Multivariate discriminant function analysis of neurologic,pain, and psychiatric patients with the MMPI

The ability of the MMPI to classify five well-defined patient groups was investigated (N = 394; control, neurologic, psychiatric, chronic pain, and random). Clinical inspection and discriminant function analyses of basic clinical and research scales could not classify groups correctly, but discriminant function analyses of 37 variables loaded with CNS items (Cripe Neurologic Symptom items) correctly classified the groups with 78% overall accuracy (70% of neurologic, 62% of psychiatric, 81% of pain, 84% of controls, and 100% random). Results indicate that differential diagnosis is not possible with clinical inspection of scales, but complex statistical analysis of the MMPI is potentially useful in diagnosis and decision making. A method for applying the discriminant function analysis to individual cases is provided.     

MMPI-derived indicators of organic brain dysfunction

For more than a decade researchers have been attempting to develop a reliable and valid way in which the MMPI can be used as an indicator of organic brain dysfunctioning. This investigation reviews the major approaches and tries all of them on a sample of 32 organic and 34 schizophrenic hospital patients. Results show that the MMPI methods used assigned patients correctly in 45 to 76% of cases. The Sc scale of the MMPI was generally superior to other scales. Most methods predicted better for women than for men. In all, the idea of making the organic-schizophrenic distinction on the basis of the MMPI appears to have been a “nice try” that ultimately did not yield the desired results.     

Prognostic validity of the MMPI

Attempted to determine the degree to which the MMPI has prognostic validity for psychiatric patients diagnosed as schizophrenic (N = 41). Patient records were analyzed with multiple regression techniques using the MMPI scales as the independent variables and length of hospital stay as the dependent measure. It was found that certain MMPI scales were highly significant and powerful predictors of length of hospital stay, which indicates that the MMPI does indeed have validity as a quantitative predictor of prognosis in clinical situations. Particularly, the two scales with the highest predictive power, Sc and D, were inversely related concerning clnical outcome.     

A validity index for the MMPI-2.

There have been numerous methods developed for the detection of valid profiles on the Minnesota Multiphasic Personality Inventory (MMPI)-2. The current study examined a method of combining seven different validity scales of the MMPI-2 into a common weighted method in assessing malingering in chronic pain patients. The weighted method was able to correctly classify 100% of nonlitigants, using a cutoff score of > or =5. The findings of this study suggest that chronic pain patients in litigation produce a different profile on the MMPI-2 validity scales than do nonlitigants. In a group of knowledgeable actors (malingerers), 86% was correctly classified. The overall finding showed 100% specificity and 86% sensitivity. The findings support the need for multiple validity scales to be examined in determining a valid profile. The weighted validity scales method was robust enough to account for "emotional distress" and still identify invalid MMPI-2 performance.     

Further evidence that M-F is bipolar and multidimensional

Conducted a study to identify a set of cognitive skills tests on which there are significant sex differences. The main study determined the comparability of several bipolar M-F scales and the relations between the full range of scores on each of those bipolar M-F scales and self-rated sexual orientation; 110 females and 113 males completed six cognitive tests, the Bem Sex Role Inventory (BSRI), the Minnesota Multiphasic Personality Inventory (MMPI) and the Kinsey Sexual Orientation scale (KSO). Correlations in the combined-sex group were substantially higher than in either of the single-sex groups. Each of the several M-F measures was related significantly to sexual orientation in the combined-sex group, and the combination of the MMPI MF scale and the Bem Androgyny Score accounted for a small but significantly greater portion of the variance in self-rated sexual orientation than did the MMPI MF scale alone. The data support the view that M-F may be considered to be a bipolar and probably multidimensional variable.     

Mothers of sexually abused children: an MMPI study.

Thirty-seven mothers of sexually abused children were contrasted with 41 mothers who were receiving outpatient psychotherapy and with 76 women with no history of psychiatric disturbance on the MMPI. The groups were matched on age and socioeconomic status, and significant between-group differences were noted. Outpatients and mothers of sexually abused children were significantly more elevated than the normal comparison group on the majority of the scales. Mothers of sexually abused children differed from psychiatric outpatient mothers on several scales. The 4-3 and spike 4 MMPI code-types were significantly more common in mothers of sexually abused children than in outpatient mothers. Differences in the child's sexual abuse characteristics were related to MMPI scales of the mother.     

Psychological adjustemnt and attitudes toward success and failure

Investigated the relationship of performance on the Success-Failure Inventory (SFI) and severity of psychological maladjustment (as measured by the MMPI). Ss (N = 120) were male veterans admitted to a VA Medical Center's psychiatric Day Hospital program. Ss were divided at median of SFI scores to form two groups: (a) Success-attainment (N = 60); and (b) failure-avoidance (N = 60). Multivariate analysis of the MMPI validity and clinical scales indicated that the two groups differed significantly, F(13,104) = 4.46, p <0.0000), Success-attainment group, in contrast to failure-avoidance group, scored significantly higher on K (p <.01), and significantly lower on PT (p <0.000), SC (p <.002), and SI (p <.0000). Discriminant analysis revealed that 76.67% of Ss were classified correctly into a success-attainment vs. failure-avoidance group by using the MMPI scales (p <.001).     

Utility of the MMPI Pain Assessment Index in predicting outcome after lumbar surgery

A Pain Assessment Index (PAI), computed from a weighted combination of MMPI scales scores, has been developed to assess a pain patient's tendency toward continued pain complaints after surgery (Smith & Duerksen, 1979). This study examined the ability of the PAI, determined from presurgery MMPIs, to predict outcome subsequent to lumbar laminectomy and discectomy. One hundred and six chronic low back pain patients (61 males, 45 females) were categorized as having good, fair, or poor outcome 1 year or more postsurgery. Using the cutting score proposed by Smith and Duerksen (1979), the PAI correctly classified 79% of the subjects as having good or fair/poor outcome, and 87% as to whether they were working at follow-up. However, the MMPI Hypochondriasis (HS) scale alone correctly predicted the overall outcome of 83% and the work status of 81% of the subjects. The PAI (and, to a lesser extent, the HS and Hysteria [HY] scales of the MMPI) were found to have good ability to identify patients who were doing well after surgery, but low power in predicting which patients would have a fair or poor surgical result.     

An external validity study of the MMPI Personality Disorder Scales

This study examined the external validity of the MMPI Personality Disorder (PD) scales. Patients diagnosed with personality disorders (n = 23), according to a structured interview, were contrasted with medical control subjects and psychiatric patients with no personality disorder (n = 33). The MMPI PD scales could discriminate patients with any personality disorder and patients with specific clusters of personality disorders. Convergent validity was demonstrated by high correlations between the MMPI PD scales and the comparable Millon Clinical Multiaxial Inventory scales, as well as the number of DSM-III personality disorder symptoms. This study represents a preliminary step in the external validation of the MMPI PD scales.     

MMPI correlates of menstrual distress.

The MMPI and Moos' Menstrual Distress Questionnaire (MDQ) were administered to 60 undergraduate women. Partial correlations between MMPI clinical scales and menstrual and premenstrual MDQ symptom scales were computed with intermenstrual (baseline) symptom reports and response set (Gough's F-K index) statistically controlled MMPI variables tended to correlate with some symptom scales (premenstrual pain, negative affect; menstrual behavior change) but not with others (water retention, arousal). Where correlations did occur, common MMPI scales (Sc, Hs, Hy, Pt) were involved. Results suggest that psychological factors are more closely associated with some areas of menstrual symptomatoloty than with others.     

The assessment of denial and physical complaints: the validity of the Hy scale and associated MMPI signs

Little and Fisher (1958) have demonstrated that the MMPI Hysteria (Hy) scale consists primarily of items that address the denial of psychological problems (the DN scale) and items that involve the admission of physical problems (the AD scale). The traditional interpretation of an elevated Hy score implies that both DN and AD subscales are elevated significantly. Using samples of psychiatric (N = 1,246), medical (N = 330), and chronic pain patients (N = 126), the effectiveness of the HY scale and of various combinations of MMPI scales as predictors of the simultaneous occurrence of these two characteristics was evaluated. The results suggest that when the 13 standard MMPI scales were used, simultaneous elevations on scales K, HS, and HY proved to be the best indicator of the denial of psychological factors associated with a physiological disorder. It was found unwarranted to suggest the presence of hysteroid features on the basis of an elevated HY score unless scales K and HS also were elevated. The value of the HY scale as a measure of psychological denial associated with physical complaints must be considered questionable.     

Toxoplasma antigens recognized by immunoglobulin G subclasses during acute and chronic infection

The immunoglobulin G (IgG) subclass response to Toxoplasma gondii antigens during the acute and chronic stages of T. gondii infection were studied by using immunoblots with reduced antigen (RA) and nonreduced antigen (NRA) preparations. Serum samples were from individuals with acute or chronic T. gondii infection, and sequential samples were from women who seroconverted during gestation and were treated with spiramycin. IgG1 antibodies were predominant in sera from each of the groups and recognized a large number of RA and NRA. In the latter group of patients, IgG1 and IgG3 were the first antibodies to appear in response to the infection. In all groups, an antigen with a molecular weight (MW) of 30,000 was the most intensely stained and frequently recognized by IgG1 antibodies in NRA preparations. In RA preparations, antigens of MW 35,000 and 30,000 were the most intensely stained and frequently recognized by IgG1 antibodies. An antigen with an MW of 22,000 was intensely stained in the IgG1 immunoblots of the NRA preparation and to a lesser extent in the RA preparation. In contrast to immunoblots with IgG1 antibodies, very few antigens in the RA and NRA preparations were detected by IgG2 and IgG3 antibodies; IgG4 antibodies rarely detected any antigens. Of interest was that IgG2 antibodies detected antigens distributed over the entire MW range, whereas those detected by IgG3 antibodies were located mostly below the 35,000 MW marker. Enzyme-linked immunosorbent assay results paralleled those of the immunoblots in that IgG1 antibodies were predominant.     

Tavern patrons and the MacAndrew Alcoholism Scale: self-reported drinking behavior in relation to the MMPI L and K scales

This study investigated the ability of the MacAndrew Alcoholism (MAC) Scale to distinguish between alcoholics and nonalcoholics using self-ratings from patrons of taverns (N = 128); while only about 14% of the variance in the MAC Scale was accounted for by the self-rating (alcoholic/nonalcoholic) scale, the MAC Scale correctly identified 70% of those who reported alcohol use to levels consistent with alcoholic dependency after variance in the alcohol use scale attributable to the L and K scales of the MMPI was accounted for.     

MMPI personality differences in various populations of the unwed mother

This study was designed specifically to assess the personality differences between pregnant unwed young women who had a varying number of pregnancies and sexually active, but nonpregnant females. The MMPI was administered to 42 volunteer unmarried women assigned to one of the three following groups: Primigravida (pregnant for the first time), multigravida (pregnant for at least the second time), or to a “control” group (sexually active with no pregnancies). A discriminate function analysis was employed to predict group membership on the basis of the MMPI scales, which were used as predictor variables. The first function consisted of the Hs, Pt, and Sc scales, which accounted for 64.33% of the total variance. A second function consisted of K, Hs, Hy, Mf, Pa, Pt, Sc, and MA scales, which accounted for 35.67% of the total variance. The clinical implications of the first function (used primarily because it produced a higher degree of predictability with fewer scales) were discussed in terms of its ability to distinguish personality characteristics among the three groups of young women.     

Differentiating violent and nonviolent prison inmates by use of selected MMPI scales

Used multivariate procedures to determine whether prisoners who committed violent acts while in prison could be differentiated from those who did not. Ss were 141 adult male inmates in maximum security prisons. Multiple variable profile analysis that employed 22 MMPI scales and four demographic items significantly differentiated between violent and nonviolent inmates. Discriminant analysis that used all 26 variables yielded a significant root that accounted for 34.9% of the variance between violent and nonviolent inmates. The MMPI scales that contributed most to prediction of group membership were F, PA, PT, and Sc, all of which had discriminant load values above 0.40. Finally, a discriminant function prediction equation was derived to predict the criterion variable (violent or nonviolent behavior). Applying this equation without knowledge of the S's actual group membership correctly classified 72.9% of the violent and 80.6% of the nonviolent inmates.     

Correspondence of MMPI and MMPI-2 with chemically dependent patients

One hundred (39 women and 61 men) hospitalized substance abusers were administered the MMPI and MMPI-2. The correspondence of T-Scores, Code-types, and MAC/MAC-R scores were analyzed. Results showed that the MMPI-2 produced profiles that were consistent with previous research on the MMPI with substance abusers. However, the relationship between MMPI and MMPI-2 T-scores appeared complex and varied across individual scales for both men and women. Codetype congruence was 38% for one, two, and three-point code-types. Well-defined codetypes accounted for 30% of the profiles. MAC/MAC-R raw scores were different for men but not for women. Codetypes established for the MMPI may not be fully transferable to the MMPI-2 in a substance abusing population. © 1996 John Wiley & Sons, Inc.     

Use of the MMPI to predict effective hotline workers

The present study investigated the use of the MMPI to discriminate between two groups of hotline volunteers: those who after training would become effective interviewers and those who would not. Effective interviewers were defined as those who emitted 40% or more good responses when they were responding to a crisis call. The criterion measure employed was the Therapist Error Checklist. When a specification equation based on standard MMPI scales was used, 91% of effective interviewers and 90% of ineffective interviewers were identified correctly.     

Prediction of adjustment to prison in a federal correctional population

Investigated the ability of MMPI and correctional data to differentiate between adjusted (N = 25) and rebellious federal prisoners (N = 25). There was a difference at the 0.01 level on the HY scale, while marginally significant differences were found on the Hs and D scales (p < 0.06) and with age (p < 0.07). Comparisons of other clinical MMPI scales, the AP scale, correctional data, and ethnic background were not significant. It further appeared that prediction of institutional maladjustment from any criteria would be extremely unlikely due to the naturally occurring low base rate of the event.     

Prediction of help-seeking with the MMPI: The problem of base rates

The usefulness of combined elevation of the Depression and Psychasthenia scales of the MMPI as a predictor of psychological help-seeking was investigated for two classes at a small liberal arts college for men. Counseling service use rates during the 4 years of college for all students who scored at or above T = 60 on both MMPI scales (N = 83) were compared with rates for all other students (N = 269). Students in the high Depression-Psychasthenia group were significantly more likely to seek counseling during their freshman years than were other students (p < 0.05). Among counselees the high MMPI group also made significantly more visits to the service (p < 0.05). Application of a Bayesian rule disclosed that the MMPI measure would have improved decision accuracy only with respect to the prediction that a student would come in for counseling at some time during college. In this population, the MMPI measure is both a valid and a useful predictor.     

Age differences on alcoholic MMPI scales: a discriminant analysis approach

Ninety-one male alcoholic inpatients were administered the MMPI after detoxification. Subjects were divided into four age groups. One-way analysis of variance was calculated on the 13 clinical scales and 15 experimental scales (A, R, ES, MT, LB, CA, DY, DO, RE, PR, ST, CN, AT, SO-R, AMAC). In addition, two discriminant analyses were calculated to evaluate whether the inclusion of the 15 experimental scales would increase the percent of correctly classified alcoholics according to age. Results indicated that the Psychopathic Deviant and Paranoia scales declined with increasing age. The Low Back Pain scale indicated a decline from the 50-65 group to the 20-29 group, while the Responsibility scale increased with age. Results from the two discriminant analyses suggested that inclusion of the 15 experimental scales improved the percentage of correctly classified alcoholics.