Thursday, February 08, 2018

Letter to my statistics professor

You are in Cape Town, and I’m in Columbus, Ohio.  Through the miracle of on-line learning and MOOC, I am in week 6 of your Coursera offering on Understanding Medical Research, in which you often stress the importance of statistics, choosing the right tests, and having access to the original data. I'm definitely a statistical outlier among your students--I'm 78 and retired in 2000, and have had no math since 1955. And although this class doesn't demand a heavy background in math, I even had to review fractions, percent, and square root!

So, in the Nov. 28, 2017 JAMA there is an article, "Adherence to Methodological Standards in Research Using the National Inpatient Sample," that caught my eye, and impressed upon me everything you've been teaching us. I was surprised to read that of 120 articles (the sample) of the 1082 (the population) published in a 2 year period using data from 35 million hospitalizations in the U.S., 85% did not adhere to 1 or more required practices of the AHRQ (Agency for Healthcare Research and Quality), and 62% did not adhere to 2 or more required practices. A total of 79 of the 120 in the sample (68.3%) did not account for sampling error, clustering, and stratification; 62 (54.4%) extrapolated non-specific secondary diagnoses to infer in-hospital events, and so forth.
So I began linking the people involved; not only did the researchers get it wrong, the peer reviewers missed the errors, as did the editors of the medical journals (some with a high impact factor). The articles with incorrect statistical information were used to get promotion and tenure at universities.  Possibly policies and regulations for hospitals may result from some of these published articles with careless statistics, or, a "medical journalist" will attempt to translate the results for a layman, and it will be reported in the media.

In the U.S. we have been subjected to a horribly expensive electronic medical records law which all physicians and hospitals had to incorporate (I think it began around 2010).  There was no evidence this would improve medical care, but it was apparently "needed" so researchers had more data to mine.  I still have to ask for my medical visits from Dr. A to Dr. B, two miles apart, to be faxed because these systems don't talk to each other. Now with all this data ready to be mined, someone will mess up the statistics!

So, Dr. Juan Klopper, you are definitely needed, and will never run out of things to talk about.

No comments: