Poonam K Pardasaney1, Pengsheng Ni2, Mary D Slavin2, Nancy K Latham2, Robert C Wagenaar3, Jonathan Bean4, Alan M Jette2. 1. Department of Rehabilitation Sciences, Sargent College of Health and Rehabilitation Sciences, Boston University, Boston, MA; Health and Disability Research Institute, School of Public Health, Boston University, Boston, MA. Electronic address: pardasaney@rti.org. 2. Health and Disability Research Institute, School of Public Health, Boston University, Boston, MA. 3. Department of Physical Therapy and Athletic Training, Sargent College of Health and Rehabilitation Sciences, Boston University, Boston, MA; Rudolf Magnus Institute of Neuroscience and Department of Rehabilitation, Nursing Sciences and Sports, University Medical Center Utrecht, Utrecht, The Netherlands. 4. Department of Physical Medicine and Rehabilitation, Spaulding Rehabilitation Hospital Network, Harvard Medical School, Boston, MA.
Abstract
OBJECTIVES: To build an item response theory-based computer adaptive test (CAT) for balance from 3 traditional, fixed-form balance measures: Berg Balance Scale (BBS), Performance-Oriented Mobility Assessment (POMA), and dynamic gait index (DGI); and to examine whether the CAT's psychometric performance exceeded that of individual measures. DESIGN: Secondary analysis combining 2 existing datasets. SETTING: Community based. PARTICIPANTS: Community-dwelling older adults (N=187) who were aged ≥65 years (mean age, 75.2±6.8y, 69% women). INTERVENTIONS: Not applicable. MAIN OUTCOME MEASURES: The BBS, POMA, and DGI items were compiled into an initial 38-item bank. The Rasch partial credit model was used for final item bank calibration. CAT simulations were conducted to identify the ideal CAT. CAT score accuracy, reliability, floor and ceiling effects, and validity were examined. Floor and ceiling effects and validity of the CAT and individual measures were compared. RESULTS: A 23-item bank met model expectations. A 10-item CAT was selected, showing a very strong association with full item bank scores (r=.97) and good overall reliability (.78). Reliability was better in low- to midbalance ranges as a result of better item targeting to balance ability when compared with the highest balance ranges. No floor effect was noted. The CAT ceiling effect (11.2%) was significantly lower than the POMA (40.1%) and DGI (40.3%) ceiling effects (P<.001 per comparison). The CAT outperformed individual measures, being the only test to discriminate between fallers and nonfallers (P=.007), and being the strongest predictor of self-reported function. CONCLUSIONS: The balance CAT showed excellent accuracy, good overall reliability, and excellent validity compared with individual measures, being the only measure to discriminate between fallers and nonfallers. Prospective examination, particularly in low-functioning older adults and clinical populations with balance deficits, is recommended. Development of an improved CAT based on an expanded item bank containing higher difficulty items is also recommended.
OBJECTIVES: To build an item response theory-based computer adaptive test (CAT) for balance from 3 traditional, fixed-form balance measures: Berg Balance Scale (BBS), Performance-Oriented Mobility Assessment (POMA), and dynamic gait index (DGI); and to examine whether the CAT's psychometric performance exceeded that of individual measures. DESIGN: Secondary analysis combining 2 existing datasets. SETTING: Community based. PARTICIPANTS: Community-dwelling older adults (N=187) who were aged ≥65 years (mean age, 75.2±6.8y, 69% women). INTERVENTIONS: Not applicable. MAIN OUTCOME MEASURES: The BBS, POMA, and DGI items were compiled into an initial 38-item bank. The Rasch partial credit model was used for final item bank calibration. CAT simulations were conducted to identify the ideal CAT. CAT score accuracy, reliability, floor and ceiling effects, and validity were examined. Floor and ceiling effects and validity of the CAT and individual measures were compared. RESULTS: A 23-item bank met model expectations. A 10-item CAT was selected, showing a very strong association with full item bank scores (r=.97) and good overall reliability (.78). Reliability was better in low- to midbalance ranges as a result of better item targeting to balance ability when compared with the highest balance ranges. No floor effect was noted. The CAT ceiling effect (11.2%) was significantly lower than the POMA (40.1%) and DGI (40.3%) ceiling effects (P<.001 per comparison). The CAT outperformed individual measures, being the only test to discriminate between fallers and nonfallers (P=.007), and being the strongest predictor of self-reported function. CONCLUSIONS: The balance CAT showed excellent accuracy, good overall reliability, and excellent validity compared with individual measures, being the only measure to discriminate between fallers and nonfallers. Prospective examination, particularly in low-functioning older adults and clinical populations with balance deficits, is recommended. Development of an improved CAT based on an expanded item bank containing higher difficulty items is also recommended.
Authors: Stephen M Haley; Alan M Jette; Wendy J Coster; Jill T Kooyoomjian; Suzette Levenson; Tim Heeren; Jacqueline Ashba Journal: J Gerontol A Biol Sci Med Sci Date: 2002-04 Impact factor: 6.053
Authors: Jonathan F Bean; Dan K Kiely; Sharon LaRose; Evelyn O'Neill; Richard Goldstein; Walter R Frontera Journal: J Gerontol A Biol Sci Med Sci Date: 2009-05-04 Impact factor: 6.053