers but not by the complete interview. In each case, the patient denied problematic drinking at interview, but alcohol-related problems were identified by clinical records, reports from families and communities, and staff consensus. In other words, 26.3 percent of those with current alcohol use disorders (5 of 19) denied or significantly minimized their alcohol-related problems at interview. There were three disagreements on the distinction between abuse and dependence, and all were
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
Measure
Alcohol problems n (%)
measure of association (Kendall's T-C), these relationships were quite strong. With the case managers' 5-point scale collapsed into DSM-III-R diagnostic categories (no disorder, abuse, and dependence), the case manager and interview ratings agreed on 66 of 75 subjects (88.0 percent) for current diagnoses and on 57 of 75 subjects (76.0 percent) for lifetime diagnoses. Kendall's T-C correlations were 0.41 (p < 0.001) and 0.61 (p < 0.001), respectively.
SCHIZOPHRENIA BULLETIN
62
Table 2. Intercorrelations of alcohol measures Measures of lifetime alcohol use
Measures of current alcohol use Case Consensus manager DSM-III-R diagnosis rating
Consensus Interview Case DSM-III-R manager DSM-III-R diagnosis rating diagnosis
Hospital Identification
CAGE
MAST
0.262 0.30' 0.31' 0.31'
0.26* 0.222 0.32' 0.29'
0.35' 0.36' 0.40' 0.39'
0.41' 0.35' 0.45' 0.40'
0.36' 0.36' 0.49' 0.46'
0.41' 0.33' 0.49' 0.43'
0.44'
0.48' 0.66'
0.41' 0.60' 0.79'
0.44'
0.46' 0.68' 0.76' 0.801 0.79'
Current measures
ADS
0.46'
Interview Case manager Consensus
0.51' 0.44'
0.45' 0.46' 0.60'
Lifetime measures Hospital identification CAGE MAST Interview Case manager
0.60' 0.66' 0.65'
Note.—Full scales were used for all analyses. All correlations determined by Kendall's T-C. ADS - Alcohol Dependence Scale. MAST - Michigan Alcoholism Screening Test. ' p < 0.001. »p
Diagnosis of Alcohol Use Disorders in Schizophrenia
by Robert E. Drake, Fred C Osher, Douglas L. Noordsy, Stephanie C Hurlbut, Gregory B. league, and Malcolm S. Beaudett
Abstract
Reprint requests should be sent to Dr. R.E. Drake, West Central Services, 2 Whipple Place, Lebanon, NH 03766.
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
Alcohol use disorders are common comorbid conditions in schizophrenia, and their presence is associated with poor adjustment and poor treatment response. Standard alcohol assessment instruments have not been' validated for use with schizophrenic patients, and several authors have questioned the validity of these patients' selfreports. A reliable and valid screening procedure for assessing alcohol use is needed. The present study used the following three methods to evaluate a rural sample of 75 outpatients with DSM-IU-R schizophrenia or schizoaffective disorder: (1) clinical records; (2) research interviews using standard alcohol assessment instruments; and (3) case managers' ratings. In addition, consensus diagnoses, determined by combining information from all three methods with intensive case reviews, were used to determine the sensitivity and specificity of the other approaches. As expected, clinical evaluations frequently missed alcohol problems. Research interviews and case managers' ratings differentiated between alcoholic and nonalcoholic schizophrenic patients and were highly correlated. Case managers' ratings, which incorporated longitudinal observations of behavior and collateral reports as well as interview data, were more sensitive measures of current alcohol use disorders thanresearchinterviews. Subjects frequently manifested alcoholrelated problems that interfered with community adjustment without the full dependence syndrome, suggesting that schizophrenic patients may be particularly vulnerable to negative effects of alcohol.
The course of schizophrenia is frequently complicated by alcohol use. Freed (1975) reviewed studies published before 1975 and found that the reported prevalence of alcohol abuse in schizophrenia ranged from 3 to 63 percent. More recent U.S. studies have determined that between 14 and 47 percent of treated schizophrenic patients have alcohol use disorders (McLellan and Druley 1977; Alterman et al. 1981; OTarrell et al. 1983; Barbee et al. 1989; Drake et al. 1989) or are frequent, heavy users of alcohol (Test et al. 1985, 1990). The Epidemiologic Catchment Area (ECA) studies have documented the high comorbidity of alcohol use disorders and schizophrenia in community samples (Boyd et al. 1984). Aggregated data from the five ECA sites showed that schizophrenia occurred four times as frequently in alcoholics as in nonalcoholics (Helzer and Pryzbeck 1988). Though some recent studies highlight the tendency of schizophrenic patients to abuse psychotomimetic drugs (Schneier and Siris 1987), alcohol is probably the dmg they most frequently abuse (Barbee et al. 1989; Drake et al. 1989; Test et al. 1989). Alcohol use disorders in schizophrenia have been associated with several aspects of poor adjustment and poor outcome including delusions (Barbee et al. 1989), hallucinations (Noordsy et al., submitted for publication), depressive symptoms (Drake et al. 1989), disruptive behaviors (Alterman et al. 1980; Drake et al. 1989), assaultiveness (Yesavage and Zarcone 1983), poor self-care (Alterman et al. 1980; Drake et al. 1989), housing instability and
SCHIZOPHRENIA BULLETIN
58
longitudinal behavioral observations and collateral information to assess substance abuse in psychiatric patients. Confidence in assessing alcohol use is generally heightened by using several approaches concurrently and establishing concordance among them (Skinner 1984), a procedure that has thus far been neglected in studies of schizophrenia. More specifically, studies have not yet compared standard alcohol assessment instruments and case managers' ratings. Diagnosis is widely acknowledged to be difficult when both substance abuse and psychiatric symptoms are present (Lehman et al. 1989; Weiss and Mirin 1989). Several authors have attempted to validate alcohol use assessments in schizophrenic patients. Alterman et al. (1984) examined patients with clinically determined dual diagnoses of schizophrenia plus alcoholism in relation to research diagnostic interviews. Of patients with clinically determined dual diagnoses, 81 percent met lifetime diagnostic criteria for alcoholism and 69 percent met criteria for both disorders. Using a 5-point scale based on the severity of alcohol-related problems, Drake et al. (1989) found that trained clinicians were able to rate the alcohol use of schizophrenic patients with high interrater reliability and that their ratings were correlated with previous hospital diagnoses of alcoholism. Test et al. (1989) compared schizophrenic patients' selfreports of amount and frequency of alcohol use, and their judgments of whether their drinking was problematic, with assessments made by their case managers. Patients were less likely to report frequent drinking than their case managers were (27 vs. 46 percent), and heavy drinkers among them were less
likely than case managers to label their drinking as problematic (50 vs. 93 percent). For the most part, standard alcohol assessment instruments have been validated with primary alcoholics and not with dually diagnosed psychiatric patients. One exception, the Michigan Alcoholism Screening Test (MAST; Selzer 1971), has been used extensively with psychiatric patients. Although generally quite reliable and valid as a screening tool for primary alcoholics, it has been found to have good sensitivity (88-98 percent) but poor specificity (36-89 percent) with psychiatric patients (Hedlund and Vieweg 1984). Its overall accuracy may therefore be unacceptably low whenever the ratio of nonalcoholics to alcoholics is high. Recently, Toland and Moss (1989) found that the MAST failed to differentiate between alcoholic and nonalcoholic schizophrenic patients. Although their alcoholic group scored higher on the MAST (mean ± SD - 16.7 ±12.8 for alcoholic schizophrenic patients vs. 8.1 ± 9.1 for nonalcoholic schizophrenic patients), the difference was not statistically significant in their small sample (20 in each group). Several false positives were attributed to confusion on specific items, such as whether hallucinations and hospitalizations were due to heavy drinking or schizophrenia, and to the instrument's failure to differentiate between recent and remote alcohol problems. The validity of clinical diagnosis of alcoholism as the criterion variable was not addressed. As part of our clinical research program on dual diagnosis (Teague et al. 1989), we compared several methods for assessing alcohol use in schizophrenia: clinical evaluations during hospitalizations, direct
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
homelessness (Drake et al. 1989; Osher et al., in preparation), treatment noncompliance (Alterman et al. 1980; Drake et al. 1989), increased rates of other drug abuse (Barbee et al. 1989; DTake et al. 1989; Test et al. 1989), and increased rates of rehospitalization (Drake et al. 1989). The cross-sectional nature of existing studies and the strong associations of other drug abuse and medication noncompliance with alcohol abuse preclude causal inferences. Comparing studies is difficult because of the widely differing methodologies used. One reason for the variability in methods and findings is the lack of standardization in assessing alcohol use among schizophrenic patients. A reliable and valid assessment procedure is needed for clinical and research purposes (Ridgely et al. 1987; Toland and Moss 1989). In the absence of such a procedure, many studies have relied on clinicians' ratings, nonstandard instruments, or self-reported amounts of drinking, while others have used standard instruments without validation. Clinical diagnoses are generally insensitive measures of alcohol use disorders when compared with structured research interviews (McLellan and Druley 1977; Ananth et al. 1989). However, neuropsychological deficits decrease the accuracy of interview reports (Skinner and Sheu 1982), and schizophrenic patients may not be competent reporters of problems occasioned by alcohol. Two studies have found that structured interviews of schizophrenic patients failed to detect problematic drinking that was observed longitudinally by clinicians (Alterman et al. 1984; Test et al. 1989). Some authors (Safer 1987; Drake and Wallach 1989) have advocated using
59
VOL. 16, NO. 1, 1990
Methods Sample. The subjects for this study were schizophrenic outpatients treated by West Central Services, a comprehensive community mental health center in rural New Hampshire that is affiliated with Dartmouth Medical School. All active outpatient cases during 1987 with a clinical diagnosis of schizophrenia or schizoaffective disorder were considered for the study. Two research psychiatrists (R.E.D. and F.C.O.) screened hospital and mental health center records and interviewed clinically diagnosed schizophrenic patients to verify their diagnoses by DSM-III-R (American Psychiatric Association 1987) criteria. To increase confidence in the diagnoses of schizophrenia, we excluded patients for whom psychotic symptoms appeared only in association with substance abuse. Fifteen patients with inaccurate or questionable
diagnoses were eliminated. The remaining 79 patients with definite DSM-III-R schizophrenia or schizoaffective disorder were asked to participate in an interview about the role of alcohol and other drugs in their lives. Four patients, two of whom were considered alcohol abusers by their case managers, declined to participate. The final study sample of 75 patients had a mean age (± SD) of 43.6 (± 14.3) years. The sample was 48.0 percent male and 52.0 percent female. Only 14.7 percent of the sample were married; 60.0 percent were single, and 25.3 percent were separated, divorced, or widowed. Only 12.0 percent were competitively employed. Past hospitalization history, measured as total time spent in any mental hospital before the interview, ranged from a few months to 40 years (mean ± SD - 34.0 ± 77.6 months). The distribution of DSM-III-R diagnoses was as follows: schizophrenia, 89.3 percent (n = 67); and schizoaffective disorder, 10.7 percent (n = 8). Procedures and Measures. Alcohol use was independently assessed from three parallel perspectives: (1} clinical records, (2) research interviews using standard assessment instruments, and (3) ratings by clinical case managers. Information from the three perspectives was combined with intensive case review to produce consensus diagnoses. Clinical records. All available hospital discharge summaries were examined to determine whether alcohol-related problems had ever been diagnosed or detected during a hospital admission. Previous hospital discharge summaries were used as a measure of clinical practice
instead of mental health center records because hospital records are relatively independent of the case manager's perspective. Since hospitalizations occurred throughout the course of illness, clinical records were considered a lifetime measure. Five patients without discharge summaries were omitted from the analyses using this variable. The other 70 patients had a mean (± SD) number of discharge summaries available of 4.2 (± 3.1). We used a 3-point scale to indicate the rate of identification of alcohol use as a problem: never, once, and more than once. From previous work (Helzer et al. 1978; Robins et al. 1982; Ananth et al. 1989), we expected clinical records to underestimate secondary diagnoses of alcohol use disorders. Ginical records were therefore used to estimate the relative sensitivity of more structured approaches. Research interviews. Each subject was interviewed for approximately 1 hour by one of the authors who was without knowledge of other assessments. The completed research interviews, recorded verbatim, were reviewed independently by two raters to assign DSM-III-R diagnoses: no disorder, alcohol abuse, or alcohol dependence. Interrater agreement between the two independent raters on a subset of 15 randomly selected interviews was 100 percent for current and lifetime diagnoses. Research interviews included the following structured sections: 1. Alcohol Dependence Scale (ADS):
This 47-point scale consists of 25 multiple-choice questions and emphasizes the dependence syndrome, including compulsive drinking, loss of control, use of external supports to stop drinking, and frequent
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
patient interviews using standard alcohol assessment instruments, and case managers' ratings. Our sample of rural schizophrenic patients used alcohol much more frequently than other drugs (Osher et al., in preparation) and thus afforded an opportunity to examine alcohol use uncomplicated by the polydrug abuse that frequently occurs among urban schizophrenic patients (Breakey et al. 1974; Drake et al. 1989). Another advantage was that the sample was followed closely by clinical case managers who had low caseloads, training in substance abuse assessment, opportunity to make frequent behavioral observations in the community, and access to collateral information from families, community contacts, and other caregivers.
SCHIZOPHRENIA BULLETIN
60
al., submitted for publication). We also asked patients if they considered themselves alcoholic and if they thought that alcohol caused them problems. Ratings by clinical case managers. A 5-point case manager rating scale was developed as a research instrument for clinical case managers to assess the extent of alcohol-related problems among severely ill psychiatric patients in the community (Drake et al. 1989; Drake and Wallach 1989). An earlier version showed high interrater reliability and concurrent validity (Drake et al. 1989). Ratings of none (- 1), mild (= 2), moderate (- 3), severe (= 4), and extremely severe (=5) are anchored descriptively on the basis of severity of alcohol-related problems (see Appendix A). Mild indicates nonproblematic drinking; moderate indicates problematic drinking and corresponds to DSMHl-R alcohol abuse; and severe and extremely severe, which correspond to DSM-IH-R alcohol dependence, denote more serious symptoms and consequences. Case managers made current and lifetime ratings on the basis of knowledge of their clients' alcohol use during the previous 12 months and over their lifetimes, respectively. The k coefficients of interrater reliability, established by comparing ratings by clinical case managers and a team psychiatrist on a subset of 15 patients, were 055 (current) and 0.72 (lifetime). Consensus diagnosis. To establish a reference standard, we explicitly examined diagnostic disagreements between the case managers' ratings and direct interview ratings. Our research team determined the sources of disagreement and reached consensus diagnoses by first reviewing all clinical and research records, and then reviewing the cases with the entire treatment team. This procedure was used to assign current
and lifetime diagnoses according to DSM-U1-R. Data Analyses. Because several of our measures of alcohol use consist of rank-ordered scales with small numbers of rating categories, we used Kendall's T-C to measure the strength of association between scales. Kendall's T-C provides a conservative but valid measure of association for these types of data (Norusis 1987). The distinction between alcohol abuse and dependence has been controversial (Schuckit et al. 1985), and definitional criteria for abuse and dependence have been changed substantially from DSM-HI to DSMIII-R (Rounsaville and Kranzler 1989). The current distinction has not been validated in schizophrenia. Our data (Osher et al., in preparation) show few differences in adjustment between schizophrenic patients with alcohol abuse and those with alcohol dependence. Furthermore, for the purpose of planning treatment, identifying patients with alcohol use disorders (abuse or dependence) is probably the critical task (Osher and Kofoed 1989). We therefore focused our analyses on the detection of alcohol use disorder, defined as a DSM-III-R diagnosis of abuse or dependence.
Results Identification of Problem Drinkers. Table 1 summarizes rates of identification of problem drinkers (not all scales yield diagnoses) from different perspectives and using different scales. Depending on the approach, between 7 and 53 percent of the sample were identified as having a drinking problem. Rates of identification vary less when separated into measures of current and lifetime alcohol use.
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
withdrawal symptoms (Horn et al. 1984). The ADS refers to the previous 12 months and does not yield diagnoses. Instead, scores on the ADS are interpreted using the following cutoff points: no evidence of dependence (0), low level of dependence (1-13), moderate level of dependence (14-21), substantial level of dependence (22-30), and severe level of dependence (31-47). 2. CAGE: This 4-point scale consists of four questions that define the mnemonic: Have you felt you ought to Cut down on your drinking? Have people Annoyed you by criticizing your drinking? Have you ever felt bad or Guilty about your drinking? Have you had a drink first thing in the morning to steady your nerves or to get rid of a hangover (Eye-opener)? (Mayfield et al. 1974). Subjects who answer positively to two or more items are highly likely to have a lifetime alcohol problem (King 1986). 3. Michigan Alcoholism Screening Test (MAST): The MAST is a 24-item, 53-point scale developed as a brief screening instrument (Selzer 1971). As with the CAGE, the MAST does not refer to a specific time period and therefore gathers lifetime information. A score of 5 or more on the MAST indicates alcoholism; a score of 4 is suggestive of alcoholism; and a score of less than 4 indicates nonproblematic drinking. 4. Other: The interview included questions about the use of other drugs (marijuana, cocaine, speed, barbiturates, d-lysergic acid diethylamide, phencyclidine, and narcotics) in the past 12 months, use and misuse of medications prescribed by physicians, and current and lifetime amounts of consumption of alcohol and other drugs. We asked in detail about how nonpsychotic and psychotic symptoms were affected by alcohol and other drugs (Noordsy et
61
VOL. 16, NO. 1, 1990
Table 1. Frequencies of alcohol problems Cases above and below threshold No alcohol problems n (%)
Current alcohol use ADS Interview DSM-III-R rating Case manager rating Consensus DSM-III-R rating
5 14 18 19
6.7 18.7 24.0 25.3
70 61 57 56
93.3 81.3 76.0 74.7
20 29 40 33 32 38
28.6 38.7 53.3 44.0 42.7 50.7
50 46 35 42 43 37
71.4 61.3 46.7 56.0 57.3 49.3
Lifetime alcohol use Hospital identification CAGE MAST Interview DSM-III-R rating Case manager rating Consensus DSM-III-R rating Total n
75
75
Note.— ADS - Alcohol Dependence Scale. MAST - Michigan Alcoholism Screening Test. The following cutoffs were used: ADS > 14, DSM-III-R ratings > 2, case manager rating > 3, hospital identification > 2, CAGE > 2, and MAST > 5.
According to consensus diagnoses, 19 patients (25.3 percent) had a current DSM-III-R diagnosis of alcohol use disorder; 10 (13.3 percent) of these met criteria for dependence and 9 (12.0 percent) for abuse. Few subjects were identified as having moderate dependence by the ADS. The complete interview identified nearly three times as many problem drinkers as the ADS, and case managers identified an even larger proportion of subjects as problem drinkers. According to consensus lifetime diagnoses, 38 patients (50.7 percent) met DSM-III-R criteria for alcohol use disorders; 27 (36 percent) had been dependent, and 11 (14.7 percent) had been abusers. For subjects with lifetime consensus diagnoses, only half were identified as having an alcohol problem by even one
hospital discharge summary. The CAGE, case managers' ratings, and complete interviews identified more problem drinkers and yielded similar rates. The MAST identified the highest rate of alcohol use disorders. Agreement Between Measures. Table 2 shows that our various measures of alcohol use were highly interrelated. As expected, measures of current use were strongly correlated with each other (Kendall's T-C correlations between 0.44 and 0.60), and measures of lifetime use were highly correlated (correlations between 0.41 and 0.80), as compared with lower but still significant correlations between measures of current and lifetime use (correlations between 0.22 and 0.49). Thus, despite the use of a conservative
Consensus Diagnoses. Our investigations to determine the sources of disagreements and to establish consensus diagnoses revealed different results for current and lifetime assessments. For current alcohol use disorder, there were six disagreements on the distinction between no disorder and abuse, and all were resolved in consensus diagnoses of abuse. Only one patient's drinking was underestimated by the case manager but identified as abuse by interview; the case manager for this case was new, and both case records and more experienced team members quickly identified the patient as an abusive drinker. Five patients were identified as alcohol abusers by case managers but not by the complete interview. In each case, the patient denied problematic drinking at interview, but alcohol-related problems were identified by clinical records, reports from families and communities, and staff consensus. In other words, 26.3 percent of those with current alcohol use disorders (5 of 19) denied or significantly minimized their alcohol-related problems at interview. There were three disagreements on the distinction between abuse and dependence, and all were
Downloaded from http://schizophreniabulletin.oxfordjournals.org/ by guest on May 7, 2015
Measure
Alcohol problems n (%)
measure of association (Kendall's T-C), these relationships were quite strong. With the case managers' 5-point scale collapsed into DSM-III-R diagnostic categories (no disorder, abuse, and dependence), the case manager and interview ratings agreed on 66 of 75 subjects (88.0 percent) for current diagnoses and on 57 of 75 subjects (76.0 percent) for lifetime diagnoses. Kendall's T-C correlations were 0.41 (p < 0.001) and 0.61 (p < 0.001), respectively.
SCHIZOPHRENIA BULLETIN
62
Table 2. Intercorrelations of alcohol measures Measures of lifetime alcohol use
Measures of current alcohol use Case Consensus manager DSM-III-R diagnosis rating
Consensus Interview Case DSM-III-R manager DSM-III-R diagnosis rating diagnosis
Hospital Identification
CAGE
MAST
0.262 0.30' 0.31' 0.31'
0.26* 0.222 0.32' 0.29'
0.35' 0.36' 0.40' 0.39'
0.41' 0.35' 0.45' 0.40'
0.36' 0.36' 0.49' 0.46'
0.41' 0.33' 0.49' 0.43'
0.44'
0.48' 0.66'
0.41' 0.60' 0.79'
0.44'
0.46' 0.68' 0.76' 0.801 0.79'
Current measures
ADS
0.46'
Interview Case manager Consensus
0.51' 0.44'
0.45' 0.46' 0.60'
Lifetime measures Hospital identification CAGE MAST Interview Case manager
0.60' 0.66' 0.65'
Note.—Full scales were used for all analyses. All correlations determined by Kendall's T-C. ADS - Alcohol Dependence Scale. MAST - Michigan Alcoholism Screening Test. ' p < 0.001. »p
![[Diagnosis and management of alcohol-use disorders].](/assets/img/hold.png)
