Utility of the RT3 triaxial accelerometer in free living: An investigationof adherence and data loss

ARTICLE IN PRESS Applied Ergonomics xxx (2009) 1–8

Contents lists available at ScienceDirect

Applied Ergonomics journal homepage: www.elsevier.com/locate/apergo

Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss Meredith A. Perry a, *, Paul A. Hendrick a, Leigh Hale a, G. David Baxter a, Stephan Milosavljevic a, Sarah G. Dean b,1, Suzanne M. McDonough c, Deirdre A. Hurley d a

Centre for Physiotherapy Research, University of Otago, PO Box 56, Dunedin, New Zealand, New Zealand Rehabilitation Teaching and Research Unit, University of Otago, Wellington, New Zealand, New Zealand Health & Rehabilitation Sciences Research Institute, University of Ulster, Northern Ireland d School of Physiotherapy and Performance Science, University College Dublin, Ireland b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 14 June 2009 Accepted 5 October 2009

There is strong evidence for the protective effects of physical activity on chronic health problems. Activity monitors can objectively measure free living occupational and leisure time physical activity. Utility is an important consideration when determining the most appropriate monitor for specific populations and environments. Hours of activity data collected, the reasons for activity hours not being recorded, and how these two factors might change over time when using an activity monitor in free living are rarely reported. This study investigated user perceptions, adherence to minimal wear time and loss of data when using the RT3 activity monitor in 21 healthy adults, in a variety of occupations, over three (7 day) repeated weeks of measurement in free living. An activity diary verified each day of monitoring and a utility questionnaire explored participant perceptions on the usability of the RT3. The RT3 was worn for an average of 14 h daily with 90% of participants having complete data sets. In total 6535.8 and 6092.5 h of activity data were collected from the activity diary and the RT3 respectively. An estimated 443.3 h (6.7%) of activity data were not recorded by the RT3. Data loss was primarily due to battery malfunction (45.2%). Non-adherence to wear time accounted for 169.5 h (38.2%) of data loss, of which 14 h were due to occupational factors. The RT3 demonstrates good utility for free living activity measurement, however, technical issues and strategies to manage participant adherence require consideration with longitudinal and repeated measures studies. Ó 2009 Elsevier Ltd. All rights reserved.

Keywords: Adherence Usability Physical activity Data loss Triaxial accelerometer

1. Introduction There is increasing evidence for the benefits of both occupational physical activity (PA) and sport and leisure time PA on a range of health outcomes (Hildebrandt et al., 2000; Bauman, 2004; Zhang et al., 2006; Probert et al., 2008) including benefits in cardiovascular disease, diabetes, stroke, mental health and musculoskeletal disorders. Occupational activity (or the physical requirements of the job) and occupational inactivity have been shown to influence the risk of

* Corresponding author. Tel.: þ64 4 3855357; fax: þ64 4 43855427.

E-mail addresses: meredith.perry@otago.ac.nz (M.A. Perry), paul.hendrick@ otago.ac.nz (P.A. Hendrick), leigh.hale@otago.ac.nz (L. Hale), physio.dean@otago.ac. nz (G.D. Baxter), stephan.milosavljevic@otago.ac.nz (S. Milosavljevic), sarah.dean@ otago.ac.nz, sarah.dean@pms.ac.uk (S.G. Dean), s.mcdonough@ulster.ac.uk (S.M. McDonough), deirdre.hurleyosing@ucd.ie (D.A. Hurley).

1 Permanent address: Sarah Dean, Peninsula Medical School, St Luke’s Campus, Universities of Exeter and Plymouth, United Kingdom.

musculoskeletal work related disorders (Chan et al., 2004; Marras et al., 2009). Musculoskeletal work related disorders are thought to arise from prolonged periods of exposure to repetitive specific high or low load tasks causing tissue fatigue (Westgaard and Winkel, 1996), coupled with a combination of other psychological and sociological factors (Marras et al., 2009). Exploring the accuracy and utility of methods to measure occupational activity is necessary for understanding the relative contribution of this specific work related factor to occupational physical disorders.

Activity monitors are a common objective measure employed to assess PA in a range of populations and occupational settings (Busser et al., 1998; Ainsworth et al., 1999; Estill et al., 2000; Heil, 2002; Cuthill et al., 2008; Dall and Kerr, 2009) and in a variety of patient populations such as people with cardiovascular disease, neurological disability, and cancer survivors (Steele et al., 2000, 2003a,b; Balogh et al., 2004; Hertzog et al., 2007; Hale et al., 2008; Sloane et al., 2009; Jerome et al., 2009). More recently, activity monitors have been used to assess the PA of people with musculoskeletal

0003-6870/$ – see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.apergo.2009.10.001

Please cite this article in press as: Perry, M.A., et al., Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss, Applied Ergonomics (2009), doi:10.1016/j.apergo.2009.10.001

ARTICLE IN PRESS 2

M.A. Perry et al. / Applied Ergonomics xxx (2009) 1–8

disorders, including those with low back pain (Verbunt et al., 2005; Ryan et al., 2009); this is timely, as lower back complaints are a prevalent occupational injury (Marras, 2000; N.R.C, 2001; Waddell and Burton, 2001). The importance of objectively measuring activity or inactivity in large population groups has led to the design and development of a number of commercially available activity monitors (Westerterp, 1999a, 1999b; Freedson and Miller, 2000; Ward et al., 2005). Determining the most appropriate monitor for a specific study design and occupational group requires accurate and easily accessible information on aspects of reliability, validity and utility in a variety of environments and populations to enable informed decision making (Berlin et al., 2006).

Utility not only encompasses the reliability and validity of an outcome measure or instrument, but also how easy it is for the user to interact with the measure in their usual activities of daily living (ADL), the emotional connection between the user and the instrument which incorporates design aesthetics (Seva et al., 2007), and whether the users’ and researchers’ expectations are met (Chamorro-Koc et al., 2009). Ideally, such measures should be simple, easily understood, unobtrusive and accurate (Law, 1987; Freedson and Miller, 2000; Trost et al., 2005; Ward et al., 2005). Utility efficacy is a combination of monitor factors that include technical limitations (such as malfunction of hardware, incomplete data transfer and sensitivity to noise); and participant factors (such as adherence, experience, social context and cognitive understanding) (Steele et al., 2003b; Slaven et al., 2006; Chamorro-Koc et al., 2009).

Data loss from activity monitors due to adherence or technical issues negatively impacts daily and weekly estimates of occupational and leisure time PA, and consequently necessitates an increase in the number of days of data collection (Trost et al., 2005) therefore increasing the economic and participant burden involved in longitudinal free living PA studies. Furthermore, such data loss can mask important or changing associations between specific populations and their participation in work, home and social PA (Conn et al., 2000; Paul et al., 2008). It is important therefore to establish cause and magnitude of data loss, along with the effectiveness of any strategies employed to minimise such loss, as these insights are important for developing study designs for specific occupational environments and populations, and for interpreting results (Conn et al., 2000).

Data loss due to technical failure (including battery faults) and participant non-adherence is periodically reported in free living PA observational studies (Williams et al., 1989; Sirard et al., 2000; Steele et al., 2000, 2003b; Verbunt et al., 2005). Estill et al. (2000) reported data quality and data loss issues during several specific ergonomic tasks but concluded that the activity monitor being investigated was user friendly and able to distinguish activity load between different groups. Few other studies have reported on the potential effects of occupation on activity monitor utility, and we are unaware of any research which has purposefully investigated hours of activity data collected, the reasons for activity hours not being recorded, and how these two factors might change over time when using an activity monitor in free living. In addition, rarely are workers’ opinions sought on the design and utility of the activity monitor. The purpose of this study was to explore the magnitude and reasons for observed RT3 data loss within a healthy population in free living over three discrete periods of 7 days of data collection, and to explore participants’ perceptions of the monitor’s acceptability and utility using a qualitative approach.

2. Method Data for this study were gathered as part of a pilot repeated measure’s observational study that has investigated the free living stability of the RT3 activity monitor (Hendrick et al., 2008).

2.1. Participants A convenience sample of 24 participants was recruited by public and University campus advertising over a four week period. All participants were of good health with no current or past medical conditions limiting PA. Prior to participation all participants read and signed an informed consent form approved by the University of Otago, School of Physiotherapy Ethics Committee.

2.2. Procedures Each participant’s sex, height and weight were downloaded onto the RT3 via the Stayhealthy software. The RT3 was clipped onto their belt or waistband in the centre of the lower back, and participants were advised to keep the monitor in this position during all waking hours, apart from water-based activities and sporting activities which precluded the use of an activity monitor such as contact sports, for example rugby. If the monitor caused discomfort in this position participants were advised to shift it to the lateral right pelvis, to note the change of position in the activity diary, and to return the monitor to its original position when appropriate. We advised that the monitor should be placed in a prominent and clearly observable position overnight to avoid the participant forgetting to wear it the next morning.

Each participant was instructed to record the primary activity for each waking hour in an activity diary; participants were also advised to record the time and reason for removal of the RT3 in the diary. Researchers contacted each participant twice (via electronicmail, short message service/text, or telephone) during the week to determine any clinical utility issues with the RT3 and to encourage adherence to the protocol.

After one week the RT3 was collected and data downloaded to a computer using the Stayhealthy software. Wearing of the activity monitor for 7 consecutive days, in conjunction with completion of a new activity diary, was repeated three and seven weeks later (i.e., weeks four and eight). A utility questionnaire, exploring participants’ perceptions of the activity monitor’s acceptability, was completed at the end of week eight. Following data collection and download, completed data sets were then cleaned and categorised for statistical analysis.

2.3. Measurement 2.3.1. RT3 triaxial accelerometer The RT3 (Stayhealthy, Inc., Monrovia, CA) is a small (71 56 28 mm), lightweight (65.2 g), battery (AAA) powered device that is calibrated to 5.3 Hz and has a dynamic range of 0.05–2.00 g (Powell et al., 2003). The accelerometer in the RT3 is sensitive to three orthogonal axes: vertical (x), anteroposterior (y), and mediolateral (z) and has four modes of recording and storing data. Mode 4 calculates an average vector magnitude (VM) count over a one-minute epoch for each axis ([X2 þ Y2 þ Z2]0.5).

2.3.2. Utility questionnaire The utility questionnaire (Hale et al., 2008) asked participants to comment on the convenience and acceptability of the RT3 and any difficulties associated with wearing the activity monitor. The questionnaire consisted of four statements where the level of agreement with the statement was marked on a 100 mm anchored line. Two closed question asked if participants would be agreeable to wear the RT3 again for future research projects and if they felt the monitor was user friendly. A final open ended question asked participants for any further comments.

Please cite this article in press as: Perry, M.A., et al., Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss, Applied Ergonomics (2009), doi:10.1016/j.apergo.2009.10.001

ARTICLE IN PRESS M.A. Perry et al. / Applied Ergonomics xxx (2009) 1–8

2.4. Data treatment and statistical analysis Trost et al. (2005) recommend that the amount of data collected should be sufficient to reliably estimate the participant’s usual activity status and Ward et al. (2005) recommended that the minimum hours of wear time should be set prior to data collection.

Failure to collect the set number of valid days typically results in participants being excluded from statistical analyses. In this study no participants were excluded from the analysis on the basis of valid wear time, however a theoretical threshold of a minimum of 10 h of activity data on 5 of the 7 days (including one weekend day) as recommended by Gretebeck and Montoye (1992) was set. This baseline point was used to evaluate how many of the participants in this current study would have met an inclusion threshold and if the number of participants meeting the threshold decreased over the repeated weeks of testing.

Hours of RT3 wear time was calculated by subtracting the hours of non-wear from 24 h (Troiano et al., 2008). Non-wear time was defined as periods during which excessively low RT3 counts were identified (60 min of continuous VM counts of 10 counts/min or below) (Buchheit et al., 2007). Hours of RT3 wear time were then compared to the number of hours recorded as ‘awake’ in the activity diary. Periods of RT3 non-wear time were cross referenced to the activity diary and the hours and reasons for RT3 removal were manually transferred onto an Excel sheet for descriptive analysis. Activities which lasted less than 60 min and required removal of the RT3, such as washing and bathing, were not accounted for. Hours of sleep, as recorded in the activity diary, were not considered to be activity data and were therefore not included in any missing data analyses. Similarly, hours of RT3 data which were missing despite the participant recording that the monitor was worn, were also estimated from the hours marked ‘awake’ in the activity diary.

Descriptive statistics were used to analyse: 1) wear hours from the RT3; 2) the reasons for data loss; and 3) to summarise responses from the utility questionnaire. A subsequent Bland and Altman (1986, 1999) analysis was performed to investigate agreement between week four and week eight daily hours of wear time for participants with complete RT3 data sets. In addition, the Bland– Altman analysis was used to confirm the estimated hours of weekly wear time for participants missing a whole week of RT3 data.

3. Results Twenty four participants were recruited but three participants withdrew over the first 2 days of week one due to work commitments. In total 21 participants (13 women, eight men) completed the study. The 21 participants had a mean (SD) age of 35.5 (13.8) years; body weight of 70.8 (7.7) kg; height of 172.0 (6.8) cm and BMI of 23.9 (2.2) kg/m2 completed the study. On average participants worked 35.0 (14.3) h a week. Table 1 shows participant’s occupation, percentage of time spent sitting during work hours, and the frequency/week involved in sport.

In total 6535.8 h of activity data as recorded in the activity diaries, were collected from the participants (n ¼ 21). The RT3 contained no data for an estimated 443.3 (6.7%) hours of time marked as ‘awake’ and active in the activity diary. RT3 activity data loss increased from 42.0 h in week one, to 85.0 in week four to 316.3 h in week eight. Most of the non-recorded activity hours occurred in weeks four and eight (91.0%). Fig. 1 illustrates that monitor (48.4%) and participant factors (52.8%) contributed equally to the 443.3 h of data loss. Specifically, technical malfunction in week eight resulted in two participants obtaining no RT3 data despite wearing the monitor and accounted for 200.5 h (45.2%) of the 443.3 total hours of data loss. Forgetting to wear the RT3

3

Table 1 Occupational and sport characteristics (n ¼ 21).

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Occupation

Percentage of time spent sittinga

Primary activity participation

Days per week involved in organised activity

Administration Assistant Secretary Physiotherapist Student Student Student Student Student Physiotherapist Beauty Therapist Personal Trainer Physiotherapist/ Education Administrator Medical Receptionist Solicitor Childminder Business Analyst Physiotherapist Network Engineer

75%

Yoga

3

84% 7% Not listed Not listed Not listed Not listed Not listed 7% 43% 36% 43%

Dancing Running Netball Swimming, gym Badminton, gym Football, running Gym None None Cycling, gym Golf

2 7 5 1 2 4 3 0 0 5 1

75% 87% 84% 36% 86% 7% 74%

Walking, gym Walking, pilates Gym Netball Running Pilates, walking Cycling, squash, dancing Walking Squash, cycling

3 3 4 2 3 7 6

20 Mother/Carer 21 Network engineer

33% 74%

5 4

a Occupational percentage of time spent sitting is classified by O*NET, http:// online.onetcenter.org/.

accounted for 169.6 h (38.2%) of the total data loss. Three female participants reported removing the monitor due to the RT3 not blending in with their clothing. This accounted for 19.5 h of RT3 data loss over the three measurement weeks and was categorised as ‘Appearance’ in Fig. 1. Data loss due to participation in sports and water-based activities were fairly uniform over the three recorded weeks and accounted for 18 h (4.2%) of data loss. This occurred with: Soccer (6 h); Netball (4 h); Running (3.5 h); Cycling (2.5 h); Dancing (1 h) and Swimming (1 h). In weeks four and eight a total of 7 h of data (1.6%) were lost due to fear of losing the RT3 when engaged in social activities.

The mean (SD) hours of weekly wear time per individual per week was reasonably consistent over the repeated three weeks with means of 101 (7.3), 100 (10.3), and 98 (10.0) h of data recorded for week one, four and eight respectively; however, the variability of wear time increased from week one to weeks four and eight.

Table 2 presents the hours of RT3 wear time containing activity data (week one and four n ¼ 21, week eight n ¼ 19). All participants met the theoretical inclusion criteria threshold of a minimum of 5 days with 10 h of RT3 data collected in weeks one and four. The loss of two participants RT3 data due to a technical failure in week eight meant that 91% of participants (n ¼ 19) achieved the theoretical threshold in week eight. Table 2 presents the hours of RT3 wear time containing activity data (week one and four n ¼ 21, week eight n ¼ 19). The number of participants obtaining at least 10 h of data on all days of the week declined from week one to week eight. In week one adherence was 100% with all (n ¼ 21) participants obtaining at least 10 h of data on all 7 days. In week four 18/21 participants obtained 10 h of data on all 7 days, while in week eight this number had decreased to 13/19 participants with 100% adherence (Table 2).

The Bland–Altman analysis demonstrated that the average difference (SD) in wear hours from week four to week eight for the 19 participants with complete RT3 data sets was 0.57 h (1.4).

The 95% confidence interval for the average difference in wear hours was 0.6 to 1.2. The limits of agreement around the

Please cite this article in press as: Perry, M.A., et al., Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss, Applied Ergonomics (2009), doi:10.1016/j.apergo.2009.10.001

ARTICLE IN PRESS 4

M.A. Perry et al. / Applied Ergonomics xxx (2009) 1–8

Fig. 1. Total activity hours not recorded by the RT3, by reason, for all participants (n ¼ 21) over the three weeks. Black based bars represent monitor factors for data loss and light grey bars represent participant factors for data loss.

average hours of daily wear time was 14.3 h 3.3 (Fig. 2). From the activity diary, we estimated that 93, (13.3 average hours of daily wear time), and 107.5 h, (15.4 average hours of daily wear time), were lost for the two participants missing all RT3 data in week eight.

The utility questionnaire demonstrated that the median score and inter quartile (IQR) range along a 100 mm line for agreement with the statement ‘‘The RT3 was acceptable to wear for 7 days’’ was 83 mm (72, 100). The median (IQR) for the statement ‘‘It is easy to remember to wear the RT3 daily’’ was 100 mm (78, 100) (Table 3). Sixteen participants responded to the open question: in total 37 comments were made. Discomfort due to the position of the RT3 on the back, especially with sitting at work or driving, was reported by 12 of the 16 (75.0%) responding participants, and was the strongest theme to emerge from the open question. Five of the 16 responding participants (31.3%) mentioned shifting or removal of the RT3 during specific manual aspects of their job for periods of up to 1 h (network engineer, physiotherapist, personal trainer and two students). Two other participants (business analyst and solicitor) removed the monitor when seated at work due to discomfort. Three participants (18.8%) felt that the monitor limited the clothing they could wear. Other themes were: catching and breaking the monitor clip on a chair when arising (n ¼ 3); excessive movement of the monitor with high intensity activities (n ¼ 3); catching of the monitor on a satchel/ backpack (n ¼ 4); one participant felt that the RT3 was too big, one felt that it interfered with lifting children, and another worried that the RT3 might fall into the toilet.

Table 2 Total weekly and mean (SD) daily hours of RT3 wear time of collected activity data, and the number of participants with 100% adherence.

Week

Total hours/ Mean hours/ No. of participants No. of participants week wear day wear with 10 h of with 10 h of data on time time (SD) data on 5 day all 7 day

1 (n ¼ 21) 4 (n ¼ 21) 8 (n ¼ 19) Total

2134 2101 1858 6093

14.5 (1.0) 14.3 (1.5) 14.0 (1.4)

21 19 19

21 18 13

4. Discussion This study assessed the utility of the RT3 activity monitor by investigating the magnitude of data loss compared to the total number of activity hours collected, explored the reasons for RT3 data loss, and investigated participant perceptions of the monitor.

Results indicate that most participants found the RT3 acceptable to wear for the 7 days, corroborated by the high hours of daily wear.

Total RT3 data loss was estimated to be 6.7% (443.3) of the 6535.8 h of data collected from the activity diary. Technical factors (48.4%) and participant factors (52.6%) were equally attributable to causes of RT3 data loss. While the percentage hours of RT3 data loss is relatively small, the combined loss of data due to either forgetfulness (adherence) or battery malfunction increased over time.

Data loss due to battery connection faults in the final week of data gathering caused 45.2% of the total data loss and accounted for 93.3% of the activity monitor related factors. New batteries were used for each week of monitoring and all RT30 s were purchased three months prior to study commencement and had not previously been used except for laboratory based calibration tests.

Therefore, it is unlikely that this result was due to either battery fault or ‘wear and tear’ on the monitors. The lack of a data memory system in the RT3 meant that any loss of battery connection caused an irretrievable loss of all data accumulated to that point in time.

This was despite diligent wearing by the participant as documented in the activity diary.

The complete loss of a week’s data for the two participants in this study due to an inexplicable technical fault is a serious utility concern as it necessitates statistical manipulation on the noncomplete data to minimise group effects (Coleman and Epstein, 1998; Catellier et al., 2005; Slaven et al., 2006). Previous studies have also reported ‘technical faults’ or battery failure with the RT3.

In a study investigating people with low back pain, Verbunt et al.

(2005), reported a similar percentage loss of participants’ data (10%); Hertzog et al. (2007), in a repeated measures study with 65 cardiac patients, reported the need to replace 20 monitors from a pool of 72 RT30 s over a three month period; and Sloane et al.

(2009) reported that 13% (15/115) of participants data at baseline was unusable in an activity activation programme in cancer

Please cite this article in press as: Perry, M.A., et al., Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss, Applied Ergonomics (2009), doi:10.1016/j.apergo.2009.10.001

ARTICLE IN PRESS M.A. Perry et al. / Applied Ergonomics xxx (2009) 1–8

5

Fig. 2. Hours of RT3 daily wear time measured for week four and week eight.

survivors. While Chen et al. (2009) did not provide an exact figure on the amount of data loss they did state the battery failure was the most common cause of data loss in a weight loss study comprising 1685 participants. Technical problems have also been reported in an occupational setting, but not with the RT3. Estill et al. (2000) found that 8% of data files were unusable (technical issues such as data quality, wear time and instrument quality were cited) from a wrist worn accelerometer used for 1 h over 4 days in office and line production workers. It is likely that the rate of technical failure in the current study is related to the number of repeat measures and the number of monitors used (9) relative to the number of participants (21), as well as monitor design limitations, and is not necessarily related to the population being investigated.

The current study found high levels of wear time with 90% (14.5 h), 89% (14.3 h) and 87.5% (14.0 h) of a standard 16 h day containing activity data over the three separate weeks of study. A 16 h day represents the likely maximum hours of wear time if an activity monitor is to be removed while sleeping (Macfarlane et al., 2006). Similar percentages and hours/day of wear time have been reported previously, within a free living population, over a single 7 Table 3 RT3 utility questionnaire data (n ¼ 21).

Question

Mean (SD) Score on 100 mm linea

Median (IQR)

Min, Max

1. The RT3 was acceptable to wear for 7 days 2. It was easy to remember to wear the RT3 daily 3. The RT3 interfered with daily activities 4. The RT3 was annoying to wear 5. Would you wear the RT3 again for research 6. Was the RT3 user friendly

82 (18)

44, 100

24 (20)

83 (72, 100) 100 (78, 100) 22 (6, 33)

23 (20)

22 (0, 39)

0, 56

89 (17)

39, 100 0, 67

Yes ¼ 21 (100%) Yes ¼ 17 (81%), Maybe ¼ 4 (19%)

a A high score indicates increasing agreement with the statement and low score indicates increasing disagreement except for questions 5 and 6 which required a yes, no or maybe answer.

day period with a waist mounted activity monitor (Macfarlane et al., 2006; Matthews et al., 2008; Troiano et al., 2008). Troiano et al. (2008) also found that as age of the population increased, the mean hours/day wear time likewise increased. In people with chronic obstructive pulmonary disease a lower average (13.1) h/day wear time was reported over a 3 day time period (Steele et al., 2000). Hale et al. (2008) reported an average of 11 h/day with the RT3 monitor over two periods of 7 days in patients with a mixture of neurological disorders. The lower average hours/day wear time may arise from increased hours of rest due to the underlying health condition. However, few studies report on actual wear hours and to the authors’ knowledge, no studies have reported on the change in wear hours in a repeated measurement design. Best practice guidelines for field use of activity monitors recommend that minimal wear time for inclusion and wear days are reported (Mathie et al., 2004; Ward et al., 2005), and also that the percentage of participants meeting minimum wear time and the reasons for data loss be routinely reported (Masse et al., 2005). Furthermore, Mathie et al. (2004) suggested that consideration and correction of unequal wear time of activity monitors should be considered.

Therefore, the reporting of wear hours and the variability (mean SD), in addition to current practice guidelines, would improve standardization of field study practice and also allow for a better comparison of results between studies.

Most participants reported that the RT3 was acceptable to wear; application was easy to remember and did not interfere with daily activities. However, forgetting to wear the RT3 increased over the three repeated weeks and was responsible for an estimated 169.5 h (38.2%) of total RT3 data loss. Despite the twice weekly contact with participants, as well as use of an instruction sheet and an activity diary, this was the primary cause of participant related data loss (74.2%).

Forgetting to wear the monitor has been reported extensively in the literature (Kochersberger et al., 1996; Steele et al., 2003b). Van Coevering et al. (2005) noted that only 50% of adolescent participants collected a continuous 7 days of data in a one week study, whilst Conn et al. (2000) found that 28% of participants (average age 74 years) did not complete a full complement of 7–9 days of

Please cite this article in press as: Perry, M.A., et al., Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss, Applied Ergonomics (2009), doi:10.1016/j.apergo.2009.10.001

ARTICLE IN PRESS 6

M.A. Perry et al. / Applied Ergonomics xxx (2009) 1–8

data collection. In our study placement site and bulk of the RT3 was most problematic for the two participants whose occupations (Solicitor, and Business Analyst) involved sitting for a high proportion of their day. Five participants removed the monitor during more manual aspects of their job or studies. Removal of the monitor during work time accounted for a relatively small percentage of data loss in predominantly sedentary occupations; however, this factor could potentially result in an underestimation of participant physical inactivity recognised as an important factor in musculoskeletal work related disorders (Chan et al., 2004; Marras et al., 2009).

The results from our study suggest that, regardless of age and the use of cues there is decreased adherence to monitor wear over time. Previously, Van Coevering et al. (2005) and Trost et al. (2005) have suggested that a variety of frequent cues and incentive payments may increase participant adherence. In our study no incentive payment was provided; however, a cue was delivered twice via the participant’s stated preferred method of communication. It is possible that the cues were not varied or frequent enough to encourage adherence, as time progressed. Our findings also suggest that the additional use of an activity diary is beneficial to not only provide more detail of daily activity, but also to act as a cross reference to the activity monitor output (Buchheit et al., 2007) and provide a prompt to wear the activity monitor daily.

Other reasons for non-recording of data included misplacement, removal for sporting activity, discomfort, appearance, and fear of losing the activity monitor. Forgetting to reattach the RT3 also occurred after prior removal for sport (e.g. swimming, netball and soccer) or after showering/bathing post activity. These hours of data loss are therefore linked to the sporting activities of the participant; consequently, as a participant becomes more active the opportunity for more data to be lost due to removal of the monitor, may at times exceed the actual participation in the sport or leisure activity. In a recent study of preadolescent girls an estimated 10% data loss due to PA participation was adjusted by imputation of an estimation of activity energy expenditure for the activity that was not monitored (Buchheit et al., 2007). A monitor which could be worn during all sporting and occupational activities may minimise this loss; however as many contact sports discourage the wearing of any hard attachments, further consideration of size and construction of activity monitors is required before this may be realised.

Results from the utility questionnaire highlighted that sitting or driving with the monitor situated on the central lumbar spine was uncomfortable and necessitated removal in some occupational situations; caused the monitor to get caught under the backrest of chairs; to get knocked off during sit to stand activities, and to fall off when going to the bathroom. This resulted in a fear of losing the RT3 which accounted for 7 h of data loss. Hale et al. (2008) previously explored various psychometric properties of the RT3 and also found, despite the differences in population groups, major themes of discomfort and fear of losing the monitor. These results may not have occurred if the monitor had been placed on the belt line or waistband over the lateral pelvis. However, the monitors were placed in a more posterior position as previous research by Steele et al. (2003b) had noted that too lateral a placement over the pelvis could result in either the monitor getting knocked off or breakage of the clip (under the armrest of the chair) when participants went from sitting to standing. Activity monitor placement may therefore be a compromise between participant comfort and the position in which calibration equations to interpret the activity monitor output were derived (Welk, 2005). Placement of the monitor to minimise loss of data due to occupation or sporting activity requires careful consideration. This study determined that placement in the centre of the lower back would not be acceptable for those complaining of

lower back pain and for those in occupations which require high periods of time spent sitting. Leg, waist and arm mounted monitors are all available but it is unlikely that one monitor will be suitable for all job related tasks and all occupations. Alternative methods of monitor attachment and the aesthetics of the monitor may address this mechanism of data loss. While data loss due to activity participation is important to recognise, the majority of non-recorded activity hours in this current study was not linked to participation in a particular sporting or occupational activity.

There are several limitations to this research. We did not compare the RT3 to any other activity monitor and thus it is difficult to determine if our population would have found the RT3 preferable or easier to remember to wear than any other monitor. In addition the small number of participants means our results should be considered carefully; however our findings appear to be consistent with larger studies and a range of populations.

It is difficult to accurately establish the cause and magnitude of data loss. We cross referenced the RT3 data to the activity diary however participants were only asked to record their main activity for each hour. Consequently, we only noted RT3 data as missing when 60 min of continuous VM counts of 10 counts/min or below were recorded. Therefore, numerous activities of less than 60 min in duration but also requiring removal of the RT3 were possibly unaccounted for. Increasing the frequency of activity diary input to every 30 min, for example, might have yielded further categories of data loss. Diaries are however subject to recall bias and, depending on the frequency of recording required, diaries can also place considerable burden on participants which can further affect their reliability (Stone et al., 2002). Therefore, while we acknowledge that the results presented in this paper may not accurately reflect the participant factors for data loss because the 60 min interval of recording was so wide, a pragmatic decision to not increase the frequency of diary recording was made.

The Bland–Altman analysis found that the average daily hours of wear time for weeks four and eight was 14.3 h with limits of agreement of 3.3 h for the 19 participants with complete RT3 data sets. The Bland–Altman plot demonstrated that the average difference in wear hours from week four to week eight was evenly spread above and below the zero intercept and that the average difference in daily wear hours from week four to week eight was not statistically significant. There was an increase in variability, shown by an increase in the scatter of the differences, as the magnitude of measurement dropped below 13 h of daily wear time.

However, as the number of participants for this analysis was small these findings should be treated cautiously until future studies have investigated the change in wear hours from week to week in larger sample sizes and in a range of populations.

In addition, the Bland–Altman plot was conducted to confirm the estimation of data loss for two participants. From the activity diary we estimated that 93, (13.3 average hours of daily wear time), and 107.5 h, (15.4 average hours of daily wear time), of RT3 activity hours were lost due to a technical fault. The estimation of this data loss was found to be consistent with the weekly variation demonstrated by the remaining 19 participants with complete data and lies within the Bland–Altman limits of agreement.

5. Recommendations and conclusions While reliability and validity aspects of activity monitors are extensively documented within the literature, utility issues are not so widely published. Accurate documentation of adherence issues with specific activity monitors as well as technical limitations may influence study design and monitor selection with respect to occupational groups and environment.

Please cite this article in press as: Perry, M.A., et al., Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss, Applied Ergonomics (2009), doi:10.1016/j.apergo.2009.10.001

ARTICLE IN PRESS M.A. Perry et al. / Applied Ergonomics xxx (2009) 1–8

This study shows that extensive data were lost due to battery connection faults, which provision of an internal memory system within the RT3 would address. Increasing non-adherence to wear over time should also be taken into account when planning further studies, particularly for determining power calculations.

We would recommend careful consideration of the requirements of specific populations and occupational groups when determining minimally acceptable hours of wear time and placement site of the monitor prior to data collection. Pilot testing of activity monitors, in a repeated measures design, should be considered, to ascertain any specific participant or technical issues in the population of interest. Effective protocols to maintain adherence to wear time require further development, and estimated hours of data loss over time should be routinely reported.

Reporting on the mean hours/day wear time and the cause and magnitude of data loss is particularly important when accelerometry is being employed to assess PA change over time, as it enables effective comparisons between data collection points. It is hoped this report may contribute to further development of this technology by manufacturers in response to user commentary, and also encourage other researchers to be more transparent as to causes of data loss.

Utility, reliability and validity are important issues when logistically determining where to invest resources. The intent of this study was to formally report on the magnitude of data loss due to technical issues and adherence to wear time with the RT3 activity monitor in a healthy population over three repeated 7 days of measurement. This study found that the RT3 was acceptable to wear in a healthy population over 21 days in a variety of occupational groups. Missing data were equally due to monitor and participant factors and the ratio of hours of data collected to hours of data loss requires consideration in future studies. Further studies on usability issues with specific occupational groups and settings are warranted.

Acknowledgments The authors would like to thank Dr Mark Weatherall for his statistical advice. This research was supported by the Centre for Physiotherapy Research, University of Otago.

References Ainsworth, B.E., Richardson, M.T., Jacobs Jr., D.R., Leon, A.S., Sternfeld, B., 1999.

Accuracy of recall of occupational physical activity by questionnaire. J. Clin.

Epidemiol. 52, 219–227.

Balogh, I., Orbaek, P., Ohlsson, K., Nordander, C., Unge, J., Winkel, J., Hansson, G.A., 2004. Self-assessed and directly measured occupational physical activities–influence of musculoskeletal complaints, age and gender. Appl. Ergon. 35, 49–56.

Bauman, A.E., 2004. Updating the evidence that physical activity is good for health: an epidemiological review 2000–2003. J. Sci. Med. Sport 7, 6–19.

Berlin, J.E., Storti, K.L., Brach, J.S., 2006. Using activity monitors to measure physical activity in free-living conditions. Phys. Ther. 86, 1137–1145.

Bland, J.M., Altman, D.G., 1986. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1 (8476), 307–310.

Bland, J.M., Altman, D.G., 1999. Measuring agreement in method comparison studies. Stat. Methods Med. Res. 8, 135–160.

Buchheit, M., Platat, C., Oujaa, M., Simon, C., 2007. Habitual physical activity, physical fitness and heart rate variability in preadolescents. Int. J. Sports Med. 28, 204–210.

Busser, H.J., De Korte, W.G., Glerum, E.B.C., Van Lummel, R.C., 1998. Method for objective assessment of physical work load at the workplace. Ergonomics 41, 1519–1526.

Catellier, D.J., Hannan, P.J., Murray, D.M., Addy, C.L., Conway, T.L., Yang, S., 2005.

Imputation of missing data when measuring physical activity by accelerometry.

Med. Sci. Sports Exerc. 37, S555–S562.

Chan, C.B., Ryan, D.I.A.J., Tudor-Locke, C., 2004. Health benefits of a pedometer-based physical activity intervention in sedentary workers. Prev. Med. 39, 1215–1222.

7

Chamorro-Koc, M., Popovic, V., Emmison, M., 2009. Human experience and product usability: principles to assist the design of user-product interactions. Appl.

Ergon. 40, 648–656.

Chen, C., Jerome, G.J., LaFerriere, D., Young, D.R., Vollmer, W.M., 2009. Procedures used to standardize data collected by RT3 triaxial accelerometers in a largescale weight-loss trial. J. Phys. Act Health 6, 354–359.

Coleman, K.J., Epstein, L.H., 1998. Application of generalizability theory to measurement of activity in males who are not regularly active: a preliminary report. Res. Q. Exerc. Sport 69, 58–63.

Conn, V.S., Minor, M.A., Meh, D.R., Burks, K.J., 2000. Recording activity in older women with TriTrac. Am. J. Health Behav. 24, 370–378.

Cuthill, J.A., Fitzpatrick, K., Glen, J., 2008. Anaesthesia a sedentary specialty?

Accelerometer assessment of the activity level of anaesthetists while at work.

Anaesthesia 63, 279–283.

Dall, P.M., Kerr, A., 2009. Frequency of the sit to stand task: an observational study of free-living adults. Appl. Ergon. 41, 58–61.

Estill, C.F., Macdonald, L.A., Wenzl, T.B., Petersen, M.R., 2000. Use of accelerometers as an ergonomic assessment method for arm acceleration-a large-scale field trial. Ergonomics 43, 1430–1445.

Freedson, P.S., Miller, K., 2000. Objective monitoring of physical activity using motion sensors and heart rate. Res. Q. Exerc. Sport 71, 21–29.

Gretebeck, R.J., Montoye, H.J., 1992. Variability of some objective measures of physical activity. Med. Sci. Sports Exerc. 24, 1167–1172.

Hale, L.A., Pal, J., Becker, I., 2008. Measuring free-living physical activity in adults with and without neurologic dysfunction with a triaxial accelerometer. Arch.

Phys. Med. Rehabil. 89, 1765–1771.

Heil, D.P., 2002. Estimating energy expenditure in wildland fire fighters using a physical activity monitor. Appl. Ergon. 33, 405–413.

Hendrick, P.A., Perry, M.A., Hale, L., Milosavljevic, S., Bell, M.L., McDonough, S.M., Hurley, D.A, Baxter, G.D., 2008. Clinical reliability and utility of an RT3 activity monitor in the measurement of free-living physical activity. In: New Zealand Society of Physiotherapists conference proceedings, 18–20 April, Dunedin, p. 76.

Hertzog, M.A., Nieveen, J.L., Zimmerman, L.M., Barnason, S.A., Schulz, P.M., Miller, C.L., Rasmussen, D.A., 2007. Longitudinal field comparison of the RT3 and an activity diary with cardiac patients. J. Nurs. Meas. 15, 105–121.

Hildebrandt, V.H., Bongers, P.M., Dul, J., van Dijk, F.J.H., Kemper, H.C.G., 2000. The relationship between leisure time, physical activities and musculoskeletal symptoms and disability in worker populations. Int. Arch. Occup. Environ.

Health 73, 507–518.

Jerome, G.J., Young, D.R., Laferriere, D., Chen, C., Vollmer, W.M., 2009. Reliability of RT3 accelerometers among overweight and obese adults. Med. Sci. Sports Exerc.

41, 110–114.

Kochersberger, G., McConnell, E., Kuchibhatla, M.N., Pieper, C., 1996. The reliability, validity, and stability of a measure of physical activity in the elderly. Arch. Phys.

Med. Rehabil. 77, 793–795.

Law, M., 1987. Measurement in occupational therapy: scientific criteria for evaluation. Can. J. Occup. Ther. 54, 133–138.

Macfarlane, D.J., Lee, C.C., Ho, E.Y., Chan, K.L., Chan, D., 2006. Convergent validity of six methods to assess physical activity in daily life. J. Appl. Physiol. 101, 1328–1334.

Marras, W.S., 2000. Occupational low back disorder causation and control. Ergonomics 43, 880–902.

Marras, W.S., Cutlip, R.G., Burt, S.E., Waters, T.R., 2009. National occupational research agenda (NORA) future directions in occupational musculoskeletal disorder health research. Appl. Ergon. 40, 15–22.

Masse, C.L., Fuemmeler, B.F., Anderson, C.B., Matthews, C.E., Trost, S.G., Catellier, D.J., Treuth, M., 2005. Accelerometer data reduction: a comparison of four reduction algorithms on select outcome variables. Med. Sci. Sports Exerc. 37, S544–S554.

Mathie, M.J., Coster, A.C.F., Lovell, N.H., Celler, B.G., 2004. Accelerometry: providing an integrated, practical method for long term, ambulatory monitoring of human movement. Physiol. Meas. 25, R1–R20.

Matthews, C.E., Chen, K.Y., Freedson, P.S., Buchowski, M.S., Beech, B.M., Pate, R.R., Troiano, R.P., 2008. Amount of time spent in sedentary behaviors in the United States, 2003–2004. Am. J. Epidemiol. 167, 875–881.

N.R.C, 2001. Musculoskeletal Disorders and the Workplace: Low Back and Upper Extremities. National Academy Press, Washington, DC. O*NET OnLine, 2008. Summary of Tasks for Job Descriptions. Available from.

Occupational Information Network [Online]. http://online.onetcenter.org/ (Accessed 5 November 2008).

Paul, D.R., Kramer, M., Stote, K.S., Spears, K.E., Moshfegh, A.J., Baer, D.J., 2008.

Estimates of adherence and error analysis of physical activity data collected via accelerometry in a large study of free-living adults. Available from. BMC Med. Res. Methodol. [online] 8 (1), 38. http://www.biomedcentral.com/1471-2288/8/ 38. Accessed 23 November 2008.

Powell, S.M., Jones, D.I., Rowlands, A.V., 2003. Technical variability of the RT3 accelerometer. Med. Sci. Sports Exerc. 35, 1773–1778.

Probert, A.W., Tremblay, M.S., Gorber, S.C., 2008. Desk potatoes: the importance of occupational physical activity on health. Can. J. Public Health 99, 311–318.

Ryan, C.G., Grant, P.M., Dall, P.M., Gray, H., Newton, M., Granat, M.H., 2009. Individuals with chronic low back pain have a lower level, and an altered pattern, of physical activity compared with matched controls: an observational study.

Aust. J. Physiother. 55, 53–58.

Seva, R.R., Duh, H.B.-L., Helander, M.G., 2007. The marketing implications of affective product design. Appl. Ergon. 38, 723–731.

Please cite this article in press as: Perry, M.A., et al., Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss, Applied Ergonomics (2009), doi:10.1016/j.apergo.2009.10.001

ARTICLE IN PRESS 8

M.A. Perry et al. / Applied Ergonomics xxx (2009) 1–8

Sirard, J.R., Melanson, E.L., Li, L.I., Freedson, P.S., 2000. Field evaluation of the Computer Science and Applications, Inc. physical activity monitor. Med. Sci. Sports Exerc. 32, 695–700.

Slaven, J.E., Andrew, M.E., Violanti, J.M., Burchfiel, C.M., Vila, B.J., 2006. A statistical test to determine the quality of accelerometer data. Physiol. Meas. 27, 413–423.

Sloane, R., Snyder, D.C., Demark-Wahnefried, W., Lobach, D., Kraus, W., 2009.

Comparing the 7-day physical activity recall with a triaxial accelerometer for measuring time in exercise. Med. Sci. Sports Exerc. 41, 1334–1340.

Steele, B.G., Belza, B., Cain, K., Warms, C., Coppersmith, J., Howard, J., 2003a. Bodies in motion: monitoring daily activity and exercise with motion sensors in people with chronic pulmonary disease. J. Rehabil. Res. Develop 40, 45–58.

Steele, B.G., Belza, B., Hunziker, J., Holt, L., Legro, M., Coppersmith, J., Buchner, D., Lakshminaryan, S., 2003b. Monitoring daily activity during pulmonary rehabilitation using a triaxial accelerometer. J. Cardiopulm. Rehabil. 23, 139–142.

Steele, B.G., Holt, L., Belza, B., Ferris, S., Lakshminaryan, S., Buchner, D.M., 2000.

Quantitating physical activity in COPD using a triaxial accelerometer. Chest 117, 1359–1367.

Stone, A.A., Shiffman, S., Schwartz, J.E., Broderick, J.E., Hufford, M.R., 2002. Patient non-compliance with paper diaries. BMJ 18, 1193–1194.

Troiano, R.P., Berrigan, D., Dodd, K.W., Masse, L.C., Tilert, T., McDowell, M., 2008.

Physical activity in the United States measured by accelerometer. Med. Sci. Sports Exerc. 40, 181–188.

Trost, S.G., McIver, K.L., Pate, R.R., 2005. Conducting accelerometer-based activity assessments in field-based research. Med. Sci. Sports Exerc. 37, S531–S543.

Van Coevering, P., Harnack, L., Schmitz, K., Fulton, J.E., Galuska, D.A., Gao, S., 2005.

Feasibility of using accelerometers to measure physical activity in young adolescents. Med. Sci. Sports Exerc. 37, 867–871.

Verbunt, J.A., Sieben, J.M., Seelen, H.A., Vlaeyen, J.W., Bousema, E.J., van der Heijden, G.J., Knottnerus, J.A., 2005. Decline in physical activity, disability and pain-related fear in sub-acute low back pain. Eur. J. Pain 9, 417–425.

Waddell, G., Burton, A.K., 2001. Occupational health guidelines for the management of low back pain at work: evidence review. Occup. Med. 51, 124–135.

Ward, D.S., Evenson, K.R., Vaughn, A., Rodgers, A.B., Troiano, R.P., 2005. Accelerometer use in physical activity: best practices and research recommendations.

Med. Sci. Sports Exerc. 37, S582–S588.

Welk, G.J., 2005. Principles of design and analyses for the calibration of accelerometry-based activity monitors. Med. Sci. Sports Exerc. 37, S501–S511.

Westerterp, K.R., 1999a. Assessment of physical activity level in relation to obesity: current evidence and research issues. Med. Sci. Sports Exerc. 31, S522–S525.

Westerterp, K.R., 1999b. Physical activity assessment with accelerometers. Int. J. Obes. 23, S45–S49.

Westgaard, R.H., Winkel, J., 1996. Guidelines for occupational musculoskeletal load as a basis for intervention: a critical review. Appl. Ergon. 27, 79–88.

Williams, E., Klesges, R.C., Hanson, C.L., Eck, L.H., 1989. A prospective study of the reliability and convergent validity of three physical activity measures in a field research trial. J. Clin. Epidemiol. 42, 1161–1170.

Zhang, Y., Cantor, K.P., Dosemeci, M., Lynch, C.F., Zhu, Y., Zheng, T., 2006. Occupational and leisure-time physical activity and risk of colon cancer by subsite. J. Occup. Environ. Med. 48, 236–243.

Please cite this article in press as: Perry, M.A., et al., Utility of the RT3 triaxial accelerometer in free living: An investigation of adherence and data loss, Applied Ergonomics (2009), doi:10.1016/j.apergo.2009.10.001