Slashdot Mirror
Interesting Enemies For a Diagnostic Database
dlh writes: "Boston.com is carrying an article about Dr. Lawrence L. Weed's Problem Knowledge Coupler software. Apparently the medical profession is not exactly thrilled at the idea." Seems access to information is a positive thing, but certain doctors seem to feel threatened by this sort of database.
Is that people tend to live the symtoms that their medical complaint suggests. That's why you have to run blind and double blind tests, to weed out people who unconciously fake what they know to be the symptoms.
Something like this could comprimise the blind tests.
[On the other hand, a lot of subtle bugs in software come from analysing the blind elements. Ie, trying to understand subtle behaviour.]
OS/2 - because choice is a terrible thing to waste.
As the article points out clearly and several times, doctors are (usually) humans. This means they have personality traits that affect they way they accomplish their work.
In this case, it's ego. Of course no one wants to see a printout handed to them by someone who isn't a professional in the field saying "hey, this computer said you're wrong!" For chrissakes!! I wouldn't either. Of course there should be some level of interest and consession by the professional to review the information and test its validity. A doctor with an ego problem should be avoided just like a network engineer/administrator who thinks he already knows everything he needs to know about any given subject.
So yeah, it's fun to take the immortals down a notch back to Earth reminding them that they're still human. But it should also serve as a reminder to anyone who lives in the ever-growing world of science and technology (this does include medical science) that there is always something new to learn and never to stop challenging the "facts" that have been layed out before us. Oddly, there is no "spontaneous generation" as was once suspected and those "wandering stars" (aka, planets) aren't like other stars for more reason than the fact that they don't move like the rest.
And of course, let us never forget that "science" isn't about proving anything "right" so much as it is about proving things to be wrong. It's never easy to know the truth. But we get closer every time we eliminate that which is untrue.
It's actually frequently worse than this. (I work currently with 8 doctors, four pediatricians, three internists, and one FP. I've worked with.... Problem two dozen over the last few years. What I say doesn't necessarily reflect on the ones I currently work with.)
Doctors have bad egos. Really bad. Many of them refuse to acknowledge the shear drudgery of their average daily workings. Like another poster said, every one of her kids visits were the same. Yeah, no kidding. 90-95% of the visits to our office are within a few categories. Colds, heart problems, diabetes, and checkups of various sorts. (There are a couple of others, but not many). Yet it is not at all unusual for doctors to rail at this sort of technology for even these cases. They cling to an absurd belief that each patient is different. That, to put it bluntly, is bullshit. Most patients are the same. This sort of tool would make the routine stuff go MUCH faster, and would help narrow down the weird stuff to where you are doing real tests to really differentiate between two different (or five, or whatever) diagnoses. Of course, when those weird cases are programmed...
Others in this topic have mentioned that docs embrace new tech. Kinda. They embrace new diagnostic tools that they can play with. But they are not as in love with decision making/helping tools. It undermines their education. It undermines the fact that most of them just have incredibly good 'wetware' databases.
I would also discount the actions/thoughts/ideas of younger docs. They frequently change by the time they hit their mid-30's. I've seen it before, I'll see it again. They love the idea when in school/fresh out, but come to believe in their own manifest godhood over time. No mere pile of silicon could be greater than I.
Another problem that I actually do sympathize with is that this is grounds for serious lawsuits. You could claim your doctor did nothing/wrong thing based on what some stupid machine said to do. Any rational person knows the doc shouldn't automatically trust what the machine spits out, but you and I also know that there will be at least one or two docs who, when these things gain wider use, will take an extra martini at lunch, counting on the machine to catch his stupidity, ignoring the fact that the man and machine have to work in concert.
Given the decision support software (the drug interaction databases are one example. The only problem is that EVERY reaction is typically flagged, so you need to know what's going on to interpret the data. Kinda like the discussion of SQUID and other NIDS the other day) I can forsee this making strides. But it will be some time. Twenty years? There are two scenarios where this will happen faster:
First, HMO's and other insurance companies use this software or something similar to find out how quickly their docs are zeroing in on diagnoses. If they find something that lets them diagnose in one visit instead of four, they'll use it. And that's good for them and good for the patient (cheaper, quicker, more accurate care). The other scenario is one wherein the government mandates this sort of testing. Likely it would manifest similar to the HMO model, and be used to cut costs of state-subsidized healthcare. Again, not a bad thing.
The better docs I have spoken with (and being raised by a physician, I've likely spoken with more physicians than the average slashdotter knows) wouldn't mind getting to deal with the tough or fringe cases. That is a challenge. That's interesting. Pap smears and kid shots are rote drudgery.
I hope we'll see this gain prevalence soon, but don't count on it. And, as the article says, docs are more likely than ANYONE to dig in their heels.
Jesus was all right but his disciples were thick and ordinary. -John Lennon
I lost both my parents because of physician error, and I can assure that malpractice lawsuit payouts do not ease the pain, nor did the physicians testimony that they in one case they hadn't considered the correct diagnosis because it was so unlikely, or in the other case, that they hadn't considered the correct diagnosis because they weren't familiar with the condition.
I've been aware of this software for a long time, due to research done in order to show that a doctor was guilty of gross negligence in his misdiagnosis, and I never cease to be amazed at the number of doctors who really, truly believe that they can get everything right, every time, or that having a computer help with diagnosis is somehow different than reading an article in JAMA.
I have a serious question. Will you feel guilty when a patient dies, who would've been saved had you consistently used a system such as this, or will you be glad that you didn't waste 10 minutes to consult a computer?
I may be posting anonymously, but I'm not trolling. I understand there's more to medicine than diagnosis, but I don't understand why doctors can't admit that the 'I'll just remember everything' system that's currently in use is criminally irresponsible.
You seem to be a GP. I would assume your patients routinely report with nonspecific back pain, or with headaches.
The data I've seen suggests that these two symptoms in particular are both pervasive in the patient population and routinely undiagnosed or misdiagnosed. It would be interesting to run a double-blind comparative study of diagnostic efficiency of physicians and laypeople with and without the database...
It may come as a surprise to most people, but diagnosis is not the hard part of medicine...
Indeed, too much focus on just reaching a proper diagnosis can lead to poor care or worse.
Although I'm not an MD, I once had a very interesting case study in an operations research class (management science, statistics, expert systems, etc.). We used decision trees to study different diagnostic and treatment procedures in an actual (although somewhat simplified) healthcare setting. The model considered the results of appropriate/inappropriate diagnosis/treatment, the cost and latency of tests, false positives and negatives, and the differing goals of each key stakeholder.
From memory, the doctors were assumed to be most worried about correct diagnosis and treatment regimen, patients were most interested in the safest and most effective prognosis, and the hospital administrators were concerned about costs and legal liability. We found that optimizing the model for these different goals produced very different outcomes.
The results were somewhat counterintuitive: increasing the accuracy of diagnosis or ordering the most tests did not necessarily increase the probability of a cure, and could even increase the probability of death depending on the role of false positives/negatives, waiting periods, and drug side-effects.
That being said, the software in question seems to be more than a just a simple diagnostic tool. Combined with a patient centric outlook, I see medical expert systems becoming obvious (and essential) reference tools as long as they provide the MD with an ability to tweak the level of detail for minor ailments and to consider clinical experience, risks, local expertise, access to equipment, patient preferences, and so on.
My next sig will be ready soon, but friends can beat the rush!
Anyway, of those four diagnoses, three were wrong. Based on that, it sure doesn't sound like diagnoses are easy to me! Add to that the fact that I'm pretty good at troubleshooting and I'm one of the few that I know of. Most people flail about trying things at random and, as far as I know, training isn't much help for most of those people. Yes, it's easy to memorize a few pat answers to the most common problems, which is why many people who visit the "doctor" wind up seeing a PA, who forwards to the doctor only those cases whose diagnosis is in doubt, but that's exactly why this sort of thing is important. As time goes on, doctors are going to be less and less likely to see the simple cases.
You mention psychiatric diagnosis, so I'll talk about those. A quick check of my local DMDA chapter shows that some 70% of those suffering from some serious mental illness were misdiagnosed at least once. I can't help but think that a computer program that prompts the asking questions about typical symptoms of mania and schizophrenia would reduce that because most of the misdiagnoses start as a diagnosis of depression because it's what people complain about. I know the doctors don't ask those questions because in the sample that I have (8 so far) none have asked the right questions to make what we (that is, myself and the ill person) now believe is the correct diagnosis.
In any case, since visiting a doctor (and I spend a lot of my time in doctors' waiting rooms so I know this quite well) is something like an hour waiting to see the doctor followed by maybe 10 minutes of answering questions I don't understand, filling out the forms while I wait can't do me any harm even if the diagnosis is not remotely in doubt or irrelevant to the treatment, can it? I mean, it becomes part of the patient history just like the temperature and blood pressure check you're going to do whether I come in with a fever or with a splinter, right?
Heck, I suggest you put terminals tied into that database in the ER waiting room so there'll be something to do while you're waiting the 4-6 hours (on average) it'll take to get to the head of the line.
When my kid had a broken arm in the 80s -- a crappy greenstick fracture I could have set myself -- the total bill was well over $1000. I have no idea how much a doc would charge today for reducing and casting a simple fracture, but I bet it would be huge.
The bills are so high now, because they need to offset the lowball payments that the HMO's give them. I just had a visit to the ER two months ago, and got a notice in the mail from BlueCross that the hospital bill for that day (had an EKG, CT scan, and an MRI) was over $5,000, though they only paid $1,200. Subsequent tests over the next few weeks were paid at even lesser rates (than the 20% from the ER visit).