Using parameter perturbation to facilitate transparency in measure development

Excerpt

Transparency and openness in sharing scientific results is increasingly becoming a hallmark of research. However, concerns exist regarding the unrestricted disseminations of data or products. Firstly, it may sometimes conflict with the desire of private entities to disseminate their research while maintaining keeping proprietary information unpublished. Furthermore, unrestricted dissemination of information poses challenges to upholding scientific integrity without adequate safeguards in place. In Quality of Life Research, if a study presents the development of an item bank, the item parameters need to be published as part of the paper, which can conflict with, for example, the proprietary nature of item parameters used for scoring PROMIS(R) measures. Under an agreement with the US Department of Health and Human Services (DHHS), Northwestern University is obligated to promote appropriate use and prevent misuse and abuse. In general, data use agreements can be used to obtain proprietary parameters, enabling their use under restricted circumstances; in the PROMIS case, calibrated parameters can be requested by contacting the authors. During the resubmission process of a recent article ([1], this issue) to Quality of Life Research, we reached an impasse wherein only this restriction precluded the article's publication. The solution we reached was to apply random transformations, or perturbations, to the item parameters, masking their true values. Below we provide some details on the strengths and weaknesses of this approach, and some recommendations for authors whose measures include proprietary item parameters, for reviewers, and for readers who represent consumers of this research. …