Physical activity at discharge

During the discharge transition phase, patients return to familiar social and physical surroundings. Physical surroundings influence a person's PA,15 and as a transitional challenge, changes in PA may be expected after changes in the physical environment. PA has been measured before and after discharge for people with ABI.10^,14 However, no previous studies have explored patterns of PA specifically during discharge.

Setting

Consecutive recruitment occurred from September 2020 to September 2021 in parallel with usual care and treatment at a national Danish neurorehabilitation center.

Participants and recruitment

The inclusion criteria were age 18 years or older, ability to speak and understand Danish language, and absence of severe cognitive deficits that might limit the patients’ understanding of instructions, as assessed by a member of the patients’ rehabilitation team. Participants were required to be able to walk 10 meters independently to ensure valid classification by the applied algorithm.17 See flow chart in figure 1. To ensure minimal interference with daily rehabilitation interventions, a staff representative from the patients’ rehabilitation team scheduled a meeting with the primary investigator (PI) after having assessed patient eligibility. The data collection was planned to take place 1 week prior to discharge for all patients, but the discharge date had not always been established 1 week in advance. Data collection then took place as close to this time point as possible. After inclusion, the PI introduced the patients to the study and invited them to volunteer their participation. If agreed, a 30-45-minute survey interview was conducted on anxiety and depression, fatigue, pre-morbid PA level, and health-related quality of life, and participants were invited to wear an accelerometer.

Open in a separate window

Fig 1

Flowchart presenting the recruitment and participation of patients.

Measures and materials

Demographics on age and sex, functional level, and type of ABI were obtained from medical journals. Depression and fatigue have been found to be associated with activity performance and PA after ABI.18 These potentially influential factors were collected as self-reported assessments with interviewer-supported questionnaires. Level of functioning was assessed with the modified Rankin scale (mRs)12^,19; motor and cognitive levels with the functional independence measure (FIM). Level of functioning was rated on a 7-point scale from 1 = total assistance to 7 = complete independence for the 2 motor and cognitive subscale domains of 13 and 5 items, respectively. Scores ranged from 13 to 91 points for motor scores; from 5 to 35 points for cognitive scores.20 Self-reported pre-morbid PA level was classified with the Saltin-Grimsby Physical Activity Level Scale,21 fatigue with the multidimensional fatigue inventory (MFI).22 Anxiety and depression levels were assessed with the hospital anxiety and depression scale (HADS)23 and health-related quality of life with the EuroQol index value calculated from the EQ-5D-5L.24

Accelerometer measurements

A thigh-worn triaxial accelerometer (Axivity AX3, UK) was patched with adhesive bandage 10 cm above the apex patella, pointing downward. Patients were instructed to wear the accelerometer in all daily life activities, including bathing and sleeping, for 14 consecutive days to avoid non-wear time. Furthermore, patients were advised to contact the PI in case of any unforeseen events. After 14 days of wear time, patients were to return the accelerometer in a pre-labeled envelope. To ensure uniform time distributions, a day was defined to start at 6:00 AM. Any data from the first day prior to this time point were discarded. Likewise, data from the last day were defined to end at 6:00 AM. Data were stored using REDCap electronic data capture tools hosted at Aarhus University, Denmark25^,26 and exported to Stata15 (StataCorp LLC, College Station, USA) for statistics and graphs.

Physical activity classifications

Classification of activity was based on a machine learning algorithm developed and validated specifically for people with ABI by Honoré et al.17 The algorithm was updated to comprise a broad range of activities: standing, walking, running, cycling (riding a bicycle), stairs (ascending or descending), ambulation (wheelchair ambulation and car driving), and sedentary (sitting or lying down).27

To study sedentary versus active time, the categories: standing, walking, running, cycling, and stairs were combined to form an “active” category, whereas ambulation and sedentary were combined to form a “sedentary” category.

Data completion

Among the 45 patients who met the inclusion criteria, 10 (22%) participated and 8 (18%) supplied accelerometer data for the study. No participants dropped out after inclusion, but we completely discarded data from 1 patient who wore the accelerometer for less than 24 hours because the data were insufficient for analysis (participant 9). All data were lost from 1 accelerometer because of technical problems (participant 10). For the remaining 8 participants, data from 2 to 13 days (median 8 days) were analyzed, yielding data from a total of 66 days of which 8 were discharge days. For 31 days, data were collected prior to discharge; for 30 days, after discharge. Total active time for the eight participants was 21,502 min (22.6%) of total wear time with a daily mean (95% CI) of 349 (266-433) min per participant, equivalent to 5:49 (4:26-7:13) hours. Participants 1, 3, and 8 removed and returned the accelerometer before their discharge because of sensory discomfort.

Activity before and after discharge

Five participants (2, 4, 5, 6, and 7) wore the accelerometer in both hospital and home settings. For those participants, it was of particular interest to study the average daily PA time before and after discharge. Pre-discharge activity time ranged from 185 to 515 minutes and post-discharge activity from 246 to 660 minutes. Three participants (2, 4, 6) increased their average active time and 2 decreased their average activity,5^,17 making statistical testing for trends inappropriate. An overview of wear days and average active time is provided in table 2.

Exploring factors influencing physical activity

With data points showing total average active time per participant plotted against the FIM subscales for cognitive and motor dimensions in figure 2A and B, a reversed pattern was discovered. For the FIM cognitive score, a high functional level indicated a lower daily average active time; opposite, for the motor score, a higher activity level matched higher scores. However, multiple data points were outliers, especially for the motor dimension.

Open in a separate window

Fig 2

Average active time for each participant (represented by a dot) plotted against the cognitive (A) and motor (B) subscales of the FIM. The dotted line shows a linear regression of the investigated variable (x-axis) on average daily active time (y-axis).

For plots of average active time versus fatigue measured with the MDFI20 in figure 3A, B, D, and E, no systematic linear distribution between scores was observed along the fitted regression line. For the dimension “Reduced Activity” in figure 3C, there may be a uniform tendency between low scores (implicating less severe fatigue) and high activity levels, which indicated a connection between high levels of fatigue and low levels of PA.

Open in a separate window

Fig 3

Average active time for each participant plotted against the 5 dimensions of fatigue in the Multidimensional Fatigue Inventory. The dotted line shows a linear regression of the investigated variable (x-axis) on average daily active time (y-axis).

Two patients self-reported a high symptom level for depression; the depression dimension of the HADS score is shown in figure 4D. The plot showing active time against HADS depression revealed an individual performance of approximately 280 and 520 minutes for these patients. This indicates the absence of a tendency toward a uniform influence from depression on active time in the dataset. Health-related quality of life indicated a positive relation with PA in figure 4B, but an outlier made the trend ambiguous. Only premorbid PA levels reflected an even distribution along the fitted regression line between measurements (figure 4A).

Open in a separate window

Fig 4

Average active time for each participant plotted against the premorbid PA level (A), quality of life score from the EQ5D-5L (B) and anxiety (C) and depression sub scores (D) from the HADS. The dotted line shows a linear regression of the investigated variable (x-axis) on average daily active time (y-axis).

Limitations

Though accelerometer-based activity monitoring has the advantage of eliminating bias from any self-reported monitoring or use of observational methods31^,32 and may optimize patient outcome at a low cost,33 a decline rate of 78% should always trigger considerations of appropriate methodology and selection bias. The challenges related to obtaining accelerometer data were larger than expected. Decline rates from previous accelerometer studies were 4-25% for in-patients, and a full day of wear time was an indicator of a declination rate of 55% in a study of elderly rehabilitation patients.34 With a requested wear time of 14 days in the present study, participation would likely be declined by patients without the mental and physical surplus needed to engage in extra measures alongside their usual rehabilitation efforts.

Selection of patients toward individuals with a high functional level was substantiated by the fact that 80% of the participants were classified as employed. Furthermore, 80% were co-habiting with another adult, which may enable support toward health-promoting behavior such as increased PA owing to positive social norms.35 In all probability, this would leave a surplus of patients with a social and active lifestyle, which may bias the average activity time reported in the present study. The data that were lost from participants 9 and 10 were unrelated to the patients' functional level and probably did not inflict bias, but the data loss reduced the number of participants substantially, by 20%.