Schizophrenia Bulletin, Vol. 28, No. 4, 2002
D.F. Levinson et al.
Table 1. Preliminary version of the Lifetime Dimensions of Psychosis Scale—Continued For each item (P, P-1, P-2, etc.), with exceptions noted below, separate ratings were made for the following:
Duration (0—absent, 1—less than 2 wks but at least hrs, 2—2+ wks, 3—2+ episodes or 2+ mos, 4—6+ mos, 5—2+ yrs, 6—5+ yrs).
Severity (0—absent, 1—minimal, 2—moderate, 3—severe, 4—very severe) defined more specifically in the manual in terms of preoccupation and/or interference with function.
Certainty that one or more of the symptoms were present (0—absent, 1—possible, 2—likely, 3—very likely, 4—defi- nite).
Documentation (checkbox for inadequate documentation for each item).
Exceptions: DE-2 and M-2 scored as the maximum number of symptoms; C1, duration only; C2, severity only; C3, reversed scale.
and disorganized factors, and some support differentiating bizarre psychotic symptoms (Lindenmayer et al. 1994; Peralta et al. 1994a; Toomey et al. 1997) and excited and depressive factors (Kay 1991). Items were added for psy- chosocial and interactional impairment, course of illness (deterioration and remission), temporal relationship between psychotic and mood symptoms, depressive and manic features, classical mood-congruent psychotic symp- toms, and complicating factors such as substance abuse, neurological disorders, manipulation, and dissociative symptoms. Somewhat redundant items were included to determine which proved most reliable and useful.
2. Lifetime perspective. In research practice, one typi- cally obtains information from a semistructured interview, family members, and written records spanning years or decades. The LDPS provides a dimensional profile over time, although it is not "longitudinal" in that it is a single retrospective rating. We considered using a single contin- uum of anchor points for each item that defined combina- tions of time course and severity, but raters reported that it was easier to judge time course and severity separately. It is difficult to obtain reliable details about the variation of symptoms over time. We concluded that time course could be adequately summarized by an estimate of the total time that a particular symptom had been present since the onset of illness (total lifetime duration), supplemented by the separate "deterioration" rating of interepisodic function- ing. Increased weight was given to the occurrence of a symptom during two different episodes of illness, given that recurrence is often a meaningful predictor of progno- sis and of familial risk. The number of duration anchor points was reduced in the final version, as described below. For severity, we found it most practical to rate "typ- ical" severity (the most severe level that characterized a significant portion of the illness).
In the preliminary version, each item was also rated for clinical "certainty" (how certain is the rater that the feature has ever been present) and adequacy of documen- tation. Also, a global rating was made for each domain to allow us to consider the usefulness of global ratings.
Design of the Reliability Study. The preliminary version of the scale was designed as described above, and a man- ual was written. Thirty-six schizophrenia spectrum and mood psychosis cases were utilized for the preliminary study, including 17 collected (by B.J.M.) for the Australia/U.S. schizophrenia linkage study (Levinson et al. 1998), 5 evaluated in Philadelphia for this exercise, and 14 collected for a genetic study of schizophrenia in Costa Rica (M.A.E. and D.F.L.). A minority were inpa- tients, and most had schizophrenic or schizoaffective diagnoses by either DSM or RDC systems. They had a mean duration of illness of 14.9 (± 9.5) years with a range of 1 to 38 years. Seven raters participated in the exercise: D.F.L. rated 30 cases; B.J.M. rated 17; M.A.E. rated 9; and the other four rated 6 cases, 5 cases, 4 cases, and 1 case, respectively. In each case, two experienced research clinicians who had reviewed the manual (always including D.F.L. and/or M.A.E.) were provided with all available material from a structured research interview (Diagnostic Interview for Genetic Studies [DIGS, Nurnberger et al. 1994] or Schedules for Clinical Assessment in Neuropsychiatry [Wing et al. 1990]), psychiatric records, and informant reports, and made independent ratings. To approximate real-world conditions, where researchers often do not undergo specific training for every rating scale used in a study, no other training was provided.
Statistical Analysis. To assess interrater reliability, intra- class correlation coefficients (ICCs) were calculated, con- sidering as item scores each rater's certainty + severity + duration ratings, and as dimension scores the sum of severity + duration for all items within each dimension; analyses of separate duration and severity scores yielded similar results. Then, using the average of the two raters' scores (assuming that averages were more accurate), tau-b correlation coefficients were computed between all pairs of items (the sum noted above, and separately for severity and duration) to identify highly redundant items that