Schizophrenia Bulletin, Vol. 28, No. 4, 2002
toms are common. For example, 40 percent of a commu- nity-based sample of subjects with psychotic symptoms received Research Diagnostic Criteria (RDC) (Spitzer et al. 1978) schizoaffective diagnoses based on structured interviews (J. McGrath, personal communication to D.F.L., 1998).
Diagnostic categories remain the best predictors of familial risks as well as treatment response. But there is widespread interest in developing a complementary dimensional perspective to permit a richer understanding of the relationships between phenotypes and genotypes, and between symptoms and treatment response. For exam- ple, schizophrenic, schizoaffective, mood, and atypical psychotic disorders each show some familial coaggrega- tion with one or more of the other disorders, and clinical features overlap (Maier et al. 1993), suggesting that addi- tional methods are needed for clinical characterization.
The LDPS grew out of discussions about how to quantify "how schizophrenic" or "how affective" each case was, with each dimension independent of the other. It is based largely on previous clinical schedules, scales, fac- tor analyses, and diagnostic criteria sets, none of which offers all of the features of the LDPS: (1) profiling symp- toms over the entire course of illness (lifetime perspective) rather than at a single time point; (2) separating ratings of time course ("duration") from a symptom's typical sever- ity; (3) incorporating multiple domains of psychotic and mood-related symptoms into a single scale; (4) differenti- ating between "generic" positive symptoms and those most closely associated with schizophrenia; and (5) cap- turing the presence of mood-congruent psychotic symp- toms.
By comparison, semistructured interview schedules establish the presence of categorical diagnoses without quantifying each symptom dimension and its time course. The Scale for the Assessment of Positive Symptoms (SAPS) (Andreasen 1990) and the Scale for the Assess- ment of Negative Symptoms (SANS) (Andreasen 1989, 1990) are more comprehensive in their range of specific symptoms and have been used successfully in studies of biological variables (Andreasen 1990) and of genetic link- age (Brzustowicz et al. 1997), but they lack coverage of mood disorders or a lifetime perspective. Similarly, the Positive and Negative Syndrome Scale (Kay et al. 1987; Peralta et al. 1994) is cross-sectional. The Operational Cri- teria Checklist for Psychotic Disorders (OPCRIT) system (McGuffin et al. 1991) covers a broad group of symptoms relevant to categorical diagnosis and has been used suc- cessfully for factor analytic studies (Cardno et al. 1996, 1997) but does not make quantitative or explicitly lifetime ratings in most areas. Kendler's Multiple Symptoms of Schizophrenia scale, which rates psychotic and mood symptoms on a lifetime basis, has shown excellent reliabil-
D.F. Levinson et al.
ity in one study and has defined clusters of probands with different patterns of familial disorders (Kendler et al. 1998). The LDPS might have certain advantages, includ- ing separate ratings of duration and severity, additional mood psychosis and schizophrenia syndrome items, and demonstration of reliability without extensive training.
Finally, a dimensional approach might improve both research and clinical assessments of psychotic cases. First, the categorical focus leads to obtaining the minimum information necessary to assign a diagnosis, without deter- mining the pattern of all relevant symptoms over time, as is required by a lifetime-dimensional approach. Indeed, items with lower interrater reliability are often those that are less well documented in clinical and research records. Second, research teams often develop tacit biases that enhance within-group but not across-group reliability. For example, a team might define schizophrenia narrowly if it believes that "pure" cases are needed for a biological study, but a team recruiting a very large sample might define it broadly. These biases can be based on the nature of the research, a theoretical perspective, or simply a ten- dency to accept colleagues' customary styles of inquiry and documentation. Within each team, there is covert pres- sure to find what is expected, and in research collabora- tions, there is often tension over whose diagnoses are "bet- ter." Dimensional ratings partially address these problems: the issue is no longer whether the "right" diagnosis has been made but whether each dimension has been fully and reliably evaluated.
Design of the Preliminary Version
The preliminary version of the LDPS (table 1) included 30 clinical items in 9 symptom domains or dimensions, plus 3 items on course of illness and ratings of atypical or comor- bid features. The scale was intended to incorporate all of the following features:
1. Multiple psychotic and mood dimensions. We reviewed factor analytic studies of schizophrenia (Liddle 1987; Arndtetal. 1991; Peralta etal. 1992, 1994a, 1994£; Silver et al. 1993; Thompson and Meltzer 1993; Linden- mayer et al. 1994; Salokangas 1997; Toomey et al. 1997) and the criteria for psychotic disorders in the RDC, DSM-III-R, and DSM-/V. Closely related symptoms were combined, and four sets of items selected: (1) positive symptoms seen in all psychoses; (2) "schizophrenia syn- drome" symptoms historically considered most specific to schizophrenia ("bizarre'Vimpossible and "Schneiderian" symptoms); (3) negative symptoms, emphasizing, in the preliminary version, the deficit syndrome (Carpenter et al. 1988; Kirkpatrick et al. 1989); and (4) disorganized symp- toms. Most factor analyses support the positive, negative,