Evaluation of Physical Activity in Patients with Chronic Obstructive Pulmonary Disease Using an Accelerometer with Tapestry-Style Display Capability

*1)Division of Respiratory Medicine, Juntendo University Faculty of Medicine, Tokyo, Japan, *2)Department of Respiratory Medicine, Juntendo University Graduate School of Medicine, Tokyo, Japan, *3) Department of Respiratory Medicine, Saiseikai Kawaguchi General Hospital, Saitama, Japan, *4)Divisions of Neurology, Juntendo University Graduate School of Medicine, Tokyo, Japan, *5)Research Institute for Disease of Old Age, Juntendo University Graduate School of Medicine, Tokyo, Japan


Introduction
Physical activity (PA) is an important parameter in disease management for patients with chronic obstructive pulmonary disease (COPD).Reduced PA is related to COPD exacerbation, hospitalization, as well as mortality 1)-3) .For these reasons, the international guideline, the Global Initiative for Obstructive Lung Disease (GOLD) emphasizes the enhancement or maintenance of PA as an objective for COPD management (GOLD Reports 2017 http: //goldcopd.org/gold-reports/) and the importance of objective evaluation of PA.
Low PA levels are observed even in COPD cases at stage I or II 4) 5) , whereas many patients with dyspnea regard their severity as mild 6) .
Accordingly, it is difficult to predict the extent of a patientʼ s PA based on respiratory function and subjective symptoms alone.
Existing methods for PA evaluation include selfreport questionnaire, step-count calculations, metabolic monitoring, and multiaxial accelerometers, but no standard method has been established, so far 7) .During the recent years, there were several reports describing the utility of accelerometers for PA evaluation.PA levels and daily step count measured by accelerometer are more useful as predictors of mortality in COPD patients than St. Georgeʼs Respiratory Questionnaire (SGRQ) activity score 1) .Accelerometers with triaxial motion sensors have greater accuracy than mono-or biaxial equivalents 7) .In this study, we investigated the utility of a newly developed triaxial accelerometer, the HT-PB3 (HITACHI; Tokyo, Japan), for evaluation of PA in COPD patients.

Materials and Methods
Study subjects were 10 COPD patients and four age-matched individuals without COPD between June 2011 and April 2014.
HT-PB3 accelerometers were used for PA evaluation (Figure -1).The HT-PB3 triaxial accelerometer is a light (20 g) wristwatch-type device that can hold data from continuous recording periods of at least three months.It is also capable of measuring other indicators of activity such as step count, exercise intensity, exercise frequency, and quality of sleep.The HT-PB3 displays daily PA in a tapestry-style format, and this readily intelligible visualization of data is one of its main features.
Study subjects wore the HT-PB3 on the wrist of the non-dominant arm, and the device continuously recorded PA.The study subjects also made a record of their actual daily physical activity during the measurement period.The COPD patients in the study completed SGRQ and COPD assessment test (CAT) questionnaires, and their respiratory function was examined.We obtained tapestry displays corresponding to the HT-PB3 motion sensorcaptured activity levels and exercise frequency.Other data obtained included daily step count and exercise time distribution by activity level.
HT-PB3-generated tapestry displays were recorded and compared with actual self-report records of physical activity, to ascertain whether differences in PA level were visually detectable.We also investigated associations between motion sensor-derived parameters (step count and time distribution by activity level) with SGRQ and CAT scores.Mean values were used in motion sensorderived parameter analysis for all measurement days.This study was approved by the local ethics committee of Juntendo University (Approval No. 26-535).
SPSS(IBM, New York, USA) version19 was used for statistical analyses.The Kruskal Wallis H test was used for three-way comparisons by COPD stage and patient baseline characteristics.Spear-manʼ s rank correlation coefficient was used for evaluation of two-way associations.Two-tailed p-values below 0.05 were regarded as statistically significant.Data are expressed as mean ± SD and 95% CI unless otherwise stated.

Patient characteristics (Table-1)
The 10 COPD patients had a median age of 74 years, the same as control subjects.The numbers of patients with GOLD stages I, II, III, and IV were 2, 3, 3, and 2, respectively.Comorbidities were hypertension in five patients, diabetes mellitus in four patients, and heart disease in three patients.None of the patients had complicating bronchial asthma.An assessment of comorbidities was made with the Charlson Co-morbidity Index (CCI), which revealed one patient with a CCI score of 1, five patients with CCI scores of 2, and two patients each with CCI scores of 3 and 4. Eight patients were receiving concomitant treatment with a long-acting muscarinic antagonist and a long-acting β2-agonist, and two patients were receiving single-agent therapy with a long-acting muscarinic antagonist.
In assessments of quality-of-life (QOL) questionnaires, the analysis set for SGRQ total score and  The distribution of time by PA level can also be graphically represented, with time distribution (hr) on the vertical axis and date on the horizontal axis.In this graphical display, blue represents the time spent at less than 1.1 METs, green represents the time spent at 1.1 to 1.5 METs, yellow represents the time spent at 1.5 to 3.0 METs, orange represents the time spent at 3.0 to 6.0 METs, and red represents the time spent at more than 6.0 METs.
Actual tapestries for control subjects and COPD patients at different GOLD stages and with different activity levels are presented below (Subject No. 1 to 5, Figure -2). 1) Subject No. 1 (a 71-year-old man, control subject) Tracking recorded daily movements against tapestry display and revealed the following pattern: The indicated color changed from blue (representing sleep) to yellow after waking, and yellow-toorange coloration was frequently indicated after the start of work (a cleaning job).Light blue coloration was indicated at lunch and rest times, and the indicated color changed to yellow by the time of returning home and to blue after going to bed for the night.Changes in PA were clearly captured in the daily activity tapestry display.This subject was a recognizably highly active individual with conspicuous areas of yellow-to-orange and red coloration corresponding to his well-regulated activities.
2) Subject No. 2 (a 78-year-old man, GOLD I, CAT 12 points) This patientʼs CAT score suggested that COPD was having little effect on his daily life.Yellowto-orange coloration was indicated in the tapestry display for time bands when he was exercising and playing with his dog in mornings and evenings, and we found that his PA was relatively well maintained.The yellow-to-orange coloration was indicated at a comparatively similar time each day, and we found that dog walking and looking after were his daily routine.
3) Subject No. 3 (a 72-year-old woman, GOLD I, CAT 24 points) The patientʼs CAT score suggested that COPD was having a large effect on her daily life, even though the disease was at GOLD Stage I. Her tapestry display indicated a small proportion of yellow-to-orange coloration (indicating an activity level above light work), and she was considered to show reduced PA level.We also found a variance in her waking and sleeping times.In addition to COPD, this patient had comorbid depression and breast cancer, and these were considered as factors influencing the decrease in PA. 4) Subject No. 4 (a 67-year-old man, GOLD III, CAT 16 points) This patientʼ s CAT score suggested relatively well-maintained activity, although he had a high degree of airflow obstruction.His tapestry display actually indicated a large proportion of yellow-toorange coloration.5) Subject No. 5 (a 70-year-old man, GOLD III, CAT 26 points) This patientʼs COPD was at GOLD stage III and his CAT score was high (score of 26).Furthermore, he worked as a car driver and thus spent a lot of time in a sedentary position.Therefore, he had low level of day-time activity.Areas of green coloration, indicating a sedentary position, were also widespread on his tapestry display.
The above results showed a large variation in PA due to individual lifestyles and comorbidities, even at similar levels of airflow obstruction.PAs were visually and chronologically discernible in the tapestry display of the newly developed HT-PB3

Correlations between PA and each parameter
We compared the total step count and durations of activity (time distribution by METs) between control subjects, COPD patients at GOLD Stage I or II, and COPD patients at GOLD Stage III or IV, and found no statistically significant differences between these three groups (total step count, p = 0.445; durations of activity: less than 1.1 METs, p = 0.893; 1.1-1.5 METs, p = 0.329; 1.5-3.0METs, p = 0.398; 3.0-6.0METs, p = 0.552; and more than 6.0 METs; p = 0.069).
The total step count was significantly negatively correlated with the total SGRQ score (p = 0.002, rs = 0.88).Analysis based on SGRQ components showed that the total step count was negatively correlated with activity and impacts scores (p = 0.036, rs = 0.70, and p = 0.004, rs = 0.82, respectively), but was not correlated with the symptoms score (p = 0.128, rs = 0.52).
Besides, duration of light activity (exercise time distributed to 3.0 to 6.0 METs) was negatively correlated with the total SGRQ score (p = 0.003, rs = 0.86).An evaluation by SGRQ components showed that duration of light activity was negatively correlated with symptoms, activity, and impacts scores (p = 0.044, rs = 0.64; p = 0.008, rs = 0.81;

Discussion
Many motion sensors are currently commercially available.The Actimaker ® (Panasonic; Osaka, Japan) and the Welsupport ® (Nipro; Osaka, Japan) are triaxial accelerometers like the HT-PB3; how-ever, both of these devices are waist-attachable, which the wearer may feel uncomfortable compared with wrist-attachable ones.In contrast to those devices, the Actiwatch 2 ® (Philips Respironics Inc., USA) and the Lifecorder ® GS4 (Suzuken Co; Aichi, Japan) are used as monoaxial accelerometers.The Actiwatch 2 is worn on the wrist and the Lifecorder ® GS4 attaches to the waist, and these devices differ in the number of measurable axes and location of attachment.The triaxial HT-PB3 accelerometer used in this study can be worn on the wrist in the same manner as a watch.Therefore, wearing of the HT-PB3 does not burden the subjects.Noteworthy characteristics of the HT-PB3 include its capacity for continuous two-week operation after charging and in-device storage of data collected over approximately three months.
The HT-PB3 can also display recorded motions in a colored tapestry format.This major feature of the HT-PB3 is not shared with other accelerometers.In this study, we compared patientsʼ stated records of actual daily life with the recording-based color tapestry display generated by the HT-PB3.We found that changes in the tapestry display reflected those in the stated record of actual daily life, and we considered that the HT-PB3 can recognize daily PA in COPD patients, many of whom have low PA levels.The number of subjects in the present study was too small, and further validation study is needed; however, similar to previous reports 8) -10) , this accelerometer has potential utility for evaluation of PA in COPD patients.
PA is naturally influenced by the other factors other than COPD as each lifestyle is different even in patients with COPD at the same stage.PA may be reduced when the disease is exacerbated, and this can be recognized visually from recorded tapestry displays of day-to-day PA.The color tapestry display has a bigger visual impact, and allows patients, who may not have specialist knowledge of their condition, to easily recognize the state of their current PA.Accordingly, the HT-PB3 could be used as an aid in patient education aiming at PA improvement.Moreover, HT-PB3 could be useful for evaluating efficacy of pharmacological or physical therapy through the capacity for pre-and post-intervention.
Minakata Y, et al reported that duration of activity (exercise time distribution by METs) was significantly shorter in COPD patients than individuals without COPD 8) .The duration of activity (by METs) was also negatively correlated with respiratory function parameters, such as FVC %predicted, FEV1 %predicted, and DLco/VA %predicted, the Age-Dyspnea-Obstruction (ADO) index, and modified Body mass-Obstruction-Dyspnea-Exercise (BODE) index 8) .In another reported study, significant reduction in step count was noted with progression in COPD disease stage 11) .However, in this study, step count and duration of activity (exercise time distributed by METs) did not significantly differ between control subjects and COPD patients or among different stages of COPD.These results may be explained by the small number of subjects and the large, lifestyle-driven variation in individual activity levels of patients.Further research with a larger number of study subjects is required.Weather, season, and measurement time are also suggested to be other factors that can influence physical activity level.For example, people may not go outdoors when it is rainy or the temperature is low or high.Duration of activity was significantly lower on rainy than non-rainy days 9) .A survey of the London COPD cohort showed that a proportion of patients going outdoors was decreased as daily mean temperatures fell in the range of 2.5 to 20.5 (low to high values).A decrease in going outdoors was also reported with temperatures below 2.5 and those above 20.5 12) .Potential differences in PA between working and non-working days have also been identified for members of the labor force 7) .In the present study, PA was monitored for all time bands (except bathing time) without any restrictions set for weather or day of the week in order to achieve continuous recording through a fixed period.We cannot exclude this configuration as a factor in PA variation.We suggest that attention needs to be paid to factors including weather, measurement stage, and measurement time in future analyses.
We found the total step count correlated with SGRQ total score, activity score, impact score, and CAT score.Only symptom score showed no correlation with the total step count.This finding suggested that evaluation of PA based on subjective symptoms in COPD patients was not reliable.
In the current study, strong correlation with CAT and SGRQ total score was indicated, same as previously 13) .Although SGRQ scores are frequently used in clinical studies, their implementation in actual clinical practice is problematic, and the use of the simpler CAT is accordingly widespread in actual clinical setting.CAT score was also suggested to have potential utility for PA evaluation, based on the results of the present study.
The HT-PB3 triaxial accelerometer can record PA data over a prolonged period and display them in tapestry format, which has huge visual impact for both clinicians and patients.We conclude that the HT-PB3 has potential utility for evaluation of PA in COPD patients.

Figure- 1
Figure-1 The triaxial accelerometer (HT-PB3)The HT-PB3 is a light wristwatch-type device that can hold data from continuous recording periods of at least three months.It is capable of measuring indicators of activity such as step count, exercise intensity, exercise frequency, and quality of sleep.The HT-PB3 displays daily PA in a tapestry-style format (Source: Reproduced from the instruction manual for the HT-PB3).

Figure- 2
Figure-2 Tapestry display of physical activityThe tapestries for control subjects and COPD patients at different GOLD stages and with different CAT scores were presented.In the graphs showing the distribution of time by PA level, blue represents the time spent at less than 1.1 METs, green represents the time spent at 1.1 to 1.5 METs, yellow represents the time spent at 1.5 to 3.0 METs, orange represents the time spent at 3.0 to 6.0 METs, and red represents the time spent at more than 6.0 METs.The results indicated a large variation in PA due to individual lifestyles and comorbidities.

Figure- 3
Figure-3 Correlations between the total step count and QOL questionnaire scoreThe total step count was negatively correlated with SGRQ total, activity and impacts scores.The total step count was also negatively correlated with the CAT score.

Figure- 4
Figure-4 Correlations between the duration of light activity distributed to 3.0 to 6.0 METs and QOL questionnaires scoreThe duration of activity distributed to 3.0 to 6.0 METs was negatively correlated with SGRQ total scores and three SGRQ components.The duration of activity distributed 3.0 to 6.0 METs was also negatively correlated with the CAT score.

Table - 1
Patient characteristics activity score consisted of nine patients because one patient did not respond to multiple items on the SGRQ.The total SGRQ score in the COPD patients was 24.1 ± 13.6 and the scores for the SGRQ components were 31.2 ± 22.7 for symptoms, 38.7 ± 23.0 for activity, and 18.2 ± 14.0 for impacts.The mean CAT score was 11.9 ± 8.6.Tapestry displays are plotted with time of the day on the horizontal axis and date on the vertical axis, with different colors representing PA levels and exercise frequencies.Specifically, blue represents sleep [1.0 metabolic equivalents (METs)], light blue represents resting in a sedentary position (1.0 to 1.2 METs), green represents sedentary occupation (such as deskwork; 1.2 to 1.8 METs), yellow represents light activity (1.8 to 3.0 METs), and red represents activity (more than 3.0 METs).
2. Tapestry display of PA : p<0.01 Values are Spearman correlation coefficients.Abbreviations used are the same as those in Table-1, except for the followings; CAT, COPD assessment test; METs, metabolic equivalents; pred, predicted; SGRQ, St. Georgeʼs Respiratory Questionnaire.

Table - 2
Correlations of physical activities with pulmonary functions and QOL questionnaires in patients with COPD Patient characteristics