2005 HSR&D National Meeting Abstract
3038 — Development of a Computerized Adaptive Self-Report Functional Measure
Velozo CA (North Florida/South Georgia Veterans Health System)
Johnson A (University of Florida)
Byers K (North Florida/South Georgia Veterans Health System)
Piantieri S (North Florida/South Georgia Veterans Health System)
Pomeranz J (North Florida/South Georgia Veterans Health System)
Lehman L (University of Florida)
Wang Y (North Florida/South Georgia Veterans Health System)
Wang J (University of Florida)
Wen P (North Florida/South Georgia Veterans Health System)
Gray D (Washington University)
Global functional measures, such as the Functional Independence Measure, lack sensitivity and are burdensome to both the clinician and patient. The purpose of this study is to demonstrate how Item Response Theory (IRT) and computer technology can be used to develop more precise and efficient measures of functional ability.
An item bank based on the Activity dimension of the International Classification of Functioning Disability and Health (ICF) was modified by consumer focus group input (n=30), professional panel consultation and verified through cognitive interviews (n=32). Two hundred sixty-three items were field-tested on 413 veteran and civilian community-dwelling participants with movement disabilities due to low back pain, upper extremity, lower extremity, and spinal cord injury. Rasch analysis, differential item functioning (DIF) and principal components analyses were used to identify constructs and select items for the final item bank. A computer algorithm was developed to tailor item administration thereby reducing respondent burden. Finally, industry standard apache web server and PHP server side scripting technology was used to develop a web-based computerized adaptive instrument.
A final bank of 257 items covering six constructs were empirically derived: 1) positioning/transfers, 2) gross upper extremity, 3) fine hand, 4) walking/climbing, 5) wheelchair/scooter and 6) self care. All constructs showed person separation reliability (analogous to Cronbach’s alpha) between .90-.97. When applied to relevant diagnostic groups, three of the constructs showed slight ceiling effects (11-17%). DIF results supported the separation of gross upper extremity and fine hand items into two constructs. Five to fifteen items per construct were required when conducting computer simulations of the measure at 0.3 standard error criterion.
The above study demonstrates the feasibility of developing computerized adaptive functional measures that have good psychometrics and are efficient to administer.
IRT and computer technology can lead to functional measures that advance health services research while reducing respondent burden of veteran health care consumers. Precise and efficient measures of functional ability are critical in advancing care for veterans with chronic illnesses.