It’s been a while since I wrote one of these summaries. Chalk it up to new projects, new opportunities, and a lack of time to simply sit down and do the analyses. If you’ve enjoyed these and found them helpful, please let us know in the comment below or reach out to us at email@example.com
If you’ve never read this before, here’s the overview. There is a lot of literature floating around out there. Most people never get to it. I currently have a reading list over 50 articles long that grows constantly.
This is a problem. There is great work being done, but people don’t have the time to really dig in. So it sits in journals without being used.
What I try to do is use AI methods to identify and summarize the important and interesting pieces every month or so and provide some commentary around it all.
To do this, I first pulled all the articles from a few health equity-focused journals over the past month, along with any and all other articles with the health equity keyword. In just the last month, this yields 365 unique articles across 204 unique journals. You would have to read ten articles per day to just stay current!
The first thing I look at is words. Words can be meaningful just by themselves. This first figure is a word cloud. A word cloud is a group of words that are in a picture. The bigger the word, the more times it was mentioned in the articles.
This next figure is a network plot that shows the relationships between words. The bigger circles are words that occur more frequently, and the thicker lines are relationships that occur more frequently.
It’s always beneficial to read review articles. They combine the study of a particular issue and attempt to draw generalizations about many separate discoveries. I put together an excel file with the 24 review articles that have come out over the past month. If nothing else, these are always good to peruse. Click the link below to download.
Picture from Wikimedia Commons
Next comes topic modeling. When we use topic modeling, we use the content of the articles to identify the clustering of the articles. We’ve been moving to a new method where we use MeSH terms to better organize the articles
MeSH allows for a consistent approach to locating material with diverse names but similar ideas. Rather than being textual, articles are organized in a hierarchical structure that allows for more focused searches. Another benefit of this structure is that it enables searchers to search for suitable keywords by navigating through the site. The content is refreshed by subject experts in a range of topics. Hundreds of new ideas are added each year, and thousands of changes are made.
We’re hoping this pays dividends in more effective queries, but let us know in the comments!
Using MeSH, we can then cluster the main topics and the number of articles that correspond to that topic. The graph below shows how many articles are on each topic. This shows the different content areas in which health equity is being discussed.
Taking this one step further, I then made a correlation matrix between topics. This shows how these different clusters are related. So while the comprehensive, language, and sleep topic was the largest topic overall, we can see the very strong relationships between several other topics like obesity, asthma, and opioids.
I also flagged the articles that are most representative of each topic and put the citation into another excel sheet. So, if you have even more time, you might want to check this 15 out to get a good sense of the landscape.
This last figure is more exploratory than anything else. I trained a list of word vectors to look for relationships between the most frequently used keywords. This looks at what words are related to one another in how they are used in the text. In this figure, each dot is a word. Since we don’t see any major patterns in the figure, this suggests some broad consensus in language and terms.
So there you have it–the last 31 days of health equity research summarized in this article. If you want to stay up-to-date on the latest news and publications, it’s worth your time to check out some of these resources. There are 24 review articles that have come out over the past month alone, each with different insights into how best to address issues these critical public health issues!
We also use MeSH terms as an organizing tool for articles so that searching becomes more streamlined. Topic modeling is another way for us to explore what topics are most common. We then saw how different topics are connected to one another. And finally, we conducted an advanced topic modeling experiment that pulls together the top 30 words from the vector space model and created a graph to show how these words are related. I hope you find this as valuable (and fun) as I do!
But wait, there’s more!
We’re formally getting ready to launch our 101 Days of Health Equity newsletter that will go much more into depth, pull in many more sources of information, and provide you with what you need to know, right in your inbox. So stay tuned!