RSS
Twitter
Doctors Know Very Little About The Devices They Shove Inside Of You
from the well-that-makes-me-feel-great dept
In a story that probably isn't going to make many people feel particularly safe, the NY Times notes that, doctors have less safety information about the devices they stick into people than consumers have about the safety ratings on their cars. It turns out that the information isn't really collected, and what information is collected isn't available to the doctors who make the decision about what devices like pace makers and hip replacements to use. That means it takes a lot longer to realize when a certain product has serious safety issues -- putting many more people at risk. Also, if the data were more publicly available, it would force the makers of these devices to take safety issues a bit more seriously. As it stands right now, the FDA does get some info, but they don't seems to share it publicly. The story starts off with an anecdote about an attempt to collect much more data for these purposes... but, which got blocked for a few reasons, including that Medicare would have to change their forms and software (and, also, the fact that collecting such data is illegal right now). Given the strict regulatory efforts in the healthcare space, you would certainly expect a bit more attention to be paid to such things.
Reader Comments
(Flattened / Threaded)
A Poisson Regression Model
We can assume that device failures will follow a Poisson distribution, a standard distribution used for modeling rare events.
The difference-of-brand hypothesis would say that devices from brand A have a higher value of lambda (Poisson parameter) than brand B. The null hypothesis would say that lambda is the same for both brands.
Whether or not the hypothesis is true, the data will have plenty of outliers. We would want to fit a GEE (Generalized Estimating Equation) for Poisson regression, assuming there is compound symmetry within clusters. In other words, some patients in poorer or more variable health will tend to have devices replaced more often over time than others. Thus, we would take this auto-correlation into account when conducting a repeated measures study, measuring the number of times individuals have their devices replaced over time. We would, of course, have to take into account the subjects' age, exact health condition, and other random variables such as location.
It's unlikely that most techies on Techdirt will understand what I just said. Would we want politicians getting involved in these issues?
(reply to this comment) (link to this comment)
Re: A Poisson Regression Model
dorpus, we rarely understand what you say. I had assumed you were a politician...
(reply to this comment) (link to this comment)
Add Your Comment