Activity Monitoring in Assessing Activities of Daily Living

Abstract

Individuals who have active lifestyles appear to reap substantial benefits. It is therefore of interest to assess level of activity and to determine whether interventions are capable of altering activities of daily life. Questionnaires are often employed because of their simplicity, but objective measures are sought. Long-term assessment of energy expenditure, either through doubly labeled water analysis or through measurements of expired gases are expensive and often impractical. Activity monitors include pedometers, heart rate monitors, accelerometers and integrated multisensor systems. Rapidly advancing activity monitor technology has enabled long-term use and facilitated downloading of recordings to computers where sophisticated analysis of activity patterns can be made. Accelerometer-based systems have received the most attention. When applied to chronic obstructive pulmonary disease patients, accelerometric monitors have demonstrated low levels of activity; those using long-term oxygen and those having exacerbations are particularly inactive.

Keywords::

• Accelerometer
• Pedometer
• COPD
• Energy expenditure

INTRODUCTION

A major precept of modern health maintenance strategies is that an active lifestyle has substantial benefits. In the healthy elderly and in patients with chronic disease, those individuals who remain active generally have a better quality of life and survive longer. Though less well established, the idea that interventions that improve activity levels are likely to improve survival is generally accepted.

In both research applications and in clinical practice, it is often desired to obtain a measure of activity level of an individual. Two methods that are often employed can be seen to have serious flaws. One is to ask the individual to estimate his or her activity level—either informally or by questionnaire. Generally, however, this method yields inaccurate results and an overestimate of activity level is often obtained ([1], [2]). The second method involves obtaining a measure of exercise tolerance, either in the exercise laboratory or with field tests, and to assume that higher exercise tolerance translates to higher levels of activity. In fact, correlational analysis confirms this concept: those with higher exercise tolerance tend to have higher levels of accustomed activity (e.g., Pitta et al. ([3])). However, it may be wondered whether observed correlations between higher exercise tolerance and survival ([4], [5], [6], [7]) represent cause-and-effect or whether identification of those with higher exercise tolerance simply identifies a subgroup with higher activity level. Are beneficial effects mechanistically linked to higher everyday activity levels or to an increased physiological ability to exercise? If the former is true, then an assessment of exercise capacity can be seen as a mere proxy-variable for activity level.

There is developing interest in methodology to directly assess activity performed in the home and community during daily life. Several approaches have been proposed and two major classes of measurements have been reported. The first involves recording energy expenditure and the second involves direct mechanical assessment of movement.

Assessments of energy expenditure

These methods carry the disadvantages of being both technically demanding and not providing a direct measure of activity. The relation between energy expenditure and activity depends on variables such as basal metabolic rate, body weight (and composition) and work efficiency. One method, doubly labeled water analysis, is an ingenious technique capable of estimation of energy expenditure over an extended period. In this technique, the subject drinks water containing a known quantity of water with non-radioactive isotopic labels on both the hydrogen and oxygen atoms. At some later time (generally about 2 weeks later) body water (in urine or plasma) is sampled and the relative abundance of labeled hydrogen and oxygen is determined.

Over the period between ingestion and sampling, both labeled hydrogen and oxygen atoms leave the body. The hydrogen atoms leave as water. The oxygen atoms, however, leave both as water and as carbon dioxide. The higher the level of CO₂ excretion, the lower will be the level of labeled CO₂ relative to labeled hydrogen remaining in the body at the end of the assessment period. CO₂ leaves the body mainly in the exhaled breath and is a measure of metabolic rate and, thereby, energy expenditure. The doubly labeled water method can be seen to allow an integrated assessment of the total CO₂ output over the period of observation. However this method has disadvantages. Doubly labeled water is expensive and the analytic methods for isotopic quantitation are not widely available. The temporal resolution of the assessment (weeks) is quite limited. Further, CO₂ output is not an unequivocal measure of energy expenditure.

Oxygen uptake is more closely related to substrate utilization. The ratio of CO₂ production to O₂ consumption depends on the respiratory quotient of the substrate consumed (0.7 for fats and 1.0 for sugars). Moreover, during heavy exertion, lactic acid is produced; buffering this strong acid consumes sodium bicarbonate and results in superimposed CO₂ production unrelated to oxygen consumption. Nevertheless, doubly labeled water methodology has been utilized in research investigations to estimate activity levels of study participants and has yielded important insights ([8]).

Alternately, energy expenditure can be estimated by direct analysis of the respiratory gases to determine oxygen uptake. Technologic advances have allowed miniaturization of the apparatus required for measurement of the time course of oxygen uptake—this allows for measuring metabolic rate during activities of daily living ([9]). This approach yields accurate assessment with high temporal resolution, but it is intrusive and not suitable for long periods of observation.

Direct measurement of activity

The technology underlying devices for activity monitoring is advancing at an impressive pace. Only a few years ago these devices relied on body motion to trigger mechanical devices that would then increment mechanical counters. Today's devices employ sophisticated sensors, support electronic data analysis, feature electronic memory often capable of recording for weeks at a time and allow downloading to computers for extensive analysis. Four general types of activity monitoring devices are currently available: pedometers, heart rate monitors, accelerometers and integrated multisensor systems.

Pedometers are usually mounted on the ankle or calf and sense step rate, which is a limited measure of activity. Bicycle riding, for example, will be poorly represented. Current devices range from simple to complex. Those that report only the total steps recorded in a given period of time will have limited temporal resolution, but this may be adequate for some purposes. More recent devices allow electronic recording of the time course of step rate (for up to 2 months with 1 device) and, when these data are downloaded to a computer, temporal displays can be generated ([10]). Some investigators have found that step count is underestimated by some devices when walking speed is slow, as may occur in debilitated patients ([11]).

Heart rate monitors can be utilized as semi-quantitative measures of activity, as heart rate rises and falls with metabolic rate. The Fick relationship dictates that oxygen uptake is, in fact, the product of heart rate, stroke volume and arterio-venous oxygen content difference. In many situations, activity-induced increases in oxygen uptake are predominantly accommodated by the associated change in heart rate. Heart rate may be recorded in the ambulatory subject through electrocardiographic leads, although the current generation of pulse oximeters often track heart rate as well. Heart rate monitors do not seem to have been used extensively as measures of activity in patient groups, especially in debilitated patients. Such groups have widely varying basal heart rates, may have cardiac arrhythmias such as atrial fibrillation, sometimes have chronotropic incompetence and may be prescribed beta-blockers; all these factors would complicate the relation between the level of activity and heart rate. Heart rate has, however, been found of use in multivariable activity monitor configurations (see later).

Accelerometers are electronic devices that are often worn on the waist or the arms. They can best be thought of as being vibration sensors, as smooth movements record less than abrupt movements. For example, when worn by a person riding in an automobile, an accelerometer will register appreciable “activity,” but the main source will be from the vibration of the car as it travels down the road, not from the acceleration as it speeds up and slows down. Accelerometers are classified as uni-axial or tri-axial, depending on whether they sense acceleration in one or three orthogonal directions. Though the tri-axial configuration has been found to be preferable by some, in many applications the movements in the three axes are not analyzed separately. In one application, the accelerometer output is expressed as a vectorial sum (the square root of the sum of the squares of activity recorded on each of the 3 axes) of the activity counts generated over the period of 1 minute.

Table 1 Methods for activity assessment

• Questionnaires

• Energy expenditure measures

ˆ Doubly labeled water

ˆ Portable metabolic measurement systems

• Activity monitors

ˆ Pedometers

ˆ Heart rate monitors

ˆ Accelerometers

ˆ Integrated multi-sensor monitors

Accelerometer readings have been correlated with the associated metabolic cost of the activity (measured in calories), but such correlations are highly dependent on a host of factors, including the subject's body weight, the type of activity and the way in which the activity is performed. To easily understand the tenuous nature of the correlation between activity counts and metabolic cost, consider that walking at a given speed on a level surface and up a steep ramp will generate similar accelerometer activity counts, but engender greatly different metabolic costs.

Interpreting accelerometer readings are complicated by other factors.

• Traveling in a motor vehicle generates activity counts that are similar to those generated during moderate exertion. Thus records obtained from a subject who spends an appreciable time in an automobile will tend to overestimate activity levels, unless steps are taken to exclude such periods ([2], [12]).

• Subjects are often not perfectly compliant with wearing their monitors. Monitors generally continue recording when not being worn and it is necessary to distinguish sedentary periods from those in which the monitor is not being worn. As non-zero readings are often seen even when the accelerometer is resting on a stationary surface, special attention is required to make this distinction.

• Devices designed to measure a phenomenon sometimes influence the event they are designed to assess. Thus, knowledge that activity is being assessed can induce a greater level of activity during the measuring period. This can be partially offset by monitoring over a long time period. Matthews and colleagues ([13]) found that reliable measures of activity behaviors require at least 7 days of monitoring. A similar consideration involves day-by-day variation in activity. For example, for many people, weekends are a period of relative inactivity ([13]). Either excluding weekends from the monitoring period or monitoring over a long enough period so that such periods are uniformly represented in the recordings of all monitored subjects seems advisable.

• The way in which the monitor is worn can influence its readings. For example, a device designed to be worn on the waist may register different activity readings depending on how it is affixed to the belt, which hip it is worn on and (in obese subjects) how abdominal fat vibrates as the subject ambulates. Further, some types of activities, e.g., those involving upper extremity exercise ([14]), may be poorly represented in activity monitoring recording.

• Even with good manufacturing practices, variability among accelerometers of a given model has been detected ([15], [16]) and can influence recorded activity levels. Whether it is important that serial measurements in a given subject (e.g., before and after an intervention) employ the same unit remains to be determined.

Despite these limitations, accelerometers have proven to be quite useful in examining activity patterns in health and disease. Recently, enhanced models have been introduced that may prove even more useful. The Dynaport device (McRoberts BV, The Hague, Netherlands) locates two accelerometer devices on the body: one in a pack worn on the waist, the other attached to a belt encircling one thigh ([1], [3], [17]). Computer analysis is used to determine body position; time spent in the lying, sitting, and standing positions can be distinguished. Clearly, this ability adds a new dimension to the assessment of a subject's accustomed activity. Limitations of this device include a relatively short recording period (approximately 1 day) and a more intrusive configuration (weight is several-fold greater than simpler accelerometers).

Integrated multisensor systems employ an accelerometer but assess other variables as well in an attempt to refine activity assessments. Several investigators have combined heart rate with accelerometry and determined that estimation of physical activity energy expenditure is improved ([18], [19], [20]). Recently, a multisensor device has been introduced and has been validated ([21], [22], [23]). The Sensewear Armband (HealthWear Bodymedia, Pittsburgh, PA) is designed for extended periods of wear on the upper arm. It employs five primary measurements: a bi-axial accelerometer, galvanic skin response (electrical resistance of skin beneath the sensor), skin temperature, near-body temperature and heat flux (skin temperature minus near-body temperature). Proprietary algorithms integrate these signals to yield estimation of a large number of variables including energy expenditure, step rate, body position (upright vs. supine) and sleep time. It has been found, however, that exercise-specific algorithms may be necessary to achieve accurate estimation of metabolic variables ([22]).

The use of accelerometers as activity monitors has recently been reviewed in depth in a published conference proceeding ([24], [25], [26], [27], [28], [29]). The use of both activity monitors and questionnaires has been comprehensively reviewed by Pitta et al. ([30]).

Studies of activity monitoring in COPD

Reports of use of activity monitoring in COPD patients are beginning to reach the literature and are briefly reviewed here.

• Ringbaek and Lange ([31]) assessed survival over an 8-year period after administering a mobility questionnaire to 226 Danish patients receiving long-term oxygen therapy (LTOT). Self-reported performance status and outdoor activity were found to be independent predictors of survival.

• Garcia-Aymerich and colleagues ([32]) administered a physical activity assessment questionnaire to 346 COPD patients in Barcelona. One third of COPD patients reported walking less than 15 minutes/day; low quality of life and use of LTOT were associated with low levels of self-reported activity.

• Utilizing a questionnaire, Donaldson et al. ([33]) found that time spent out of doors decreased after a COPD exacerbation.

• Schonhofer et al. ([34]) employed a pedometer which was worn for a week by COPD patients with and without respiratory failure and by healthy controls. Good repeatability was demonstrated and the movement count recorded in the healthy controls was threefold greater than in either patient group. The movement count more than doubled when respiratory failure patients were treated with nocturnal mechanical ventilation.

• Steele et al. ([2]) evaluated a triaxial accelerometer in 47 COPD patients over a 3-day monitoring period and found high correlations with exercise capacity and FEV₁, but not with self reported activity, suggesting that this monitor provided a better measure of activity than self-report.

• In a study of 63 outpatients with COPD, Belza et al. ([12]) recorded daily activity level over a 4-day period with an accelerometer; activity level was closely correlated with 6-minute walk distance, but not with self-report of functional status.

• Singh and Morgan ([35]) studied 11 COPD patients and 9 age-matched controls who each wore a uni-axial accelerometer for a 2-day period. Subjects also completed detailed activity diaries. Brisk walking was detectable as a distinct activity in both groups.

• In COPD rehabilitation participants, tri-axial accelerometer recordings demonstrated that elevated activity levels were restricted to days in which the patients came to the rehabilitation center ([36], [37]).

• A uniaxial accelerometer was worn for 7 days by a group of 29 COPD patients and 10 controls ([38]). Activity levels were lower in severe COPD patients receiving LTOT than in severe COPD patient not receiving LTOT.

• Pitta et al. ([17]) utilized the Dynaport accelerometer device to assess activity in 17 COPD patients experiencing a disease exacerbation requiring hospitalization. Twelve-hour recordings were made after 2 and 7 days of hospitalization and 1 month after discharge. Median walking time during the period of recording was found to be less than 10 minutes during hospitalization and had recovered to only about 15 minutes a month after discharge; this was less than half the walking time observed in a group of stable COPD patients.

• Pitta et al. ([1]) studied 10 COPD patients in a one hour measuring period protocol and 13 COPD patients in a 1-day measuring period protocol. In the shorter protocol, recordings of the Dynaport accelerometer were found to closely match determinations from video observation; self-reported estimations were much inferior. In the longer protocol, self-report overestimated the time spent walking, as determined by the Dynaport monitor.

• Pitta et al. ([3]) studied 50 COPD patients and 25 age-matched healthy subjects with 12-hour assessments with the Dynaport device. Patients demonstrated less walking and standing time than the healthy group. Walking time was highly correlated to the 6-minute walking distance.

There is growing interest in activity monitoring as rapidly evolving technology allows more and more sophisticated assessments. We seem to be poised at the beginning of an era when our suppositions regarding the effect of interventions on the level of everyday activity levels in COPD patients can be tested and important questions addressed. Do bronchodilators, that have been demonstrated to improve exercise tolerance in the laboratory, actually increase everyday activity in these patients? Does providing lightweight ambulatory oxygen supplies to LTOT patients actually increase their out-of-doors activity? Does rehabilitation promote a long-lasting increase in activity? Answering these questions will improve the evidence base regarding the value of the therapies we utilize.

Dr. Casaburi occupies the Alvin Grancell-Mary Burns Chair in the Rehabilitative Sciences.