Lifetime Dimensions of Psychosis Scale
might be dropped. Tau-b correlations are particularly powerful when score distributions have long tails (O'Gorman and Woolson 1995). To determine which sub- ratings provided independent information, we also exam- ined tau-b correlations between (1) same-item duration and severity scores, and (2) the certainty rating and the sum of duration + severity.
We also carried out a principal components factor analysis on averaged total scores (certainty + duration + severity) for each item. We interpret this preliminary analysis with caution because there were too few subjects for this many items, and because of the limitations of the sample: it will be more useful to study factor structure in a sample representing the full range of schizophrenic, mood, and atypical psychoses. We performed the preliminary analysis to determine whether the apparent factor structure met expectations based on the previous studies of schizo- phrenia that guided the design of the scale.
Reliability Study of the Modified (Final) Version. Based on these analyses and discussion with the inter- viewers, the number of items and subitems was reduced as described below (version 2). A second reliability study was carried out using 32 schizophrenia spectrum and mood disorder cases from the Washington University site of the National Institute of Mental Health Schizophrenia Genetics Initiative (Cloninger et al. 1998) (with all identi- fiers removed), based on DIGS interviews, narrative sum- maries, and family history reports for each case. Ten experienced research clinicians from eight centers partici- pated; the two raters for each case were from different centers. Only D.F.L. and B.J.M. participated in both exer- cises. Training was limited to reading a brief revised man- ual. ICCs were computed for each item (severity + dura- tion ratings). SYSTAT (versions 7.0 and 8.0, SPSS 1998) was used for all statistical analyses.
First Reliability Study. Table 2 shows the correlations of the average of the two raters' severity and duration ratings for global items. (Similar correlations were observed for individual items.) A decision was made to retain separate severity and duration ratings because they were poorly correlated for psychosis, schizophrenia syndrome, nonaf- fective psychosis, and impairment, and because raters considered it easier to rate them separately. However, cer- tainty ratings were dropped because there were modest (mean 0.62) tau-b correlations between the certainty and duration + severity scores (not shown) and high correla- tions (mean 0.88) between certainty and certainty + dura- tion + severity, and because raters found it cumbersome to make a separate certainty rating for each item.
Schizophrenia Bulletin, Vol. 28, No. 4, 2002
Table 3 shows the ICCs for pairs of raters for dimen- sions using severity + duration for all items within the dimension (e.g., PI through P6 for psychosis), or certainty
severity + duration for key items. The mean ICC was
The positive symptom and mood scores had gener-
ally excellent reliability, with most scores in the range of 0.76 to 0.87. Negative, disorganized, and course of illness ratings had generally lower but acceptable reliability. For individual severity and duration items, the mean ICC was 0.59.
Factor analysis (table 4) yielded a rotated solution of six factors that explained 80 percent of the variance, inter- preted as delusions, mania, disorganization, depression, deterioration, and hallucinations. Negative symptom items were part of the deterioration factor. There was no separate "bizarre" psychosis factor, but these symptoms might load separately in samples with a higher proportion of non- schizophrenic psychoses characterized by nonbizarre posi- tive symptoms.
For tau-b coefficients for pairs of items, items inter- correlated at 0.8 or higher included any delusions (P-l) with preoccupation with delusions or hallucinations (P-6); auditory voices (P-4) with any hallucinations (P-3) and hallucinations characteristic of schizophrenia (S-3); and at 0.7 or higher, reduced motivation (N-2) with reduced self- expression and emotion (N-l) and deficit syndrome (N-3); deterioration (C2) with total impairment (1-1 + 1-2) and social role impairment (1-1); mania (M-tot) with manic syndrome (M-l), maximum number of manic fea- tures (M-2), and classical manic features (M-3); and max- imum number of manic features (M-2) with classical manic features (M-3). Depression and depressive syn- drome items were substantially intercorrelated, as were global ratings and the sums of individual items.
Selection of Items for Version 2 of the LDPS. On the basis of these results, and input from raters, the following changes were made:
1. Global ratings, and items P-4 (auditory voices), P-5 (concurrent delusions and hallucinations), and P-6 (preoccupation with delusions or hallucinations) were eliminated (intercorrelated with other items).
2. Based on the relatively low reliability of the nega- tive symptom items, and the observation that longitudinal course of negative symptoms is difficult to rate retrospec- tively (Arndt et al. 1995), only items for observable nega- tive symptoms were retained (blunted affect and poverty of speech), along with the global deterioration item, which was highly correlated with negative symptoms in the fac- tor analysis. The remission item was dropped because it was strongly negatively correlated with deterioration.
3. Inappropriate affect (D-2) and impaired attention (D-3) were eliminated (poor reliability).