Background: There is a lack of objective and reliable tools to measure walking performance in children with cerebral palsy (CP).

Objective: To evaluate the reliability of inertial measurement units (IMUs) measuring daily life walking performance and physical activity (PA) in children with CP and healthy controls.

Methods: Algorithms were developed to analyse data collected with IMUs during 2 standard school days of the same week and 1 weekend day in 15 children with CP and 14 controls. Additionally, within a clinical trial, 10 children with CP were measured twice, on the same weekday 2 to 4 weeks apart. Relative and absolute reliabilities of PA (% time walking, standing, sitting/lying) and gait parameters (e.g., velocity, cadence) were evaluated by using the intraclass correlation coefficient (ICC) and minimal detectable change (MDC95), comparing 2 school days of the same week, a school day with a weekend day, and the same weekday 2 to 4 weeks apart.

Results: For the 15 children with CP (mean [SD] age 13.5 [3.4] years), ICCs were very high (0.70-0.98) when comparing gait parameters for 2 school days. ICCs were lower when comparing 2 school days for 14 control children (mean [SD] age 13.9 [3.0] years) and lowest when comparing a school day with a weekend day for both CP and control children. ICCs for PA were 0.90-0.91 when measuring the same weekday 2 to 4 weeks apart but were very low when comparing 2 school days of the same week or a school day with a weekend day. MDC95 values were high for both groups and all comparisons but comparable with findings of in-lab studies of similar parameters.

Conclusions: Our IMU and algorithm setup appears to be a reliable tool to measure daily life gait parameters in children with CP when repeatedly measured on 2 school days. PA was also reliably assessed but when measuring the same school day some weeks apart. However, the high MDC95 values question whether the setup can be used as a responsive outcome measure of interventions. (C) C 2019 Elsevier Masson SAS. All rights reserved.