Routine functional assessment in clinical practice requires advances in the construction of health status measures that are intended for use at the individual-patient level.
The key question underlying this project was: can Item Response Theory (IRT) as a measurement paradigm yield a pool of items of sufficient breadth of measurement that will be useful for individual-patient assessment and monitoring? The short-term objective was to evaluate whether use of IRT as a measurement paradigm can generate measures of physical-health status that would be appropriate for use at the individual-patient level. The efficiency and precision of practice-based functional health assessment at the individual-patient level would be enhanced if one could develop an adaptive measurement paradigm with an underlying item structure tailored to the functional proficiency level of each individual respondent. This can be achieved through the conjoint use of IRT and computerized adaptive testing. However, before a computerized-adaptive framework can be designed, the item bank must exist and item properties, particularly item difficulty, must be determined. This project achieved this necessary, but not sufficient, prerequisite for computer-adaptive functional health assessment.
This project used IRT to develop platforms of measurement for functional health status. In Phase 1, we used secondary data from the Asset and Health Dynamics among the Oldest Old (AHEAD) Study, a nationally representative panel study of elderly in the U.S, to equate three different modules of activities of daily living (ADLs) in 4,655 respondents. In Phase 2, we developed an item bank of physical-health items based on the content of existing instruments. The goal in developing this item bank was to construct a pool of items that would spread across the entire underlying continuum of functional ability. We also conducted focus groups with community-dwelling elderly to glean their understanding of the items and response scales. Three forms (versions) of the survey were developed and they were equated using IRT in a sample of 3,358 older persons.
In Phase 1, secondary data analysis from the AHEAD data set equated three distinct sets of ADL measures (AHEAD, Longitudinal Study of Aging (SOA), and National Long Term Care Survey (NLTCS)). A two-parameter model for dichotomous items was used for linking the three modules of items. Using this model, both sets of supplemental items were successfully linked to the common items, allowing the placement of all items on the same underlying measure of ability. The small-muscle IADLs were the easiest of all the items for respondents to perform. The item on walking from the Longitudinal Study of Aging was the most difficult for respondents to perform. These analyses furthered our understanding of the practical and methodological challenges in using IRT to equate functional health status measures.
In Phase 2, 162 published articles, books, and book chapters that focused on the measurement of physical functioning and functional status were abstracted. A total of 1,588 items were abstracted form published literature and entered into an item banking program. Elimination of redundancies resulted in 206 candidate items for Health of Seniors Survey. The 206 items cross-cut toileting, bathing, cooking, dressing, mobility, household and community activities, and recreation. The items were successfully equated to a common scale of function. The dressing items were the most discriminating, followed by bathing, toileting, mobility, cooking/eating, and household and community activities. The five most discriminating items were put underclothes on, manage clothes after toileting, move between rooms, take pants/slacks off, and get into bed. Most of the items were located on the easier end of the ability continuum. Only six would classify as being very difficult.
IRT linking methods are a useful way to overcome test dependency and to place items on a common metric even if different respondents answer different sets of items. This project furthered the understanding of practical and methodological issues in using IRT for test equating and linking.
External Links for this Project
None at this time.