Design
This study used a prospective observation design, measuring physical activity continuously with two activity monitors for three to seven consecutive days during hospitalization.
Study setting
All patients were treated in an evidence-based comprehensive stroke unit that emphasizes a multidisciplinary approach and early rehabilitation. The treatment focused on independence in daily life. All patients, regardless of participation, received routine medical treatment, including rehabilitation, in accordance with Norwegian guidelines for treatment and rehabilitation after stroke [27].
Study population
Patients were recruited during three time periods, October–December 2013, May–December 2014, and March–August 2016. Patients admitted to Trondheim University Hospital, Norway, with diagnosed first ever or recurrent acute ischemic or hemorrhagic stroke were eligible for inclusion if the onset of stroke had been within seven days, they spoke fluent Norwegian, and if they gave written consent. In line with the Norwegian regulations for informed consent, patients who were unable to consent were included if their next of kin not opposed participation. Patients with terminal illness, other health conditions severely affecting their ability to walk, or expected discharge within three days after inclusion were excluded.
Measurements
The ActivPAL Professional sensor system (from PAL Technologies Ltd) consists of a three-axis accelerometer that collects continuous data with a sampling frequency of 10 Hz, and with a battery capacity of up to 14 days. The primary outcome for this study was whether time spent lying, sitting or upright changed during the hospital stay. It has previously been shown that this activity monitor provides valid data for time spent lying, sitting, standing and walking [28, 29] and that placement on the thigh and sternum provides good validity for postures and transitions compared to video observations [26]. The 24-h period was measured from the time the activity monitors were attached. Duration of sitting bouts was estimated according to the following ratio: time spent in sitting/number of transitions from lying to sitting, while duration of upright bouts was estimated correspondingly. A time threshold for transitions was set at 1.5 s to eliminate unreliable event records.
The Short Physical Performance Battery (SPPB) (ranging from 0 to 12, where 12 is the best score) was used to assess physical function. The test consists of three different mobility tasks, and has been found to be both valid and reliable for assessing physical function amongst elderly people [30]. Gait speed was calculated based on the walking task of the SPPB. An experienced nurse or physiotherapist performed the test.
Global function was defined using the Modified Rankin Scale (mRS) (ranging from 0 to 6, where 0 is normal function and 6 denotes death) [31] at admission and after the first day in hospital. The mRS measures independence in activities of daily living.
The National Institutes of Health Stroke Scale (NIHSS) was used to measure severity of the stroke. The scale is widely used and has proved both valid and reliable [32].
Both mRS and NIHSS were scored by an experienced clinician within the first day after the patient’s admission and recorded together with age, gender, number of days from first symptom to admission date, and number of days at the hospital, which were collected from the medical record.
Procedure
All patients accepting participation had two ActivPAL activity monitors attached, one at the sternum and one at the unaffected thigh, to distinguish between lying, sitting and upright position. If neither of the lower extremities were affected by the stroke, the activity monitor was attached to the right thigh. Patients were instructed to follow the standard rehabilitation routines without paying attention to the equipment. Both activity monitors were removed 14 days post stroke unless the patient had been discharged earlier. Data from the activity monitors were transferred via a USB docking station, and processed in Matlab R2015 B. The valid body and leg data had an individual start and stop time defined and used to measure a count for a full 24-h day, and the activity was coded in reference to the body position. To distinguish the different positions, data from the two sensors were synchronized and recoded according to body position (lying, sitting, and upright).
State durations were calculated by accumulating time intervals for lying, sitting, and upright and state transitions between these positions per 24-h measurement. Time upright was calculated by merging time standing and time walking. A validation procedure was performed to control for possible time drifting of the sensors.
Statistical analysis
Demographic data were reported as mean values and standard deviation (SD) for all patients, and for three subgroups categorized by time from onset of symptoms to inclusion. Baseline characteristics were compared for those included within the first two days after stroke, those included 3 to 4 days after stroke and those included 5 to 7 days after stroke. This was done using a one-way ANOVA or Kruskal-Wallis test, depending on normality of residuals, which was judged by visual inspection of Q-Q plots. To assess changes in activity levels, we used linear mixed models (LMM) with time sitting, time upright, duration of sitting bouts and duration of upright bouts, respectively, as dependent variables, patients as random factor, and days since stroke as covariate. Next, we included the following covariates, one at a time: NIHSS, age, gender, mRS pre-stroke, mRS one day post stroke, and SPPB. Because the time spent lying, sitting and upright totaled 24 h, only time spent sitting and upright were used in the mixed model analysis. Two-sided p-values less than 0.05 were considered statistically significant, and 95% confidence intervals (CI) are reported where relevant. Statistical analyses were done in SPSS 23.