Analysis and Interpretation of Accelerometry Data in Older Adults: The LIFE Study

Abstract

Background.

Accelerometry has become the gold standard for evaluating physical activity in the health sciences. An important feature of using this technology is the cutpoint for determining moderate to vigorous physical activity (MVPA) because this is a key component of exercise prescription. This article focused on evaluating what cutpoint is appropriate for use with older adults 70–89 years who are physically compromised.

Methods.

The analyses are based on data collected from the Lifestyle Interventions and Independence for Elders (LIFE) study. Accelerometry data were collected during a 40-minute, overground, walking exercise session in a subset of participants at four sites; we also used 1-week baseline and 6-month accelerometry data collected in the main trial.

Results.

There was extreme variability in median counts per minute (CPM) achieved during a controlled bout of exercise (n = 140; median = 1,220 CPM (25th, 75th percentile = 715, 1,930 CPM). An equation combining age, age², and 400 m gait speed explained 61% of the variance in CPM achieved during this session. When applied to the LIFE accelerometry data (n = 1,448), the use of an individually tailored cutpoint based on this equation resulted in markedly different patterns of MVPA as compared with using standard fixed cutpoints.

Conclusions.

The findings of this study have important implications for the use and interpretations of accelerometry data and in the design/delivery of physical activity interventions with older adults.

Accelerometry has been an important methodological advancement in the health sciences that enables researchers to capture objective data on patterns of physical activity behavior (1,2). A key feature of these devices is that the acceleration data captured per unit of wear time is then used to determine time spent in moderate to vigorous intensity physical activity (MVPA) (3). There has also been increasing interest in time spent in light activity (4,5) and sedentary behavior (6,7). Because published physical activity guidelines for older adults recommend moderate levels of physical activity that are relative to each individual and based upon perceived exertion (8), it is critical to understand the relationship between clinical exercise prescription conducted at a moderate level of effort and established cutpoints used in the literature to define accelerometry-based MVPA. If the threshold in counts per minute (CPM) is too high when evaluating accelerometry data, then participants will not be credited for engaging in MVPA even when being adherent to their prescriptions; conversely, using a cutpoint that is too low would falsely elevate levels of MVPA.

The challenge of selecting a cutpoint for defining accelerometry-based MVPA in older adults is formidable. For example, the most commonly cited cutpoint for use with adults, 1952 CPM, was published by Freedson and colleagues (9). Their study protocol involved testing 50 men and women in their 20s exercising for 6-minutes on a treadmill at three different speeds: 4.8 (a slow walk), 6.4 (a fast-paced walk), and 9.7 km/h (jogging). There was a linear relationship between CPM and METs (r = .88) with moderate levels of physical activity (3.0–5.99 METs) corresponding to a range of from 1,952 to 5,724 CPM. In contrast, applying this protocol to a healthy cohort of older men and women (able to walk on a treadmill at a fast face and free of medications that limited their exercise response) with an average age of 69.7 (10) (n = 38), resulted in identifying a MVPA threshold of 1041 CPM. It is important to emphasize that in this latter study speeds had to be set at 2.4, 3.2, and 4.8 km/h to accommodate the decline in fitness with age, and the relationship between CPM and METs was lower than that reported by Freedson and colleagues (r = .60 vs .80). Finally, to complicate matters even further, a study from Canada (11) conducted secondary analyses on data from David Bassett’s laboratory who had participants with a mean age in their 40s (range from 19 to 74 years) perform activities in a real world setting that were known to elicit either a light or moderate MET demand. The authors concluded that the best threshold to use as a cutpoint to define moderately demanding tasks of daily living for older adults is 760 CPM.

Because so little is known about using accelerometers with older adults who have compromised function and are often encountered in clinical settings, we take advantage of data collected in connection with a recent multi-center trial of physical activity—the Lifestyle Intervention and Independence for Elders (LIFE) study (12). Specifically, the current investigation has three aims. We first describe the heterogeneity in exercise intensity in CPM achieved by older adults, age 70–89, when engaging in a bout of walking exercise in a supervised setting. Second, we develop a predictive equation for estimating individually tailored cutpoints based on the median CPM achieved during exercise based on simple participant characteristics. And third, using baseline and 6-month accelerometry data from the LIFE study, we examine the clinical utility and interpretation of accelerometry data among this subpopulation of older adults employing the three aforementioned cutpoints (9–11), along with an individually tailored cutpoint based on our predictive equation.

Methods

Participants

From February 2010 to December 2011, 14,831 participants were screened for the LIFE study at eight different field centers (see Supplementary Appendix); 1,635 of these potential participants were eligible and randomized to treatment, 818 to physical activity (PA) and 817 to health education (HE). Details regarding screening, recruitment yields, and baseline characteristics have been published (13) as has the CONSORT Diagram and the main outcomes of the trial (12). The LIFE study eligibility criteria were designed to target older persons (age 70–89) who were: (a) sedentary; (b) at risk for mobility disability (SPPB score of ≤9); (c) able to walk 400 m in ≤15 minutes without sitting, using a walker, or needing the help of another person; and d) able to safely participate in the intervention. Persons with a SPPB score ≤7 were preferentially enrolled to enrich the sample with individuals at higher risk for major mobility disability (MMD). The study protocol was approved by the institutional review boards at all participating sites and the trial is registered at ClinicalsTrials.gov with the identifier NCT01072500.

For the first two study aims, a random list of potential participants from the PA intervention groups at each of four clinic sites (Wake Forest University, Northwestern, Tufts, and the University of Florida) was generated by the coordinating center stratifying the sample by three age groupings (70–75, 76–80, 80+ years) and sex. At each site, participants were selected from this list until 35 were identified for a total target n of 140. We limited testing at any one site to reduce staff burden.

Procedure and Measures

Aim 1

To describe heterogeneity in the pattern of physical activity of older adults during exercise, participants wore an ActiGraph GT3X (Actigraph, Pensacola, FL) accelerometer on two separate occasions during center-based walking exercise within the maintenance phase of the intervention. Walking intensity was prescribed as a 13 (moderately hard) using the Borg Rating of Perceived Exertion scale (14). Two repeated assessments were made on each individual separated by an interval of at least 1 week but not more than 2 weeks. Because the relationship between CPM for the two assessments was high (r = .90, p < .0001), we used data from the first session of valid results in our analyses.

The accelerometer, worn on the right hip, produced output that was digitized by a 12-bit analog to digital convertor at a rate of 30 Hz. Once digitized, the signal passed through a digital filter limiting the frequency range from 0.25 to 2.5 Hz. Each 1 second sample from the vertical axis was aggregated over 60 seconds to create a data file in CPM. Participants wore the monitor for the entire walking session without additional monitoring from the staff. The device was shaken vigorously both at the beginning and completion of the exercise session so that start and stop times were clearly marked. Participants were instructed to exercise as usual—that is, according to their exercise prescriptions used for all exercise activity—and to include rest stops if needed. There was no indication that participants changed their behavior as a result of wearing the device.

Aim 2

To develop a predictive equation for estimating individually tailored cutpoints based on median CPM achieved during exercise training, we employed conceptually relevant predictor variables from assessment visits in the main LIFE study. We used data from visits that occurred as close as possible to the timing of the ancillary study. Variables selected were age and sex; biometric/functional data including (a) body mass index (BMI), (b) height (to possibly account for step length), (c) self-reported disability using the Pepper Assessment Tool for Disability (PAT-D) (15), (d) the short physical performance battery (SPPB) (16), (e) an overground 400 m walking test performed at participants’ usual pace on a 20 m course (40 m/lap) (17), (f) gait speed calculated from the 4 m walking test of the SPPB (16); and disease-related burden including the presence of (a) heart disease (congestive heart failure, stroke or myocardial infarction), (b) diabetes, and (c) the presence of joint stiffness in the back, hips or legs.

Briefly, the PAT-D consists of 19-items that yield three subscales: basic ADLs, instrumental ADLs, and mobility. The mobility subscale was used in the current analysis (15). The items assess difficulty—using 5-point Likert type scales—with verbal anchors ranging from 1 (usually did with no difficulty) to 5 (could not do). The PAT-D and the subscales have acceptable levels of reliability and validity (15).

The SPPB is a summary performance measure consisting of three increasingly difficult standing balance tests, usual pace walking speed over a 4 m distance, and time for five repeated chair stands performed as quickly as possible. Each test is assigned a categorical score from 0 to 4 that are added together to create a summary score ranging from 0 to 12. Support for the measurement properties of the SPPB has been provided by Guralnik and colleagues (16).

The 400 m walk test is a modified version of a fast-pace mobility walking test originally developed by Newman and colleagues (17). Participants walk at their usual walking pace for 400 m (10 laps of a 20 m course). The maximum time allowed is 15 minutes; participants are allowed to stop and stand to rest and may use a cane, but they are not allowed to lean against any object to support their weight nor are they allowed to use a walker or to seek help from another person (9).

Aim 3

Using the prediction equation developed in aim 2, we examined how using individually tailored cutpoints and three published cutpoints for MVPA—760 (11), 1041 (10), 1952 (9),—influenced the volume of MVPA assessed at baseline and 6-months in conjunction with the PA and HE interventions from the main LIFE study (n = 1,448 at either assessment). Participants were instructed to wear the Actigraph GT3X on their hip for seven consecutive days except during sleep, showering/bathing, and swimming. Movement was captured along the vertical axis and expressed in 1min epochs; non-wear time was defined as 90min of consecutive zero counts. In general, an outlier in CPM was defined as a value that exceeded the nearest value for the day by 1,000 and exceeded the median of the 2nd highest daily value by 3,500. For our analyses, participants were required to have worn the device for 600 minutes per day for at least 5 days.

Statistical Methods

Descriptive statistics were used to characterize the ancillary study sample with linear regression being used to examine the prediction of median CPM achieve during the structured exercise sessions. Median CPM was log-transformed to allow the residuals from this model to approximate the normal distribution and to account for heterogeneity of variance when analyzing untransformed counts. We used a backwards elimination regression procedure to produce a parsimonious model and then investigated for interactions between predictors in this model. We standardized age because standardization of age prevents a rise in the variable inflation factor with both age and age² in the model. Statistical significance was set at the 0.05 level. Finally, for aim 3, constrained mixed models (18) were used to evaluate differences between randomized groups using sex and clinic site (used as stratifying variables during randomization) as covariates. For randomized trials, constrained mixed-models can provide more efficient estimates of postrandomization treatment differences when either baseline or postrandomization measures are missing (18). Least-squares (LS) means of median CPM were obtained from these models fit to log-transformed median CPM. Prior to log transformation, 0.5min/week was added to the median CPM to permit taking logs for participants with 0min/week median CPM. All analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC).

Results

Table 1 provides descriptive data on the subsample that participated in the monitored exercise session (n = 140) along with data from participants that provided accelerometry data at baseline or 6 months from the larger cohort of the LIFE study (n = 1,448). The mean (±SD) age of the subsample was 78.73 (±5.16), 55.0% were women, 21.43% were African American and there was considerable socioeconomic diversity. Despite their advanced age, the average BMI was 29.94kg/m². The leading comorbidity was diabetes at 30.22% followed by arthritis at 17.39%. Note that the descriptive statistics on the subsample are strikingly similar to those for the entire cohort. There are minor exceptions because the subsample was stratified by age and sex to facilitate the development of the predictive equation.

Table 1.

Descriptive Data on Participants

Characteristic	Mean (SD)/N(%), N = 140 Subsample	Mean (SD)/N(%), N = 1,448 LIFE Participants
Age	78.73 (5.16)	78.86 (5.22)
Female	77 (55.00%)	964 (66.57%)
Body mass index	29.94 (5.69)	30.08 (5.97)
Race
Caucasian/White	106 (75.71%)	1110 (76.66%)
African American/Black	30 (21.43%)	241 (16.64%)
Hispanic	3 (2.14%)	56 (3.87%)
All other	1 (0.71%)	41 (2.83%)
Education
Elementary school (K–08)	3 (2.16%)	39 (2.70%)
High school/equivalent (09–12)	40 (28.78%)	432 (29.92%)
College (13–16)	53 (38.13%)	564 (39.06%)
Post graduate	36 (25.90%)	360 (24.93%)
Other	7 (5.04%)	49 (3.39%)
Comorbid conditions
Arthritis	24 (17.39%)	274 (19.03%)
Diabetes	42 (30.22%)	359 (24.91%)
Myocardial Infarction	8 (5.76%)	115 (7.99%)
Stroke	6 (4.32%)	98 (6.79%)
Pain and/or stiffness in the knees	25 (17.99%)	201 (13.92%)
400 m gait speed (m/s)	0.85 (0.15)	0.83 (0.16)

Describing Patterns of Exercise Behavior in the LIFE Cohort

Figure 1 presents a histogram of the median CPM achieved during supervised exercise with counts (number of people) plotted on the Y-axis. Vertical lines are drawn within the histogram to depict the position of the observed median CPM (1220 CPM; 25th, 75th percentile = 715, 1930 CPM) and three published cutpoints for defining MVPA in adults—760 (11), 1041 (10), and 1952 (9) CPM. In addition, kernel density plots are overlaid on this histogram for each of three age groups: 1st group = 70–74 years; 2nd group = 75–79 years, and 3rd group = 80–90 years.

Figure 1.

Histogram and density plots of median count per minutes during structured exercise by age group.

Two points warrant emphasis. First, the large variability in median CPM achieved by this group of older adults during a center-based, supervised exercise session is striking. Twenty-five percent of the sample never reached the lowest threshold of 760 and less than 25% exceeded an intensity of 1952 CPM. And second, the density plots by age groups illustrates the rather dramatic shift to the left that occurs in median CPM among the oldest participants.

Predicting Exercise Intensity

As a second aim, we examined whether conceptually relevant demographic variables, biometric information, and comorbid conditions would predict median CPM achieved during structured, supervised exercise. Table 2 provides the results of a linear model for log-transformed counts that includes all variables considered.

Table 2.

Linear Model for Log-Transformed Median Counts per Minute: All Predictors (R ² = .63)

Variable	Unstandardized Coefficients	p-Value
Intercept	6.48367	<.0001
Age (standardized)	−0.20236	<.0001
Standardized age squared	−0.11626	.0086
Body mass index	−0.01833	.0331
Height	−0.00671	.3104
Sex	−0.09671	.4756
Joint stiffness*	−0.08175	.4268
Diabetes	0.08365	.3866
Heart disease†	−0.24311	.0786
Race‡	−0.00556	.9556
SPPB total score§	0.04268	.1376
SPPB 4 m gait speed	−0.22204	.6018
PAT-D mobility disability||	0.06795	.3304
400-m gait speed	2.38056	<.0001

N = 136 due to missing covariate data.

*Stiffness indicated in spine, knees, hip, or back: coded 0 = No or 1 = Yes.

^†Heart disease including heart failure, heart attack, or stroke.

^‡White versus non-white.

^§Short physical performance battery (SPPB).

^||Pepper Assessment Tool for Disability: Mobility Subscale.

A backwards elimination procedure identified three variables (age, age² and 400 m gait speed) that explained 61% of the variability in log-transformed median CPM (Table 3). Inspection of partial r ² values revealed that the strongest predictor was 400 m gait speed, which explained 53% of the variability by itself (bivariate r _s between 400 m gait speed and CPM was 0.73). Age and age² explained 25% of the variability when entered into the model independently.

Table 3.

Parameters Estimates from Final Linear Regression Model on Log Transformed Counts per Minute (R ² = .61; N = 136)*

Variable	Unstandardized Coefficient	Standard Error	p-Value	Standardized Coefficient
Intercept	4.74375	0.23642	<.0001
Age standardized	−0.16230	0.04120	<.0001	−0.2221
Standardized age2	−0.08504	0.04025	.0364	−0.12207
400m gait speed	2.70993	0.24893	<.0001	0.63696

*Investigators that may want to use our equation to predict CPM targets for older adults need to standardize age using the mean and standard deviation for age found in Table 1.

Using Cutpoints to Describe MVPA: Baseline and 6-Month LIFE Assessments

As a final aim, we applied the four cutpoints to the accelerometry data collected at baseline within the LIFE study: 760 CPMs, 1041 CPM, 1952 CPM, and individually tailored cutpoints that were computed using the formula in Table 3. A cap of 1952 was placed on the calculation since this is the standard cutpoint use for MVPA in adults. There was a total n = 1,448 participants when considering data at both baseline and 6 months: 901 participants providing both baseline and 6 month data, 317 with just baseline, and 230 with 6 month data only. The loss of baseline data was due to a brief delay in collecting accelerometry at several sites. The constrained mixed model can obtain more precise estimates based on incomplete data and also accounts for “missingness” that may be related to the values of the observed outcomes. Parenthetically, in unreported sensitivity analyses, we found no evidence that these three subgroups differed from one another on a variety of baseline demographic, biometric, and disease-related variables. Moreover, the estimates of MVPA from ANCOVA on the complete data (n = 901), were always within 3min/week of the LS means on the n = 1,448 presented in Table 4.

Table 4.

Minutes/Week of Moderate to Vigorous Intensity Physical Activity by Treatment Group for Four Different Cutpoints at Baseline and 6-Months in the LIFE Study: Means (SD), Medians, and 6-Month Least-square (LS) Means (N = 1,448)

Cutpoint	Visit Averages*	Treatment Group
		Health Education	Physical Activity
	Baseline mean (SD)	136.9 (149.5), Median = 86.6	127.9 (153.5), Median = 79.5
Tailored (calculated)	6-mo mean (SD)	126.5.0 (129.2), Median = 83.4	165.1 (162.0), Median = 122.8
	6-mo LS mean (95% CI)†	70.6 (65.4, 76.2)	108.4 (100.4, 117.0)
	Baseline mean (SD)	200.7 (184.7), Median = 147.7	195.7 (161.1), Median = 161.5
760 CPM	6-mo mean (SD)	189.9 (155.7), Median = 153.0	230.9 (161.5), Median = 197.6
	6-mo LS mean (95% CI)†	131.6 (124.2, 139.4)	173.1 (163.4, 183.4)
	Baseline mean (SD)	110.5 (127.8), Median = 67.8	105.8 (105.9), Median = 75.4
1041 CPM	6-mo mean (SD)	105.3 (101.9), Median = 75.5	140.6 (118.6), Median = 111.5
	6-mo LS mean (95% CI)†	66.4 (62.0, 71.1)	95.0 (88.7, 101.7)
	Baseline mean (SD)	22.1 (45.7), Median = 6.7	20.9 (41.4), Median = 7.4
1952 CPM	6-mo mean (SD)	21.9 (35.8), Median = 7.9	40.0 (64.8), Median = 12.7
	6-mo LS mean (95% CI)†	9.1 (8.1, 10.1)	14.4 (12.9, 16.0)

*Unadjusted mean (SD); median.

^†Adjusted means (95% CI) obtained using constrained mixed models for randomized prepost designs, adjusting for gender and clinical site (both used to stratify randomization) and constraining pre-randomization means to be equal in the two intervention groups. Analyses were performed on log-transformed values and then estimates were back-transformed.

Table 4 provides the unadjusted average (and median) min/wk and adjusted estimates obtained using constrained mixed models at baseline and 6 months for MVPA for both PA and HE using these four cutpoints. Other than the high and low min/wk for MVPA using either the 760 or 1952 cutpoint, the data using the tailored cutpoint are consistently higher than 1041; albeit, the cutpoint of 1041 was clearly closer to the tailored cutpoint then either 760 or 1952. The constrained model revealed that differences between HE and PA for MVPA at 6-months differed from one another using either the tailored or 1041 cutpoint: the LS means for 6 month MVPA by group using the tailored cutpoint was 70.6min/wk for HE and 108.4min/wk for PA, p < .0001; using 1041 CPM, it was 66.4min/wk for HE and 95.0min/wk for PA, p < .0001.

Despite the similarity of MVPA in Table 4 using the tailored and 1041 cutpoints, we hypothesized that these average values might be masking important differences. Table 5 provides data on the PA group to illustrate how MVPA differed as a function of the tailored and 1041 CPM cutpoints when partitioning the data by the three age groups: 70–74, 75–79, and 80–90 years. Note that, when using the tailored cutpoint at 6 months, the 1st age group had fewer min/wk of MVPA than the 3rd age group, 150.2min/wk versus 204.2min/wk, respectively. The opposite was true when using the 1041 cutpoint; that is, the 1st age group had higher MVPA than the 3rd age group, 177.2min/wk versus 114.2min/wk, respectively. Fitting a linear regression to log-transformed MVPA in minutes/week based on tailored cutpoints resulted in a positive slope for age, beta = 0.04 (95% CI 0.03–0.06), whereas the estimated beta for log-transformed MVPA in min/wk using the 1041 cutpoint was −0.05 (95% CI −0.07 to −0.04).

Table 5.

Mean (SD), Median (Mdn) Minutes/Week of Moderate to Vigorous Intensity Physical Activity by Baseline Age Groups Defined by Using Tailored and 1,041 CPM Cutpoints for the PA Group Only

Visit	Tailored Cutpoint (min/wk): by Age Group			1041 CPM Cutpoint (min/wk): by Age Group
	1st	2nd	3rd	1st	2nd	3rd
	70–74 y	75–79 y	80–90 y	70–74 y	75–79 y	80–90 y
Baseline	106.7 (115.0), Mdn = 65.9, N = 182	89.8 (108.0), Mdn = 57.3, N = 178	173.2 (191.9), Mdn = 112.6, N = 246	136.6 (114.7), Mdn = 104.8, N = 182	111.8 (111.5), Mdn = 80.4, N = 178	79.5 (87.5), Mdn = 53.3, N = 246
6 months	150.2 (142.0), Mdn = 109.8, N = 154	125.4 (117.2), Mdn = 90.0, N = 166	204.2 (191.2), Mdn = 151.9, N = 240	177.2 (114.3), Mdn = 150.5, N = 154	145.6 (130.5), Mdn = 114.0, N = 166	114.2 (106.9), Mdn = 77.0, N = 240

Discussion

These findings have important implications for the analysis of accelerometry data in older adults and for methodological steps that are needed in study designs to permit such analyses to be conducted. Furthermore, the current data suggest that, as a field of study, some rethinking is warranted regarding clinical exercise prescription for older populations. It also casts doubt on conclusions being reached with large data sets such as NHANES. For example, in a recent publication on NHANES data (19), the threshold used to define MVPA among 4,000 adults aged 18–65 years was 1952 CPM (9). Although their sample did not include older adults aged 70–89, many adults in their 50s and 60s are compromised by chronic health conditions and/or are sedentary. As we argue below, there is no physiological or behavioral rationale for using 1952 to define MVPA at the population level.

The heterogeneity in median CPM achieved during supervised walking exercise underscores the problem associated with applying a fixed cutpoint to define MVPA. Using a threshold of 1952 CPM would result in concluding that 75% of LIFE participants in the physical activity intervention had been noncompliant to the intensity of treatment. Conversely, using 760 CPM would mean that many participants would be credited for engaging in MVPA, when in fact they were not doing so. In our initial data analyses (Table 4), the 1041 CPM was closest to using individually predicted cutpoints; however, when examined by age (Table 5), the patterns of MVPA during the first 6 months of the LIFE study for those in the PA intervention were markedly different using 1041 versus the tailored cutpoint. That is, the oldest group was engaging in only 114.2min/wk when using the 1041 cutpoint, yet 204.2min/wk when using the tailored cutpoint.

Fixed or Individually Tailored Cutpoint

An important question is whether a fixed or individually tailored cutpoint, capping the threshold at 1952 for predicted cutpoints, should be the threshold of choice when prescribing and evaluating exercise intensity in an older, mobility impaired, population. Extant research in exercise prescription (20), published guidelines for promoting physical activity among older adults (8), and the data presented in the current study support the use of an individually tailored cutpoint. Within the context of cardiac rehabilitation, there is a long history of prescribing exercise relative to patients’ symptom limited capacities (21). Thus, the critically important role that participants’ functional capacities play in defining a physiological meaningful dose of activity is not new, although it seems to have been largely ignored in the realm of processing and interpreting accelerometry data. Rather, the approach has been to simply lower the absolute cutpoint originally published by Freedson and colleagues (9) based on population specific characteristics (10,11). Although this strategy for the current data would have resulted in using the 1041 CPM, our findings reveal the flaw in such a decision. Specifically, inspection of Figure 1 reveals that over 40% of participants were unable to achieve an intensity ≥1041 CPM and would be classified as being nonadherent to the intervention. Interestingly, in the LIFE main outcomes paper (12), the oldest participants, 80+ years, derived as much benefit from the exercise intervention as those 70–79 years.

When using the tailored cutpoint, readers may be perplexed by the pattern of MVPA at both baseline and 6 months (Table 5). At baseline, the youngest group was credited for 106.7min/wk versus 173.2min/wk for the oldest group; at 6-months MVPA increased to 150.2min/wk in the youngest group and 204.2min/wk in the oldest group. However, the data presented in Figure 1 offer a compelling explanation. That is to say, the median CPM for those participants to the far left of the histogram are so low that these individuals achieve MVPA by simply walking more in their free-living environments. Not so for the youngest group who, because of their maximal capacity, must work relatively harder compared to their normal free-living behavior to achieve MVPA. Again, despite the low median CPM that defines MVPA for older participants, they clearly were responsive to the intervention (9).

Implications for Accelerometry, the Design of Interventions and Public Health Messaging

There are important lessons inherent in these data. First, researchers using accelerometry need to calibrate the devices so they have individually tailored benchmarks from which to evaluate change and provide meaning feedback regarding intensity of activity behavior. This could be accomplished by having participants wear accelerometers during the performance of standardized walking tests such as the 400-m walk or the 6-min walk test. Alternatively, one could conduct brief physical activity sessions to derive meaningful targets for each individual.

Second, it is quite clear that 400 m gait speed and age are important predictors of CPM achieved by the LIFE study participants during moderate physical activity performed at an RPE of 13. Schrack and colleagues (22), using data collected in conjunction with the Baltimore Longitudinal Study of Aging on adults with an average age of 67 (range 32–93y), found that age and 6 m gait speed were important correlates of overall physical activity as were BMI and working status. In our analysis, the potential variance due to BMI was captured by gait speed during the 400 m walk. Also, in unreported analyses, when using gait speed in the 4 m walk in place of the 400 m walk, BMI was an independent predictor of CPM, albeit the overall model (age, age², BMI, and 4m gait speed) accounted for 46% as opposed to 61% of the variance in CPM. Clearly, gait speed assessed during the 400 m walk taps into a functional ability that is not represented by shorter performance-based tests of gait speed.

Third, in a previous publication (23) we demonstrated that some older adults are incapable of exercising for even 10-min bouts without taking a break. Because older adults who often have multiple comorbid conditions cannot achieve high volumes of activity in a single session, and often are forced to take frequent periods of rest, interventions for older adults should emphasize a high weekly frequency of physical activity (5–7 days) and multiple bouts across the day as compared with emphasizing extended sessions, a conclusion raised in 2007 with the publication of exercise guidelines for older adults (8). This perspective is consistent with research by Costello (24) who, using qualitative methods, reported that individually tailoring treatment was essential to promoting physical activity behavior of inactive seniors.

Limitations of a Perceptual Definition for MVPA in Older Adults

One might question whether it is appropriate to define a cutpoint for MVPA based on individuals’ perception of moderately demanding walking behavior. Clearly, the best approach would be to assign intensity based on some relative physiological demand such as 70% of heart rate reserve. However, recently published guidelines recommend that establishing moderate levels of physical activity for older adults should be guided by each individual’s level of perceived effort while walking or performing other large muscle activities (8). The reason is that there are multiple inputs determining what these older adults are capable of doing. For example, in the case of older adults with peripheral artery disease, moderate levels of CPM will be dictated by discomfort in the lower extremities, whereas sarcopenia and/or balance problems may dictate what constitutes a moderate demand for others. We realize the limitation in a subjective definition of MVPA when prescribing activity or evaluating accelerometry data, but nonetheless, it is the clinical reality of working with older adults and other special populations that are compromised by chronic disease and or physical impairments.

Summary

In summary, we observed large variability in what constitutes MVPA when older adults are monitored in a structured exercise setting, underscoring the fact that MVPA is a relative construct. It is inadvisable from a behavioral perspective to use fixed cutpoints to (a) set physical activity goals, (b) provide feedback to participants, and (c) determine adherence to treatment. Hence, these findings have important implications for the use and interpretations of accelerometry and the design/delivery of physical activity interventions for older adults.

Supplementary Material

Supplementary material can be found at: http://biomedgerontology.oxfordjournals.org/

Funding

The Lifestyle Interventions and Independence for Elders Study is funded by a National Institutes of Health/National Institute on Aging Cooperative Agreement #UO1 AG22376 and a supplement from the National Heart, Lung and Blood Institute 3U01AG022376-05A2S, and sponsored in part by the Intramural Research Program, National Institute on Aging, NIH. The first author was supported, in part, from the following mechanisms: (a) National Heart, Lung, and Blood Institute grant (R18 HL076441), (b) National Institutes for Aging grant (P30 AG021332), and (c) General Clinical Research Center grant (M01-RR007122).

Conflict of Interest

None.