MPADD
Measurement of Physical Activity in older adults through Data Donation
Engagement in physical activity (PA) is the foundation of a healthy lifestyle, elevated immune and psychological function, and decreased mortality, especially for aging populations. The accurate measurement of PA is key to identifying determinants of health and developing appropriate interventions. To measure PA, most population studies use self-report. However, self-reports are usually limited to global measures of PA (e.g., average daily hours of moderate/vigorous activity, sedentary behavior) and suffer from misclassification (e.g., walking the dog not considered PA). More fine-grained self-report needs day-reconstruction methods, which are limited to short reference periods (e.g., yesterday or last week), burdensome for respondents, and prone to recall error. As an alternative, researchers are providing study participants with wearable devices that passively track PA, which reduces reactivity and error due to forgetting. However, issues of this approach pertain to non-compliance and high device costs.
In this project, we propose an innovative approach to collecting PA data from adults 50 years and older through data donation. Study participants are asked to download PA data from devices they already own (e.g., smartphones, smart watches, fitness bracelets) and share them with the researchers. This approach leverages the advantages of passive data collection by providing detailed, high-frequency information. In contrast to studies where loaner devices can only be given to participants for a short period, data donation allows to study PA in a true longitudinal setting. Data donation is also a cost-efficient way of collecting data, given that participants use their own devices for the study. However, very little is known about the biases in PA estimates that arise from selective nonparticipation in data donation among aging populations. The project addresses three aims in a data donation study on PA among older adults.
Aim 1. Investigate determinants of consent and selection bias. We will conduct a baseline online survey with 2,000 adults 50+ in a probability-based online panel in the Netherlands (LISS). Smartphone owners in that sample will be asked to download PA data from their devices and share them via a privacy-preserving data donation infrastructure that will be adapted for PA data specifically. Based on a comparison between those who do and do not donate their data, we will (1A) identify correlates of consent to data donation. Self-reported measures of PA from the survey will allow us to (1B) estimate bias in these measures.
Aim 2. Assess the quality of donated PA data. We compare self-reported PA data from the survey with the PA measures in the donated data to (2A) identify differences in the two data sources. Using an Multitrait-Multimethod (MTMM) approach, we will (2B) jointly study measurement error in the two data sources. To identify threats to validity of donated PA data (2C), we will examine patterns of smartphone use and how they correlate with the outcome measures.
Aim 3. Use multi-source PA data to predict health outcomes. To determine the predictive power of the different sources of PA data (donated: Apple Health, Samsung Health, Google Location History; self-report: survey), we regress health outcomes from the LISS core health module on the measures generated from the different PA sources.
Understanding PA patterns of older adults in their natural setting will help inform better, more targeted public policies. Government agencies collect information on PA to determine whether citizens adhere to the WHO norms. Donated data can achieve cost savings and better quality compared to traditional self-reports. This study contributes to developing future-proof methods of collecting high-quality PA data.
Collaborators: Bella Struminskaya
Funding: Network for Innovative Methods in Longitudinal Aging Studies (NIMLAS)
Return to Research page
In this project, we propose an innovative approach to collecting PA data from adults 50 years and older through data donation. Study participants are asked to download PA data from devices they already own (e.g., smartphones, smart watches, fitness bracelets) and share them with the researchers. This approach leverages the advantages of passive data collection by providing detailed, high-frequency information. In contrast to studies where loaner devices can only be given to participants for a short period, data donation allows to study PA in a true longitudinal setting. Data donation is also a cost-efficient way of collecting data, given that participants use their own devices for the study. However, very little is known about the biases in PA estimates that arise from selective nonparticipation in data donation among aging populations. The project addresses three aims in a data donation study on PA among older adults.
Aim 1. Investigate determinants of consent and selection bias. We will conduct a baseline online survey with 2,000 adults 50+ in a probability-based online panel in the Netherlands (LISS). Smartphone owners in that sample will be asked to download PA data from their devices and share them via a privacy-preserving data donation infrastructure that will be adapted for PA data specifically. Based on a comparison between those who do and do not donate their data, we will (1A) identify correlates of consent to data donation. Self-reported measures of PA from the survey will allow us to (1B) estimate bias in these measures.
Aim 2. Assess the quality of donated PA data. We compare self-reported PA data from the survey with the PA measures in the donated data to (2A) identify differences in the two data sources. Using an Multitrait-Multimethod (MTMM) approach, we will (2B) jointly study measurement error in the two data sources. To identify threats to validity of donated PA data (2C), we will examine patterns of smartphone use and how they correlate with the outcome measures.
Aim 3. Use multi-source PA data to predict health outcomes. To determine the predictive power of the different sources of PA data (donated: Apple Health, Samsung Health, Google Location History; self-report: survey), we regress health outcomes from the LISS core health module on the measures generated from the different PA sources.
Understanding PA patterns of older adults in their natural setting will help inform better, more targeted public policies. Government agencies collect information on PA to determine whether citizens adhere to the WHO norms. Donated data can achieve cost savings and better quality compared to traditional self-reports. This study contributes to developing future-proof methods of collecting high-quality PA data.
Collaborators: Bella Struminskaya
Funding: Network for Innovative Methods in Longitudinal Aging Studies (NIMLAS)
Return to Research page