Weight-Biased Language across 30 Years of Australian News Reporting on Obesity: Associations with Public Health Policy
Round 1
Reviewer 1 Report
The manuscript by Grant el al., is an interesting study in which they exhaustively utilized a very large database generated for almost 3 decades to analyze the representation of obesity and public health policy and its association with gender, healthiness, social status, and negative stereotypes by using machine learning and computational language analysis approach. This study highlights a very relevant topic of obesity and weight stigma in the society, particularly when the research have shown that “metabolically obese normal weight” people still exhibit significant risk of developing cardiovascular and other metabolic related disorders, regardless of having normal weight and BMI. This study deals with question that needs important discussion for obesity-related public health policy development.
Author Response
Reviewer 1:
General comments: The manuscript by Grant el al., is an interesting study in which they exhaustively utilized a very large database generated for almost 3 decades to analyze the representation of obesity and public health policy and its association with gender, healthiness, social status, and negative stereotypes by using machine learning and computational language analysis approach. This study highlights a very relevant topic of obesity and weight stigma in the society, particularly when the research have shown that “metabolically obese normal weight” people still exhibit significant risk of developing cardiovascular and other metabolic related disorders, regardless of having normal weight and BMI. This study deals with question that needs important discussion for obesity-related public health policy development.
No comments to address (‘Yes’ ticked for all assessment criteria). We thank Reviewer 1 for their encouraging feedback.
Reviewer 2 Report
The authors present a manuscript which evaluates weight-biased language across 30 years of Australian news reporting on obesity: Associations with public health policy. This is interesting topic that, although based on restricted area exploration, may have much broader significance in analyses that reveal worldwide social trends, as a whole.
I would like the authors to emphasize the impact of the analyzed trends by means of mental disorders, at least to those with significant clinical and medical importance. Therefore, I suggest exploring, using the same methodology, the data regarding the associations of weight-biased language with eating disorders (the incidence of anorexia nervosa and bulimia) and mood disorders (the incidence of depression and anxiety, etc.).
Legend for Figure 2 is missing.
Author Response
Reviewer 2:
General comments: The authors present a manuscript which evaluates weight-biased language across 30 years of Australian news reporting on obesity: Associations with public health policy.
Comment 1: This is interesting topic that, although based on restricted area exploration, may have much broader significance in analyses that reveal worldwide social trends, as a whole.
Response: Thank you for your encouraging feedback on the potential reach of the findings. We have added at statement to this effect in the Discussion:
Although based on a restricted area of exploration, Australian print news media only, our findings may have much broader significance for worldwide social trends and prompt the need for ongoing analysis of media reporting of obesity and weight-related public health policy. Future research could also extend our word embedding analysis to policy texts themselves, to draw direct correlations between media and policy data sources.
Comment 2: I would like the authors to emphasize the impact of the analyzed trends by means of mental disorders, at least to those with significant clinical and medical importance. Therefore, I suggest exploring, using the same methodology, the data regarding the associations of weight-biased language with eating disorders (the incidence of anorexia nervosa and bulimia) and mood disorders (the incidence of depression and anxiety, etc.).
Response: We agree that analysis of associations between weight-biased language in news media and the prevalence of eating disorders and other mental disorders correlated with weight stigma would be informative to the weight stigma literature, however our input data for the natural language processing (NLP) were textual data only and do not take into account other data modalities such as images and/or clinical data (floating numbers). In other words, NLP is a text-analytical tool to understand the nuances of human language about a certain topic (obesity in our case). This is done by capturing the contextual relationships between words and sentences in text corpus.
Furthermore, this suggestion goes beyond the scope of the present paper, which focuses on associations between language biases tied to individual and structural dimensions of obesity and changes in public health policy rather than associations between language biases and changes in mental disorder prevalence. An alternative approach with the aid of NLP, would be to add mental disorders as a dimension in the analyses, but this would require a comprehensive literature review to make sure the mental disorder keywords, and their dichotomous mappings, were inclusive. Given the turnaround time for the revision (10 days), unfortunately we cannot extend the analysis in this way, but we absolutely agree with Reviewer 2 that this is an important and interesting direction for future research that can be achieved with further application of the techniques we have developed for this paper. We have acknowledged investigation of the association between news reporting on obesity and mental disorders as a fruitful direction for future research in the Discussion as follows:
It is also important to examine relationships between news media reporting of obesity and health outcomes over time given, for example, medium to large meta-analytic associations between weight stigma and mental health disorders such as anxiety, depression, eating disorders and other psychopathological symptoms [new citation to be added and numbered accordingly – see below].
Emmer, C., Bosnjak, M., Mata, J. The association between weight stigma and mental health: A meta-analysis. Obesity Reviews 2020, 21:e12935. doi:10.1111/obr.12935
Comment 3: Legend for Figure 2 is missing.
Response: We have provided new versions of the figure as a separate file to support editing/reproduction.
Reviewer 3 Report
The manuscript reports a study about weight-biased language in the Australian media across the last decades. The topic is interesting, and the manuscript is clear. The methods applied are innovative and allow to underline new interesting evidence.
I do not have particular concerns about the paper. I have some comments for the authors about the structure of the manuscript that should take into consideration:
- please move the research questions from the methods to the introduction
- last paragraph of the methods: why did you report here your conclusions? Please move this part from this section.
- Please revised Figure 1 because it's very difficult to read.
- A recent paper has pointed out that postbariatric patients, using a novel approach based on words, presented a weight bias regards their body, but it was not present for other bodies (see http://dx.doi.org/10.1007/s11695-020-05166-z). I think this aspect is interesting and linked to your results, showing a connection with clinical data.
- You report that an automatic approach was used to check all the papers. Is there any possible limit with this approach in the selection of the papers?
Author Response
Reviewer 3:
General comments: The manuscript reports a study about weight-biased language in the Australian media across the last decades. The topic is interesting, and the manuscript is clear. The methods applied are innovative and allow to underline new interesting evidence. I do not have particular concerns about the paper. I have some comments for the authors about the structure of the manuscript that should take into consideration.
Comment 1: Please move the research questions from the methods to the introduction.
Response: The research questions have been moved to the Introduction.
Comment 2: Last paragraph of the methods: why did you report here your conclusions? Please move this part from this section.
Response: This text was intended as an overview of what follows in the Results. We have moved it to the beginning of the Results and rephrased it as follows:
In this section, we show the associations between obesity-related terms and the gender, healthiness, social status, and stereotype dimensions. These associations are subsequently cross-matched with the obesity policy timeline in the Discussion, to help interpret the context of change in biases over time.
Comment 3: Please revise Figure 1 because it's very difficult to read.
Response: Figure 1 has been removed, along with the following associated text:
Data extraction and analysis processes are illustrated in Figure 1.
Comment 4: A recent paper has pointed out that post-bariatric patients, using a novel approach based on words, presented a weight bias regards their body, but it was not present for other bodies (see http://dx.doi.org/10.1007/s11695-020-05166-z). I think this aspect is interesting and linked to your results, showing a connection with clinical data.
Response: We have added the following text from this paper in the first paragraph of the discussion as it seems to fit best with our results:
Such entrenched weight biases, persistent in the media, may lead to internalised or self-stigma among individuals with overweight and obesity that persist even after weight loss. A recent study [insert numbered citation] performed a semantic evaluation of body shapes in obesity surgery patients and overweight/obesity controls and found that both groups were more willing to accept positive adjectives as a match when BMI was low and negative adjectives as a match when BMI was high.
Comment 5: You report that an automatic approach was used to check all the papers. Is there any possible limit with this approach in the selection of the papers?
Response: We have now added a Limitations and Future Research section to the Discussion to address this point and others:
There are two limitations in our data curation process, the automated approach we used to check and select papers. Firstly, automatic classifiers of any sort can include some irrelevant or false positive articles. Due to the large amount of articles in our dataset, in Step 2 of our methodology, we developed a machine learning binary classifier – a support vector machine - with 87.56% accuracy to automatically identify relevant articles (accuracy is the number of correct predictions made, divided by the total number of predictions made, and then multiplied by 100 to convert it into a percentage). As a rule of thumb, accuracy of a predictive model that is above 80% is very commonly used to summarise the performance of that model. Still, it doesn’t exclude the possibility of irrelevant articles in our final dataset, but that possibility is less than 12.44 percent and this has to be considered against the benefit of scale and efficiency that this method allows.
We also acknowledge that our analysis limited is in not being able to consider visual content visual content, images within articles (known as image framing), which have been shown to carry stigmatising elements [48]. This is something that could be added to the approach by including image classification along with additional measures. Furthermore, even though the Dow Jones is one the largest news databases, it might still miss some articles or news sources (although this doesn’t relate to the automatic approach as such). This applies to social media, even though this would not span as long a timeframe.
Round 2
Reviewer 2 Report
The revised version of the manuscript is appropriate for publishing.
