The goal of the study was to develop and apply a predictive model approach to reduce the number of items collected for scales that yield a total summary score. A parsimonious subset of items from the 21-item Quality of Life Scale (QLS) that can accurately predict the total scale score was sought and evaluated in 198 patients with schizophrenia, using a statistical modeling approach. Two additional data sets were used for model validation: the subset of 101 patients used in the model construction who had the QLS administered approximately 1 year later and a new sample of 37 patients. Using only seven QLS items as predictors, the correlation was 0.9831 between the predicted and true QLS totals. Applying the model based for these seven QLS items, the correlations from the first and second validation data sets were 0.9791 and 0.9637, respectively. The study demonstrates that a small subset of items of the QLS predicts the entire 21-item scale with high accuracy. Two validation samples have confirmed the finding. This reduces the effort associated with scale administration and is likely to increase the assessment of an important functional domain. Such models can guide efforts for item reduction in other rating instruments.