Reporting outcomes of the Assisting Hand Assessment: what scale should be used?

The Assisting Hand Assessment (AHA) is now a well-established and frequently used evaluative tool. It measures how effectively children with unilateral hand dysfunction actually use their involved hand collaboratively with their well-functioning hand to perform bimanual tasks. Since the first publication about the AHA in 2003,1 several studies have reported on different psychometric properties, which have important implications for interpretation of outcomes.2–4 However, significant variation in the reporting of results in intervention studies using the AHA emphasizes a need for clarification. This paper aims to provide recommendations for how to report AHA outcomes, where improved consistency will facilitate comparison of data across studies. The AHA produces two types of ordinal level scores at the score form: raw scores and a percentage score. The AHA version 4.4 includes 22 test items each rated on a 4-point scale, giving a raw score range of 22 to 88 points, with the higher number indicating higher ability. These raw scores are also presented as a percentage score, simply indicating the percentage of the possible score the child has attained. The use of ordinal scores however has long been known to be invalid for descriptive statistical measures such as mean and standard deviations, and for parametric statistical analyses. Furthermore, simply adding up raw scores of unknown intervals may cause misrepresentation of results. To use a metaphor; which is more, two big apples or three small ones? Counting ‘how many’ provides a sum, but it does not tell how much since that can only be determined by putting them on a scale. For raw scores, changes at the lower or upper ends of a scale may understate the real increase of ability as compared to changes in the middle of the scale. Thus, the magnitude of change in an observed raw score will reflect a different increase depending on the starting point along the scale.5 Interval level measures are recommended for reporting outcomes, especially regarding evaluation of change.6 Since the AHA was developed by use of Rasch measurement analysis, interval level measures in unit logits (log-odds probability units) were found to range from −10.18 to 8.70 logits.2 Earlier, AHA logit data was only attainable from a Rasch-conversion performed by the original authors. Following further scale development and recommendations,6 a conversion table showing the relationship between the different AHA outcome scales is hereby provided (see Table I). The logit scale which ranges from negative to positive values may however be difficult to interpret. Therefore a more user-friendly 0 to 100 scale is now introduced, called the logit-based 0 to 100 AHA-unit scale (AHA-units). This scale is still Rasch-based and thus presents interval level data. By converting raw scores using the information in Table I, AHA results can now easily be reported on the logit-based 0 to 100 AHA-unit scale, which is strongly recommended. To ensure this scale is not confused with the ordinal level percentage scale, the recommendation is to clearly indicate that the reported data concerns logit-based AHA-units. Furthermore, to interpret whether a change seen in AHA measures is a true change, information is needed on the degree of error variance the measure contains. To consider both rater inconsistency and by chance variations in children’s behaviors, test–retest data can be used to calculate the smallest detectable difference (SDD). This has been calculated for the AHA.4 The SDD for AHA raw scores is 3.89; thus a change of 4 raw score points can be regarded a true change, exceeding the error variance within the measure. A raw score change of 4 points was also found to be a significant change, with p=0.046.4 For logit measures the corresponding SDD is 0.97 logits, and for the logit-based 0 to 100 AHA-units scale the SDD is estimated to five AHA units. The question about what magnitude of change would indicate a clinically meaningful difference is more complicated and has not yet been investigated for the AHA. Although, as a rule of thumb, at an individual level a change has to be equal to, or bigger than the SDD level to be considered a real change. However, also small changes can be very meaningful for an individual, if they imply functional gains. For the AHA, this needs to be further evaluated. As a concluding remark I would like to highlight that AHA measures are valid and reliable when used with children aged between 18 months and 12 years who have unilateral cerebral palsy or obstetric brachial plexus palsy. There are several ongoing projects developing the scale for younger and older age groups, and other diagnostic groups. The Mini-AHA for children with unilateral cerebral palsy in the age range 8 to 18 months is the next one in progression.7 Further information and continuous updates about the AHA can be found at http://www.ahanetwork.se.