Follow Slashdot stories on Twitter


Forgot your password?

Comment Re:About damn time (Score 5, Interesting) 126

It's harder than you're probably giving it credit for, especially for miles-long freight trains, where a hill can mean one segment of the train is accelerating while another is decelerating. We're just about there, though, insofar as we have software that automatically drives throttle, brakes, and other controls. Link:

Norfolk Southern, an American rail operator, now pulls roughly one-sixth of its freight using locomotives equipped with "route optimisation" software. By crunching numbers on a train's weight distribution and a route's curves, grades and speed limits, the software, called Leader, can instruct operators on optimum accelerating and braking to minimise fuel costs. Installing the software and linking it wirelessly to back-office computers is expensive, says Coleman Lawrence, head of the company's 4,000-strong locomotive fleet. But the software cuts costs dramatically, reducing fuel consumption by about 5%. That is a big deal for a firm that spent $1.6 billion on diesel in 2012. Mr Lawrence reckons that by 2016 Norfolk Southern may be pulling half its freight with Leader-upgraded locomotives. A competing system sold by GE, Trip Optimizer, goes further and operates the throttle and brakes automatically.

Comment About damn time (Score 5, Insightful) 126

The article's a bit short on details, but this is where I expect autonomous driving to take off first--long-haul trucking. Controlled-access highways present fewer complications like pedestrians, four-way stops, and the like, and I imagine automating that would take care of 80% of the driving. Even if you still needed a human driver to reel 'er in at the warehouse gates or even the city limits, it still strikes me as a huge improvement.

Laws and liability are going to be the biggest limiting factor to commercial deployment, especially if they boil down to "a human must be ready to intervene at any time," but I think there are fudges around that. You could have one human operator in a remote control center "driving" multiple trucks, kind of like a cross between drone pilot and remote ICU monitoring.

Not that even a human sitting in the seat with hands on the wheel would be likely to intervene effectively should something go wrong after eight hours of idle monotony. But, having a human somewhere supervising in some capacity would soothe the more irrational fears that also serve as part of the reason we still keep human pilots flying planes, while still yielding the benefits that come with automation--self-driving trucks are much less compelling if each one still needs a full-time human driver to comply with laws.

Comment Re:Give me a choice (Score 1) 122

It doesn't do ANYTHING you are talking about

EHRs absolutely do do all of those things, including checking drug interactions, allergies, and pregnancy and lactation warnings, and there absolutely is scholarly evidence of their effectiveness if you care to Google. Here's one concluding barcoding and CPOE are "vital."

If your EHR doesn't do those things, then that's a defect peculiar to whatever software you're using.

The medication lists are ALWAYS wrong or misleading as they are huge and hard to read, harder to update

You think they were any shorter, easier, or more up to date on paper? A bad computer system can make med rec harder, but even a good one can't make anyone give a shit about the patient's PTA meds. Nor can a hospital losing Joint Commission accreditation, apparently.

results from another EMR is always incompatible, so you have to scan it in

Yes, that's very common. It doesn't have to be the case, but it usually is.

Have you read an EMR progress note?

Yes. I have read lots and lots of EMR progress notes. I've seen notes filled with pages of stale labs because computers make it easy to copy forward the entire chart. I've seen notes flagged by the built-in plagiarism tools because copy-pasting an old note (and getting paid when you file it!) is easier than actually rounding on the patient. I've seen SOAP notes filed with weeks-old vitals because one group of residents preferred copying things into and out of Microsoft Word, which didn't exactly have the latest telemetry. I've seen notes with assessments that weren't done, filed on patients that weren't rounded on, because easy money.

But you know what? You can read them. And so can the doctor who sees the patient after you.

Comment Re:Give me a choice (Score 1) 122

Certain kinds of errors have decreased dramatically. Computers reduce wrong patient/wrong medication errors dramatically, especially in systems that require you to scan the patient's barcode (to make sure you have the right patient), and then scan each medication's barcode (to make sure you've got the right meds). There's a lot of scholarly research available if you search for EHR medication errors, but this is one of the first non-paywalled things that pops up in Google.

Other issues are more challenging. You might have grabbed the right IV bag, and the computer might have confirmed that it had the right barcode, but that alone doesn't prevent pharmacy from filling the bag with the wrong fluid in the first place, or putting the wrong barcode on the wrong bag. It also doesn't stop a harried nurse from failing to mix both diluent and active ingredient of a vaccine when filling a syringe, or from using the same insulin pen on multiple patients; nor do computers keep toddlers from digging fentanyl patches out of the trash, overdosing on painkillers, and dying.

TL;DR most kinds of medical errors are decreasing, and computers tend to help--they're good at checking meds, and good at counting how many sponges you had in the OR field before and after cutting the patient up. Others are still around--handwashing compliance can still be flakey, silk neckties encounter years of sick people yet are never laundered, and "nosocomial infections" are still troublesome. Pick your favorite kind of medical error and give it a Google for recent stats.

Comment Re:Give me a choice (Score 3, Informative) 122

I hear you--even within a hospital system, and even where standards exist, it's a pain. Ultrasound machines (for those that aren't imaging informaticists) are supposed to speak DICOM, but some do it creatively--one technically sent DICOM messages over the network, but most of what they contained was wrapped inside a proprietary XML blob rather than standard DICOM fields. What standard fields were implemented were implemented strangely, waffling between spelling out measurements ("centimeters") or using their abbreviations, mixing case, and reporting measurements to absurd precision (dozens of zeroes after the decimal point, for a bone measured in millimeters).

Sharing charts between hospitals is a mire of politics. There's the government's own Direct standard, which they mandated every hospital use to send charts, without any indication of what the recipient is supposed to do--a lot pipe them to /dev/null, because the vaguely defined content of the message is often useless and redundant with existing methods of communication. They're now working on legalese to require that you "do something" with the messages you receive, but exactly what that is (and how to objectively prove that you did it) they're still figuring out.

Then there are organizations like Commonwell, trying to monetize a data-sharing "standard" not even their founding members could be bothered to implement. They haven't sent a single chart as far as I know, but that doesn't stop them from issuing press releases praising their "interoperability" with the same frequency AT&T issues press releases praising their gigabit fiber.

Then there are HISPs (centralized, sometimes quasi-public, repositories of patient information). Some have managed to legislate themselves as mandatory middlemen, and, having done so, have proceeded to extract monopoly rents over the transmission of outdated and incorrect patient information. Even better is provider look-up--if they give you the wrong fax number for a physician, you are responsible for the HIPAA violation when a random gas station gets someone's medical information. This causes them to care as much as you'd expect about the integrity of the data they peddle (and that you're required to buy).

It's frustrating, because medical information has to be shared for it to be of use--there's no use having a mammography if no one will read the results, or if the people treating you can't access the study and have to order their own.

Comment Re:clueless management (Score 1) 122

Being licensed profession will stop clueless management from force stuff to be so easy to hack / not willing to pay the costs to have be done right.

If you're going to make it illegal for literally anyone else to write software, then maybe. I'd love to see you square your favorite licensing regime with anything resembling open source development.

Comment Re:Give me a choice (Score 5, Insightful) 122

I wish I could request paper records.

You really don't. I've shilled for EHRs before, but the TL;DR is

  • Paper charts kill people. They don't check for drug interactions; they don't double-check that you've got the right patient when you're operating or administering medications; in the case of a recall, they can't tell you who received a bad batch of a vaccine; and they certainly can't tell a first responder that unconscious you is allergic to blue dye, unless they already happen know your regular clinic and have a fax machine in the ambulance.
  • Paper charts are useless for patient care. The hospitalist trying to reconcile what you were taking at home with what they want to give you in the hospital can't actually determine whether they're about to kill you if the cardiologist treating your heart attack happened to take the only copy of the chart to enter his notes. If they made a second copy for the cardiologist, there's no guarantee his notes and medications will ever get entered into the hospitalists copy, or into pharmacy's copy, who might also wonder why two different doctors plus your PCP are trying to dose you on blood thinners, or into your regular doctor's copy, who might be totally unaware of the cardiologist's findings
  • Paper charts are expensive. If nobody knows that you already had a lab or an X-Ray, they're going to order it again. If they do know you had one of the above, you're going to have to wait for a fax, or for them to mail negatives. Because handwriting and general disorganization, especially over a long admission, tends to make them write-only, it's much harder to know exactly what they gave you and why, which makes it harder to justify to the government or an insurance company why they should pay your tab.

That doesn't mean the electronic versions don't have terrible, even maddening, flaws, but even the worst are better than paper.

Comment Re:clipboards? (Score 1) 60

I won't disagree that medical billing is still a nightmare, but it's not the fault of CPT codes. No insurance company will sign a blank check and ask the doctor to fill in the amount; they all have a maximum they'll reimburse for, say, a broken leg, and they'll reimburse "broken leg" differently for a simple fracture than an unexpected amputation that took a crack team of surgeons 32 hours to reattach.

The codes are just a standard way to quantify exactly what was done. The "standard" part is important since most docs wouldn't be able to deal with more than a single insurance company if they all spoke a different code set. Absent a code set, it's much harder to determine how much your treatment should have when all you have are subjective descriptions from multiple doctors about what each thought they did. Codes remove that ambiguity, which insurance companies otherwise use to delay paying you and the hospital for as long as possible.

Generating the codes in the first place is complicated in the clipboard world, where a medical coder has to pore over sometimes hundreds of pages of scribbles to find out whether your vitals ever spiked into a danger zone that would have required more frequent nursing attention (at a higher cost), whether you were ever given certain kinds of medication, and, more subjectively, trying to divine what you actually had from the doctor's free-form notes, and what codes best describe that particular malady. Since this requires examining linear yards of flowsheet annotations (where an individual, critical cell might be missed) and subjective interpretation (does this description of a fracture qualify for Break-1 or Break-2?), different medical coders can end up coding the same chart differently.

Computers, in theory, can solve this. The computer knows exactly which medications you were given, how often, and at what cost. It knows exactly how often nursing took care of you, and can unambiguously determine your acuity. It knows every human in the operating theater, how long they were there, how much they cost, and every roll of gauze they used. Even free text notes are often templated, with doctors filling in blanks on standardized forms with multiple-choice answers, each of which can be unambiguously evaluated. Computers can instantly and unambiguously determine who too care of you, exactly how much it cost to take care of you, what codes best describe those services, and can communicate that information electronically to your insurers.

Where this falls down is a lot of computer systems grew organically--first a program for scheduling, then a program for billing, then a program for tracking clinical notes, then a program for monitoring vitals, etc., rather than some comprehensive monolith springing forth fully-formed right at the start. This means that to bill for a pregnancy, a medical coder might have to hunt through a dozen different systems: An ADT suite to determine where the patient was roomed and for how long, since a regular bed costs much less than intensive care. A specialty labor and delivery system, to gather the myriad of data states and to defend against lawsuits. An often separate fetal heart monitoring suite. An often separate ultrasound modality, with its own operating system that may or may not speak industry standards, that may or may not pair with a separate vendor-specific image viewer on computer workstations. A separate OR suite for doing documenting personnel, and doing counts and checks (to make sure you're billed for the roll of gauze they did use, not billed for the one they didn't, and that neither were left inside you.) A still-separate anesthesia suite, to make sure they don't gas you to death. An inpatient system for tracking medications and vitals, which may or may not be missing the medications and vitals collected by the labor and delivery suite, which may or may not have the vitals collected by the fetal heart monitor, which may or may not have the vitals collected by the OR suite for your C-section, which may or may not have collected anything at all if they hooked the monitor to the anesthesiology suite instead. Complete billing requires searching multiple systems for the medications given and supplies used, others to determine how long you stayed and where, individual ultrasound carts (or, if you're lucky, a single RIS suite) to determine if any ultrasonography was performed, others to determine who was present and for how long, and yet another set of systems to find out all of the above for your new youngling(s). Labs are often third-party and are even more likely to have a separate computer system than any of the above specialties, so billing for tests and their interpretation requires more branching.

Interfacing the cross product of software applications is difficult and expensive, even when they comport to speak industry standards, and even when they don't involve third-party labs, so it often happens imperfectly if at all. This means the same manual, subjective, error-prone, hunt-for-the-charge style of billing that made clipboards suck. This also ignores the fact that most doctors aren't hospital employees--the surgeon has his own practice, and merely has privileges to operate at the hospital--and therefore bill separately for their own labor.

The best computer systems make all this go away--all the documentation is accurate and in one place, bills are generated automatically and unambiguously, and include professional fees (which the hospital knows how to remit). Those systems really do make billing better for hospital, physician, and patient, if not the insurance company. The worst computer systems make things worse, with dozens of individual programs grafted into a Frankenstein and dropped into the laps of an outsourced third-party coder to deal with. It's that stuff that leads to the nightmare I'm sorry you and yours have to deal with.

Comment Re:clipboards? (Score 4, Informative) 60

What was so bad about clipboards again?

Clipboards have a bunch of known deficiencies. They're effectively write-only, especially if no one else can read the doc's handwriting.

Then, they're hard to duplicate. Should you end up in the hospital (heaven forbid), hopefully you're conscious enough to explain your drug allergies to the EMT, because it'll take a while to find out which clinic you normally see and get a copy of their clipboard. Then the copy of the clinic clipboard ends up in the hospital's clipboard, but the stuff in the hospital clipboard probably won't make it back to the clinic clipboard.

There's also only one copy of the hospital clipboard, so the cardiologist treating your heart attack can't put notes in your clipboard if the hospitalist took it to figure out what meds you were (or should be) on. If they do make copies, someone has to make sure the cardiologist's annotations make it into all of them without error. Those charts then have to be stored in a giant bunker somewhere, forever.

Clipboards are also bad at medication safety. When you're giving millions of med administrations to millions of patients, eventually you end up giving the wrong drug to the wrong one. Clipboards can't verify that you nabbed the right patient or the right drug, which kills people once you scale up the mistakes that would have happened to a national level.

Even before the nurse gives the meds, a clipboard can't tell the doctor that one of the medications he's ordering will interact with the medications someone else ordered. That also kills people. If one lot of those medications was tainted and recalled, it's also really, really hard to find out who was affected if all your administrations are documented on paper.

Finally, it's really hard to bill correctly if all of your documentation is on paper. If the coder going over the clipboard misses a charge, the hospital loses out on money. If the coder invents a charge, you lose out on money. If the coder can't find whatever documentation a kafkaesque insurance company demands to justify a procedure, you both lose out on money. Also harder to reject a claim for not being written in blue pen with block caps when the claim is electronic.

There's a bunch of other ways clipboards suck, and a bunch of ways the clipboard-replacements suck, but the former tends to suck a lot more than the latter.

Comment Re:Equally suspect (Score 1) 306

Unlike price elasticity, autorectogenesis is entirely responsible for that tortured non-expression and the verbification of "pence." Idea was that there hasn't been that much shilling since before decimalization.

Comment Re:Equally suspect (Score 1) 306

Exactly! Or from seeing Valve's success with their Steam sales, or Apple's success with lower iTunes prices, or from any other number of things obvious to you and I and everyone but John Scalzi.

"Conversion, fastidious Goddess, loves blood better than brick, and feasts most subtly on the human will." -- Virginia Woolf, "Mrs. Dalloway"