It’s the 20th anniversary of The Tale of Despareux by Kate DiCamillo, and this is a book the world needs. I regularly wonder how much of what I’ve published is needed by the world.
Research publications are our product, and since the public funds a lot of our research, our product should be accessible to them. Also, our findings can help others avoid re-inventing the wheel. Our impact is largely measured by the number of publications we have, and our publication record is a determining factor in whether we get research funding. Because quantity matters in healthcare and technology fields, we tend to submit the “least publishable unit”. Does it have to be that way?
My colleagues in economics tell me someone can get tenure from one outstanding paper. They value a paper that contains substantial, comprehensive research rather than a single portion of the research. Writing one paper may take years as they present an idea, get critical feedback, further develop the idea, do more modeling, and get more feedback. One high quality paper is worth more than a dozen mediocre papers.
I wonder how to engender this principle in the researchers I mentor. I am now a full professor, so it may seem easy for me to say this in hindsight. But here are some principles that have guided me from day dot (I was looking for a place to use this Australian phrase) or have evolved throughout my career:
Let the problem solving drive you--not the papers. Papers will come.
Don't publish anything you don’t think is worth publishing.
Measure your impact in more ways than publications. Most universities now support a broader view of impact--it’s your case to make in your promotion materials. Have faith the review panels will respond.
Those guided my career, but if I ask myself the following questions, I don’t come out as strong:
What is your purpose for submitting research to a conference? Is it to show off? Should you rush to meet a conference deadline you want to attend by submitting a project not ready to be published? My mentor Greg Cooper told me he selects research that he wants to share with the audience at the conference so that he can get quality feedback that will help him in the next phases of the project. To be honest, I have seldom met that bar.
Go back to the ‘80’s and watch a great talk about writing papers by Leslie Lamport On Writing Papers. Two points that resonated with me:
Good writing requires good thinking…bad thinking can never create good writing. You have to have something to say, and it’s very important to be clear in your mind about what it is you want to say, and to whom.
Good thinking precedes good writing, and the best way to think about things is in terms of examples. Your examples tell a story, and they test your theories.
Let me share three stories from my research. I list the number of citations of these papers to contrast the meaning the research had to me with the contribution it made to the world.
My most influential paper: A Simple Algorithm for Identifying Negated Findings and Diseases in Discharge Summaries. 1156 citations. I did not create this algorithm (a grad student named Will Bridewell did), but I built the story of why it was better than a baseline and why it was needed when we had just been scooped by a more sophisticated algorithm. With some trepidation, I requested to be first author. It charted a 10-year path for my career, and the work I continued to develop with colleagues has been implemented in many languages by researchers and industry.
The paper I had the most fun writing: Developing Syndrome Definitions based on Consensus and Current Use. 33 citations. In this paper, I was truly able to tell a story. Everyone deployed syndromic surveillance in the mid 2000’s with different definitions that couldn’t be compared or shared. I brought them all to Pittsburgh to create consensus definitions. I was terrified my efforts would fail as everyone protected their turf, and in the first hour of the workshop, it looked grim. But we came up with a framework that achieved perfect harmony. Here are two slides from my presentation at the International Society for Disease Surveillance conference where Farzad Motashari, before becoming the national coordinator for health information technology at the U.S. Department of Health and Human Services, told the audience it was the most important paper of the conference.
 |  |
The paper I am most proud of: Anaphoric Reference in Clinical Notes: Characteristics of an Annotated Corpus. 18 citations. This is the story of a woman with a degree in linguistics and no satisfactory career path who followed her husband into medical informatics for the lure of clinical NLP. In this project, I got to combine my knowledge of linguistic theory, such as discourse models, linguistic form, and salience, with the practical but hugely effortful team-based task of creating an annotated corpus I hoped would be used to train and develop anaphoric reference algorithms. It wasn’t used much beyond our team as far as I know.
Comments