BACKGROUND:Sensory and cognitive function can impact physical performance, but the relationship of multiple sensory impairments (SIs) with mobility in older adults is not well understood. We hypothesized that severity and number of SIs would be associated with worse timed physical mobility performance, and that cognitive processing speed would mediate the association. METHODS:Participants (N = 832) were older adults (mean age 76.3 ± 5.0 years; 59.4% women; 84.2% non-Hispanic White) who completed tests of physical performance, cognitive function, and multiple sensory domains. Separate linear regression models examined the association of SI with 400-m walk, expanded Short Physical Performance Battery (eSPPB), 4-square step test (FSST), and stair climb test. Cognitive measures of executive function/processing speed (Digit Symbol Coding [DSC] and Trail Making Test [Trails] B) were tested as potential mediators of the relationship between SI and physical performance. RESULTS:Each 1-point decrement in SI scale was associated with slower 400-m walking speed (β = -0.01 m/s, p = .03), lower eSPPB score (β = -0.05 points, p < .001), and longer FSST time (β = 0.20 seconds, p = .01), but there was no association with stair climb time. Using a causal mediation approach with DSC and Trails B as potential mediators, 47.9% of the association of SI with 400-m walk, 43.8% of the association of SI with eSPPB, and 56.7% of the association of multiple SI with FSST were mediated. CONCLUSIONS:Greater SIs were associated with worse physical performance in older adults, and the association was partially mediated by measures of cognitive processing speed and executive function. Future studies should investigate the temporal relationship between SIs, cognitive function, and physical function.

Background: Reduced functional capacity (FC) is associated with adverse surgical outcomes in older adults. Current FC assessments rely on questionnaires; however, it remains unclear whether accelerometer-measured daily activity provides a more accurate evaluation. Our primary aim was to identify accelerometer-based variables associated with reduced FC. Methods: We conducted a secondary analysis of the Study of Muscle, Mobility and Aging (SOMMA) cohort. Participants were community-dwelling adults (non-surgical) aged ≥70 yr and recruited between the years 2019 to 2021 at the University of Pittsburgh (Pittsburgh, PA, USA) and Wake Forest University School of Medicine (Winston-Salem, NC, USA). Participants were included if they completed cardiopulmonary exercise testing and had valid wear time (≥3 days) for two accelerometers used in the SOMMA study (ActiGraph GT9X and activPAL4). We applied classification and regression tree and random forest models to accelerometry-derived metrics. For comparison, we constructed a logistic regression model using modified Duke Activity Status Index 4-Question (M-DASI-4Q) scores extrapolated from the Community Healthy Activities Model Program for Seniors questionnaire. Results: The final cohort included 640 participants (57.2% [366/640] women; mean age 76.3 [5.0] yr), of whom 18% (114/640) had reduced FC (peak oxygen uptake [VO2peak] <16 ml kg−1 min−1). Participants with adequate FC had higher daily step counts (5843.9 [2950.4] vs 2988.3 [1757.2] steps per day; P<0.001) and more time in moderate-to-vigorous physical activity (118.0 [62.2] vs 59.9 [42.4] min day−1; P<0.001) compared with those with reduced FC. The accelerometer-based random forest model (AUC 0.79) did not significantly outperform the M-DASI-4Q model (AUC 0.72; P=0.16). Conclusion: Among community-dwelling older adults, daily step count and time in moderate-to-vigorous activity were most associated with FC, but the accelerometer-based model showed only fair discrimination to identify participants with reduced FC. Validation in surgical populations is needed.

Fatigability, a phenotype of poor energy regulation, is associated with lower physical activity in older adults, but independent associations with sedentary behavior are unknown. We examined whether sedentary behavior was associated with fatigability using cross-sectional data from the Study of Muscle, Mobility and Aging. Mean sedentary time, sedentary bout length, and sedentary breaks/day were measured using 7-day waking hour data collected from a thigh-worn accelerometer. Fatigability was assessed using the Pittsburgh Fatigability Scale Physical subscale (PFS, score 0–50, higher = greater fatigability) and the Pittsburgh Performance Fatigability Index (PPFI), a percentage decline of observed cadence to maximal cadence from a wrist-worn accelerometer captured during a usual-paced 400 m walk (range 0–100%, higher = more performance deterioration). The participants (N = 663; 76.4 ± 5.1 years, 58% women, 54% high PFS, median PPFI 1.4%) were sedentary for 614.8 ± 111.7 min/day, with a mean 15.0 ± 5.5 min/day bout length and mean 46.1 ± 13.2 sedentary breaks/day. Higher total sedentary time was associated with greater PFS Physical score (β = 0.71, p = 0.0368), but this association was not independent of step count/day. After adjusting for step count/day, higher sedentary time was associated with lower PPFI score (β = −0.44, p = 0.0039). Sedentary bout length and breaks/day were not associated with perceived or performance fatigability. Future studies should aim to better understand the inter-relatedness of these behaviors.

Tryptophan (TRP) metabolites along the kynurenine (KYN) pathway (KP) have been found to influence muscle. Proinflammatory cytokines are known to stimulate the degradation of TRP down the KP. Given that both inflammation and KP metabolites have been connected with loss of muscle, we assessed the potential mediating role of KP metabolites on inflammation and muscle mass in older men. Five hundred and five men (85.0 ± 4.2 years) from the Osteoporotic Fractures in Men cohort study with measured D3-creatine dilution (D3Cr) muscle mass, KP metabolites, and inflammation markers (C-reactive protein [CRP], alpha-1-acid glycoprotein [AGP] and a subsample [n = 305] with interleukin [IL-6, IL-1β, IL-17A] and tumor necrosis factor-α [TNF-α]) were included in the analysis. KP metabolites and inflammatory markers were measured using liquid chromatography-tandem mass spectrometry and immunoassays, respectively. 23%-92% of the inverse relationship between inflammatory markers and D3Cr muscle mass was mediated by KP metabolites (indirect effect p < .05). 3-hydroxyanthranilic acid (3-HAA), quinolinic acid (QA), TRP, xanthurenic acid (XA), KYN/TRP, 3-hydroxykynurenine (3-HK)/3-HAA, QA/3-HAA, and nicotinamide (NAM)/QA mediated the AGP relationship. 3-HAA, QA, KYN/TRP, 3-HK/XA, HKr ratio, 3-HK/3-HAA, QA/3-HAA, and NAM/QA mediated the CRP. KYN/TRP, 3-HK/XA, and NAM/QA explained the relationship for IL-6 and 3-HK/XA and QA/3-HAA for TNF-α. No mediation effect was observed for the other cytokines (indirect effect p > .05). KP metabolites, particularly higher ratios of KYN/TRP, 3-HK/XA, 3-HK/3-HAA, QA/3-HAA, and a lower ratio of NAM/QA, mediated the relationship between inflammation and low muscle mass. Our preliminary cross-sectional data suggest that interventions to alter D3Cr muscle mass may focus on KP metabolites rather than inflammation per se.

Purpose: Cardiorespiratory fitness (CRF) measured by peak oxygen consumption (VO2peak) declines with aging and correlates with mortality and morbidity. Cardiopulmonary exercise testing (CPET) is the criterion method to assess CRF, but its feasibility, validity, and reliability in older adults are unclear. Our objective was to design and implement a dependable, safe, and reliable CPET protocol in older adults. Methods: VO2peak was measured by CPET, performed using treadmill exercise in 875 adults >= 70 yr in the Study of Muscle, Mobility and Aging (SOMMA). The protocol included a symptom-limited peak (maximal) exercise and two submaximal walking speeds. An adjudication process was in place to review tests for validity if they met any prespecified criteria (VO2peak <12.0 mLkg(-1)min(-1); maximum heart rate <100 bpm; respiratory exchange ratio <1.05 and a rating of perceived exertion <15). A subset (N = 30) performed a repeat test to assess reproducibility. Results: CPET was safe and well tolerated, with 95.8% of participants able to complete the VO2peak phase of the protocol. Only 56 (6.4%) participants had a risk alert and only two adverse events occurred: a fall and atrial fibrillation. Mean +/- SD VO2peak was 20.2 +/- 4.8 mLkg(-1)min(-1), peak heart rate 142 +/- 18 bpm, and peak respiratory exchange ratio 1.14 +/- 0.09. Adjudication was indicated in 47 tests; 20 were evaluated as valid and 27 as invalid (18 data collection errors, 9 did not reach VO2peak). Reproducibility of VO2peak was high (intraclass correlation coefficient = 0.97). Conclusions: CPET was feasible, effective, and safe for older adults, including many with multimorbidity or frailty. These data support a broader implementation of CPET to provide insight into the role of CRF and its underlying determinants of aging and age-related conditions.

BACKGROUND:The relationship between amino acids, B vitamins, and their metabolites with D3-creatine (D3Cr) dilution muscle mass, a more direct measure of skeletal muscle mass, has not been investigated. We aimed to assess associations of plasma metabolites with D3Cr muscle mass, as well as muscle strength and physical performance in older men from the Osteoporotic Fractures in Men cohort study. METHODS:Out of 1 425 men (84.2 ± 4.1 years), men with the lowest D3Cr muscle mass (n = 100), slowest walking speed (n = 100), lowest grip strength (n = 100), and a random sample (n = 200) serving as a comparison group to the low groups were included. Metabolites were analyzed using liquid chromatography-tandem mass spectrometry. Metabolite differences between the low groups and random sample and their relationships with the muscle outcomes adjusted for confounders and multiple comparisons were assessed using t-test/Mann-Whitney-Wilcoxon and partial correlations, respectively. RESULTS:For D3Cr muscle mass, significant biomarkers (p < .001) with ≥10% fold difference and largest partial correlations were tryptophan (Trp; r = 0.31), kynurenine (Kyn)/Trp; r = -0.27), nicotinamide (Nam)/quinolinic acid (Quin; r = 0.21), and alpha-hydroxy-5-methyl-tetrahydrofolate (hm-THF; r = -0.25). For walking speed, hm-THF, Nam/Quin, and Quin had the largest significance and fold difference, whereas valine (r = 0.17), Trp (r = 0.17), HKyn/Xant (r = -0.20), neopterin (r = -0.17), 5-methyl-THF (r = -0.20), methylated folate (r = -0.21), and thiamine (r = -0.18) had the strongest correlations. Only hm-THF was correlated with grip strength (r = -0.21) and differed between the low group and the random sample. CONCLUSIONS:Future interventions focusing on how the Trp metabolic pathway or hm-THF influences D3Cr muscle mass and physical performance declines in older adults are warranted.

Background: The Study of Muscle, Mobility and Aging (SOMMA) aims to understand the biological basis of many facets of human aging, with a focus on mobility decline, by creating a unique platform of data, tissues, and images. . Methods: The multidisciplinary SOMMA team includes 2 clinical centers (University of Pittsburgh and Wake Forest University), a biorepository (Translational Research Institute at Advent Health), and the San Francisco Coordinating Center (California Pacific Medical Center Research Institute). Enrollees were age =70 years, able to walk =0.6 m/s (4 m); able to complete 400 m walk, free of life-threatening disease, and had no contraindications to magnetic resonance or tissue collection. Participants are followed with 6-month phone contacts and annual in-person exams. At baseline, SOMMA collected biospecimens (muscle and adipose tissue, blood, urine, fecal samples); a variety of questionnaires; physical and cognitive assessments; whole-body imaging (magnetic resonance and computed tomography); accelerometry; and cardiopulmonary exercise testing. Primary outcomes include change in walking speed, change in fitness, and objective mobility disability (able to walk 400 m in 15 minutes and change in 400 m speed). Incident events, including hospitalizations, cancer diagnoses, fractures, and mortality are collected and centrally adjudicated by study physicians. Results: SOMMA exceeded its goals by enrolling 879 participants, despite being slowed by the COVID-19 pandemic: 59.2% women; mean age 76.3 +/- 5.0 years (range 70-94); mean walking speed 1.04 +/- 0.20 m/s; 15.8% identify as other than Non-Hispanic White. Over 97% had data for key measurements. Conclusions: SOMMA will provide the foundation for discoveries in the biology of human aging and mobility.

Background: Parents carry their infants, toddlers, and young children every day. An ergonomic aid to carry (ie, babywearing) has been used for generations by caregivers of young children worldwide. While the benefits of close physical contact for infants are well documented, little is known on how this additional load impacts the health of the caregiver. Objective: An understanding of how babies are carried during their early years, especially how this behavior affects the musculoskeletal and mental health of the caregiver, is the first step to understanding this dynamic and is the objective of this research. Study Design: Cross-sectional observational study. Methods: A survey was designed to provide insight into current practices in the United States and the self-perceived physical and mental health benefits or challenges to babywearing. Results: A total of 3758 babywearing enthusiasts with a high level of experience and frequent babywearing responded. Respondents reported babywearing to allow for multitasking (97%) and for bonding/attachment (87%). Increased babywearing frequency was associated with improvements in fatigue, insomnia, and interest in sex among caregivers. Most respondents had experienced back pain (82%). Urinary incontinence and pelvic organ prolapse appear more prevalent than other research reports, although strong relationships were not found with babywearing. Finally, respondents had mild symptoms of stress, anxiety, and depression. Surprisingly, no relationships were identified between mental health scales and babywearing frequency or experience. Conclusions: Taken together, this data provides a better understanding of physical and mental health of caregivers in the United States, especially as they relate to babywearing. See the Video, Supplemental Digital Content A (available at: http://links.lww.com/JWHPT/A72).

Background: For breast cancer survivors, moderate to vigorous physical activity (MVPA) is associated with improved survival. Less is known about the interrelationships of daytime activities (sedentary behavior [SB], light-intensity physical activity, and MVPA) and associations with survivors’ health outcomes. This study will use isotemporal substitution to explore reallocations of time spent in daytime activities and associations with cancer recurrence biomarkers. Methods: Breast cancer survivors (N = 333; mean age 63 y) wore accelerometers and provided fasting blood samples. Linear regression models estimated the associations between daytime activities and cancer recurrence biomarkers. Isotemporal substitution models estimated cross-sectional associations with biomarkers when time was reallocated from of one activity to another. Models were adjusted for wear time, demographics, lifestyle factors, and medical conditions. Results: MVPA was significantly associated with lower insulin, C-reactive protein, homeostatic model assessment of insulin resistance, and glucose, and higher sex hormone-binding globulin (all P < .05). Light-intensity physical activity and SB were associated with insulin and homeostatic model assessment of insulin resistance (both P < .05). Reallocating 18 minutes of SB to MVPA resulted in significant beneficial associations with insulin (−9.3%), homeostatic model assessment of insulin resistance (−10.8%), glucose (−1.7%), and sex hormone-binding globulin (7.7%). There were no significant associations when 79 minutes of SB were shifted to light-intensity physical activity. Conclusions: Results illuminate the possible benefits for breast cancer survivors of replacing time spent in SB with MVPA.

Objectives: We aimed to quantify the agreement between self-report, standard cut-point accelerometer, and machine learning accelerometer estimates of physical activity (PA), and exam- ine how agreement changes over time among older adults in an intervention setting. Methods: Data were from a randomized weight loss trial that encouraged increased PA among 333 postmenopausal breast cancer survivors. PA was estimated using accelerometry and a validated questionnaire at baseline and 6-months. Accelerometer data were processed using standard cut-points and a validated machine learning algorithm. Agreement of PA at each time-point and change was assessed using mixed effects regression models and concordance correlation. Results: At baseline, self-report and machine learning provided similar PA estimates (mean dif- ference = 11.5 min/day) unlike self-report and standard cut-points (mean difference = 36.3 min/ day). Cut-point and machine learning methods assessed PA change over time more similarly than other comparisons. Specifically, the mean difference of PA change for the cut-point versus machine learning methods was 5.1 min/day for intervention group and 2.9 in controls, whereas it was ≥ 24.7 min/day for other comparisons. Conclusions: Intervention researchers are facing the issue of self-report measures introducing bias and accelerometer cut-points being insensi- tive. Machine learning approaches may bridge this gap.

Accelerometers are person-worn sensors that provide objective measurements of movement based on minute-level activity counts, thus providing a rich framework for assessing physical activity patterns. New statistical approaches and computational tools are needed to exploit these densely sampled time-series data. We implement a functional principal component mixed model approach to ascertain temporal activity patterns in 578 overweight women (60% cancer survivors) and summarize individual patterns with unique personalized principal component scores. We then test if these patterns are associated with health by performing multiple regression of health outcomes (including biomarkers, namely, insulin, C-reactive protein, and quality of life) on activity patterns represented by these scores. Our model elucidates the most important patterns/modes of variation in physical activities. Results show that health outcomes including biomarkers and quality of life are strongly associated with the total volume, as well as temporal variation in activity. In addition, associations between physical activity and health outcomes are not modified by cancer status. Our findings suggest that employing a multilevel functional principal component analysis approach can elicit important temporal patterns in physical activity. It further allows us to study the relationship between health outcomes and activity patterns, and thus could be a valuable modeling approach in behavioral research.

Physical inactivity and unhealthy diet are modifiable behaviors that lead to several cancers. Biologically, these behaviors are linked to cancer through obesity-related insulin resistance, inflammation, and oxidative stress. Individual strategies to change physical activity and diet are often short lived with limited effects. Interventions are expected to be more successful when guided by multi-level frameworks that include environmental components for supporting lifestyle changes. Understanding the role of environment in the pathways between behavior and cancer can help identify what environmental conditions are needed for individual behavioral change approaches to be successful, and better recognize how environments may be fueling underlying racial and ethnic cancer disparities. This cross-sectional study was designed to select participants (n = 602 adults, 40% Hispanic, in San Diego County) from a range of neighborhoods ensuring environmental variability in walkability and food access. Biomarkers measuring cancer risk were measured with fasting blood draw including insulin resistance (fasting plasma insulin and glucose levels), systemic inflammation (levels of CRP), and oxidative stress measured from urine samples. Objective physical activity, sedentary behavior, and sleep were measured by participants wearing a GT3X+ ActiGraph on the hip and wrist. Objective measures of locations were obtained through participants wearing a Qstarz Global Positioning System (GPS) device on the waist. Dietary measures were based on a 24-h food recall collected on two days (weekday and weekend). Environmental exposure will be calculated using static measures around the home and work, and dynamic measures of mobility derived from GPS traces. Associations of environment with physical activity, obesity, diet, and biomarkers will be measured using generalized estimating equation models. Our study is the largest study of objectively measured physical activity, dietary behaviors, environmental context/exposure, and cancer-related biomarkers in a Hispanic population. It is the first to perform high quality measures of physical activity, sedentary behavior, sleep, diet and locations in which these behaviors occur in relation to cancer-associated biomarkers including insulin resistance, inflammation, impaired lipid metabolism, and oxidative stress. Results will add to the evidence-base of how behaviors and the built environment interact to influence biomarkers that increase cancer risk. ClinicalTrials.gov NCT02094170 , 03/21/2014.

This study compared five different methods for analyzing accelerometer-measured physical activity (PA) in older adults and assessed the relationship between changes in PA and changes in physical function and depressive symptoms for each method. Older adult females (N = 144, M-age = 83.3 +/- 6.4yrs) wore hip accelerometers for six days and completed measures of physical function and depressive symptoms at baseline and six months. Accelerometry data were processed by five methods to estimate PA: 1041 vertical axis cut-point, 15-second vector magnitude (VM) cut-point, 1-second VM algorithm (Activity Index (AI)), machine learned walking algorithm, and individualized cut-point derived from a 400-meter walk. Generalized estimating equations compared PA minutes across methods and showed significant differences between some methods but not others; methods estimated 6-month changes in PA ranging from 4 minutes to over 20 minutes. Linear mixed models for each method tested associations between changes in PA and health. All methods, except the individualized cut-point, had a significant relationship between change in PA and improved physical function and depressive symptoms. This study is among the first to compare accelerometry processing methods and their relationship to health. It is important to recognize the differences in PA estimates and relationship to health outcomes based on data processing method.

BACKGROUND: Advancements in accelerometry have led to different methods to process the data for physical activity (PA) in older adults. PURPOSE: The purpose of this study was to: 1.) Compare five different methods for analyzing PA in older women; 2.) Assess the relationship between changes in PA and changes in physical function and depressive symptoms over six months for each analysis method. METHODS: Older adult females (N=144, Mage = 83.3 ± 6.4yrs) wore a hip accelerometer for 6 days and completed measures of physical function and depressive symptoms at baseline and 6 months. Accelerometry data were processed by 5 different methods to estimate PA: a 1041 vertical axis cut point, a 15-sec vector magnitude (VM) cut point (Evenson), a 1-sec VM algorithm (Activity Index) , a machine learned (ML) algorithm from 39 features, and an individualized cut point derived from the median counts of rapid 400-meter walk. Generalized estimating equations and a confusion matrix were used to compare and contrast PA minutes/day. Linear mixed models for each processing method tested the associations between changes in PA and changes in physical function and depressive symptoms. RESULTS: Baseline comparisons between methods for minutes/day of PA and for each minute of PA are in Table 1. There were significant differences between some methods but not others, and methods estimated 6-month change in PA from 4 minutes to over 20 minutes. All methods, except the individualized cut point had a significant positive relationship between change in PA and improved physical functioning. There was also a significant relationship between changes in PA and decreased depressive symptoms for all methods except the individualized cut point. CONCLUSIONS: Time spent in PA differs by the choice of data processing method. Results from individualized cut points are counter to methods that use absolute cut points. Additional research is needed to understand these discrepancies.Table 1: Confusion Matrix for each method with percent (%) overlap between methods for each minute of PA and differences (*) between methods for total minutes/day at baselineNote. * indicate significant differences based on total minutes/day (p<.05); cpm: counts per minute; ML: Machine Learned; MW: meter-walk

As the US population ages, there is an increasing need for evidence based, peer-led physical activity programs, particularly in ethnically diverse, low income senior centers where access is limited.

PURPOSE:Walking for health is recommended by health agencies, partly based on epidemiological studies of self-reported behaviors. Accelerometers are now replacing survey data, but it is not clear that intensity-based cut points reflect the behaviors previously reported. New computational techniques can help classify raw accelerometer data into behaviors meaningful for public health. METHODS:Five hundred twenty days of triaxial 30-Hz accelerometer data from three studies (n = 78) were employed as training data. Study 1 included prescribed activities completed in natural settings. The other two studies included multiple days of free-living data with SenseCam-annotated ground truth. The two populations in the free-living data sets were demographically and physical different. Random forest classifiers were trained on each data set, and the classification accuracy on the training data set and that applied to the other available data sets were assessed. Accelerometer cut points were also compared with the ground truth from the three data sets. RESULTS:The random forest classified all behaviors with over 80% accuracy. Classifiers developed on the prescribed data performed with higher accuracy than the free-living data classifier, but these did not perform as well on the free-living data sets. Many of the observed behaviors occurred at different intensities compared with those identified by existing cut points. CONCLUSIONS:New machine learning classifiers developed from prescribed activities (study 1) were considerably less accurate when applied to free-living populations or to a functionally different population (studies 2 and 3). These classifiers, developed on free-living data, may have value when applied to large cohort studies with existing hip accelerometer data.

Life-logging devices are becoming ubiquitous, yet still processing and extracting information from the vast amount of data that is being captured is a very challenging task. We propose a method to find discriminative regions which we define as regions that are salient, consistent, repetitive and discriminative. We explain our fast and novel algorithm to discover the discriminative regions and show different applications for discriminative regions such as summarization, classification and image search. Our experiments show that our algorithm is able to find discriminative regions and discriminative patches in a short time and extracts great results on our life-logging SenseCam dataset.

Accelerometer cut points misclassify important behaviors. In addition, the intensity paradigm is not helpful for promotion of specific behaviors for public health. Advances in computational techniques to classify accelerometer data into behaviors have been limited by training and testing within controlled observational settings. Previous studies have shown that accelerometers may not work well in older or obese populations, but these studies were limited to a single accelerometer feature, counts per minute. PURPOSE: To compare machine learned algorithms from accelerometer features developed across multiple days of data to estimate minutes of sitting, standing, walking, and biking. METHODS: We collected 268,325 minutes (523 days) of hip worn accelerometer and GPS data in a sample of scripted activities and in two free living adult cohorts with annotated person worn image data as the ground truth. One cohort was cyclists (N=40, 70% male, mean age 36, mean BMI 23.4). The other cohort was overweight and obese women (N=36, mean age 55, mean BMI 32.0). A Random Forest technique and Hidden Markov Model smoothing were employed to predict minute level behaviors from over 40 accelerometer features. Leave one out cross validation was applied. RESULTS: The algorithm, trained and tested on the scripted activities, performed with a mean accuracy of 92.7%. When applied to the cyclists, it performed with 70.9% accuracy. The algorithm, trained and tested on the cyclists, performed with a mean accuracy of 91.3%. The algorithm, trained on the cyclists and applied to the overweight and obese women, performed with 73.4% accuracy. The algorithm, trained and tested on the women, performed with a mean accuracy of 87.2%. The standard intensity cut points performed with 36.5% accuracy for walking, 8.7% accuracy for bicycling, and 77.2% for sedentary behaviors. The accelerometer features varied by age and obesity status. CONCLUSION: Standard accelerometer cut points greatly misclassify key behaviors such as walking and bicycling. Algorithms developed in a controlled observational setting do not accurately predict free living behavior over multiple days. Algorithms developed on one population may not apply to another group with different demographic and health characteristics. Supported by grant # U54 CA155435