On item difficulty in the modified Iowa Level of Assistance Scale

Dear Editor,

We read with interest the paper by Elings et al. [1] investigating the modified Iowa Level of Assistance Scale (mILAS) in patients after hip replacement. We are concerned, however, about their recoding and computing procedures, which could have introduced serious flaws in the results.

• 1.

  Treating ‘not tested’ answers (scoring option six) as missing responses (and then excluding these cases) produces mathematical problems and could lead to significantly biased item and person parameter estimates [2], particularly when high percentages are missing and/or there are so-called ‘missing not at random’ data [3]. Indeed, the missing values relate to two distinct categories: 1) patients not tested for reasons of safety (the highest level of dependence according to Shields et al. [4]); 2) patients who did not have to climb stairs at their preferred discharge destination, and thus were not assessed (see Ref. [1,Appendix B]). We do not know how many there were in this second category.

• 2.

  ‘Stair climbing’ was the item showing highest difficulty in both Tables 2 and 4 (note “a more negative logit score indicates more difficult items” [1,Appendix D]). Please also refer to our Rasch map in Fig. 1 (regarding 99 subjects who underwent total hip arthroplasty within the six previous days; (for Materials and Methods please see [5]), where the hierarchy of ILAS item difficulty shows that ‘stair climbing’ was one of the most difficult items (after ‘walking speed’). Nonetheless, Elings et al. state that stair climbing “was the easiest item to complete for the majority of patients…”. This statement is in harsh contrast with the literature on this topic [4, 5, 6].

  

  Fig. 1

  Map of person-ability (left-hand column; each “#” indicates two patients) and item-difficulty (right-hand column; each item is indicated by its number in the scale) for the ILAS. The vertical line represents the measure, in linear logit units (on the extreme left-hand side). By convention, the average difficulty of items in the test is set at 0 logits (and indicated with “M”). Accordingly, a candidate with average ability is indicated with M. “S” markers are placed one sample standard deviation away from the mean, and “T” markers are placed two sample standard deviations away from the mean.

   

  View Large Image | View Hi-Res Image | Download PowerPoint Slide

• 3.

  We wonder if their operational definition of ‘stair climbing’ is comparable with that in the original ILAS [5]. During ‘stair climbing’ patients used ‘infrastructural assistance of a guardrail’, while in the original ILAS the item ‘climbing-up and down three steps’ can be performed with sticks, crutches, or frames: no mention of a handrail.

• 4.

  In order to reduce the early floor effect in the ILAS scale measuring the assistance needed during functional activities, the ILAS authors [5] added a complementary scale, related to the need for assistive devices, not used in the present study.

For these reasons, we wonder whether Elings et al. could kindly provide further information about these aspects of their study, to clarify the interpretation and help towards a possible refinement of the psychometric properties of this outcome measure.

On item difficulty in the modified Iowa Level of Assistance Scale