Identifying non-affective psychosis in first admission patients: MMPI-2, structured diagnostic interview, and consensus lifetime best estimate

Highlights

• The MMPI-2 classified first episode psychosis patients more accurately than the SCID.

• The results point to increased diagnostic accuracy if the MMPI-2 is added to the SCID.

• The SCID demonstrated high specificity but low sensitivity for identifying non-affective psychosis in a first-admission sample.

• The Goldberg Index, revised for the MMPI-2, was validated as a measure distinguishing psychotic from non-psychotic psychiatric patients.

Abstract

Purpose

We aimed to evaluate the Minnesota Multiphasic Personality Inventory-2 (MMPI-2) as a potential means of improving on the SCID's diagnostic efficacy.

Methods

76 first-admission patients were assigned DSM-IV consensus diagnoses by two experienced psychiatrists using all available information, then dichotomized into non-affective psychosis and other mental illness groups. The patients were also given the SCID and the MMPI-2. The diagnostic performance of the MMPI-2 was compared to that of the SCID to assess both diagnostic accuracy and incremental validity.

Results

MMPI-2 scales 8 (Schizophrenia) and BIZ (Bizarre Mentations) correctly identified 58% and 56% respectively of non-affective psychotic patients. The Goldberg Index had an overall correct classification rate of 70%, but only identified 49% of the psychosis group. The SCID had a correct classification rate of 66% but correctly identified only 25% of the non-affective psychosis patients. Three MMPI-2 scales combined with the SCID resulted in an overall correct classification rate of 73%, and identification of 66% of the non-affective psychosis patients.

Conclusion

The results suggest that the MMPI-2 may identify early psychosis at least as well as the SCID. Furthermore, using a combination of the MMPI-2 and the SCID shows incremental validity over using the SCID alone.

Introduction

The last decades have seen worldwide efforts and research targeting early detection of schizophrenia-spectrum psychosis. This effort was initiated in Australia (Yung et al., 2003) and was motivated by the recognition of the fact that many help-seeking individuals with subtle psychotic features failed to meet the full criteria for the schizophrenia diagnosis. We have elsewhere discussed in detail the high diagnostic threshold for schizophrenia in DSM-IV, DSM-5 and ICD-10 (Parnas, 2002, Parnas, 2012).

In the studies on early detection a variety of diagnostic approaches have been used, ranging from self-report measures to structured interviews. Kendler et al. (1996) observed in the US National Comorbidity Survey that the structured interview has limited utility in detecting psychosis. In our own study of 100 consecutive first admission patients, we found that a fully structured psychiatric interview, the Structured Clinical Interview for DSM-IV-TR or SCID, displayed high specificity but low sensitivity for schizophrenia and other non-affective psychotic disorders (Nordgaard et al., 2012). Other evidence has shown an overall low level of agreement between the SCID and clinicians’ diagnoses (Steiner et al., 1995).

We have consistently advocated for thorough psychiatric assessment by experienced clinicians as the “gold standard” for all schizophrenia-related diagnostic efforts (Leckman et al., 1982, Nordgaard et al., 2012) including early detection (Nordgaard et al., 2013, Parnas, 2015). Core criteria include the use of multiple sources of information such as historical records, psychological tests, informants’ observations, and following the patient over time (Meehl, 1959). We recognize the practical reality that in some clinic settings this gold standard may not be feasible, either due to a lack of economic resources or staff expertise. In this paper, we explore the possibility that using the MMPI-2 (Minnesota Multiphasic Personality Inventory-2) as an adjunct to the SCID can improve its diagnostic accuracy for schizophrenia-related psychosis at first admission. Since the MMPI-2 is a self-complete instrument that can be administered with limited staff time and training, its addition would not pose an undue burden on clinic resources.

One impediment to its use in early detection paradigms is that the MMPI-2 cannot be used “right off the shelf” to assign psychotic diagnoses like the SCID can. Although there is a traditional clinical practice of profile interpretation, where scale elevations above a standard cut point are sorted to identify characteristic 2- or 3-point code types, this practice has not yet produced a set of validated differential diagnosis decision rules using current diagnostic criteria.

Instead, the research literature has advanced several alternative approaches for identifying patients with schizophrenia¹ among mixed psychiatric populations. One method is to use the Clinical scales, which reflect diagnostic categories in use at the time the MMPI was developed in the 1940s; the most consistent results for schizophrenia patients are a pattern of higher scores on Scale 8 (Schizophrenia) and lower scores on Scales 2 (Depression) and 7 (Psychasthenia) (Bagby et al., 2005, Ben-Porath et al., 1991, Graham, 2006, Greenblatt and Davis, 1999, Wetzler et al., 1998). Alternatively, the more recently constructed Content Scales (Butcher et al., 1990) may be used; here, the most consistent results for schizophrenia patients are a pattern of higher scores on the Bizarre Mentation (BIZ) scale and lower scores on the Depression (DEP) and Anxiety (ANX) scales (Bagby et al., 2005, Ben-Porath et al., 1991, Butcher et al., 1990, Graham, 2006, Munley et al., 1997, Wetzler et al., 1998). Another approach has been to construct an index using an arithmetic combination of several different scales. Goldberg developed an index using the original MMPI that outperformed expert clinicians in predicting psychotic vs. non-psychotic discharge diagnoses (Goldberg, 1965, Goldberg, 1968); an MMPI-2 update of the Goldberg Index was successful in discriminating schizophrenia from major depression (Egger et al., 2003). It is worthwhile to note that although Goldberg's Index made more efficient use of MMPI scale data than clinicians’ judgments, the gold-standard criterion that both attempted to predict was the discharge diagnosis assigned by the treating clinician.

Our overall aim in the current study is to test the ability of the above MMPI-2 measures to identify non-affective psychosis in a sample of first-admission psychiatric patients, and then to determine if the best-performing MMPI-2 measure can add to the diagnostic accuracy of the SCID alone.