Choosing negative symptom instruments: issues of representation and redundancy.

Assuming that the negative syndrome in schizophrenia may be multidimensional, this study examines how conclusions about the structure of negative symptoms may be influenced by the particular rating scale used, the level of data reduction used (such as total, subscale and individual item scores), and also the type of data analyses used to compare scales. Forty-seven subjects with RDC schizophrenia were rated on three instruments: the negative symptom subscale of the BPRS (BPRS-WR); the negative symptom subscale of the PANSS (PANSS-NS); and the SANS. Comparisons were made of different levels of data reduction and different methods of analysis, which included bivariate correlation, bi-multivariate canonical correlation and redundancy analysis. We found that while the total scores from all three scales were highly correlated and therefore highly redundant, both the individual items and subscale scores from the SANS contained information independent of the BPRS-WR, and also, to a lesser extent, of the PANSS-NS. When the BPRS-WR was correlated with either the SANS or the PANSS-NS, one strong canonical variate (CV) emerged, on which all or most items loaded, particularly the affective items. When the SANS and PANSS-NS were correlated, this component again emerged along with three less strong but interpretable components. When examining the non-symmetrical redundancy, we found that the BPRS-WR variates explained 40% of the SANS variance, while conversely the SANS variates explained 80% of the BPRS-WR variance. The PANSS-NS variates were found to explain 58% of the SANS variance, while the SANS variates explained 85% of the PANSS-NS variance. Finally, the PANSS-NS variates explained 79% of the BPRS-WR variance, while conversely the BPRS-WR variates explained 54% of the PANSS-NS variance. AH three scales appear to measure a single general 'affective' component of the negative syndrome, while the PANSS-NS and the SANS also cover additional components which identify cognitive, anergic and social dimensions. This extra information is lost, however, if inappropriate data reduction and/or statistical analyses are used. The fact that the three instruments predicted the various dimensions of the negative syndrome to different degrees suggests that the best choice of a negative scale depends on the type of information required. Nevertheless, further examination of how negative symptom scales cover the multi-dimensional nature of the negative syndrome is required.