Disability: a model and measurement technique. (Williams, Johnston, Willis, & Bennett, 1976)

Williams, R. G., Johnston, M., Willis, L. A., & Bennett, A. E. (1976). Disability: a model and measurement technique. British Journal of Preventive & Social Medicine, 30(2), 71-78.

Abstract: Current methods of ranking or scoring disability tend to be arbitrary. A new method is put forward on the hypothesis that disability progresses in regular, cumulative patterns. A model of disability is defined and tested with the use of Guttman scale analysis. Its validity is indicated on data from a survey in the community and from postsurgical patients, and some factors involved in scale variation are identified. The model provides a simple measurement technique and has implications for the assessment of individual disadvantage, for the prediction of progress in recovery or deterioration, and for evaluation of the outcome of treatment regimes.
http://jech.bmj.com/cgi/content/abstract/30/2/71

My notes:

• Katz et al. (1963)—hypothesize that the relearning of very basic abilities such as continence, feeding, and washing followed the same pattern taken by learning in infancy
• Carroll (1962) explored similar orderings in the recovery of stroke patients
• Harris and Luck (1972) explored cumulative orderings of activities

The authors' work assumed that disability would fit the cumulative model called Guttman Scaling (Guttman, 1950)

Example:

1. Getting up
2. Going out of the house
3. Going back to work
   

With 0 means nondisabled and 1 means disabled

Scale Type 111 011 001 000

Nonscale Type 101 110 100 010

Different types of plausible error: measurement error, sampling error and random variation among individuals

Two coefficients are used to represent the conventional level of acceptable errors which allows us to draw inferences that there exists a valid cumulative and unidimensional Guttman scale:

• Coefficient of reproducibility: Acceptable when >.90. It reflects the variation in the proportion of disabled people under each item
• The coefficient of scalability: Acceptable when > .6. It tells us the proportion of the remaining response that could be correctly predicted using our hypothesis.

The authors then went on analyzing two sets of data using the Cuttman scale hypotheses.
For me, the most interesting part of their finding is in how the ordering of occurrences for certain disabilities is dependent on gender and SES. For instance, men were slower in relinquishing their independent mobility than cooking for themselves… while women, the contrary… 8-O

According to the authors, the usefulness of this model is that it helps us to make prediction about the deterioration and recovery processes.

Please see the following link for more information about Guttman Scale: http://en.wikipedia.org/wiki/Guttman_scale