Research
WIT develops statistical methods for the very large, high-resolution, and heterogeneous data produced by wearable and implantable sensors, and applies them in population and clinical studies. Below are the group’s main research directions, with examples of recent work.
Walking, step counting, and fingerprinting in NHANES
Can an accelerometer tell whether someone is walking, and how? This question started the group. Our current work applies it at national scale using the raw 80 Hz wrist accelerometry from the National Health and Nutrition Examination Survey (NHANES) 2011–2014, worn for up to seven days by nearly 15,000 participants.
Step counting. We benchmarked open-source and proprietary step counting algorithms on sub-second wrist accelerometry, first against publicly available data sets with ground-truth steps, then across the full NHANES sample.
- Koffman L, Muschelli J. Evaluating step counting algorithms on subsecond wrist-worn accelerometry: a comparison using publicly available data sets. Journal for the Measurement of Physical Behaviour, 2024. doi:10.1123/jmpb.2024-0008
- Koffman L, Crainiceanu C, Muschelli J. Comparing step counting algorithms for high-resolution wrist accelerometry data in NHANES 2011–2014. Medicine & Science in Sports & Exercise, 2025. doi:10.1249/MSS.0000000000003616 (preprint)
Walking fingerprinting. Each person’s walking has a signature. We identify individuals from short segments of wrist accelerometry during walking, first in controlled studies and then in free-living NHANES data.
- Koffman L, Crainiceanu C, Leroux A. Walking fingerprinting. Journal of the Royal Statistical Society Series C, 2024.
- Koffman L, Muschelli J, Crainiceanu C. Walking fingerprinting using wrist accelerometry during activities of daily living in NHANES. Preprint, 2025.
Physical activity and health in large cohort studies
Objectively measured physical activity is among the strongest predictors of health. We analyze accelerometry from national surveys, population cohorts, and trials to quantify how activity, sleep, and daily rhythms relate to mortality, dementia, Parkinson’s disease, falls, and depression.
UK Biobank
A subset of UK Biobank participants wore a wrist accelerometer for up to seven days.
- Leroux A, Xu S, Kundu P, Muschelli J, Smirnova E, Chatterjee N, Crainiceanu C. Quantifying the predictive performance of objectively measured physical activity on mortality in the UK Biobank. Journals of Gerontology, Series A, 2021. Among 82,304 participants over 50, accelerometry measures predicted mortality better than every traditional risk factor except age.
- Zhao A, Cui E, Leroux A, Lindquist MA, et al. Evaluating the prediction performance of objective physical activity measures for incident Parkinson’s disease in the UK Biobank. Journal of Neurology, 2023. Total acceleration was the best single predictor of incident Parkinson’s disease, ahead of age.
- Zhao A, Cui E, Leroux A, Zhou X, et al. Objectively measured physical activity using wrist-worn accelerometers as a predictor of incident Alzheimer’s disease in the UK Biobank. Journals of Gerontology, Series A, 2025. Among 42,157 adults aged 65 and older, moderate-to-vigorous activity was the second-best predictor of Alzheimer’s disease after age.
NHANES
The National Health and Nutrition Examination Survey used hip-worn accelerometers in 2003–2006 and high-resolution wrist accelerometers in 2011–2014. See also our NHANES step counting work.
- Leroux A, Di J, Smirnova E, McGuffey EJ, et al. Organizing and analyzing the activity data in NHANES. Statistics in Biosciences, 2019. A processing pipeline and tutorial for NHANES accelerometry.
- Smirnova E, Leroux A, Cao Q, Tabacu L, Zipunnikov V, Crainiceanu C, Urbanek JK. The predictive performance of objective measures of physical activity derived from accelerometry data for 5-year all-cause mortality in older adults: NHANES 2003–2006. Journals of Gerontology, Series A, 2020.
- Agarwala N, Zipunnikov V, Crainiceanu C, Leroux A. Quantifying the time-varying association between objectively measured physical activity and mortality in US older adults over a 12-year follow-up period. BMJ Evidence-Based Medicine, 2024. The association weakens over follow-up, so baseline activity is best suited to short- and medium-term risk prediction.
- Leroux A, Cui E, Smirnova E, Muschelli J, Schrack JA, Crainiceanu CM. NHANES 2011–2014: objective physical activity is the strongest predictor of all-cause mortality. Medicine & Science in Sports & Exercise, 2024. In 3,653 adults aged 50–80, the top three mortality predictors were all activity measures.
ARIC
Older participants in the Atherosclerosis Risk in Communities Study wore wrist-worn ActiGraph devices at 80 Hz, and others wore the chest-worn Zio XT ECG patch, which contains an accelerometer.
- Gao S, Zhou X, Koffman L, Wanigatunga AA, Schrack JA, Crainiceanu CM, Muschelli J. Comparing step-counting algorithms for high-resolution wrist accelerometry data in older adults in the ARIC study. Journals of Gerontology, Series A, 2025. Open-source step counting algorithms were highly correlated but gave very different step totals.
- Wang W, Wanigatunga AA, Etzkorn LH, et al., including Zipunnikov V and Crainiceanu C. Association between circadian rest-activity rhythms and incident dementia in older adults: the ARIC study. Neurology, 2026. Weaker day–night contrast in activity was associated with higher dementia risk.
- Etzkorn LH, Davoudi A, Dooley EE, Gabriel KP, et al. Physical activity cut-points for older adults using the Zio XT onboard accelerometer. BMC Digital Health, 2024.
- See also the ARIC sleep and dementia study using the Zio XT accelerometer.
BLSA
The Baltimore Longitudinal Study of Aging collects 24-hour, minute-level accelerometry in community-dwelling adults.
- Schrack JA, Zipunnikov V, Goldsmith J, Bai J, et al. Assessing the “physical cliff”: detailed quantification of age-related differences in daily patterns of physical activity. Journals of Gerontology, Series A, 2014.
- Wanigatunga AA, Di J, Zipunnikov V, Urbanek JK, et al. Association of total daily physical activity and fragmented physical activity with mortality in older adults. JAMA Network Open, 2019. More fragmented activity predicted mortality, while total activity did not.
- Di J, Spira A, Bai J, Urbanek J, Leroux A, et al., Zipunnikov V. Joint and individual representation of domains of physical activity, sleep, and circadian rhythmicity. Statistics in Biosciences, 2019.
STURDY
The Study to Understand Fall Reduction and Vitamin D in You (STURDY) is a vitamin D trial in adults aged 70 and older at elevated risk of falls, who wore wrist accelerometers 24 hours a day for seven days.
- Urbanek JK, Roth DL, Karas M, Wanigatunga AA, et al. Free-living gait cadence measured by wearable accelerometer: a promising alternative to traditional measures of mobility for assessing fall risk. Journals of Gerontology, Series A, 2023. Higher free-living walking cadence was associated with fewer falls, while clinic-based gait tests were not.
- Schrack JA, Cai Y, Urbanek JK, Wanigatunga AA, et al. The association of vitamin D supplementation and serum vitamin D levels with physical activity in older adults: results from a randomized trial. Journal of the American Geriatrics Society, 2023.
Other cohorts
- CoLaus|PsyCoLaus (depression): Kang SJ, Leroux A, Guo W, Dey D, et al., Zipunnikov V. Integrative modeling of accelerometry-derived sleep, physical activity, and circadian rhythm domains with current or remitted major depression. JAMA Psychiatry, 2024.
- Multicenter AIDS Cohort Study: Etzkorn LH, Liu L, Urbanek JK, Heravi AS, et al. Patterns of objectively measured physical activity differ between men living with and without HIV. AIDS, 2022.
Software: actimetrics, mapnhanespa, actiquantiles, svyfosr
Functional data methods
Minute-level and sub-second sensor data are naturally functions of time. We develop functional data analysis methods that scale to tens of thousands of participants and respect the structure of the data.
- Li, Lindquist, Gunning, Crainiceanu. Functional moments regression. Preprint, 2026.
- Lu, Zhou, Cui, Rogers, Crainiceanu, Wrobel, Leroux. Generalized conditional functional principal component analysis. Preprint, 2024.
Continuous glucose monitoring
Continuous glucose monitors record blood glucose every few minutes. Standard summaries such as mean glucose and time in range miss the shape of glycemic variability, so we develop measures and models that capture it.
- Sartini, Fang, Rooney, Selvin, Coresh, Zeger. Glucose Color Index: development and validation of a novel measure of the shape of glycemic variability. Journal of Diabetes Science and Technology.
- Sartini, Matabuena, et al. Beyond scalar metrics: functional data analysis of postprandial continuous glucose monitoring in the AEGIS study. BMC Medical Research Methodology.
- Sartini J, Zhou X, Selvin E, Zeger S, Crainiceanu CM. Fast Bayesian functional principal components analysis. Journal of Computational and Graphical Statistics, 2026. A fully Bayesian FPCA (FAST) that accounts for uncertainty in the estimated components, applied to mealtime glucose variability in the DASH4D continuous glucose monitoring study. (preprint)
Sleep, circadian rhythms, and wear time
Wearables measure sleep and daily rhythms passively over many days. We develop methods to estimate time in bed, sleep, and wear time from devices worn on the wrist, hip, and chest, and relate these measures to long-term health.
- Chamberlin KW, Etzkorn LH, Spira AP, Jiang K, Chen LY, Crainiceanu C, Zipunnikov V, Schrack JA. Novel sleep metrics derived from chest-worn accelerometers are associated with dementia risk: the Atherosclerosis Risk in Communities neurocognitive study. SLEEP Advances, 2026. An open-source method estimates time in bed and sleep from the accelerometer in the chest-worn Zio XT ECG patch, using posture to separate sedentary time from time in bed. In 2,316 ARIC participants, time in bed and total sleep time had U-shaped associations with dementia risk.
- Lin Z, Reid MJ, Davoudi A, Nestadt P, Urbanek J, Di J, Zipunnikov V, Rojo-Wissar DM, Spira AP. Chronotype and all-cause mortality in US middle-aged and older adults: results from the NHANES. Journal of Sleep Research, 2026. Behavioral chronotype, estimated from the midpoint of time in bed in hip-worn accelerometry, and its association with mortality in 2,261 NHANES participants aged 50 and older.
Hemodynamics in cardiac surgery
During cardiac surgery, monitors record mean arterial pressure (MAP), central venous pressure (CVP), and cardiac index every minute. With clinicians in Johns Hopkins Anesthesiology and Critical Care Medicine, we study how the time patients spend in joint ranges of these measures relates to complications after surgery. WIT students work directly with the surgical teams to understand how the data are collected.
- Goeddel LA, Koffman L, Zhou X, Hernandez M, Zhao A, Pasrija C, Saha AK, Muschelli J, Parikh CR, Crainiceanu CM, Brown C, Khanna AK, Faraday N. Low cardiac index during periods of arterial hypotension and risk of acute kidney injury in cardiac surgery. British Journal of Anaesthesia, 2026. In 1,272 coronary artery bypass patients, each 5 minutes of hypotension combined with a low cardiac index (≤2 L/min/m²) was associated with an 11% higher risk of acute kidney injury and longer ICU and hospital stays. Hypotension with a normal cardiac index was not.
- Goeddel LA, Gomez J, Hernandez M, Koffman L, Murphy Z, Wayson G, Whitman G, Bandeen-Roche K, Muschelli J, Crainiceanu CM, Faraday N, Brown CH. Association of postoperative delirium with haemodynamic determinants of cerebral perfusion pressure during cardiac surgery: a retrospective cohort study. British Journal of Anaesthesia, 2026. In 1,759 patients, time in a zone of low cerebral perfusion pressure (low MAP with high CVP) was associated with postoperative delirium.
Software: Hemodynamics-and-Urinary-Biomarkers
Other physiologic sensing
We also work with ECG patches and other physiologic waveforms, for example classifying posture from the accelerometer inside chest-worn ECG patches.
Software: wfdb
Open, reproducible software
We release the tools behind our research so others can inspect, reuse, and extend them, from reading raw device files to computing activity metrics. See About & Projects and the jhuwit GitHub organization.