Disclaimer: English Kinda Thing

The sole purpose of the "English Kinda Thing" is to document my attempts to correct my own mistakes in standard English usage and to share the resources I find. In no way do I attempt to teach nobody English through these blurbs--just as I intend not to teach nobody to be a neurotic and psychotic handicap in Ratology Reloaded or Down with Meds! :-)

Friday, May 15, 2009

Combining Activities of Daily Living with Instrumental Activities of Daily Living to Measure Functional Disability (Spector & Fleishman, 1998)

Spector, W. D., & Fleishman, J. A. (1998). Combining Activities of Daily Living with Instrumental Activities of Daily Living to Measure Functional Disability. Journal of Gerontology: Social Sciences, 53B(1), S46-S47.

Nagi's framework: differentiating between

  1. Impairments: Incontenence such as bowel and bladder incontinence (Jagger, Clarke,& Davies, 1986; Linacre et al., 1994)
  2. Functional limitations: walking up the stairs etc (e.g.,bending and reaching; Clark et al., 1997; Haley, McHorney, & Ware, 1994; Jette, 1980; Linacre et al., 1994; (Wolinsky & Johnson, 1991))
  3. Disability: focusing only on a core set of ADL and IADL activities (Clark et al., 1997; (Fitzgerald, Smith, Martin, Freedman, & Wolinsky, 1993)).

3 Approaches about ADL and IADL measures:

  1. Treating ADL and IADL as distinct concepts and maintaining separate ADL and IADL measures (e.g., Jette et al., Stern, 1995; Sloane, Hoerger, & Picone, 1996
  2. Include information on IADL only for those without ADL disabilities (e.g., Altman & Walden, 1993; Lagorge, Spector, and Sternberg, 1992, Tennstedt, Crawford, & McKinlay, 1993; Spector & Kemper, 1994).
  3. Combining ADL and IADL into one measure

The issues to be addressed when considering whether it is a legitimate to combine ADL and IADL

  1. Unidimensionality and local independence
  2. How to combine responses and turn them into a composite score

Research on dimensionality

  1. The multidimensionality of functional disability (Clark, Stump, & Wolinsky, 1997; (Fitzgerald et al., 1993); (Wolinsky & Johnson, 1991))
  2. The unidimensionality (Kempen & Suurmeijer, 1990;Spector et al., 1987; Suurmeijer et al., 1994)

IRT model: The authors did a good job in providing an introduction for IRT models

  1. Some other research done using IRT to assess health status measure (Haley, McHorney, & Ware, 1994; Granger et al., 1993; Linacre et al., 1994; Silverstein et al., 1992; Teresi, Cross, & Golden, 1989)
  2. Previous studies include items for ADLs, IADLs, impairments, and functional limitations together. This study only looks at ADLs and IADLs.
  3. Θ or theta as the latent dimension of disability
  4. ICCs or item characteristic curves shows the probability of positive response as a function of theta
  5. The advantage of IRT over CTT

    1. IRT item parameters are invariant to the population distribution of the trait being measured
    2. When analyzing dichotomous models, nonlinear models such as IRT models are preferable.
    3. IRT provides reliability estimate for each of the items
    4. IRT provides a framework to assess item bias
  6. IRT model and the parameters

    1. Beta: difficulty and location
    2. Alpha: discriminating ability
  7. Existing studies using 1 PL without double-checking its appropriateness (Haley, McHorney, & Ware, 1994; Heinemann et al., 1993)


  • Data are from the 1989 National Long-Term Care Survey (NLTCS)

    • Purpose of the Survey: to provided a national sample of functional disabled people
    • To estimate the change in functional disability between 1998 and the previous years
  • Measures of functional disability include Katz's Activities of Daily Living Scale (1963) and the Instrumental Activities of Daily Living Scale by Lawton and Brody (1969)
  • Positive or disabled response:

    • ADL: the receipt of human help to perform task
    • IADL: respondents did not perform the task, could not perform the task and their disability is a result of a health problem.
  • Katz et al (1969): the differentiation between dependence and independence
  • Bilog-MG (Zimowski et al., 1996) is used to perform IRT analyses.

    • The item parameters are estimated using the maximum likelihood procedure
    • the person parameters are estimated using Bayesian (EAP: expectation a posteriori) approach (Bock & Aitkin, 1981) The EAP method could provide estimation for people with all correct or all incorrect response.


  • Sample data: age, frequency of positive response
  • Exploratory and confirmatory factor analysis:

    • Tetrachoric correlations
    • The scree method (Rummel, 1970)
    • If the first eigen value is relatively larger than that for the second one, the item is relatively unidimensional (Lord, 1980)
    • The number of eigenvalues greater than 1
    • Bentler's fit index (1990)=> crit=.90
    • Root Mean Square Error of Approximation (RMSEA) RMSEA <.05 indicates an acceptable model (1993)
  • IRT Analysis

    • 1 PL and 2 PL
    • Model comparison: differences in λ2 (calculated as the -2 Log Likelihood) and the ratio of the difference over λ2 for 1 PL
    • Large sample size makes it easier to obtain significant λ2 statistics
    • Criteria to evaluate model fit

      • The meaningfulness of λ2 test
      • The size of the ratio
      • The differences in the magnitude of the slops
      • The correlation between the estimated theta for 1 PL and 2 PL
      • Adding more parameters adds more complexity
    • Item location Parameter

      • the breadth
      • How well an item differentiate between ability levels within a given range
      • Large gaps identify the point at which the scale is less precise and suggest for the inclusion of new items within the range
    • DIF found in gender


  • Unidimensionality
  • 1 PL is sufficiently good
  • People with al zeros: we need to have items that are more sensitive or milder
  • Lawton and brody (1969) suggest IADLs to be more complicated than ADLs. As a result, the IADLs should have a lower value in the location paramter.
  • Studies showing that ADL and IADL overlaps (Spector et al., 1987, Suurmeijer et al., 1994, Kempen, Myers, and Powell, 1995)

Fitzgerald, J. F., Smith, D. M., Martin, D. K., Freedman, J. A., & Wolinsky, F. D. (1993). Replication of the multidimensionality of activities of daily living. Journal of Gerontology, 48(1), s28-s32.

Wolinsky, F. D., & Johnson, R. J. (1991). The Use of Health Services by Older Adults. Journal of Gerontology, 46(6), S345-S357.

Spector and Fleishman (1998) used exploratory and confirmatory factor analyses as well as Item Response Theory to analyze data of the combined ADL/IADL scale, which included 16 items and was collected through the NLTCS project. This study used only data from individuals who reported disabled in at least one of the disability indicators. Results of exploratory and confirmatory factor analyses provided supportive evidence for the unidimensionality of the combined scale. In addition, the authors made a comparison between the goodness-of-fit of the one-parameter and that of the two-parameter IRT model. It was found that the one-parameter model yielded a sufficient good fit, which is used as the supporting evidence for the feasibility of using a composite score to summarize the ADL/IADL data.

No comments:

Post a Comment