Purpose: Psychometric scales often change over time, complicating comparison of scores across different versions. The Impact of Cancer (IOC) scale was developed to measure quality of life of long-term cancer survivors. We sought to develop a method for scoring the earlier version, IOCv1, to obtain scores comparable to IOCv2, which is the recommended version.
Methods: Data from 1,828 cancer survivors who had completed a questionnaire including all IOCv1 and IOCv2 items were randomly split into training, validation and test sets. The training and validation sets were used to develop and validate linear regression models for predicting each IOCv2 item missing from IOCv1. The models were then applied to the test set to obtain pseudo-IOCv2 scores, which were compared to observed scores to assess predictive performance of the models in independent data.
Results: Observed and pseudo-IOCv2 scale scores were highly correlated in the test sample and had mean differences near zero. The models performed especially well in predicting summary scale scores, with correlations exceeding 0.98.
Conclusions: The approach facilitates comparison across samples of survivors surveyed using different versions of the IOC and may be useful to other investigators trying to compare participants surveyed using different versions of the same instrument.