B -- Multi-dimensional and Hierarchical Psychometric Modeling for Cancer Outcomes Measurement

The National Cancer Institute intends to negotiate, on a sole source basis, with the Center for Health Statistics at the University of Illinois at Chicago, 1601 W. Taylor, Chicago, IL 60614, for services for the - Evaluation of the Added-value of Multi-dimensional and Hierarchical Psychometric Modeling for Cancer Outcomes Measurement. Under 41 USC 253(c)(1) and FAR citation 6.302-1, the Center for Health Statistics at the University of Illinois at Chicago (UIC) is the only source with the pre-requisite experience to complete this procurement. The reason for this justification for other than full and open competition is due to UIC's unique expertise in applied multi-dimensional and hierarchical (including the bi-factor model) methods for modeling health outcomes data (with ordinal response items) for multiple applications including computerized-adaptive testing. In addition, they have developed the only software program that estimates the full-information bi-factor model parameters for ordinal response data. The primary purpose of this contract is to assess the value of utilizing multi-dimensional models and hierarchical psychometric models for improving measurement of cancer patient-reported outcomes (PROs) to create both a profile of scores as well as an overall summary score, as compared to more standard approaches used in PRO measurement. The two overarching goals of the work proposed are to examine the added-value, over traditional uni-dimensional approaches, of: 1) multi-dimensional item response theory models to improve the precision of measurement of PRO sub-domains borrowing information from other correlated sub-domains; and 2) hierarchical models (including the bi-factor model for ordinal data) to estimate a summary score that takes into account the precision of and relationship among various sub-domains underlying the higher order construct measured by the scale. This study will use both real ordinal-response data and simulated data to determine the limits of these methods in terms of evaluating the robustness and trustworthiness of item parameters and scores given varying sample sizes and correlations among domains. Further, the strengths of these complex methods will be considered for various PRO assessment strategies including computerized-adaptive testing (CAT). The contractor will produce a didactic workbook that describes the methodology, various models strengths and limitations for different datasets and PRO applications, input and output software code, and several examples of applying these methods in cancer outcomes data. Further, the contractor will deliver a publishable manuscript and presentation that summarizes key findings. This is not a request for competitive quotation. However, if any interested party believes it can meet the above requirements they may submit a statement of capabilities. The contractor must also demonstrate knowledge and research experience in the field of psychometric modeling in health outcomes research and be able to deliver the same superior products. Contractors should also reference recently published papers in peer-reviewed journals on the subject of multi-dimensional item response theory (IRT) modeling, bi-factor analysis, and CAT in health measurement. It will be important for the contractor to be able to produce both written and presentation materials to articulate the complex methods to a cancer outcomes research audience. Information furnished should not exceed 10 pages. Responses should be submitted electronically to Cynthia D. Washington, Contracting Officer, at contact@example.gov in either WordPerfect or Microsoft Word or fax to +1 (555) 000-0000 by June 20, 2005 at 2:00pm local time. The NAICS Code is 541910. See Numbered Note 22.