Gross motor performance of infants with an at-home wearable measurement and the Alberta Infant Motor Scale: A concurrent validity study.
To evaluate the concurrent validity of assessing infants' gross motor performance with an at-home wearable measurement versus the Alberta Infant Motor Scale (AIMS).
This observational study used a new multi-sensor wearable Motor Assessment of Infants with a JUmpsuit (MAIJU) for 67 at-home measurements of 42 infants (60% males; mean age 11.5 months, SD 3.7 months, range 4-18 months). The study collates a normative cohort of typically developing volunteers (n = 17, 65% males; 27 measurements, mean age 11.8 months, SD 3.1 months, range 7-18 months) and a clinical cohort recruited from a neurodevelopmental follow-up clinic (n = 25, 56% males; mean age 11.2 months, SD 3.8 months, range 4-17 months). We correlated the expert-assessed total AIMS score to a holistic assessment of motor development (the BABA Infant Motor Score [BIMS]) from the MAIJU measurements. Additionally, detailed MAIJU-derived metrics were compared to the AIMS scores, to train machine learning-based AIMS predictions.
There was a very strong correlation (Spearman's rank correlation coefficient [rho] = 0.97, p < 10-40) between BIMS and AIMS. Centile-based detection of low-performing individuals (cut-offs of 5% and 10%) showed high agreement between AIMS and BIMS (Cohen's kappa = 0.81). Performance of a machine learning-based AIMS prediction from the wearable data was comparable (rho = 0.96, p < 10-37) to direct use of the BIMS score.
Assessing motor performance with scalable, at-home wearable measurements is highly comparable to the widely used AIMS assessment. An objective, quantitative, and expert-independent motor assessment holds promise for providing benchmarks in geographically distributed health care or clinical trials.
This observational study used a new multi-sensor wearable Motor Assessment of Infants with a JUmpsuit (MAIJU) for 67 at-home measurements of 42 infants (60% males; mean age 11.5 months, SD 3.7 months, range 4-18 months). The study collates a normative cohort of typically developing volunteers (n = 17, 65% males; 27 measurements, mean age 11.8 months, SD 3.1 months, range 7-18 months) and a clinical cohort recruited from a neurodevelopmental follow-up clinic (n = 25, 56% males; mean age 11.2 months, SD 3.8 months, range 4-17 months). We correlated the expert-assessed total AIMS score to a holistic assessment of motor development (the BABA Infant Motor Score [BIMS]) from the MAIJU measurements. Additionally, detailed MAIJU-derived metrics were compared to the AIMS scores, to train machine learning-based AIMS predictions.
There was a very strong correlation (Spearman's rank correlation coefficient [rho] = 0.97, p < 10-40) between BIMS and AIMS. Centile-based detection of low-performing individuals (cut-offs of 5% and 10%) showed high agreement between AIMS and BIMS (Cohen's kappa = 0.81). Performance of a machine learning-based AIMS prediction from the wearable data was comparable (rho = 0.96, p < 10-37) to direct use of the BIMS score.
Assessing motor performance with scalable, at-home wearable measurements is highly comparable to the widely used AIMS assessment. An objective, quantitative, and expert-independent motor assessment holds promise for providing benchmarks in geographically distributed health care or clinical trials.
Authors
Airaksinen Airaksinen, Palsa Palsa, Hautala Hautala, Boonzaaijer Boonzaaijer, Haataja Haataja, Vanhatalo Vanhatalo
View on Pubmed