Background/Purpose
Pedometers are a common instrument used to measure walking activity, yet there remains little evidence on the measurement error associated with this device for disability populations, particularly Down syndrome (DS). Extrinsic sources of error including different pedometer models, walking speed and pedometer tilt have been identified in the general population (Crouter et al., 2005), but there may also be systematic sources of measurement error. These sources of error may also be unique to certain groups of individuals. The purpose of this study was to utilize Generalizability theory to examine systematic and random sources of error in pedometer measurements during free-walking for adults with and without DS.
Method
Participants included 18 adults with DS (10 female, 8 male) and 23 adults (13 female, 10 male) without a disability. Participants wore a total of 12 pedometers including 2 piezoelectric (Omron HJ-112) and 2 spring-levered (Yamax Digiwalker SW-200) units on each hip and 2 piezoelectric units in each front pocket. The number of steps taken was measured during a 20 minute continuous free-walking trial.
Analysis/Results
Generalizability theory is a psychometric theory that not only estimates overall reliability but also evaluates possible sources of error. A three-facet (model, placement, and unit) fully crossed G-study design revealed that for adults with DS, participants accounted for 55% of the variance. Other substantive sources of variance include the interaction between participants and models (12%), the interaction between participants and placements (6%), and the interaction of participants, placements, and units (6%). For adults without DS, participants accounted for 49% of the variance. Additional sources of variance include the “participant-placement-unit” interaction (13%), the “participant-model-unit” interaction (10%), and the “participant-placement” interaction (5%). In both analyses, 19% of the variance was unexplained. Indexes of reliability were calculated for both dependability (Φ) and generalizability (G) coefficients. The reliability coefficients for adults with DS (Φ=0.84, G=0.86) and without DS (Φ=0.87, G=0.95) denote high levels of reliability.
Conclusions
This study suggests there are systematic sources of variance in pedometer measurement and that these sources are different between adults with and without DS. The largest source of variance was from participants, yet half of the variance was attributed to other facets and interactions. Additionally, 19% of the variance was unexplained which signifies a substantial amount of random error. The high levels of reliability indicate that pedometers can consistently measure walking activity, but efforts should be taken to minimize these systematic sources of variance when used in future studies.