PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • The study was approved by the University of Oxford Central University Research Ethics Committee. Recruitment started in January 2015 and this paper presents an analysis of data collected from January to November 2015. Participants were recruited from the general public on a rolling basis, through a number of recruitment channels. Study launch was timed to coincide with a television programme on weight loss airing nationally throughout the UK; the associated website for the television programme contained advertisements for the study, which were the initial driver for recruitment. In addition, the study was publicised on social media, to members of existing studies, and via a local health improvement service. Participants were encouraged to share information on the study with family and friends. The system was open to anyone who wished to register and was over the age of 18, but to be included in the present study, participants were required to be overweight adults (BMI greater than or equal to 25 kg/m2, aged 18 or older), resident in the UK, and trying to lose weight through changes to their diet, physical activity, or both. There were no recruitment or retention incentives beyond sharing preliminary findings with participants once the analysis was completed. A bespoke, cross-platform mobile phone application and a study website (www.oxfab.org) were created to collect data. Participant consent was obtained electronically through the same system. After consenting, participant demographics (age, gender, ethnicity, highest level of education obtained) and self-reported weight and height were collected. Participants were then asked to log in episodically to indicate which specific strategies they were using and to report their weight. Participants could register to receive e-mail prompts to log in to the website on a daily, weekly, or fortnightly basis, or choose to opt out of prompts. If, at three months from baseline, participants had not yet reported a post-baseline weight they were prompted to do so (in order to continue to use the system to log other information). Both the website and the mobile phone application allowed participants to visualise their weight progression on a graph and receive simplified statistics about the strategies they reported using. The data set upon which this analysis is based is publicly available via the Oxford University Research Archive. [15] The OxFAB questionnaire was used to establish which strategies were being employed by participants to manage their weight. At baseline, participants were asked three screening questions to rule out questions that were irrelevant to them. Accordingly, participants who were not using diet to control weight were not asked questions specific to diet, participants who were not using physical activity to control their weight were not asked questions specific to physical activity, and participants who did not shop for their own food were not asked questions about food shopping. After irrelevant questions had been screened out, participants were asked all remaining questions from the OxFAB questionnaire in a randomly generated order. This questionnaire consists of 117 questions grouped into 21 domains (see Table 1) and is based on the OxFAB taxonomy, which was constructed through a qualitative analysis of existing resources and a review of existing behaviour change taxonomies and theories. The OxFAB questionnaire has been previously assessed for reliability and validity; details on its development and the full list of questions are available elsewhere. [9] For each question, participants were asked to indicate if they used the strategy most of the time, sometimes, or never/hardly ever. ‘Not relevant’ was also provided as an answer option, and participants had the facility to mark individual questions as unclear. Questions were presented in batches of 10 and participants were given the possibility to continue with the next batch or stop. Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0202072.t001 OxFAB taxonomy domains and definitions. Analyses were conducted using Stata v11. [16] Participants were excluded if they had missing data for one or more of the strategies relevant to them or if, when calculated using self-reported weight and height data, their BMI at baseline was less than 25. For analyses of associations between strategy use and weight change at three months, participants were excluded if they had not provided a self-reported weight at 75 to 105 days from baseline. Baseline differences between those who completed all questions and those who did not were calculated using unpaired t-tests for continuous variables and using Pearson’s chi square tests for binary and categorical variables. In the primary analyses, strategy use was defined dichotomously, with ‘most of the time’ and ‘sometimes’ considered as use of that strategy, and ‘never’ and ‘not relevant to me’ considered as not using that strategy. Domain use was evaluated in two predefined ways: as percentage of strategies used within each domain, and as use of at least one strategy within a domain. These two approaches were used as domains contained varying numbers of strategies. Data-driven patterns of strategy use were derived by exploratory factor analysis for the previously described binary strategy use variables using a tetrachoric correlation matrix with orthogonal rotation (varimax option) to enhance factor identification and interpretability. Factors were retained based on eigenvalues >1 and the proportion of variance explained. For each factor identified, every strategy has a factor loading which is the correlation coefficient between each strategy and the factor. Only the strategies with factor loadings (z-score)>0.2 were retained and extracted in the factors identified. [17] For each factor identified, factor scores were derived for each participant, with higher scores indicating higher adherence to that factor (e.g. higher usage of the strategies contained in each factor). Factors will be referred as patterns of strategies from now on. Model of essential strategies for weight loss: In addition to the factor analysis, we conducted a non-data driven exploratory analysis testing an a priori model which included only those strategies that were deemed essential to weight management. This strategy model was created by the authors based on existing knowledge. [18, 19] It contained nine components: food and weight targets; a strategy to enhance motivation to lose weight; advance meal planning; monitoring of food intake; swapping less healthy foods for healthier ones; keeping unhealthy food out of the house; an impulse management strategy to employ when cues to eat are presented; and self-weighing. Each participant was given a score of 0–9 based on whether or not they used a strategy within that component ‘most of the time’; each component of the model was equally weighted (use of that component = 1). A linear multivariable regression model was used to test associations between demographic characteristics and each pattern, and between demographic characteristics and domain use. A linear regression model was used to explore associations between strategy, pattern and domain use and weight at three months controlling for baseline weight. [20] We then tested whether participants’ scores for the model of essential strategies for weight loss were associated with weight at three months controlling for baseline weight using a linear regression model with a continuous scale (0 to 9; with one point for each strategy used within the model) and a binary scale (use of seven or more strategies within this model; this threshold was determined as it represented high use of relevant strategies (mean score 5.4)). For each analysis, all model assumptions for using parametric tests were checked and met. All analyses were exploratory and p values should be taken as indicative.
rdf:type