CONCURRENT VALIDITY OF THE ALBERTA INFANT MOTOR SCALE - SHORT VERSION COMPARED WITH THE ORIGINAL VERSION IN TERM AND PRETERM INFANTS

Y. Purwanto¹, Y.-J. Hsu², P.-N. Tsao³, T.-A. Yen³, W. Liao⁴, W.-J Chen⁵, S.-F. Jeng¹

¹National Taiwan University, School and Graduate Institute of Physical Therapy, Taipei, Taiwan, ²National Taiwan University, Department of Computer Science & Information Engineering, Taipei, Taiwan, ³National Taiwan University Hospital, Department of Pediatrics, Taipei, Taiwan, ⁴National Taiwan University Hospital, Department of Internal Medicine, Taipei, Taiwan, ⁵National Taiwan University, Graduate Institute of Epidemiology and Preventive Medicine, Taipei, Taiwan

Background: Preterm infants are at a higher risk of neuromotor disorder than their term counterparts. The Alberta Infant Motor Scale (AIMS) is a standardized assessment tool widely used for motor assessment in infants from birth to 18 months of age. Our research team has recently applied artificial intelligence with machine learning to identify 31 out of the 58 movement items from the AIMS with sufficient accuracy against the scoring results of three physical therapists. Before these 31 movement items can be used for motor assessment in infants who are at risk of motor disorder, their concurrent validity compared with the results of the original AIMS requires investigation.

Purpose: To examine the concurrent validity of the AIMS- Short version (AIMS-S) with 31 items for motor assessment in term and preterm infants when compared with the results of the AIMS-Long version (AIMS-L) with 58 items.

Methods: The study enrolled 44 term infants and 56 preterm infants who were prospectively administered the AIMS-L assessment during 4 to 18 months of age (corrected for prematurity). The AIMS-S consisted 9 prone items, 3 supine items, 7 sitting items, and 12 standing items. These items were designated at monthly basis within 4 to 18 months according to the scoring criteria of the AIMS-L with the items of 90% credited in normal infants and the standards of Denver Development Screening- 2^nd Edition and Taipei Development Screening-II. The number of essential items for 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, and 15 months were: 3, 4, 3, 2, 4, 4, 4, 4, 2, 3, 2 and 2 items, respectively. In the AIMS-L, motor delay was defined when the infant’s percentile was <10^th centile %; In the AIMS-S, motor delay was defined when an infant showed absence of one or more essential items at the assessment age. Validity indices included percentage of agreement, sensitivity, specificity, positive predictive value and negative predictive value using the AIMS-L results as the standards.

Results: The 100 infants were administered 235 AIMS-L assessments at 4 to 18 months of age that consisted 55 trials at 4 months, 4 trials at 5 months, 49 trials at 6 months, 6 trials at 7 months, 42 trials at 8 months, 2 trials at 9 months, 26 trials at 10 months, 3 trials at 11 months, 25 trials at 12 months, 2 trials at 13 months, 11 trials at 14 months, 9 trials at 15 months, and 1 trial at 18 months. Overall, the AIMS-S results demonstrated acceptable levels of concurrent validity compared with the AIMS-L results with a high level of percentage of agreement (0.96), sensitivity (0.87), specificity (0.96), and negative predictive value (0.99), and moderate positive predictive value (0.62).

Conclusions: The AIMS-S showed acceptable concurrent validity compared with the results of the original version in term and preterm infants.

Implications: The findings provide important basis for further validation of these movement items for motor assessment in a large sample of infants via machine learning.

Funding acknowledgements: This study was funded by National Science and Technology Council (MOST 110-2314-B-002-055-MY3)

Keywords:
Artificial intelligence
Infant motor development
Motor assessment

Topics:
Paediatrics
Innovative technology: information management, big data and artificial intelligence

Did this work require ethics approval? Yes

Institution: The National Taiwan University Hospital

Committee: Research Ethics

Ethics number: 202012089INB and 202010031RINB