Online Program

Return to main conference page

Friday, January 12
Fri, Jan 12, 8:30 AM - 10:15 AM
Crystal Ballroom B
Clinical Outcome Assessment (COA)

Optimal methods for shortening patient-reported outcome measures (303991)

*Daphna Harel, NYU 
Miyabi Ishihara, UC Berkeley 
Brooke Levis, McGill University 
Brett D Thombs, McGill University 

Keywords: Patient-reported outcome measures, short forms, optimal test assembly

Patient-reported outcome measures (PROs) - such as aspects of mental health - assess aspects of patients' lives from their own perspective. Patients enrolled in clinical trials or observational studies may be asked to respond to many different scales to provide information regarding their experiences or treatment response. Efficient measurement of PROs is thus essential to limit patient burden and research cost. However, methods to shorten these instruments are under-developed, leading to several shortened versions of the same PRO.

Optimal test assembly (OTA) is an application of linear programming used frequently for item selection in designing high-stakes educational tests that incorporates the results of an IRT model to select a subset of an item pool that best satisfies pre-specified constraints while optimally maximizing an objective function, such as total test information. This presentation shows how OTA may be used to shorten PROs in an objective, reproducible, and replicable way to produce optimal shortened forms, and compares the use of OTA to the selection of items based on factor loadings. The utility of this method then applied to the Patient Health Questionnaire – 9.