What's your reference for this? That's incredibly (read, unbelievable) high for a routine procedure.
RESULTS Among the 30,818 records identified, 82 population-based studies from 24 countries were included, involving a total of 38.5 million colonoscopies. The estimated incidence per 10,000 colonoscopies was as follows: gastrointestinal AEs, including perforation (5.15; 95% confidence interval [CI] 4.19-6.34, I2 = 99%), bleeding (18.39; 95% CI 13.53-24.99, I2 = 100%), and splenic injury (0.61; 95% CI 0.43-0.85, I2 = 93%); nongastrointestinal AEs, including cardiovascular events (52.11; 95% CI 18.67-144.59, I2 = 100%), respiratory events (4.26; 95% CI 0.73-24.99, I2 = 100%), and deaths related to colonoscopy (0.18; 95% CI 0.10-0.34, I2 = 74%). Subgroup analyses yielded partially divergent findings. The majority of the included studies exhibited a low to moderate risk of bias.
just ask any AI, i don't got time to play tic-tac-toe with the NIH.gov website gating me behind click bus images for 10 minutes
My source is not seeing one perforation each week at work.
> just ask any AI
These do not give reliable answers, as I am sure you know
i only answered the specific question of where the number "48" or the range 40-80 came from.
my cite even shows perforations are 3-5 per 10000 so i don't know what you're on me about
Yes, I am sure. Do send the actual citations.
> my cite even shows perforations are 3-5 per 10000
An implausible number for humans who have actual, non-LLM experience in this area
https://jamanetwork.com/journals/jama/fullarticle/2779987 3.5 per 10,000
take it up with JAMA and the AJG.
do you do SCREENING or DIAGNOSTIC/POLYP REMOVAL?
because there's a difference. And it has nothing to do whether i use google.com or chat.whatever.com to find that out.
You didn't give me a source before now, so I unfortunately had no other source to challenge except the LLM!
> https://jamanetwork.com/journals/jama/fullarticle/2779987 3.5 per 10,000
Different source, friend. Please note that they say 3.1, not 3.5.
That is at about 25% less than 4 (and even less than 5, let alone your previous assertions). And if either you or your LLM troubled yourself to read the article, even this is a confounded number in that we cannot determine whether low-volume, 'community' operators are worse than high-volume settings.
> do you do SCREENING or DIAGNOSTIC/POLYP REMOVAL?
We are talking about screening—once you have a target to remove, you are looking at a high-likelihood-of-cancer population. I am fully aware of this, so I don't understand why you are bringing up this difference which has not yet figured into our discussion. Is this something your LLM suggested to you?
---
I think the thing to take away is that LLMs do not yet replace human understanding and discretion.
those two numbers fall within the range of "3 - 5 per 10,000"
the thing in the first sentence of my reply is called a DOI, it's a document identifier, that you can type into google or bing and it will pull up the correct journal citation i was citing.
I understand that the JAMA paper said 3.1, but it also said a number almost 5 times higher near there for "major complication", separate from perforation.
My original reply, where i did the > block quote, is from the AJG source; 10.14309/ajg.0000000000003429
And, for all we know someone is using a roto-rooter to perform colonoscopies and throwing the statistics off
My doctor says that since Cologuard catches a large percentage of those 3-5 per 10,000 without any of the colonoscopy risk, the marginal benefits from colonoscopy really aren't justified since FIT+DNA testing is almost as good, at least for low-risk cohorts.
Very few things in medicine are zero risk. I wish more doctors would help balance the risk of doing A vs. the risk of doing B vs. the risk of doing nothing.
It's all Bayesian conditional probabilities, considering your own individual risk factors, and considering the false positive rate and false negative rate of each test.