Children with clinically apparent unilateral Cerebral Palsy (CP) have specific hand function problems. If they use their affected hand, it is always as an assisting hand. Even with only minor impairment of their affected hand, they often do not use it to its full potential in bimanual tasks. This is called developmental disregard [1–4]. Therefore, evaluation of hand function over time or after treatment should focus on the actual use of the affected hand in bimanual activities of daily life, i.e. bimanual performance, as well as on the ability to use the affected hand to its maximal potential in bimanual tasks performed in a standardized environment, which is called capacity [5, 6].

Several assessment tools have been developed for children with unilateral CP. Gilmore et al. [7] reviewed the psychometric properties and clinical utility of several upper limb measures at the International Classification of Functioning, Disability and Health (ICF) activity level [8]. They concluded that the Melbourne Assessment of Unilateral Upper Limb Function (MUUL) [9] is superior for measuring unilateral capacity, while the Assisting Hand Assessment (AHA) [10] and ABILHAND-Kids questionnaire [11] have the best psychometric properties for measuring bimanual performance. Only the AHA is considered to be sufficiently responsive [12].

In 2005 we studied the effect of botulinum toxin A (BoNT-A) injections on upper limb functional skills in children with unilateral CP [13]. We used the MUUL as primary outcome measure, because, at that time, it was the best available tool at the ICF activity level. No effect of BoNT-A could be demonstrated. In hindsight this is not surprising because the MUUL contains many items relating to target accuracy, and these are unlikely to be influenced by BoNT-A. Furthermore, it measures one hand at a time and contains tasks that are usually not done by the assisting hand. In 2003, the AHA was specifically developed to assess the effective use of the assisting hand in bimanual performance. Use of the affected hand is stimulated but not obligated in the AHA. We therefore felt the need for an instrument that measures the capacity of the affected hand in bimanual activities and so developed the Observational Skills Assessment Score (OSAS) using basic ideas of the Video Observation Aarts and Aarts (VOAA) [14]. Whereas the MUUL measures unilateral capacity and the AHA measures actual use in bimanual performance, the OSAS measures both the amount and the quality of use (capacity) of the affected hand in tasks in which both hands need to be used. Task performance, especially in young children, can be influenced by visual spatial insight, praxis and cognitive aspects. The OSAS’ tasks were therefore chosen to be appropriate for the children’s ages and their intellectual abilities to prevent these factors from affecting task performance. They involve many repetitions of actions so that quality of use can be assessed reliably. Task execution is filmed, allowing blind assessment. The amount and quality of use of the affected hand are scored every second in order to make the OSAS more sensitive to subtle differences in these features than the AHA. In the AHA, the performance that is observed most frequently during the play session, or sometimes the best performance, is scored.

The OSAS is still under development. It is intended for use in clinical practice to support choice of treatment and for treatment evaluation as well as for research purposes. Because, as a first step in its evaluation, it is important to know its ability to give the same results in repeated measurements, we assessed intra-rater, inter-rater and test-retest agreement and reliability using the Guidelines for Reporting Reliability and Agreement Studies (GRRAS) [15].

OSAS scoring protocol and data processing

The quality of use of the affected hand is scored on an ordinal scale. Four domains of use were defined: reach, grasp (position thumb and fingers, and position wrist), hold (position thumb and fingers, and position wrist), and release. Each domain has 3 to 5 quality score levels ranging from poor to good quality. The same expert group that designed the tasks also formulated the quality criteria. These criteria are described in detail in Additional file 2.

The video recordings are analyzed with a user-dedicated software program (Figure 1) based on MATLAB (MathWorks inc). The frontal and cranial views are displayed next to each other. In the same window the quality criteria for scoring are shown as a pop-up menu. The video recordings are forwarded second-by-second. Every second both the quality of use of the affected hand and the use of the non-affected hand are scored according to the previously mentioned quality criteria. Scoring the video recording of one task takes 20 minutes for an experienced rater. To prevent scoring from taking too long, each task is limited to a maximum of 2.5 minutes. All participants were able to complete the tasks within this time. At least one of the quality criteria of the affected hand has to be scored every second unless the affected hand was not used. In that case, only the use of unaffected hand is scored. The amount of use of both hands during the task is expressed as a proportion of the total time needed to complete the task. The mean quality score of each domain has to be calculated separately for each task; a higher mean score corresponds to better quality of hand function.

We designed the OSAS to measure both the amount and quality of use of the affected hand in tasks in which repetitive bimanual use is demanded. It was developed primarily to measure treatment effect in research and clinical practice. In the present study, as a first evaluation, intra-rater, inter-rater and test retest reliability and agreement were determined using ICC, standard deviation of measurement differences and SDD [17, 18].

Because the ICC provides an index that relates to distinguishing patients within a specific group its importance for clinical evaluation of change is limited. However, based on the generally high ICC values in most tasks, it can be concluded that the OSAS has good discriminative capacity in patient groups resembling the study population. For the older children, an exception in this general pattern is the amount of use of both hands for which the variation is small in all tasks, which explains the low ICCs in test-retest reliability. The mean differences between the measurements for inter- and intra-rater reliability are generally small compared to the width of their scales, which indicates good agreement of measurements.

The largest SDD for the amount of use of both hands was 14.5% in the older age group and 30.8% for the younger children. As the SDD uses the same units as the original measurement, its interpretation for clinical use is straightforward. The high amount of use in the older children, in combination with large SDDs leads to a ceiling effect, rendering this measurement non-useful for follow-up. Because the OSAS demands the use of both hands to perform the tasks, this is not surprising. In the younger age group, in which movement patterns are not yet very stable, there was less use of both hands, more variation and large SDDs. This means that amount of use of the OSAS in younger children is not suited for evaluating individual changes but may still be used to compare groups in scientific research.

The ‘quality of reach’ has very low variation, which leads to low reliability in both age groups. The lack of variation may be explained by the fact that children with unilateral CP tend not to reach with their affected hand. If they do, it is only for a second and almost always with the same pattern. For the future it would be better to score the fact that the affected hand reached without an attached quality criterion.

The ICC of quality of use is good to excellent for ‘grasp wrist’, ‘hold fingers’, and ‘hold wrist’ in all children and all tasks. This is also the case for ‘quality of release’ in all tasks in the older age group and the threading beads task in the younger children. They do not yet show very consistent release patterns, which is especially obvious during the Pop-Onz and stacking blocks tasks. The ICC of ‘quality of grasp fingers’ is not good in the Pop-Onz task of the younger, and in the small and large construction tasks (only inter-rater reliability) of the older children. This means that better coaching of the observers in this criterion will be needed. Generally, the reliability for the older age group is better than for the younger children. This may be explained by the fact that older children show more consistent movement patterns. The SDDs of the quality items are generally small, but clearly larger in the stacking blocks task of the younger children. The largest SDDs are found in the ‘quality of grasp’ and ‘hold wrist’ mean score in the stacking blocks task. Therefore this task is not very useful to measure change. Apart from ‘reach’ the SDDs of the quality criteria for the OSAS tasks are low compared to the width of their scales which makes these criteria potentially useful for assessing change in patients.

The OSAS seems to be a useful addition to existing assessments of bimanual functioning for children with unilateral hand function problems, such as the AHA. The AHA measures actual spontaneous use of the affected hand in bimanual performance. With the OSAS, the amount and quality of use of the affected hand can be measured in a precise way, as a measure of capacity. The tasks are designed to force the child to use the affected hand repeatedly, are appropriate for the age group and do not interfere with visual spatial or praxis problems. In contrast to the MUUL, the OSAS measures the affected hand as an assisting hand in bimanual functioning. The simplicity and the short duration of the tasks make the OSAS easier to administer with young children. A disadvantage is that scoring takes longer, 20 minutes per task.

In the present study, 32 children between the age of 2.5 and 16 years were included in the intra- and inter-rater reliability analyses and 26 children in the test-retest reliability analysis. This number is limited. Moreover, children aged 7–11 years old were not included in the present study. However, part of the OSAS was developed for children aged 7–16 years old. Reliability data from this age group will need to be collected.

More agreement data is needed, with adapted scoring of the reach item. Only the frequency of reaching with the affected hand during the task can be scored. The stacking blocks task, which proved to be unreliable, might be removed. Precise coaching of observers is needed, especially for the assessment of ‘quality of grasp fingers’. The next evaluation step is to measure concurrent validity. In the children aged twelve years or younger this is possible with the AHA. Because the AHA is not available for the older age group yet, the Jebsen test [19], which measures speed of movement of the affected hand, could be used instead. Concurrent validity will also be determined with the achievement of treatment goals assessed by the Goal Attainment Scaling (GAS) [20] and performance scores of the COPM [21].

In conclusion, the OSAS appears to be a reliable assessment tool, with good agreement between repeated measurements, for measuring the quality of use of the affected assisting hand in forced bimanual task execution in CP children. Some modifications as mentioned above, may improve agreement, reliability and ease of scoring. More agreement and reliability data should be gathered, and the responsiveness of the scores also needs to be tested.