Pizza for Cancer Prevention? Mining Twitter for Popular Nutrition and Cancer Myths

I am not a cancer expert.canstockphoto6433880

So when I want to get a sense of the state of research related to nutrition and cancer I turn to resources like the AICR. But many people don’t know about such resources, and this is reflected in cancer knowledge. For instance, the AICR’s own survey found that 42% of Americans don’t think that a healthy diet can reduce cancer risk, and fewer are aware of associations with specific foods such as processed meat.

What resources are people visiting, and what are they talking about? One way to get a glimpse of this is to mine Twitter, which I did for the last couple weeks, collecting about 420,000 tweets containing the word “cancer”. From these, I used the list of terms and their derivatives from this AICR/WCRF report, along with other health related terms that I came up with (82 in total) to assemble a list of 14,585 tweets for analysis of the links being posted. Continue reading


Trends in Cancer Prevention Research

Like last year, I created a Google Spreadsheet to query raw publication counts from PubMed of food/nutrition/health terms from an AICR/WCRF report. This is a way to visualize how nutrition and cancer research may be changing over time. The counts represent the number of published articles containing the term in column A plus “cancer” in the title or abstract.

Like last year, I used the following color scheme from the AICR report: the coloring reflects the strongest confidence for an association between the term and a decreased or increased risk for at least one cancer type. Terms with a yellow background were not in the document list and added by me.

The spreadsheet is sorted by column D: the percentage of papers published in 2013 compared to 2012. This will still change as PubMed will be indexing papers published in 2013 for some time, but the spreadsheet auto-updates daily so it will become more accurate over time. Column M shows the average publication count change per year over the last 6 years (2008 to 2013), which may better reflect the recent trend. Continue reading