Electroencephalography and clinical Neurophysiology, 89 ( 1993 ) 359-362 359 © 1993 Elsevier Scientific Publishers Ireland, Ltd. 0924-980X/93/$06.00

EEM 93554

Where commonly

Technical note

to record motor activity: an evaluation of used sites of placement for activity monitors

J.J. Van Hilten, H.A.M. Middelkoop, S.I.R. Kuiper, C.G.S. Kramer and R.A.C. Roos

Departments of Neurology and Clinical Neurophysiology, Academic Hospital, State University of Leiden, 2300 RC Leiden (The Netherlands)

(Accepted for publication: 30 June 1993)

Summary The aim of this study was to evaluate the role of the site of attachment of activity monitors. We compared mean diurnal and nocturnal motor activity measures as well as the daily pattern of motor activity of dominant and non-dominant wrist recordings over 5 successive days of 10 healthy right-handed and 10 left-handed subjects. In a second study we evaluated the relationship between truncal motor activity and wrist motor activity. No differences emerged between the diurnal and nocturnal measures as well as the diurnal activity pattern of the dominant and the non-dominant wrist. Additionally, no differences were found in relation to handedness. Our results show that it is inaccurate to regard wrist motor activity measurements as.representative of an invariable percentage of truncal motor activity; the relation between the measurements of these two recording sites varies across the daytime period. During the nocturnal period bilateral wrist motor activity occurred frequently in absence of truncal motor activity. The reverse situation, however, may occur sporadically. This finding most likely reflects that integrated generalized movements like postural shifts are measured at all recording sites, while the small movements that occur at the distal extremities are predominantly recorded by wrist-worn monitors.

In conclusion, it is apparent that for both diurnal and nocturnal studies on the assessment of motor activity, either wrist can be chosen as the recording site. Across the diurnal period the relation between motor activity recorded at the wrist and waist is subject to considerable variability. During the nocturnal period motor activity can best be recorded at the wrist because this recording site detects both integrated generalized movements and movements that occur at the distal extremities.

Key words: Motor activity; Recording; Activity monitor; Site of placement

The clinical utility of ambulatory activity monitoring in sleep, behavioral, epidemiological and neurological studies has broadend within recent years (Laporte et al. 1979; Aharon-Peretz et al. 1991; Van Hilten et al. 1991, 1993; Hauri and Wisbey 1992). The advan- tages of the activity monitor largely derive from its potential to record unrestrained motor activity for several days continuously, while subjects live at home. Additionally, because of its small size and weight, this monitor is minimally disturbing to individuals.

There are, however, no standardized criteria regarding the site at which motor activity should be recorded. The majority of the studies that focus on diurnal recordings use the non-dominant wrist as the site of attachment of the monitor (Renfrew et al. 1987; Brown et al. 1990). Reasons for this choice are often lacking. In some studies, however, the non-dominant wrist is chosen because of convenience for the subjects investigated. Regarding sleep, the site of attachment issue was adressed by Webster et al. (1982). Monitor placements at the wrists were shown to detect the greatest amount of motor activity compared to forehead and ankle attachment. Although the monitor placement at the non-dominant wrist measured slightly more motor activity than at the dominant wrist, the difference was considered to be of little importance.

Correspondence to: J.J. Van Hilten, M.D., Dept. of Neurology, Academic Hospital, P.O. Box 9600, 2300 RC Leiden (The Nether- lands). Tel.: 71-262104; Fax: 71-154537.

The problem remaining, however, is the site of attachment of the monitor for diurnal studies. Therefore, in this study we compared mean diurnal and nocturnal motor activity measures as well as the daily pattern of motor activity as obtained from dominant and non-dominant wrist recordings of 10 healthy right-handed and 10 left-handed subjects. In a second study we evaluated the relationship between truncal motor activity and wrist motor activity.

Subjects and methods

The study was carried out in two parts. In the first study, motor activity was recorded continuously for 5 days at both wrists of 10 left-handed (4 males; mean age 26.1 years, S.D. 8.1) and 10 right- handed subjects (7 males; mean age 30 years, S.D. 10.6). Handedness was assessed with a questionnaire (Annet 1970), ambidexter subjects were excluded. In the second study, motor activity was recorded continuously for 3 days in 10 right-handed subjects (6 males; mean age 30 years, S.D. 9.28) at both wrists and the waist to assess the relationship between truncal and wrist motor activity. To measure truncal motor activity a third monitor was suspended on a stretch belt worn around the waist, positioned in the midline approximately 4 cm below the navel. All subjects were members of the University community or were attracted by publicity describing the program. They were in good health and had no history of sleep complaints.

360 J.J. VAN HILTEN ET AL.

Monitor The characteristics of the activity monitor (Gaehwiler Electronic,

Hombrechtikon, Switzerland) used in this study have been reported previously (Borb61y 1986; Van Hilten et al. 1991). Briefly, the device counts supra-threshold motor activity (accelerations > 0.1 g) with filtering of the analog sensor signal by a bandpass filter of 0.25-3 Hz over 15 sec epochs. It stores the resulting sum as a 1-byte value in a 32-kbyte solid state memory. After completion of the recording, the data are read out in a personal computer for analysis.

For this study 9 activity monitors were used. In order to synchro- nize each set of activity monitors they were initialized within 1 min by the same computer. As the error of the crystal clock of the monitors is minimal and timing errors by loading date and time from one personal computer are at most 1 sec, the maximum synchroniza- tion error is about 1 sec.

Each monitor was calibrated on 3 occassions by a standardized bench procedure to ensure comparability within and among the devices. The monitor was mounted on top of a low frequency speaker which produced up and down movements. The speaker was connected to a low frequency oscillator which produced a sinusoidal signal at a fixed frequency of 2.0+0.001 Hz, so that the response within the bandwidth of interest may be recorded. The test mode of the monitor interface was used to determine the threshold limit at which the sine amplitude is measured. All activity monitors behaved within the 3% range of the mean amplitude of each series. Between 3 different calibration series the mean amplitude behaved within the 1.5% range, without any tendency. The intra-instrument variability of each activity monitor was less than 1%. Hence, the reliability of the activity monitors was established by consistency between movement and the monitor output signal on repeated trials. Measurements in different directions showed a loss of sensitivity from the perpendicu- lar (relative to the activity monitor groundplate) to sideways move- ments of 37-57%.

Recording procedures We measured motor activity with an activity monitor continuously

from Monday 7 p.m. until Sunday 11 a.m. (first study), and from Monday 7 p.m. until Thursday 7 p.m. (second study). Subjects were asked to maintain their habitual 24 h pattern of activities and remove the monitor only when taking a bath. During the recording period all subjects kept a log. They recorded the time they switched off the light to go to sleep, the time of definitive awakening, any naps, and the time they removed the monitors.

Healthy subjects have a motor activity pattern characterized by 4 diurnal periods and a nocturnal period with different motor activity levels: (1) the first 2 h after definitive awakening, (2) the remaining part of the morning till 13:00 h, (3) the afternoon; 13:00 h-19:00 h, (4) the evening; 19:00 h till lights out, and (5) the night; lights out till definitive awakening (Renfrew et al. 1987; Brown et al. 1990; Van Hilten et al. 1991). Therefore, for each of the 4 diurnal periods and the nocturnal period the following measures were calculated for each subject, and for each recording site.

(1) The activity level (AL), expressed as the mean number of counts per 15 sec epoch.

(2) The movement index (MI), calculated as the number of epochs with any movement (activity count > 0) expressed as a percentage of all epochs that make up the period. The movement index reflects the proportion of activity and immobility of a recorded period.

For both measures the "mean" of each of the 5 periods across 5 days was calculated.

Statistical methods The data were analyzed by a multivariate analysis of variance

(MANOVA) procedure (Statistic Package for Social Science PC + ) with a repeated measures design. The AL and MI are considered dependent variables. The factor "subjects" is nested within the fixed

80

I 6O

w

utl

.~ 40

~ 20

[ ] dominant wrist [ [ ] nondominsnt wrist I [ ] weis, J

.T_

/ ,

first 2 hrs morning awake

afternoon night evening

Fig. 1. Mean (S.E.M.) motor activity levels of each recording site for all diurnal and nocturnal periods. The graph shows a lower amount

of truncal motor activity and no differences between both wrists.

factor "recording site" and crossed with the fixed factor "diurnal period." Accordingly, we investigated: differences between the recording sites about.the mean values of the AL and MI (taken over all diurnal periods), the existence of a diurnal pattern for these parameters, and "recording site" differences about their diurnal pattern. A similar design was used to evaluate the effects of handed- ness. The results of the nocturnal measures were analyzed with an 1-way ANOVA. Differences between means were tested by paired t tests. P values < 0.05 are considered statistically significant.

Results

Dominant versus non-dominant wrist recording For the mean diurnal and nocturnal values as well as the diurnal

pattern of the AL (Fig. 1) and MI (not shown) of the right- and left-handed subjects no significant differences emerged between the dominant and non-dominant recording sites (0.08 < P < 0.8). Only in the right-handed subjects the dominant wrist recordings across all diurnal periods showed slightly higher values than the non-dominant

"G

2

11

4 16

3

i

T

T I

0.0 0 AL Idl

12

N $ _

Fig. 2. Mean (S.D.) values of AL and MI during the nocturnal period. * indicates a significant difference between wrist and truncal

motor activity values (P = 0.004).

WHERE TO RECORD MOTOR ACTIVITY 361

DOMINANT

NONDOMINANT WRIST

;i

WAIST

i . . . . . , . . . . . i

12 18 00 06 12

Fig. 3. A representative example of truncal and wrist motor activity profiles over 24 h of a healthy right-handed subject. Note the

similarity between wrist and truncal motor activity patterns.

wrist recordings. During the nocturnal period, on the other hand, the non-dominant wrist recordings measured slightly higher values than the dominant wrist recordings. No differences were found for these measures with regard to handedness (0.1 < P < 0.9).

Wrist recording versus waist recording The mean diurnal values of the AL and MI of the waist are

significantly lower than those of the wrists (both measures; P < 0.01). With regard to the diurnal pattern of the AL (Fig. 1) and MI (not shown) of both wrists recordings showed a gradual but significant (P < 0.01) decline across the diurnal periods (Fig. 1). On the other hand, truncal motor activity showed no difference (P > 0.5) across the first 3 diurnal periods but declined prominently in the evening (P < 0.05) (Fig. 1).

During the nocturnal period no differences were found for the MI between the 3 recording sites (P = 0.296). The AL values ob- tained from both wrists, however, were substantially higher (both measures; P < 0.005) than that of the waist (Fig. 2).

Fig. 3 shows a representative example of truncal and wrist motor activity recordings of one subject over 24 h. Note the similarity of the truncal and wrist patterns. During the nocturnal period bilateral wrist motor activity occurred frequently in absence of truncal motor activity. The reverse situation, however, may occur sporadically.

Discussion

Our results revealed no quantitative differences between the dominant and non-dominant wrist recordings. For both sites the overall diurnal motor activity level and the diurnal motor activity pattern were similar. We also found no quantitative differences for these recording sites with regard to handedness. It should be noted that this also illustrates an important limitation of motor activity monitoring, in that the quality of movement of both wrists cannot be addressed. McPartland et al. (1975) remarked that measurements of motor activity of the non-dominant arm reflect total body movement better because the dominant arm is more involved in performing

specific tasks. However, this issue was never appropriately addressed. Although the majority of studies probably choose the non-dominant wrist because of convenience, our results show that this recording site quantifies motor activity to the same extent as the dominant wrist.

As is to be expected, the amount of diurnal truncal motor activity is significantly lower than that measured at both wrists. However, comparisons between truncal and wrist motor activity should be made with caution. First, the most sensitive axis of the waist monitor maintains the same position during the recording, but that of the wrist monitors changes accordingly to the position of the arm. Second, the relation between the amount of motor activity measured at both recording sites varies across the diurnal periods (Fig. 1). Although not addressed in this study, this is also likely to be influenced by lifestyle differences and type of employment. Hence, it would be inaccurate to regard wrist motor activity measurements as representative of an invariable amount of truncal motor activity.

During the nocturnal period the non-dominant wrist monitor measured slightly more motor activity than the dominant wrist moni- tor, but the difference failed to be significant. This confirms the results of Webster et al. (1982). Both wrist recordings detected more motor activity (AL) than the waist recording. On the other hand, the movement index showed no significant difference between the differ- ent recording sites. This finding is most likely explained by the different characteristics of each measure. The movement index re- flects the proportion of motor activity and immobility over time, whereas the activity level reflects the level of motor activity per 15 sec. As can be observed from Fig. 3, during the nocturnal period bilateral wrist motor activity occurred frequently in absence of trun- cal motor activity. The reverse combination, however, is exceedingly exceptional. Thus integrated generalized movements like postural shifts are measured at all recording sites, but the small movements that may occur at the distal extremities are predominantly recorded by wrist-worn monitors.

In conclusion, from our results it is apparent that for both diurnal and nocturnal studies on the assessment of motor activity, either wrist can be chosen as the recording site. However, because our results are derived from 20 subjects this conclusion will have to be verified in a larger sample of subjects. Across the diurnal period the relation between motor activity recorded at the wrist and waist is subject to considerable variability. During the nocturnal period mo- tor activity can best be recorded at the wrist because this recording site detects both integrated generalized movements and movements that occur at the distal extremities.

References

Aharon-Peretz, J., Masiah, A., Pillar, T., Epstein, R., Tzischinsky, O. and Lavie, P. Sleep-wake cycles in multi-infarct dementia and dementia of the Alzheimer type. Neurology, 1991, 41: 1616-1619.

Annet, M. A classification of hand preference by association analy- sis. Br. J. Psychol., 1970, 61: 303-321.

Borb61y, A.A. New techniques for the analysis of the human sleep- wake cycle. Brain Dev., 1986, 8: 482-488.

Brown, A.C., Smolensky, M.H., D'Alonzo, G.E. and Redman, D.P. Actigraphy: a means of assessing circadian patterns in human activity. Chronobiol. Int., 1990, 7: 125-133.

Hauri, P.J. and Wisbey, J. Wrist actigraphy in insomnia. Sleep, 1992, 15: 293-301.

Laporte, R.D.E., Kuller, L.H., Kupper, D.J., McPartland, R.J., Matthews, G. and Caspersen, C. An objective measure of physi- cal activity for epidemiological research. Am. J. Epidemiol., 1979, 109: 158-168.

McPartland, R.J., Kupfer, D.J., Foster, F.G., Reisler, K.L. and Matthews, G. Objective measurement of human motor activity. Biotelemetry, 1975, 2: 317-323.

362 J.J. VAN HILTEN ET AL.

Renfrew, J.W., Pettrigrew, K.D. and Rapoport, S.I. Motor activity and sleep duration as a function of age in healthy men. Phys. Behav., 1987, 41: 627-634.

Van Hilten, J.J., Middelkoop, H.A.M., Kerkhof, G.A. and Roos, R.A.C. A new approach in the assessment of motor activity in Parkinson's disease. J. Neurol. Neurosurg. Psychiat., 1991, 54: 976-979.

Van Hilten, J.J., Hoogland, E.A., Van der Velde, E.A., Van Dijk, J.G., Kerkhof, G.A. and Roos, R.A.C. Quantitative assessment of Parkinsonian patients by continuous wrist activity monitoring. Clin. Neuropharmacol., 1993, 16: 36-45.

Webster, J.B., Messin, S., Mullaney, D.J. and Kripke, D.F. Trans- ducer design and placement for activity recording. Med. Biol. Eng. Comput., 1982, 20: 741-744.