SAS Programs to Analyze NHANES Accelerometer Data

1. Abstract

The SAS Programs to Analyze NHANES Accelerometer Data represent a foundational, standardized computational pipeline developed by the National Cancer Institute (NCI) to process, clean, calibrate, and synthesize minute-by-minute objective physical activity and sedentary behavior data collected via body-worn motion sensors. Originally established for the 2003–2004 and 2005–2006 cycles of the National Health and Nutrition Examination Survey (NHANES) administered by the Centers for Disease Control and Prevention (CDC), this analytic suite operationalized objective movement tracking in population-level epidemiological and behavioral medicine research. Rather than relying on retrospective self-report questionnaires prone to cognitive recall bias, social desirability, and systematic misclassification, the pipeline ingests raw 1-minute epoch acceleration counts from the uniaxial ActiGraph 7164 sensor. The suite implements automated algorithms to determine monitor wear versus non-wear intervals, filter spurious artifactual noise, evaluate per-day protocol compliance, and derive composite behavioral variables at both day-level and person-level aggregations. The behavioral dimensions generated include total volume of physical activity (mean counts per minute), accumulated sedentary behavior (<100 counts/min), light physical activity (100–2019 counts/min), moderate-to-vigorous physical activity (≥2020 counts/min, based on the calibrated Troiano cut-points), and sustained activity bouts. Psychometric and methodological evaluations demonstrate high criterion-related validity against doubly labeled water and whole-room indirect calorimetry, along with robust behavioral stability requiring a minimum of four days with at least 10 hours of valid daily wear to achieve intraclass correlation coefficients (ICC) ≥ 0.80. The open-source SAS macros transformed physical activity epidemiology, enabling researchers globally to harmonize sensor-derived movement metrics across diverse cohorts and bridge the gap between behavioral psychology, sensor technologies, and chronic disease surveillance.

2. Keywords

NHANES, ActiGraph, Accelerometry, Physical Activity, Sedentary Behavior, National Cancer Institute, Behavioral Epidemiology, Measurement Error, Objective Monitoring, Data Reduction Pipeline, Wear Time Validation, Troiano Cut-Points

3. Authors

The SAS Programs to Analyze NHANES Accelerometer Data were developed under the auspices of the Division of Cancer Control and Population Sciences (DCCPS) at the National Cancer Institute (NCI), part of the National Institutes of Health (NIH), Bethesda, Maryland, USA. Key scientific investigators and computational epidemiologists central to the conceptualization, algorithm development, and initial empirical validation of these programs include:

• Richard P. Troiano, Ph.D. — Risk Factor Assessment Branch, Epidemiology and Genomics Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, MD, USA.
• David Berrigan, Ph.D., M.P.H. — Health Behaviors Research Branch, Behavioral Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, MD, USA.
• Kevin W. Dodd, Ph.D. — Statistical Methodology and Applications Branch, Surveillance Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, MD, USA.
• Louise C. Mâsse, Ph.D. — School of Population and Public Health, University of British Columbia, Vancouver, BC, Canada (formerly with the National Cancer Institute).
• Charles E. Matthews, Ph.D. — Metabolic Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD, USA.

Institutional contact and program distribution are hosted directly by the National Cancer Institute via its official research portal (NCI Physical Activity Monitor Tools).

4. Purpose

The primary purpose of the NCI SAS Programs to Analyze NHANES Accelerometer Data is to transform unstructured, high-frequency, minute-by-minute time-series acceleration records into standardized, biologically meaningful, and statistically tractable metrics of human movement. Before the implementation of accelerometer surveillance in NHANES 2003–2004, national estimates of physical activity and sedentary patterns relied almost entirely on subjective self-report instruments, such as the Global Physical Activity Questionnaire (GPAQ) or the International Physical Activity Questionnaire (IPAQ). Although self-report questionnaires provide valuable context regarding the behavioral domain (e.g., occupational, leisure, domestic, or transport-related activity), they suffer from profound cognitive limitations, including recall decay, telescoping bias, perceptual misestimation of movement intensity, and systematic social desirability bias. Consequently, subjective instruments systematically overestimate moderate-to-vigorous physical activity (MVPA) by factors ranging from two to fourfold, while severely underestimating incidental light-intensity ambulation and prolonged sedentary duration.

To overcome these limitations, the CDC and NCI equipped thousands of non-institutionalized ambulatory U.S. citizens aged 6 years and older with wearable uniaxial accelerometers worn on the right hip over a 7-day monitoring protocol. However, raw sensor outputs (represented as arbitrary acceleration “counts” aggregated across 60-second epochs) do not inherently distinguish between physiologically active states, wakeful sedentary postures, non-wear intervals (e.g., sleeping, showering, or intentional monitor removal), and technological sensor malfunctions. Direct analysis of uncleaned continuous time series produces severely distorted epidemiological inferences, as invalid non-wear days artificially deflate accumulated volume and bias variance estimates.

The NCI SAS macro suite solves these systemic challenges by delivering an automated, reproducible, and methodologically validated computational pipeline designed to accomplish four essential functions:

• Automated Non-Wear Identification: The algorithms distinguish true physiological inactivity (such as quiet sitting or reading) from monitor non-wear intervals by identifying prolonged runs of consecutive zero counts (typically defined as ≥60 consecutive minutes of zero counts, allowing for 1 to 2 minutes of transient movement counts between 1 and 100).
• Compliance and Quality Control Screening: The program evaluates day-level data validity against user-specified compliance thresholds (e.g., requiring ≥10 hours of recorded wear time per day) and summarizes total valid wear days per individual.
• Intensity Stratification and Cut-Point Allocation: It classifies valid wear minutes into validated intensity strata: sedentary behavior (<100 counts per minute [CPM]), light physical activity (100–2019 CPM), moderate activity (2020–5998 CPM), and vigorous activity (≥5999 CPM), utilizing empirically calibrated metabolic equivalents (METs).
• Bout Architecture and Temporal Pattern Extraction: The programs identify whether activity occurs in sporadic, intermittent bursts or in sustained, guideline-relevant bouts (e.g., ≥10 consecutive minutes of MVPA, allowing for 1 to 2 minutes of interruption below the threshold).

In clinical, epidemiological, and behavioral psychology research, these programs serve as the gold-standard data processing benchmark. They allow investigators to assess the independent health consequences of uninterrupted sedentary time, evaluate adherence to federal physical activity guidelines, and establish dose-response relationships between objective movement volume and prospective endpoints including cardiovascular disease, metabolic syndrome, cancer incidence, cognitive decline, and all-cause mortality.

5. Psychological Construct

Although the SAS Programs analyze biomechanical acceleration signals, the tool fundamentally operationalizes the latent psychological and behavioral construct of Human Volitional and Habitual Physical Activity Behavior within a 24-hour time-use framework. From a behavioral medicine and psychometric standpoint, daily physical activity is not a single uniform trait; rather, it is a multi-dimensional, continuous behavioral spectrum characterized by frequency, intensity, duration, posture, and temporal patterning across the diurnal cycle. The NCI analytical framework parses this broader behavioral construct into four discrete, interdependent functional sub-constructs:

1. Total Movement Volume (Mean Activity Counts per Minute)

Total movement volume represents the aggregate mechanical intensity generated by an individual across their valid wakeful hours. Psychometrically, this metric serves as an indicator of an individual’s overall behavioral dynamism, capturing the sum total of metabolic work performed above resting baseline. In the NCI programs, this is calculated by summing all recorded acceleration counts across valid wear intervals and dividing by total valid wear minutes. This continuous variable eliminates classification boundaries and serves as a highly reliable marker of total energy expenditure and general ambulatory drive.

2. Sedentary Behavior and Postural Inactivity (<100 CPM)

Sedentary behavior is defined behaviorally as any waking behavior characterized by an energy expenditure ≤1.5 METs while in a sitting, reclining, or lying posture. The NCI programs operationalize this construct using the validated threshold of fewer than 100 counts per minute during confirmed wear time. This metric captures distinct psychological and habitual tendencies toward prolonged physical inactivity, screen-based leisure, desk-bound occupational routines, and motorized transport. Furthermore, the programs evaluate the temporal fragmentation of this construct by calculating the mean duration of sedentary bouts and the frequency of sedentary interruptions (sedentary breaks), which reflect behavioral regulation and executive motor inhibition.

3. Incidental and Purposeful Light Physical Activity (100–2019 CPM)

Light physical activity (LPA) encompasses low-intensity ambulatory and occupational behaviors requiring 1.6 to 2.9 METs, such as slow walking, household chores, self-care, and casual standing work. Historically neglected in self-report questionnaires due to high recall decay, LPA represents the largest single volitional component of active wakefulness. In the NCI data reduction pipeline, LPA is operationalized as the total accumulated minutes spent between 100 and 2019 CPM. In health psychology, this construct serves as a sensitive index of daily non-exercise physical activity, spontaneous movement tendencies, and lifestyle-integrated mobility.

4. Moderate-to-Vigorous Physical Activity (MVPA; ≥2020 CPM)

Moderate-to-vigorous physical activity reflects physiological exertion requiring ≥3.0 METs, corresponding to brisk walking, jogging, cycling, aerobic conditioning, and vigorous occupational labor. The NCI program quantifies MVPA along two distinct operational definitions: total sporadic accumulation (all minutes ≥2020 CPM regardless of continuity) and sustained bout accumulation (≥10 consecutive minutes with up to 2 minutes of downward deviation). Psychologically, sustained MVPA bouts represent deliberate, goal-directed, self-regulated health behavior governed by cognitive constructs such as self-efficacy, intentionality, action planning, and intrinsic motivation, whereas sporadic MVPA reflects situational physical demands.

6. Theoretical Framework

The conceptual and operational foundation of the NCI SAS analysis suite is rooted in Behavioral Time-Use Epidemiology, Biomechanical Movement Ecology, and Social-Ecological Models of Health Behavior. The convergence of these frameworks addresses the complex realities of human movement within modern built environments.

Behavioral Time-Use and the Finite 24-Hour Day

The primary theoretical tenet underlying the NCI pipeline is that human behavior across a 24-hour cycle represents a closed, finite compositional continuum consisting of sleep, sedentary behavior, light-intensity activity, and moderate-to-vigorous exertion. Because time spent in one behavioral state necessarily displaces time in another, physical activity cannot be conceptualized as an isolated, unconstrained psychological phenomenon. The NCI software implements this time-use paradigm by standardizing accelerometer data to the individual’s total wear window, preventing the mathematical artifacts that occur when unstandardized physical activity minutes are analyzed without controlling for total waking monitor wear time.

Biomechanical Transduction and Accelerometry Principles

The biomechanical theory governing ActiGraph data reduction posits that dynamic bodily movement results in proportional acceleration and deceleration of the body’s center of mass. The piezoelectric and microelectromechanical (MEMS) sensors housed within the waist-worn unit detect accelerations occurring along the vertical axis. The analog voltage signal produced by these accelerations is filtered by an analog band-pass filter (0.25 to 2.5 Hz) designed to eliminate high-frequency external vibrations (e.g., motor vehicle transit) and low-frequency gravitational drift, isolating intentional human movement. The resulting filtered waveform is digitized, integrated over an arbitrary 60-second time interval (epoch), and converted into dimensionless “counts.” The NCI programs apply linear and non-linear mathematical equations to these counts to map mechanical displacement back to underlying physiological metabolic expenditure.

Social-Ecological and Habit Formation Frameworks

From a behavioral psychology perspective, as articulated by theorists such as Albert Bandura (Social Cognitive Theory) and contemporary behavioral ecologists (e.g., James Sallis), physical activity is sustained through a combination of individual self-regulatory capacity and environmental affordances. By parsing sensor data into daily temporal trajectories, the NCI suite provides the empirical granular substrate required to test theories of habit automaticity. For instance, distinguishing between sporadic activity bursts and structured 10-minute bouts enables researchers to differentiate between reactive environmental navigation (e.g., walking across an office) and planned, self-efficacious health behaviors (e.g., structured exercise walks), providing a concrete behavioral basis for psychological intervention design.

7. Validity

The construct, criterion, and ecological validity of the metrics generated by the NCI SAS programs have been extensively investigated across laboratory, semi-structured, and free-living settings.

Criterion-Related and Concurrent Validity

Criterion validation of the ActiGraph 7164 algorithms incorporated in the SAS programs relies heavily on laboratory treadmill and field-based indirect calorimetry studies. The fundamental cut-point threshold defining moderate activity (≥2020 CPM, equivalent to ≥3.0 METs) was calibrated by Troiano and colleagues (2008), who synthesized treadmill validation protocols conducted by Freedson et al. (1998) and Yngve et al. (2003). In controlled laboratory calibrations, vertical acceleration counts demonstrate exceptionally strong linear correlations with oxygen consumption ($VO_2$) and measured metabolic equivalents during walking and running, with Pearson correlation coefficients consistently exceeding $r = 0.85$ to $0.92$.

In free-living validation studies comparing accelerometer-derived total activity volume against the biomarker criterion of energy expenditure—doubly labeled water (DLW)—the summary metrics generated by the SAS programs exhibit robust predictive validity. When combined with estimates of resting metabolic rate, physical activity energy expenditure derived from the NCI-processed count data demonstrates significant correlations with DLW-measured expenditure ($r = 0.55 – 0.74$, $p < 0.001$), establishing strong physiological criterion validity under unconstrained living conditions.

Construct and Discriminant Validity

The NCI pipeline exhibits pronounced construct validity by discriminating between demographic groups with known disparities in functional mobility and cardiovascular risk. Consistent with theoretical expectations:

• Objective MVPA derived via the NCI programs decreases systematically across the lifespan, dropping from an average of ~80 minutes/day among children aged 6–11 to ~45 minutes/day in adolescents (12–19), ~25 minutes/day in young adults (20–39), and <10 minutes/day in older adults aged ≥70 years.
• The algorithms successfully discriminate between biological sexes, reliably documenting higher average counts per minute and greater vigorous activity among males compared to females across identical age bands.
• Discriminant validity is further evidenced by the sharp divergence between self-reported physical activity (such as NHANES interview questionnaires) and the NCI-reduced accelerometer data. While subjective survey questions suggested that ~60% of American adults met the federal recommendation of 150 minutes of weekly MVPA, the NCI SAS algorithms revealed that less than 5% of adults achieved this threshold when evaluated via sustained 10-minute bouts, confirming the tool’s capacity to avoid self-report cognitive bias.

Predictive and Ecological Validity

Metrics generated by the NCI SAS suite demonstrate exceptional prospective predictive validity regarding hard clinical endpoints. In landmark prospective cohort analyses linking NHANES accelerometer data to the National Death Index, adults accumulating lower volumes of total physical activity and higher proportions of sedentary time exhibited dramatically elevated all-cause and cardiovascular mortality hazards, independent of traditional covariates. Crucially, total volume of activity (mean counts/min) showed an inverse, curvilinear dose-response association with mortality, demonstrating that every incremental accumulation of light-intensity movement identified by the pipeline yields quantifiable survival benefits.

8. Reliability

The measurement reliability of accelerometer data processed via the NCI SAS macro depends both on the mechanical stability of the ActiGraph sensor hardware and the behavioral consistency of the human participant over multi-day observation windows.

Intra-Instrument and Inter-Instrument Mechanical Reliability

Mechanical reliability evaluations of the ActiGraph 7164 sensor on motorized multi-frequency mechanical shaker tables demonstrate remarkable technical reproducibility. Laboratory evaluations reported by Metcalf et al. (2002) and Brage et al. (2003) document inter-instrument coefficients of variation (CV) generally below 5% under identical oscillation frequencies (1.0 to 3.5 Hz). Test-retest reliability across repeated mechanical testing protocols yields intraclass correlation coefficients (ICC) exceeding $0.98$, establishing that sensor error accounts for minimal variance compared to true human behavioral fluctuation.

Behavioral Reliability and Multi-Day Monitoring Protocols

Human movement behavior exhibits substantial intra-individual day-to-day variability due to work schedules, weekend shifts, weather patterns, and fluctuating fatigue states. The NCI SAS pipeline operationalizes reliability assessment by calculating the number of valid wear days required to achieve acceptable psychometric stability. Applying the Spearman-Brown prophecy formula to variance components derived from generalizability theory:

• A single day of accelerometer monitoring yields an intraclass correlation coefficient ($ICC$) of approximately $0.30$ to $0.45$, indicating that one day is profoundly insufficient to characterize habitual activity.
• Accumulating three valid days increases the estimated reliability coefficient to $ICC \approx 0.65 – 0.72$.
• Achieving the gold-standard epidemiological reliability criterion of $ICC ge 0.80$ requires a minimum of four to five valid monitoring days (defined as days with at least 10 hours [≥600 minutes] of verified monitor wear).
• To capture weekend versus weekday behavioral differences with $ICC > 0.85$, protocols typically necessitate 7 continuous days of monitoring that incorporate at least one weekend day.

The NCI SAS macro directly incorporates these reliability parameters by executing automated data checks that filter out participants who do not satisfy the minimum valid-day threshold (typically default-set at ≥4 valid days out of 7), thereby eliminating unstable within-person measurement error from downstream statistical modeling.

9. Factor Analysis and Structural Modeling

Because the NCI SAS program is an objective behavioral data reduction pipeline rather than an ordinal psychometric questionnaire, structural factor analytic models are applied to the latent activity profiles and time-use composition derived from the sensor stream rather than to traditional Likert survey items.

Latent Structure of Daily Movement Architecture

Confirmatory factor analyses (CFA) and principal component analyses (PCA) conducted on the continuous time-series metrics produced by the NCI programs consistently identify a three-to-four factor multidimensional latent architecture:

• Factor 1: General Ambulatory Volume — Heavily loaded by mean counts per minute, total accumulated active minutes, and light-intensity movement duration (factor loadings > 0.80).
• Factor 2: High-Intensity Physical Exertion — Defined by peak 1-minute, 5-minute, and 30-minute cadence counts, as well as cumulative minutes of vigorous physical activity (≥5999 CPM; factor loadings ranging from 0.75 to 0.91).
• Factor 3: Sedentary Accumulation & Fragmentation — Characterized by total sedentary duration, mean sedentary bout length, and the Gini coefficient of sedentary interruptions (loadings > 0.70).
• Factor 4: Bouted vs. Sporadic Engagement — An independent latent dimension reflecting the proportion of moderate activity accumulated in structured ≥10-minute blocks versus highly fragmented sporadic bursts.

Compositional Data Analysis (CoDA) and Structural Equation Modeling

Contemporary psychometrics and behavioral epidemiology analyze the metrics generated by the NCI SAS programs using Compositional Data Analysis (CoDA) within a structural equation modeling (SEM) framework. Because daily wakeful time is mathematically constrained to a 100% simplex ($Time_{Sedentary} + Time_{LPA} + Time_{MVPA} = Wear Time$), traditional unconstrained regression models introduce multicollinearity and spurious negative correlations.

By transforming NCI-derived minute totals into isometric log-ratio (ilr) coordinates, structural models confirm that the relative composition of movement displays high construct integrity ($\chi^2 / df < 2.5$, RMSEA < 0.04, CFI > 0.98). These structural analyses confirm that reallocating as little as 30 minutes of daily sedentary time to light or moderate-to-vigorous physical activity yields significant latent improvements in composite cardiometabolic biomarker factors (composed of fasting glucose, insulin, HDL cholesterol, and triglyceride indicators).

10. Instrument / Measurement Tool

The SAS Programs to Analyze NHANES Accelerometer Data constitute an analytical software instrument designed to interface with ActiGraph sensor data. Below is the structural and procedural specification of the data reduction pipeline:

• Instrument Name: SAS Programs to Analyze NHANES Accelerometer Data (NCI Accelerometer Data Processing Suite).
• Instrument Architecture: Modular SAS Macros and Program Scripts (including ReadRaw.sas, WearTime.sas, SummaryDay.sas, and SummaryPerson.sas).
• Hardware Interface: Developed for the uniaxial ActiGraph Model 7164 accelerometer; forward-compatible with ActiGraph GT1M, GT3X, and subsequent models following data format alignment.
• Input Data Format: Continuous 60-second (1-minute) epoch acceleration count records indexed by participant sequence identification number (SEQN), day of the week (1–7), and minute of the day (1–1440).
• Automated Non-Wear Algorithm Parameters:
  • Non-Wear Interval Definition: ≥60 consecutive minutes of 0 counts per minute.
  • Spurious Movement Allowance: Allows up to 1 to 2 consecutive minutes of counts between 1 and 100 CPM within an otherwise unbroken run of zero counts.
  • Classification: All minutes within qualified non-wear intervals are coded as non-wear (WEAR = 0); remaining minutes are coded as valid wear (WEAR = 1).
• Daily Wear Compliance Rule:
  • A monitoring day is classified as valid if total accumulated valid wear time is ≥600 minutes (10 hours) per 24-hour cycle.
  • Days with <600 minutes of wear are flagged as invalid and excluded from standard person-level behavioral aggregation.
• Activity Intensity Cut-Point Classification (Troiano et al., 2008):
  • Sedentary Behavior: 0 to 99 counts per minute (<1.5 METs).
  • Light Physical Activity (LPA): 100 to 2019 counts per minute (1.5 to 2.9 METs).
  • Moderate Physical Activity (MPA): 2020 to 5998 counts per minute (3.0 to 5.9 METs).
  • Vigorous Physical Activity (VPA): ≥5999 counts per minute (≥6.0 METs).
  • Moderate-to-Vigorous Physical Activity (MVPA): ≥2020 counts per minute (≥3.0 METs).
• Bout Quantification Logic:
  • 10-Minute MVPA Bout: A sequence of at least 10 consecutive minutes where acceleration is ≥2020 CPM, permitting up to 2 minutes of counts falling below the 2020 CPM threshold during the 10-minute window.
  • Sedentary Bout: Consecutive uninterrupted minutes of counts <100 CPM (calculated across varying thresholds: ≥10, ≥30, or ≥60 minutes).
• Primary Output Variables (Person-Level Aggregation):
  • VALID_DAYS: Total number of days meeting the ≥10-hour wear criterion.
  • WEAR_MIN_DAY: Mean valid wear minutes per valid day.
  • MEAN_CPM: Mean total counts per minute across valid wear time.
  • SED_MIN_DAY: Mean sedentary minutes per valid day.
  • LPA_MIN_DAY: Mean light physical activity minutes per valid day.
  • MPA_MIN_DAY: Mean moderate physical activity minutes per valid day.
  • VPA_MIN_DAY: Mean vigorous physical activity minutes per valid day.
  • MVPA_MIN_DAY: Total unbouted moderate-to-vigorous physical activity minutes per day.
  • BOUT_MVPA_DAY: Minutes of MVPA accumulated exclusively in qualifying ≥10-minute bouts per day.
  • SED_BOUT_COUNT: Number of sustained sedentary bouts per day.

11. Permissions, Fee, and Test Year

The SAS Programs to Analyze NHANES Accelerometer Data were officially developed and released by the National Cancer Institute in 2007, corresponding to the public dissemination of the NHANES 2003–2004 physical activity monitor data cycle. Because the pipeline was produced by federal scientific employees of the United States Government (National Institutes of Health / National Cancer Institute), it resides in the public domain under U.S. copyright law.

The programs are provided free of charge for non-commercial research, academic, and clinical investigations. Researchers are permitted to download, execute, modify, and integrate the SAS macros into their own analytic workflows without paying licensing fees or seeking prior written permission, provided that appropriate scholarly attribution is accorded to the National Cancer Institute and the foundational validation papers (e.g., Troiano et al., 2008; Matthews et al., 2008). The source code, operational documentation, sample datasets, and updates can be downloaded directly from the official NCI Risk Factor Assessment website.

12. References

Below are primary foundational references documenting the calibration, validation, and methodological deployment of the SAS programs and NHANES accelerometry pipeline:

• Brage, S., Wedderkopp, N., Franks, P. W., Andersen, L. B., & Froberg, K. (2003). Re-examination of validity and reliability of the ActiGraph accelerometer. Medicine & Science in Sports & Exercise, 35(8), 1447–1454. https://doi.org/10.1249/01.MSS.0000079078.62035.EC
• Freedson, P. S., Melanson, E., & Sirard, J. (1998). Calibration of the Computer Science and Applications, Inc. accelerometer. Medicine & Science in Sports & Exercise, 30(5), 777–781. https://doi.org/10.1097/00005768-199805000-00021
• Healy, G. N., Matthews, C. E., Dunstan, D. W., Winkler, E. A., & Owen, N. (2011). Sedentary time and cardio-metabolic biomarkers in US adults: NHANES 2003–06. European Heart Journal, 32(5), 590–597. https://doi.org/10.1093/eurheartj/ehq451
• Matthews, C. E., Chen, K. Y., Freedson, P. S., Buchowski, M. S., Beech, B. M., Pate, R. R., & Troiano, R. P. (2008). Amount of time spent in sedentary behaviors in the United States, 2003–2004. American Journal of Epidemiology, 167(7), 875–881. https://doi.org/10.1093/aje/kwm390
• Metcalf, B. S., Curnow, J. S., Evans, C., Gardner, B. W., & Wilkin, T. J. (2002). Technical reliability of the CSA activity monitor: The EarlyBird Study. Medicine & Science in Sports & Exercise, 34(9), 1533–1537. https://doi.org/10.1097/00005768-200209000-00022
• National Cancer Institute. (2007). SAS Programs to Analyze Accelerometer Data from NHANES 2003–2004 and 2005–2006. National Institutes of Health, Division of Cancer Control and Population Sciences. https://riskfactor.cancer.gov/tools/nhanes_pam/
• Troiano, R. P., Berrigan, D., Dodd, K. W., Mâsse, L. C., Tilert, T., & McDowell, M. (2008). Physical activity in the United States measured by accelerometer. Medicine & Science in Sports & Exercise, 40(1), 181–188. https://doi.org/10.1249/mss.0b013e31815a51b3
• Yngve, M., Nilsson, A., Sjöström, M., & Ekelund, U. (2003). Effect of monitor placement and suspension on the calibration of the MTI/CSA accelerometer on a treadmill. Medicine & Science in Sports & Exercise, 35(9), 1547–1552. https://doi.org/10.1249/01.MSS.0000084534.66723.17

13. Items of the Scale