Activities of daily living (ADL) and instrumental activities of daily living (IADL) items were stable over time in a longitudinal study on aging. (Finlayson, Mallinson, & Barbosa, 2005)

Finlayson, M., Mallinson, T., & Barbosa, V. M. (2005). Activities of daily living (ADL) and instrumental activities of daily living (IADL) items were stable over time in a longitudinal study on aging. Journal of Clinical Epidemiology, 58(4), 338-349.

Department of Occupational Therapy, University of Illinois at Chicago (MC 811), 1919 West Taylor Street, Chicago, IL, 60612, USA. marciaf@uic.edu

OBJECTIVE: The purpose of this analysis was to examine the stability over time of the activities of daily living (ADL) and instrumental activities of daily living (IADL) items in the Aging in Manitoba (AIM) Longitudinal Study and to evaluate the existence of differential item functioning across settings (home, nursing home).

STUDY DESIGN AND SETTING: The study used data from 607 participants of the AIM Longitudinal Study who were more than 85 years of age in 1996 and who had complete data from 1983, 1990, and 1996 for all ADL and IADL items. Rasch analysis was used to examine how the rating scale of the ADL and IADL items was used by participants, and to determine if the ordering of items remained stable across three time periods (1983, 1990, 1996) and the two different settings (home, nursing home).

RESULTS: The rating scale worked best when dichotomized into "received no assistance" and "receives assistance." Except for four items (making tea, making meals, doing nursing care, and going outside in any weather), the items were stable across administration periods, and across settings.

CONCLUSION: The AIM can be used to evaluate changes in disability over time and may have the potential to identify those at risk for transitions in care.

http://www.ncbi.nlm.nih.gov/pubmed/15862719

My notes

There were approximately 50 longitudinal studies on aging internationally as of 2005.

One reason the aging studies receive interest is in the pattern and predictors of long term care use.

Advances in technologies have shifted the definition of independence in the past 30 something years. For instance, people might no longer report having problem doing laundry, shopping for groceries or cooking because they could use washers and driers, online shopping and microwave oven to perform the tasks they otherwise might not be able to perform. In other words, advances in technologies might inevitably result in the "observed higher functioning of individual" while such observations are not the real change in the persons' ability.

Comparing TCC and Rasch model

"Rasch measurement focuses on the performance of items. Rasch measurement takes ordinal level data and converts the probability of endorsing an item to a logti or log of odds ratio."

In Rasch model, when the ordering of item difficulties are stable (β), we can use the data to analyze the ability or functional status for the individuals (θ). In this concept, "stability" does not refer to test-retest reliability in CTT. Stability has to do with how functional status is defined by the respondent.

For the Aging in Manitoba (AIM) study, in calculating a functional status score for the AIM, higher scores equal receipt of more help.

Methods of this study

• Use Winstep Program
• Testing the assumptions of Rasch model: local independence and unidimensionality
  
  • Item fit
  • Item discrimination
  • Principal component analysis
• Item fit: Item infit statistics are sensitive to sensitive to unexpected responses. Values greater than 1.0 indicate departures from unidimensionality while values smaller than 1.0 indicates redundancy and the potential for violating local independence. An item infit ranging between .8 to 1.2 is considered good while that ranging between .7 to 1.3 considered acceptable.
• Item discrimination: Item discrimination estimates describe the slope for the ICC. Rasch model assumes all ICC slopes have a value of 1.0. Linarcre showed that a value between .5 and 1.7 generally equate to the infit statistics of 1.2 to 0.8—it also suggest that "ICC in this range do not unduly impact the assumptions of the Rasch model and have no meaningful impact on measurement.
• Stability within settings: The authors assess the stability of item calibration by examining the order of items was constant over the 3 measurement periods separately for the home and nursing home setting. Items with the logit values 2 standard deviation apart are considered as significantly different.
• Stability across settings: Calculate item calibration across all three measurement period for home and nursing home. Again > 2 SE indicates significant differences

Results section

5-point scale was rescaled into a 2-step dichotomous scale due to the disordered steps

Checking Rasch model assumptions: Smith notes that infit statistics can be susceptible to test variance and sample size and recommended the use of standardized residual (zstd) in conjunction with the infit mean square (MnSq) and with common sense.

Item discriminations >.5 indicates that data for these items did not fit Rasch model assumption

The assumption of unidimensionalty was supported using PCA. A minimum of 10% contribution of additional variance accounted for is usually considered the cutoff point to suggest an additional factor in the factor structure.

The disordered step might be an artifact of the intent of AIM study focusing on measuring the unmet needs rather than functional status per se.

People who made a significant transition such as from a home to a nursing home showed a more variable pattern in how they endorse ADL and IADL items over time--- as shown by the floating squares shown on figure 3

Future researchers might wish to complete similar analyses on younger individuals

Finlayson and his associates conducted a study to examine the stability of responses for ADL and IADL items over years. The data used this this study contained a subset of data collected through the Aging in Manitoba (AIM) longitudinal Study. All participants were born in 1911 or ealier with the average age of around 90. Rasch model analysis was conducted using the WINSTEPS program. Originally, there were 5 response categories including: 1. Do it myself; 2. my spouse or other people in the household does it; 3. Someone from outside the house does it; 4 A formal service does it; and 5. Someone from a facility does it. Because results obtained through the 5-point scale led to disordered steps. As a result, items were collapsed into 2 categories (i.e., receives assistance, not receives assistance). The ensuing analyses were conducted using the two-point rating scale and the data was found to adequately met the assumptions of Rasch model (e.g., local independence, unidimensionality). In addition, the scale was fairly stable within setting (e.g., being at home without services, living in a nursing home) over time as well as cross settings.