Maintenance Notice

Due to necessary scheduled maintenance, the JMIR Publications website will be unavailable from Monday, March 11, 2019 at 4:00 PM to 4:30 PM EST. We apologize in advance for any inconvenience this may cause you.

Who will be affected?

Advertisement

Currently submitted to: JMIR mHealth and uHealth

Date Submitted: Jan 27, 2020
(currently open for review)

Data imputation and body weight variability calculation using linear and non-linear methods in data collected from digital smart scales: a simulation and validation study

  • Jake Turicchi; 
  • Ruairi O'Driscoll; 
  • Graham Finlayson; 
  • Cristiana Duarte; 
  • Antonio L Palmeira; 
  • Sofus Larsen; 
  • Berit L Heitmann; 
  • James Stubbs; 

ABSTRACT

Background:

Body weight variability is common in the general population and may act as a risk factor for obesity or disease. Correct identification of these patterns may have prognostic or predictive value in clinical and research settings. With advancements in technology allowing for frequent collection of body weight data from electronic smart scales, new opportunities to analyse and identify patterns in body weight data are available.

Objective:

The aim of the present paper is to compare multiple methods of data imputation and body weight variability calculation using linear and non-linear approaches.

Methods:

Fifty participants from an ongoing weight loss maintenance study (the NoHoW study) were used to develop the procedure. We addressed the following aspects of data analysis: cleaning, imputation, detrending and calculation of total and local body weight variability. To test imputation, missing data was simulated (i) at random and (ii) using real patterns of missingness. Ten imputation strategies were tested. Next, body weight variability was calculated using linear and non-linear approaches, and the effect of missing data and data imputation on these estimates was investigated.

Results:

Body weight imputation using structural modelling with Kalman smoothing or an exponentially weighted moving average provided the best agreement with observed values. Imputation performance decreased with missingness and was similar between random and non-random simulations. Biases in BWV estimations from missing-simulated data sets were lower than those produced by imputed data sets. Accuracy of BWV estimates were greater in linear than non-linear estimations under conditions of missing data.

Conclusions:

The decision to impute body weight data depends on the purpose of the analysis, and direction for the best-performing methods are provided. For the purposes of estimating body weight variability, data imputation should not be conducted. Linear and non-linear methods of estimating body weight variability provide reasonably accurate estimates under high proportions (80%) of missing data. Clinical Trial: The trial is registered with the ISRCTN registry (ISRCTN88405328).


 Citation

Please cite as:

Turicchi J, O'Driscoll R, Finlayson G, Duarte C, Palmeira AL, Larsen S, Heitmann BL, Stubbs J

Data imputation and body weight variability calculation using linear and non-linear methods in data collected from digital smart scales: a simulation and validation study

JMIR Preprints. 27/01/2020:17977

DOI: 10.2196/17977

URL: https://preprints.jmir.org/preprint/17977


Request queued. Please wait while the file is being generated. It may take some time.