Populations
Following approval by the Research Ethics Board of the University of Calgary (REB15-1742), two populations of participants were recruited for a case–control study from June 2016 to February 2017. Children with stroke aged 1 to 18 years with Magnetic Resonance Imaging-confirmed ischemic perinatal stroke (arterial or venous) were identified from a population-based research cohort (Alberta Perinatal Stroke Project) [29]. Additional criteria included unilateral stroke and no additional neurological conditions, severe developmental delays, or unstable epilepsy.
Typically developing participants aged 1 to 18 years were recruited from the Healthy Infants and Children Clinical Research Program (HICCUP, www.hiccupkids.ca). Typically developing participants had no neurological conditions or medications. Selection bias was minimized by recruiting comparable ages between stroke and control groups. All participants provided written informed consent/assent.
Actigraphy
Following recruitment, participants were fitted with actiwatches (Actiwatch2, Philips Respironics, Pennsylvania) on both wrists at the Alberta Children’s Hospital. Watches recorded movement data for a period of 48 h during routine life. Participants (or parents) were asked to record sleep and wake times on a standardized diary and press the event marker button at bedtime and wake-time [30]. Motion data was captured in 15-s epochs. Actiware software (Philips Respironics) generated an Activity Count for each epoch. Times were manually excluded if the participant reported in the diary removal of the watches, such as bathing, as the watches were not waterproof. Rest intervals were generated from participants pressing the watch marker button at bedtime and wake-time and were validated using reported diary sleep and wake times. Sleep intervals were automatically determined by the Actiware sleep interval detection algorithm and could only occur within a rest interval.
Data was segmented according to intervals and/or activity levels. Intervals were segmented into [a] active, [b] rest, [c] sleep, and [d] all. Activity levels were defined for each 15-s epoch by the sum of the Activity Counts for both hands: [a] very low (total of 0–30), [b] low (31–160), [c] moderate (161–524), [d] high (525–812), [e] very high (813+), and [f] all. Levels chosen matched those previously described [19] with the exception of two changes: an additional range of “very low” was added to expand analysis of low level movement, and all range limits were doubled to account for bimanual movement. Segmentation by both interval and activity level separated the data into 24 subgroups. It was therefore possible to have activity of different levels during all intervals including during sleep.
Statistics
A moment-by-moment asymmetry measure, which we label the Scaled activity difference (SAD), was calculated:
$$Scaled\,activity\,difference \left(SAD\right)= \frac{{L}_{it}- {R}_{it}}{{L}_{it}+{R}_{it}}$$
where Lit and Rit are the AC for the left and right hands in epoch of time t for individual i. SAD was zero when there was no activity in either hand. Thus, SAD values ranged from − 1 to 1 with − 1 and 1 indicating only right and left arm movement, respectively, and values of 0 indicating equal left and right arm use. A SAD value was generated for each 15-s epoch for each participant. The SAD is similar to the “Asymmetry Index” used by Beani et al. and elsewhere, but assesses asymmetry at a given epoch, rather than as an average value [23]. For each individual, the SAD statistics were graphed to create a visual representation of the SAD score distribution for that participant, after being ranked from smallest to largest (Fig. 1).
We also constructed a novel summary statistic called the Actigraphic Movement Asymmetry Index (AMAI) as:
$$AMAI = 1- \left|{\left(mean SAD\right)}_{ijk}\right|$$
for interval j and level k for individual i. The AMAI generated a single value to represent bilateral movement asymmetry in each participant for each interval j and level k. Values range from 0 to 1 with closer to 1 indicating greater symmetry, a value of 1 indicating perfect symmetry, and a value of 0 indicating completely unilateral movement (fully asymmetric). A key feature of the AMAI is that it gives equal weight to all epochs regardless of intensity of upper limb activity. This accords value to both large and small movements, which are each important in everyday life. The AMAI was the primary outcome.
Standard motor outcomes
Standardized clinical motor outcomes, the AHA, MA and BB, were obtained by experienced pediatric occupational therapists within the context of a clinical trial [9, 10]. Measuring therapists were blinded to all patient details including stroke type, size and location at the time of assessment. Only the summary score for each test was used in the current analysis. Most BB scores were obtained during the trial measures; seven additional participants who were otherwise eligible but lacked recent AHA and MA scores received BB assessments from a trained research assistant. We also calculated the “Block Ratio” statistic, as described by Raglio [31] to provide a single value for BB scores and enable comparison with the AMAI:
$$Block\,Ratio= \frac{\#\,of\,blocks \,moved \,in \,1\, min\, by\, affected\, hand}{\# \,of\, blocks\, moved \,in \,1 \,min\, by \,unaffected \,hand}$$
Statistical analysis
Associations were tested using Pearson correlation if data was normally distributed (Shapiro–Wilk test) and Spearman correlation if it was not. First, we examined correlations between different AMAI (actigraphy) intervals and levels. Second, we examined correlations between standard outcomes (AHA, MA, BB). Third, we tested for correlations between AMAI and the AHA, MA, and BB for all combinations of activity intervals (active, rest, sleep, all) and levels (very low, low, moderate, high, very high, all) for stroke participants only. No adjustments were made for participants without AHA or MA scores. Scatterplots provided visual representations of selected relationships between variables.
Box plots of the AMAI compared differences between groups across activity intervals. Welch’s t-test (unequal variances t-test) and Mann–Whitney U tests were applied depending on normality to compare groups. The tests were performed to systematically compare AMAI between participants with stroke and typically-developing participants. As our primary outcome (AMAI) has not been previously described, a formal power calculation was not possible. However, based on similar approaches in the literature, a minimum sample of 25 stroke participants was estimated. Analysis was performed using Stata (version 14.2).