Is there a linear relationship between the Brief Psychiatric Rating Scale and the Clinical Global Impression-Schizophrenia scale? A retrospective analysis

BACKGROUND: Although the Brief Psychiatric Rating Scale (BPRS) is widely used for evaluating patients with schizophrenia, it has limited value in estimating the clinical weight of individual symptoms. The aim of this study was 4-fold: 1) to investigate the relationship of the BPRS to the Clinical Global Impression-Schizophrenia Scale (CGI-SCH), 2) to express this relationship in mathematical form, 3) to seek significant symptoms, and 4) to consider a possible modified BPRS subscale. METHODS: We evaluated 150 schizophrenia patients using the BPRS and the CGI-SCH, then examined the scatter plot distribution of the two scales and expressed it in a mathematical equation. Next, backward stepwise regression was performed to select BPRS items that were highly associated with the CGI-SCH. Multivariate regression was conducted to allocate marks to individual items, proportional to their respective magnitude. We assessed the influence of modifications to the BPRS in terms of Pearson's r correlation coefficient and r-squared to evaluate the relationship between the two scales. Utilizing symptom weighting, we assumed a possible BPRS subscale. RESULTS: By plotting the scores for the two scales, a logarithmic curve was obtained. By performing a logarithmic transformation of the BPRS total score, the curve was modified to a linear distribution, described by [CGI-SCH] = 7.1497 × log(10)[18-item BPRS] - 6.7705 (p < 0.001). Pearson's r for the relationship between the scales was 0.7926 and r-squared was 0.7560 (both p < 0.001). Applying backward stepwise regression using small sets of items, eight symptoms were positively correlated with the CGI-SCH (p < 0.005) and the subset gave Pearson's r of 0.8185 and r-squared of 0.7198. Further selection at the multivariate regression yielded Pearson's r of 0.8315 and r-squared of 0.7036. Then, modification of point allocation provided Pearson's r of 0.8339 and r-squared of 0.7036 (all these p < 0.001). A possible modified BPRS subscale, "the modified seven-item BPRS", was designed. CONCLUSIONS: Limited within our data, a logarithmic relationship was assumed between the two scales, and not only individual items of the BPRS but also their weightings were considered important for a linear relationship and improvement of the BPRS for evaluating schizophrenia.