DOS Kongressen 2014 ·
217
Validation and inter-tester reliability of a tri-axial
accelerometer-based classification of daily
activities
Marianne Tjur, Kenneth Juul Laugesen, Kamilla Nygaard Jensen,
Mads Grosmann Svendsen, Inger Mechlenburg, Signe Kierkegaard
Orthopaedic Research, Aarhus University hospital; Bachelor of Applied Science
in Physiotherapy, VIA University College Aarhus
Background:
Daily physical activities (PA) are often used as outcome in the
evaluation of orthopaedic or rehabilitation interventions. Self-reported PA suf-
fers from recall bias and imprecision, thus there is a need for objective mea-
sures.
Purpose / Aim of Study:
The aim was to investigate the inter-tester reliability
and precision of the classification of PA measured with a tri-axial accelerometer,
and a MatLab-based algorithm.
Materials and Methods:
27 healthy participants were observed with video
recordings while performing 30 min of various time intervals of the activi-
ties; resting, standing, walking, cycling and running. A tri-axial accelerometer
(GCDataconcepts, US) was mounted on lateral side of right thigh while per-
forming the activities. Two blinded analysers performed the classification
of activities using the algorithm. The relative reliability was expressed with
Intraclass Correlation Coefficient (2,1)(ICC), the absolute reliability with Limits
of Agreement (LOA) and precision with LOA of the proportion of analysed activ-
ity compared to observed.
Findings / Results:
ICC was high for resting 0.99, standing 0.99, walking 0.88
and running 0.99 and acceptable for biking 0.73. LOAwere acceptable for resting
[-0.34;0.42], standing [-0.98;1.28], walking [-449;368], running [-0.39;0.32]
and biking [-368;448]. The precision of classifying PA was acceptable with LOA
[0.99:1.02] for resting, [0.94:1.10] for standing and [0.98:1.07] for biking.
Walking was underestimated (mean= 0.71 (p=0.03), LOA [0.19:2.60]) while
running was overestimated (mean= 1.39 (p=0.001), LOA [0.64:2.98]).
Conclusions:
Inter-tester reliability was high while precision of classification of
resting, standing and cycling was acceptable. More precision is necessary in the
classification of walking and running. The perspectives for monitoring daily PA
with accelerometer seem promising.
164.
