|Home | About | Journals | Submit | Contact Us | Français|
One hundred and five drug-dependent women in outpatient perinatal addiction treatment were classified by cluster analysis of the Minnesota Multiphasic Personality Inventory-2 (MMPI-2) profiles into high and low psychopathology (HP and LP) groups that differed on three validation measures. The HP group (n = 29, 27.6%) had elevations on MMPI-2 Scales F, 2, 4, 6, 7, and 8, while the LP group (n = 76, 72.4%) generated a normal range profile with elevations on Scales F and 4. Psychological outcomes differed by group. HP participants showed reduced alcohol, family, and psychiatric severity, and reduced depressive symptoms, while LP subjects showed reduced drug, self-debasing, and acting-out problems. Data suggest the need for lower intensity services for the majority of the perinatal drug dependent population with LP.
How does psychopathology impact treatment outcome in women and perinatal drug-dependent women? Drug-dependent women are significantly different from men in greater severity of psychopathology (DeLeon & Jainchill, 1991; Galen, Brower, Gillespie, & Zucker, 2000; Griffin, Weiss, Mirin, & Lange, 1989; Moras, 1998), legal and social problems including parenting demands, access to treatment, barriers to treatment, absence of abstinence-supporting social structures (Moras, 1998; Morris & Schinke, 1990), and more negative stereotyping (Erickson & Murray, 1989). Psychological distress may play a large role in a woman’s decision to seek treatment (Smith, Dent, Coles, & Falek, 1992). A recent study found that perinatal drug-dependent women with alcohol dependence scored higher in psychopathology and exhibited worse treatment outcomes (Miles, Svikis, Kulstad, & Haug, 2001), suggesting a link between dependence, psychopathology, and outcome.
Although they differ from men, drug-dependent women are a heterogeneous group, and subgroups may respond differentially to treatment. Haller and Dawson (1997) found that treatment response varied by Millon Clinical Multiaxial Inventory-II (MCMI-II) cluster groups among perinatal drug-dependent women. The first cluster had early onset and family history, similar with the Type II profile described by Cloninger (1987). The second cluster had a normal profile, except for drug dependence. The third cluster was a psychiatrically disturbed group with extreme pathology. Cluster 1 showed no response to treatment, while Cluster 2 manifested no psychopathology or addiction severity problems at follow-up. Cluster 3 improved on addiction problems but not in psychopathology. Similarly, other researchers found that perinatal drug-dependent women evidenced 3 MCMI-II clusters: low, moderate, and high risk, whose outcome varied (Howard, Espinosa, & Beckwith, 1997). Parenting gains depended on psychopathology status. Taken together, these results indicate that cluster analysis with the MCMI-II can identify distinct groups of drug-dependent women, and that outcome varies by psychopathology cluster.
The purpose of this study was to identify clinically relevant subgroups within a sample of perinatal drug-dependent women and to examine the subgroups’ psychopathology outcomes in an intensive perinatal program using the Minnesota Multiphasic Personality Inventory-2 (MMPI-2).
The participants were 105 drug-dependent women, who were pregnant or less than 6 months postpartum upon first-time admission to the treatment program, with valid profiles on the MMPI-2. Of the participants, 46 completed discharge assessments and were included in the treatment outcome component of the study. The setting was the Center for Perinatal Addiction (CPA), which provided comprehensive, multimodality day treatment for pregnant and recently postpartum women under a “one-stop shopping” treatment model. Women received 24 hours of intervention per week, including obstetric and pediatric care, psychological assessment, interpersonally focused group, family, and individual psychotherapy, life-skill groups, and child care during services. A more complete description of the program is available (Haller, Knisely, Dawson, & Schnoll, 1993).
Demographic information was collected using the Division of Substance Abuse Medicine Intake Form (DSAM Intake Form, DSAM, 1990). The Structured Clinical Interview for DSM-III-R (SCID; Spitzer, Williams, Gibbon, & First, 1988) was used to determine the presence of major Axis I disorders including mood disorders, anxiety disorders, and drug or alcohol dependence. The Structured Interview for DSM-III-R Personality Disorders (SIDP-R; Pfohl, Blum, Zimmerman, & Stangl, 1989) was used to determine DSM-III-R Axis II diagnosis. The Symptom Checklist 90-R (SCL-90-R, Derogatis & Cleary, 1977) measured psychiatric distress during the past seven days. The MMPI-2 (Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989) and MCMI-II (Millon, 1987) were used to assess personality pathology and disorders. The Addiction Severity Index (ASI; McLellan et al., 1992) composite scores were used to assess severity in seven areas: psychiatric, family/social, medical, employment, drug use, alcohol use, and legal.
The women completed an informed consent, process approved by the IRB, and provided written informed consent. Intake assessments were administered in 4-hour time blocks prior to the initiation of treatment. At the time of discharge, or 20 weeks, the participants completed discharge measures.
Hierarchical cluster analysis tested for homogeneous subgroups within the full sample using SAS Version 8 Proc Cluster (Ward, 1963). A validation method (Wieczorek & Miller, 1992) was used to determine whether the MMPI-2 clusters represented meaningful groups, rather than mere artifacts of this particular measure. Specifically, t tests compared the means of the cluster groups’ scores on the global severity index (GSI) of the SCL-90-R and the psychiatric severity composite of the ASI. Two-tailed Fisher’s Exact Tests were used to identify which diagnoses on the SCID and SIDP-R differed by MMPI-2 cluster group membership. Lastly, two logistic regression analyses were used to determine if aggregated MCMI-II Axis I and II diagnoses varied by group.
A mixed-models repeated-measures ANOVA (Laird & Ware, 1982) tested for between-time (intake vs. discharge) and between-groups simultaneously for all scales within a given instrument. If the multivariate procedure revealed interactions, follow-up univariate t-tests were used to examine the differences from intake to discharge on individual scales by group. Length of stay was compared for each group using SAS Proc Lifetest to determine whether the Kaplan-Meier survival curves differed. Retention measures, including type of discharge, and loss to follow-up were examined for each group, and rates were compared using Fisher’s Exact Test.
Table 1 shows the demographic characteristics of the total sample and the characteristics of the subgroup with outcome data versus those with intake data only. The 46 participants with outcome data did not differ from the 59 participants with intake data only.
Cluster analyses revealed two homogeneous subgroups on the MMPI-2. This solution was translated into an algorithm that classifies a participant as high psychopathology (HP) if the MMPI-2 Scales F, 2 (D, depression), 4 (PD, psychopathic deviate), 6 (PA, paranoia), 7 (PT, psychasthenia), and 8 (SC, schizophrenia) summed to 412 or above, and as low psychopathology (LP) otherwise. Plots of the mean MMPI-2 profiles for the two clusters are shown in Fig. 1. HP participants had elevations on MMPI-2 Scales F, 2, 4, 6, 7, and 8, while LP participants generated a normal range profile, with mild elevations, on Scales F and 4 only. The demographic characteristics of those classified into the HP and LP groups are provided in Table 1. The 76 LP participants (72.4% of the total sample) did not differ from the 29 HP participants (27.6% of the total sample) on any baseline variable.
The HP and LP clusters had significantly different means on two measures of psychiatric severity and Axis I and II psychopathologies from three instruments, depicted in Table 2. Psychiatric severity was significantly higher in the HP group than the LP group on both the GSI of the SCL-90-R and the ASI mean psychiatric severity composite score. Psychopathology on the SCID and SIDP-R also differed between groups. Major depression was diagnosed significantly more often in the HP (41.4%) than in the LP (18.4%) group on the SCID. The LP group had a higher rate of drug dependence diagnoses (98.7%) than the HP group did (86.2%) on the SCID. HP evidenced a higher rate of diagnosis in paranoid personality (HP 37.9% vs. LP 13.2%), dependent personality (HP 27.6% vs. LP 10.5%), borderline personality (HP 27.6% vs. LP 9.2%), and self-defeating personality (HP 44.8% vs. LP 15.8%) disorders.
For further validation, we compared the two cluster groups on their scores on Axis I and II disorders (N = 104, 29 HP and 75 LP). A logistic regression analysis modeling the likelihood of HP predicted by the nine MCMI-II Axis I disorders was significant (multivariate likelihood ratio χ2 with 9 df = 45.18, P < .0001). Similarly, a logistic regression analysis modeling the likelihood of HP predicted by the 13 MCMI-II Axis II disorders was significant (likelihood ratio χ2 with 13 df = 37.04, P < .0004). Consistent with validation analyses with the SIDP-R and SCID at intake, these analyses indicate that HP and LP participants also differed on MCMI-II disorders, aggregated into Axis I and II disorders.
Forty-six of the 105 participants had complete data for intake and discharge assessments: 13 (28.3%) were HP and 33 (71.7%) were LP. There was no difference by group in the length of stay in the program, detected in the Kaplan-Meier survival rates by group (log-rank χ2 = 0.29, ns). There was also no difference by cluster group in the type of discharge, with 31% of the HP and 33% of the LP completing the program, 34% of the HP and 42% of the LP leaving AMA, 31% of the HP and 24% of the LP administratively discharged, and 3% of the HP and 1% of the LP referred out.
Table 3 presents the outcome results for each cluster group on the ASI and MCMI-II. In contrast to no difference between groups on the SCL-90-R, the groups showed differential improvement on the other three tests examined, and univariate analyses were conducted. On the ASI, significant interactions were found between assessment times (intake and discharge) and cluster group (HP and LP) in the mean composite scores using a mixed-models repeatedmeasures analysis of variance. This interaction indicates that differences varied by group (likelihood ratio χ2 = 144.37, P < .0001). Univariate analyses revealed that the LP had a significant improvement in the drug composite, and the HP improved in the alcohol, family, and psychiatric severity composite scores.
Similarly, on the MCMI-II, differences between intake and discharge varied by group in the mixed models repeated measures analysis of variance (likelihood ratio χ2 = 152.03, P < .0001). Post hoc univariate tests revealed that both the HP and LP improved on the debasement and dysthymia scales. The HP showed additional decreases on the major depression scale, and the LP exhibited significantly decreased scores on the borderline scale.
Among perinatal drug-dependent women, we identified two levels of psychopathology, high and low, using cluster analysis. These clusters, derived from the MMPI-2, were validated with alternative measures of psychiatric severity and psychiatric diagnosis. Subjects classified in the HP group by the MMPI-2 cluster analysis had greater psychopathology, as measured by both standardized personality inventories and structured clinical interviews than participants classified in the LP group by the MMPI-2 cluster analysis. We conclude that the HP and LP clusters represent real, clinically relevant differences in the two groups. These results indicate that the MMPI-2 can identify distinct groups in perinatal drug-dependence treatment and provide an alternative to the MCMI-II. The cluster groups seem comparable with two of the three clusters previously identified by Haller and Dawson (1997) and Howard et al. (1997) in that the HP group was smaller and had a range of psychiatric problems in both Axis I and II disorders, while the LP group was the larger group, had less serious psychopathology, and more symptoms related to drug dependence. The evidence is accumulating that there are at least two types of perinatal drug-dependent women, with distinct differences in type and severity of psychopathology. It is important to note that the groups were not identifiable by examining demographic characteristics, indicating a clear role for personality assessment in treatment planning for this group.
Attrition limits our ability to determine whether the differential outcomes observed in the HP and LP clusters would have been observed in the whole perinatal drug-dependent sample. This concern is moderated because those with and without outcome measures did not differ on demographic characteristics at baseline, and the HP and LP groups showed minimal differences (only in employment) on demographic characteristics at baseline. The HP group would be expected to show improvement on a greater number of scales due either to regression to the mean or the restricted range in the LP group. However, the breadth of change was not necessarily greater in the HP group than in the LP group. Dropout from intensive perinatal programs is not uncommon (Svikis et al., 1996). However, due to this limitation, our conclusion that psychopathology level yields differential psychological outcomes should be replicated by others.
The identification of the types of psychopathology within a perinatal drug-dependent sample may be a step towards developing cost-effective treatment programs for this group, as has already been done for men (McLellan & Alterman, 1991). Although some participants manifested severe psychopathology, the majority of drug dependent women in perinatal treatment were classified as having low psychopathology and may not have required high-dose, extensive treatment. It is important to note that the rates of retention or type of discharge did not vary by psychopathology group. It is possible that pregnant and parenting women with low or moderate psychopathology might have benefited from briefer, less intensive treatment, or that an interaction among the psychopathology types and levels of service might be found similar with that identified in drug-dependent men (McLellan, 1986; McLellan et al., 1983a). The fact that 72% of the sample was LP, coupled with our high observed dropout rate from the required 24 h/week of treatment services, suggests the need for a range of less intensive treatment services for this population, matched to patient psychiatric severity and other needs (McLellan & Alterman, 1991; McLellan et al., 1983b). Comprehensive services may be best suited to those with extensive psychopathology in addition to drug problems and acting-out behavior patterns. Future research should examine whether briefer or less intensive psychosocial drug treatment, perhaps paired with obstetrical care, yields successful outcome in pregnant, drug-dependent women with lower psychopathology.
This work was supported by NIDA grant DA06094. We thank Diane Green, Katherine Candler, Beverly Carter, and Sarah Burns for their assistance with data collection and data entry, Deborah Haller for assessment of MMPI-2 validity, and Eleanor Campbell for her help with data management and analysis.