Lily Koffman
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Lily Koffman has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
Ciprian Crainiceanu
Department of Biostatistics Johns Hopkins Bloomerg School of Public Health Baltimore MD
5 shared publicationsJohn Muschelli
Department of Biostatistics, Johns Hopkins School of Public Health, Baltimore, MD, USA
5 shared publicationsAndrew Leroux
Department of Biostatistics & Informatics Colorado School of Public Health
4 shared publications- Ciprian M. Crainiceanu
Department of Biostatistics, Johns Hopkins University
3 shared publications - John Muschelli
Department of Biostatistics, Johns Hopkins School of Public Health, Baltimore, MD, USA
3 shared publications Alexander W. Levis
Department of Biostatistics, Harvard T. H. Chan School of Public Health, Boston, USA
2 shared publications
Latest Publications
- Open Access
A Head-to-Head Comparison of Three Literature Algorithms for Physical Activity Endpoints from a Wrist Accelerometer in Free-Living Conditions
Wrist accelerometers are widely used in clinical trials for continuous assessment of participants’ day-to-day physical activity and function, through derivation of metrics such as time spent in sedentary, non-sedentary, light, and moderate-to-vigorous physical activity (MVPA). Multiple algorithms have been proposed to derive these physical activity endpoints; however, the agreement across the endpoints derived from these algorithms has not been thoroughly explored. In this work, we leverage a publicly available dataset, CAPTURE-24, including wrist accelerometer data and experts’ annotations of various physical activities for 151 healthy adults, monitored at-home for one day, to compare three literature algorithms: Algorithm #1, SciKit Digital Health (SKDH), from Adamowicz et al., Algorithm #2 from Staudenmayer et al., and Algorithm #3 from Montoye et al. Algorithms’ outputs were assessed against experts’ annotations via paired t-tests, Pearson’s correlation (R), Bland–Altman analysis, accuracy, sensitivity, specificity, and F1-score. Algorithm #1 and Algorithm #2 showed stronger correlation and lower bias with experts’ annotations for time in sedentary (Algorithm #1: R = 0.91), and MVPA (Algorithm #1: R = 0.63), compared to Algorithm #3. In addition, Algorithm #1 and Algorithm #3 outperformed Algorithm #2 for time in light activity. Future work using longer monitoring periods and reference calorimetry devices should be performed to confirm the reproducibility of these findings.
Open AccessA Head-to-Head Comparison of Three Literature Algorithms for Physical Activity Endpoints from a Wrist Accelerometer in Free-Living Conditions
Wrist accelerometers are widely used in clinical trials for continuous assessment of participants’ day-to-day physical activity and function, through derivation of metrics such as time spent in sedentary, non-sedentary, light, and moderate-to-vigorous physical activity (MVPA). Multiple algorithms have been proposed to derive these physical activity endpoints; however, the agreement across the endpoints derived from these algorithms has not been thoroughly explored. In this work, we leverage a publicly available dataset, CAPTURE-24, including wrist accelerometer data and experts’ annotations of various physical activities for 151 healthy adults, monitored at-home for one day, to compare three literature algorithms: Algorithm #1, SciKit Digital Health (SKDH), from Adamowicz et al., Algorithm #2 from Staudenmayer et al., and Algorithm #3 from Montoye et al. Algorithms’ outputs were assessed against experts’ annotations via paired t-tests, Pearson’s correlation (R), Bland–Altman analysis, accuracy, sensitivity, specificity, and F1-score. Algorithm #1 and Algorithm #2 showed stronger correlation and lower bias with experts’ annotations for time in sedentary (Algorithm #1: R = 0.91), and MVPA (Algorithm #1: R = 0.63), compared to Algorithm #3. In addition, Algorithm #1 and Algorithm #3 outperformed Algorithm #2 for time in light activity. Future work using longer monitoring periods and reference calorimetry devices should be performed to confirm the reproducibility of these findings.
- 11 July 2026
- Published by Taylor & Francis in Women's Reproductive Health
- p. 1-19
Making Sense of the Cycle: Statistical Modeling Techniques for Menstrual Health Data
Approximately 1.8 billion people experience a menstrual cycle every month (Menstrual Hygiene | UNICEF, n.d.), yet gaps exist in understanding of the menstrual cycle and its influences on overall he...
- 11 July 2026
- Published by Taylor & Francis in Women's Reproductive Health
- p. 1-19
Making Sense of the Cycle: Statistical Modeling Techniques for Menstrual Health Data
Approximately 1.8 billion people experience a menstrual cycle every month (Menstrual Hygiene | UNICEF, n.d.), yet gaps exist in understanding of the menstrual cycle and its influences on overall he...
- 2 June 2026
- Published by The R Foundation
mapnhanespa: Map Quantiles for Physical Activity from 'NHANES'
Maps physical activity from the National Health and Nutrition Examination Survey ('NHANES') study into population-based quantiles.
- 2 June 2026
- Published by The R Foundation
mapnhanespa: Map Quantiles for Physical Activity from 'NHANES'
Maps physical activity from the National Health and Nutrition Examination Survey ('NHANES') study into population-based quantiles.
- Open Access
Function on Scalar Regression with Complex Survey Designs
Large health surveys increasingly collect high-dimensional functional data from wearable devices, and function on scalar regression (FoSR) is often used to quantify the relationship between these functional outcomes and scalar covariates such as age and sex. However, existing methods for FoSR fail to account for complex survey design. We introduce inferential methods for FoSR for studies with complex survey designs. The method combines fast univariate inference (FUI) developed for functional data outcomes and survey sampling inferential methods developed for scalar outcomes. Our approach consists of three steps: (1) fit survey weighted GLMs at each point along the functional domain, (2) smooth coefficients along the functional domain, and (3) use balanced repeated replication (BRR) or the Rao-Wu-Yue-Beaumont (RWYB) bootstrap to obtain pointwise and joint confidence bands for the functional coefficients. The method is motivated by association studies between continuous physical activity data and covariates collected in the National Health and Nutrition Examination Survey (NHANES). A first-of-its-kind analytical simulation study and empirical simulation using the NHANES data demonstrates that our method performs better than existing methods that do not account for the survey structure. Finally, application of the method in NHANES shows the practical implications of accounting for survey structure. The method is implemented in the R package svyfosr.
Open AccessFunction on Scalar Regression with Complex Survey Designs
Large health surveys increasingly collect high-dimensional functional data from wearable devices, and function on scalar regression (FoSR) is often used to quantify the relationship between these functional outcomes and scalar covariates such as age and sex. However, existing methods for FoSR fail to account for complex survey design. We introduce inferential methods for FoSR for studies with complex survey designs. The method combines fast univariate inference (FUI) developed for functional data outcomes and survey sampling inferential methods developed for scalar outcomes. Our approach consists of three steps: (1) fit survey weighted GLMs at each point along the functional domain, (2) smooth coefficients along the functional domain, and (3) use balanced repeated replication (BRR) or the Rao-Wu-Yue-Beaumont (RWYB) bootstrap to obtain pointwise and joint confidence bands for the functional coefficients. The method is motivated by association studies between continuous physical activity data and covariates collected in the National Health and Nutrition Examination Survey (NHANES). A first-of-its-kind analytical simulation study and empirical simulation using the NHANES data demonstrates that our method performs better than existing methods that do not account for the survey structure. Finally, application of the method in NHANES shows the practical implications of accounting for survey structure. The method is implemented in the R package svyfosr.
- Open Access
Walking Fingerprinting Using Wrist Accelerometry During Activities of Daily Living in NHANES
We propose a method for identifying individuals based on their continuously monitored wrist-worn accelerometry during activities of daily living. The method consists of three steps: (1) using Adaptive Empirical Pattern Transformation (ADEPT), a highly specific method to identify walking; (2) transforming the accelerometry time series into an image that corresponds to the joint distribution of the time series and its lags; and (3) using the resulting images to construct a person-specific walking fingerprint. The method is applied to 15,000 individuals from the National Health and Nutrition Examination Survey (NHANES) with up to 7 days of wrist accelerometry data collected at 80 Hertz. The resulting dataset contains more than 10 terabytes, is roughly 2 to 3 orders of magnitude larger than previous datasets used for activity recognition, is collected in the free living environment, and does not contain labels for walking periods. Using extensive cross-validation studies, we show that our method is highly predictive and can be successfully extended to a large, heterogeneous sample representative of the U.S. population: in the highest-performing model, the correct participant is in the top 1% of predictions 96% of the time.
Open AccessWalking Fingerprinting Using Wrist Accelerometry During Activities of Daily Living in NHANES
We propose a method for identifying individuals based on their continuously monitored wrist-worn accelerometry during activities of daily living. The method consists of three steps: (1) using Adaptive Empirical Pattern Transformation (ADEPT), a highly specific method to identify walking; (2) transforming the accelerometry time series into an image that corresponds to the joint distribution of the time series and its lags; and (3) using the resulting images to construct a person-specific walking fingerprint. The method is applied to 15,000 individuals from the National Health and Nutrition Examination Survey (NHANES) with up to 7 days of wrist accelerometry data collected at 80 Hertz. The resulting dataset contains more than 10 terabytes, is roughly 2 to 3 orders of magnitude larger than previous datasets used for activity recognition, is collected in the free living environment, and does not contain labels for walking periods. Using extensive cross-validation studies, we show that our method is highly predictive and can be successfully extended to a large, heterogeneous sample representative of the U.S. population: in the highest-performing model, the correct participant is in the top 1% of predictions 96% of the time.
- 26 November 2024
- Published by Wolters Kluwer Health in Medicine & Science in Sports & Exercise
- Vol. 57 (4) , 746-755
Comparing Step Counting Algorithms for High-Resolution Wrist Accelerometry Data in NHANES 2011–2014
Purpose: To quantify the relative performance of step counting algorithms in studies that collect free-living high-resolution wrist accelerometry data and to highlight the implications of using these algorithms in translational research. Methods: Five step counting algorithms (four open source and one proprietary) were applied to the publicly available, free-living, high-resolution wrist accelerometry data collected by the National Health and Nutrition Examination Survey (NHANES) in 2011–2014. The mean daily total step counts were compared in terms of correlation, predictive performance, and estimated hazard ratios of mortality. Results: The estimated number of steps were highly correlated (median, 0.91; range, 0.77–0.98), had high and comparable predictive performance of mortality (median concordance, 0.72; range, 0.70–0.73). The distributions of the number of steps in the population varied widely (mean step counts range from 2453 to 12,169). Hazard ratios of mortality associated with a 500-step increase per day varied among step counting algorithms between HR = 0.88 and 0.96, corresponding to a 300% difference in mortality risk reduction ([1–0.88] / [1–0.96] = 3). Conclusions: Different step counting algorithms provide correlated step estimates and have similar predictive performance that is better than traditional predictors of mortality. However, they provide widely different distributions of step counts and estimated reductions in mortality risk for a 500-step increase.
- 26 November 2024
- Published by Wolters Kluwer Health in Medicine & Science in Sports & Exercise
- Vol. 57 (4) , 746-755
Comparing Step Counting Algorithms for High-Resolution Wrist Accelerometry Data in NHANES 2011–2014
Purpose: To quantify the relative performance of step counting algorithms in studies that collect free-living high-resolution wrist accelerometry data and to highlight the implications of using these algorithms in translational research. Methods: Five step counting algorithms (four open source and one proprietary) were applied to the publicly available, free-living, high-resolution wrist accelerometry data collected by the National Health and Nutrition Examination Survey (NHANES) in 2011–2014. The mean daily total step counts were compared in terms of correlation, predictive performance, and estimated hazard ratios of mortality. Results: The estimated number of steps were highly correlated (median, 0.91; range, 0.77–0.98), had high and comparable predictive performance of mortality (median concordance, 0.72; range, 0.70–0.73). The distributions of the number of steps in the population varied widely (mean step counts range from 2453 to 12,169). Hazard ratios of mortality associated with a 500-step increase per day varied among step counting algorithms between HR = 0.88 and 0.96, corresponding to a 300% difference in mortality risk reduction ([1–0.88] / [1–0.96] = 3). Conclusions: Different step counting algorithms provide correlated step estimates and have similar predictive performance that is better than traditional predictors of mortality. However, they provide widely different distributions of step counts and estimated reductions in mortality risk for a 500-step increase.
- Open Access
Comparing Step Counting Algorithms for High-Resolution Wrist Accelerometry Data in NHANES 2011-2014
Purpose: To quantify the relative performance of step counting algorithms in studies that collect free-living high-resolution wrist accelerometry data and to highlight the implications of using these algorithms in translational research. Methods: Five step counting algorithms (four open source and one proprietary) were applied to the publicly available, free-living, high-resolution wrist accelerometry data collected by the National Health and Nutrition Examination Survey (NHANES) in 2011-2014. The mean daily total step counts were compared in terms of correlation, predictive performance, and estimated hazard ratios of mortality. Results: The estimated number of steps were highly correlated (median=0.91, range 0.77 to 0.98), had high and comparable predictive performance of mortality (median concordance=0.72, range 0.70 to 0.73). The distributions of the number of steps in the population varied widely (mean step counts range from 2,453 to 12,169). Hazard ratios of mortality associated with a 500-step increase per day varied among step counting algorithms between HR=0.88 and 0.96, corresponding to a 300% difference in mortality risk reduction ([1-0.88]/[1-0.96]=3). Conclusion: Different step counting algorithms provide correlated step estimates and have similar predictive performance that is better than traditional predictors of mortality. However, they provide widely different distributions of step counts and estimated reductions in mortality risk for a 500-step increase.
Open AccessComparing Step Counting Algorithms for High-Resolution Wrist Accelerometry Data in NHANES 2011-2014
Purpose: To quantify the relative performance of step counting algorithms in studies that collect free-living high-resolution wrist accelerometry data and to highlight the implications of using these algorithms in translational research. Methods: Five step counting algorithms (four open source and one proprietary) were applied to the publicly available, free-living, high-resolution wrist accelerometry data collected by the National Health and Nutrition Examination Survey (NHANES) in 2011-2014. The mean daily total step counts were compared in terms of correlation, predictive performance, and estimated hazard ratios of mortality. Results: The estimated number of steps were highly correlated (median=0.91, range 0.77 to 0.98), had high and comparable predictive performance of mortality (median concordance=0.72, range 0.70 to 0.73). The distributions of the number of steps in the population varied widely (mean step counts range from 2,453 to 12,169). Hazard ratios of mortality associated with a 500-step increase per day varied among step counting algorithms between HR=0.88 and 0.96, corresponding to a 300% difference in mortality risk reduction ([1-0.88]/[1-0.96]=3). Conclusion: Different step counting algorithms provide correlated step estimates and have similar predictive performance that is better than traditional predictors of mortality. However, they provide widely different distributions of step counts and estimated reductions in mortality risk for a 500-step increase.
- 29 July 2024
- Published by Oxford University Press (OUP) in Journal of the Royal Statistical Society Series C: Applied Statistics
- Vol. 73 (5) , 1221-1241
Walking fingerprinting
We consider the problem of predicting an individual’s identity from accelerometry data collected during walking. In a previous paper, we transformed the accelerometry time series into an image by constructing the joint distribution of the acceleration and lagged acceleration for a vector of lags. Predictors derived by partitioning this image into grid cells were used in logistic regression to predict individuals. Here, we (a) implement machine learning methods for prediction using the grid cell-derived predictors; (b) derive inferential methods to screen for the most predictive grid cells while adjusting for correlation and multiple comparisons; and (c) develop a novel multivariate functional regression model that avoids partitioning the predictor space. Prediction methods are compared on two open source acceleometry data sets collected from: (a) 32 individuals walking on a 1.06 km path; and (b) six repetitions of walking on a 20 m path on two occasions at least 1 week apart for 153 study participants. In the 32-individual study, all methods achieve at least 95% rank-1 accuracy, while in the 153-individual study, accuracy varies from 41% to 98%, depending on the method and prediction task. Methods provide insights into why some individuals are easier to predict than others.
- 29 July 2024
- Published by Oxford University Press (OUP) in Journal of the Royal Statistical Society Series C: Applied Statistics
- Vol. 73 (5) , 1221-1241
Walking fingerprinting
We consider the problem of predicting an individual’s identity from accelerometry data collected during walking. In a previous paper, we transformed the accelerometry time series into an image by constructing the joint distribution of the acceleration and lagged acceleration for a vector of lags. Predictors derived by partitioning this image into grid cells were used in logistic regression to predict individuals. Here, we (a) implement machine learning methods for prediction using the grid cell-derived predictors; (b) derive inferential methods to screen for the most predictive grid cells while adjusting for correlation and multiple comparisons; and (c) develop a novel multivariate functional regression model that avoids partitioning the predictor space. Prediction methods are compared on two open source acceleometry data sets collected from: (a) 32 individuals walking on a 1.06 km path; and (b) six repetitions of walking on a 20 m path on two occasions at least 1 week apart for 153 study participants. In the 32-individual study, all methods achieve at least 95% rank-1 accuracy, while in the 153-individual study, accuracy varies from 41% to 98%, depending on the method and prediction task. Methods provide insights into why some individuals are easier to predict than others.