The presence of measurement error is considered to be an inevitable condition associated with self-reported diet assessment and the role of measurement error in attenuating or distorting the association between diet and disease risk is well understood. In lifestyle intervention trials, where the goal is to change a participant's weigt or modify their eating behavior, self-reported diet is typically an outcome variable that is measured repeatedly throughout the intervention. In this setting, measurement error can affect the estimation of intervention effects by reducing the power to detect a treatment effect as well as by biasing estimates of treatment effectiveness. As a result, measurement error in intervention studies has handicapped the development of effective interventions aimed at changing and maintaining healthy behaviors. In response to PAR-09-224: Improving Diet and Physical Activity Assessment, we propose to develop a statistical framework to correct for measurement error in self-reported dietary data from longitudinal lifestyle intervention trials where objective validation data do not exist. We have obtained four validation studies that contain both self-reported and objective measures (i.e. recovery biomarkers) of dietary intake. Using these data sets, we will estimate the relationship between self-reported and objective measures of diet and borrow this information in order to correct for measurement error in longitudinal lifestyle intervention trials that only include self-reported dietary measures. Our approach uses a missing data framework that views unmeasured objective data as missing data. There are a number of advantages to this approach including: 1) It allows us draw upon the many computational and statistical methods for handling missing data; 2) It facilitates the use of sensitivity analysis to address the effect of unverifiable measurement error assumptions on subsequent inferences; 3) Corrected measurements of diet can be imputed so that users of measurement error-corrected data sets can use standard statistical methods to perform their analyses. The overall goal of this project is to develop a statistical framework for correcting for measurement error in longitudinal self-reported dietary data which makes use of external validation data.
Our specific aims are: 1) Investigate the implications of measurement error in self- reported dietary outcomes when performing longitudinal analyses to estimate treatment effects; 2) Develop and assess a statistical framework to correct for measurement error in longitudinal lifestyle interventions where outcomes are measured with error and measurement error may vary over time and can differ between treatment groups; 3) Develop and assess methods for combining external validation studies with intervention trials using propensity score methods. This work will allow researchers to more accurately and precisely measure the effects of a lifestyle intervention and its mechanisms. This information will facilitate the development of more effective interventions to improve the diet of at-risk populations.
This study will develop and test novel statistical methods to correct for measurement errors in self-reported diet data from longitudinal lifestyle interventions designed to improve eating behavior. The results of this study have the potential to make notable methodological and substantive contributions for the analysis of self-reported dietary data so that the effects of lifestyle interventions can be estimated more accurately and precisely. Given the detrimental health effects and enormous costs of a poor quality diet, such information could be of significant value to efforts to improve public health.
| Welch, Whitney A; Spring, Bonnie; Phillips, Siobhan M et al. (2018) Moderating Effects of Weather-Related Factors on a Physical Activity Intervention. Am J Prev Med 54:e83-e89 |
| Brincks, Ahnalee; Montag, Samantha; Howe, George W et al. (2018) Addressing Methodologic Challenges and Minimizing Threats to Validity in Synthesizing Findings from Individual-Level Data Across Longitudinal Randomized Trials. Prev Sci 19:60-73 |
| Siddique, Juned; de Chavez, Peter J; Howe, George et al. (2018) Limitations in Using Multiple Imputation to Harmonize Individual Participant Data for Meta-Analysis. Prev Sci 19:95-108 |
