Making sense of 1000s of hits on my recent blog post at PLOS Mind the Brain

Blogging about blogging…

I started the new year with a post at PLOS Mind the Brain with more praise and appreciation for a randomized trial that you typically hear from me: When Less is More: Cognitive Behavior Therapy vs Psychoanalysis for Bulimia.

I uploaded this blog post and two days later it already had received thousands of hits and it kept going. I was delighted but baffled a bit.

I am trying to figure out what can be learned from the experience. I am writing over on my secondary blog. It certainly does not get thousands of hits in a couple of days, but maybe other bloggers can learn with me and maybe readers can give us some feedback that can help us make sense of this experience.

The trial I discussed was modestly sized and by no means perfect. But I found lots to like about how it was conducted in how it was reported.  And it was special in other ways, like being

  • One of the few head-to-head trials of credible psychotherapies in which one seemed to best another.
  • The kind of evidence we need to make decisions about whether long-term psychotherapy is better just because it is long-term.

I did not plan on writing this PLOS Mind the Brain post

I learned about the study from the updates I received from Google Scholar. As I often do, I tweeted about that study, providing abstract. I could not provide the actual article, because of its being behind a pay wall.

Allow me to digress and complain about being able to get abstracts without being able to access the actual articles. Abstracts are often deceiving, and consumers need to be provide the basis for evaluating the information they can get from freely available abstracts.

Anyway, my posting of the link to the abstract created a lot of fuss on Twitter and I was moved to blog about it, even though I had something else planned.

Reactions to my blog post

One of my disappointments is that I can attract thousands of visits to my blog posts, fueled by “likes” and tweets. But so far I do not seem very good at getting much of a dialogue going at the post. I really wish they could be more post-posting peer review of the kind I am trying to encourage PubMed Commons for all the articles in PubMed.

Fortunately, I did get a couple of interesting exchanges going on Twitter and Facebook.

On Twitter, I particularly liked the comments of Joar Ø. Halvorsen, who describes himself as a “PhD candidate in clinical psychology. Keen interest in clinical trial methodology.” I do not know the guy, but he post enough things that I include him among  the relatively few people I follow on Twitter

  Joar Ø. Halvorsen@joarhalvorsen 7 Jan

@CoyneoftheRealm Moreover, what do you think of handling of missing data with last observation carried forward?

We are getting getting a bit technical, but the answer to this question has implications for hw we interpret results of this trial. Basically, it is difficult to treat eating disorders or to study the treatment of eating disorders because patients have such mixed feelings about their condition and often drop out of treatment. We know that accounting for all the patients that are assigned to treatments is very important for the validity of clinical trials. Think of it: suppose we offer treatment to 30 patients and 15 disappear on us. We certainly cannot consider that a random event and only analyze the data from the data from the 15 that we can still find.

So, it is very important to conduct intent to treat analyses that consider all the patients were randomized. But that is difficult when we cannot contact the patients. One technique that was once used a lot was called “last observation carried forward.” That involved take the data from when we last had contact from patients and using it in place of what was missing. The problem of course is that patients may have disappeared for different reasons. Often the patients who are doing worse are more likely to disappear, but sometimes as in smoking trials, patients disappear because they got what they want out of treatment.

So, Joar, we have a problem, particularly when more sophisticated techniques are available to deal with missing data. But this particular trial is an improvement over most studies of psychoanalysis that ignore dropouts and only analyze the data from patients who stick around.

  Joar Ø. Halvorsen@joarhalvorsen 7 Jan

@CoyneoftheRealm Interesting article. But what do you make of the low statistical power? Might increase the risk for both Type I & II error

Okay, we are getting technical again, but there are some non-technical relevant issues being raised. Like, that the trial was rather small.  I usually am suspicious of clear-cut findings emerging from such small trials. The reason? They often do not replicate, and to borrow some thoughts from John Ioannidis, findings from small trials often prove exaggerated or false in terms of anyone being able to replicate them, except for the original investigators.

So I have to say, yeah, the trial is worrisome because it is small, but most studies of treatment of eating disorders are smaller and have more obvious flaws. So it represents progress.

And one of the striking things about this trial is not that CBT did well, but that psychoanalytic psychotherapy performed so poorly. Can’t detect any signs of any trends that might become clinically significant with accumulation of more patients.

And then there was the comment of Ioana Cristea, that immediately appeared on my Facebook wall when I proudly announced the uploading of my blog post

Did you see there’s also a recent very similar Lancet study for anorexia? With different results, of course.

How annoying, Ioana, you are raining on the start of my parade. Seriously, Ioana is such an outspoken naysayer and evidence-based skeptic that I have previously recruited her to write papers do presentations with me.

So, I had to download the Lancet paper (from behind a pay wall, dammit) and I had to give serious thought about whether what was said in that paper undermined everything that I just posted on my blog.

I got really caught up in thinking about that. I already had another blog post prepared for next week and had promised it at the end of this blog post. I had to revise my plans and use this Lancet article for what I posted next. As a teaser, I am tentatively going to entitle that post,

Are cognitive behavior and psychodynamic therapy no better than routine care for anorexia?

Why did I get so many hits?

The topic obviously hit a responsive chord, and I need to analyze what it was about the topic or my blogging.

  • The blog post was a heavily trafficked site where I had already been developing a reputation.
  • This particular blog site at which I am writing does not routinely get this kind of traffic, and it is a tough task to cultivate that sort of thing in a solo blog.
  • I effectively used Twitter and Facebook to promote the blog post, often with catchy and provocative phrases. That should not be overlooked.
  • Others picked up and amplified on this effort:

Neuroskeptic@Neuro_Skeptic 3h

Trial: Five months of CBT is better than two years of psychoanalytic psychotherapy for bulimia … By @CoyneoftheReal

Maybe the length had something to do with the visits posting attracted attract it. It was only about 1300 words. The message could be “keep it short.”

If it is short length counts for its visits, that could have troubling implications. I am convinced that some “longreads” are needed to deal with the nuances of particular issues and to provide readers with the tools to make their own decisions. To what extent am I willing to sacrifice the number of visitors by posting long reads?

Can I tease and lure visitors into a longread? What do I need to do?

I get mixed advice on my Facebook wall about longreads and links.

Gozde Ozakinci Personally, I find long reads rarely give me food for thought. The other trap bloggers fall into is provide many hyperlinks in their blog. It makes it too long to finish reading and digesting a post if I have to read 5-10 links too. Not sure what the right answer is but I think after 1000 words it gets a bit tedious.

Richard Gist A blog post serves to (a) tell me something that should attract my interest, (b) tell me enough to evaluate what my level of interest is likely to be, (c) provide me a map toward increasing levels of detail as I pursue the question, and (d) help me evaluate the cost:benefit ratio of time spent in pursuit and information value likely to be attained at each successive level of the chase.

Jeanette Bartha I like a few external links. It affords me the opportunity to search outside a blog post if I want. Using links can help keep the blog post shorter. I used this technique reading the article above. I read the article and some of the links to increase my knowledge of the subject.

Jeanette Bartha remember, James C. Coyne generations after ours are less and less able, willing, or interested in reading anything longer than a tweet.

Walt Stawicki links should not replace blog word count but supply “further reading” and …god how I hate vague assertions that don’t have them! I find a lack can prejudice me against an article, but not on its own. something must already be “fishy.”. but in good trust able reads I do want Moe without the hazards of Google search making decisions.

Gozde Ozakinci It’s worth remembering that there are many blog writers now. With finite time people choose carefully what to read. I find that those blog posts that are very current and informative trump long articles in review journals.

Maybe I need to provide longreads for which visitors can still make decisions to exit, when they are announced the still get something out of the blog post.

There is something ephemeral about blogposts. They get a lot of attention for a few days or so and then they are buried by lots of other blog posts. I think the phenomenon is exaggerated, but there is something to it. Maybe ultimately I have to turn my blog posts into inexpensive e-books for wider and more enduring dissemination. If I intend to do that, will I have to lengthen my shorter blog posts or try to integrate a series of them?

Maybe I am focusing on all the wrong issues, and I should enjoy the attention this blog post got and leave it at that.

Feedback welcomed!