Hacker News Reader

Comment by hn2022hn2022 | Hacker News Reader

hn2022hn2022

Pharmacists are the last barrier for the patient to make sure the scripts written by the doctor/s don't kill the patient.

Pharmacy techs are the ones that fill the scripts, but the pharmacist is ultimately responsible that it's correct.

Pill filling robots are great if you're filling the same 10 scripts, but not very useful when dealing with complex and compounding scripts.

My father was a pharmacist and we owned an independent pharmacy and I remember how often the scripts written by the doctor was so wrong and had it not been caught it would have killed the patient.

Knowledge of drug interactions are key, but these days new pharmacists just look up online, so a pill filling robot could work.

View on HN

Where I live, pharmacists can rewrite prescriptions from doctors, write their own (in limited cases) and in some circumstances extend prescriptions without renewals from a doctor (e.g. a family member was still able to get bipolar medication even though on paper they shouldn't have). Pharmacists are given the power and responsibility to fix mistakes, their specialty is the exceptional cases that computers aren't coded for, that's why they are highly educated and paid.

How can a person possibly be better at knowing the graph of drug interactions better than a program can?

Like, isn't the human just going to have human error (like the doctor)?

Two humans looking at something are better than one.

But a program could still have a much lower error rate.

Humans tell the program how to act, it's chicken and egg.

No, it's not.

Humans tell a calculator how to calculate.

It's never wrong. Humans (without a calculator) get math wrong constantly.

https://en.m.wikipedia.org/wiki/Therac-25

Humans aren’t perfect, but computer programs are only as good as the programming.

Humans are more than capable of writing programs that exceeds the most elite humans in the world, such as in go or chess. In nature we find creations exceeding creators all the time, so we shouldn't really be too surprised.

They are also capable of releasing bad software to production.

go and chess are infinitely simpler than drugs, humans and their interactions.

it's like saying autonomous driving is easy you just have to follow the road and read the signs, we have the best minds and unlimited money working on it for a decade+ and it still suck ass

The difference being that two different people are unlikely to have the same error.

Whereas the software will gladly just make the same mistake over and over again until corrected.

The dictation on my phone is pretty good. It's still frequently wrong.

There are pretty good databases of known harmful drug interactions available from vendors like Wolters Kluwer. Those are generally already integrated with the eRx features of most modern EHRs and pharmacy systems, and will pop up an alert.

The problem is that patient charts are often incomplete, or lacking proper RxNorm or NDC codes for the current medication list. So competent clinicians have to prevent problems from slipping through the cracks using their judgment and intuition. This is not something which can ever be automated (short of a true AGI).

> The problem is that patient charts are often incomplete, or lacking proper RxNorm or NDC codes for the current medication list.

The first seems like a problem - but not something a top 5% wage earner needs to do. The second can be handled by machines just fine.

Wrong. It can't be reliably handled by machines. If you can figure out how to do it then you can make a fortune.

Addressing gaps and errors in a patient chart is a highly complex task. Low skilled workers don't even know what to look for, or what questions to ask. A lot of this is tacit knowledge which we don't know how to reduce to software.

Lord, please grant me the confidence of this HN poster.

"Driverless cars are easy, just follow the rules of the road!"

> How can a person possibly be better at knowing the graph of drug interactions better than a program can

Than it should warn the human doctor before he write the script.

As I mentioned in another comment, this has already been happening for 10+ years. The eRx modules in all certified EHRs do automatically check for potentially harmful drug interactions.

https://www.healthit.gov/test-method/drug-drug-drug-allergy-...

The problem is that the patient's chart may be erroneous or incomplete. So the system can only detect harmful interactions on entries that are fully populated with RxNorm (or NDC) concept codes. Human doctors are expected to manually reconcile the medication list before writing a new prescription. But sometimes they skip that step, or patients don't even remember what other drugs they're taking. In some cases it's possible to query the patient's current medication list from a service like Surescripts but even that won't necessarily be complete and some level of manual reconciliation is still needed.

The other issue with drug interaction warnings is alert fatigue. For certain drugs there are a lot of known interactions, but those are considered minor acceptable risks due to lack of a better alternative. So clinicians get into a habit of ignoring all warnings and miss the ones that are actually serious. From a UX perspective there's no good way to solve this problem in a way that satisfies both the clinicians and the lawyers.

Alert fatigue is real and sets in quickly -

That's so simple it's coded into all prescribing systems and even written in paper prescribing manuals. The issue is that while some interactions are harmful, many are harmless and some even beneficial, but knowing the context that influences this range of effects is based on more than the contents of the prescription. Judgement and wider knowledge of the patient is required. People are still good at that but not infallible. I'm an ex pharmacist. Drug interactions, overdoses and plain wrong drug mistakes abound even in the age of computerised prescribing.

One of my bosses has a photo of her and a child's prescription fixed to the wall of a paediatric consultant who would have a killed a child if a pharmacist, my boss, had not queried a prescription.

Checking each other's work is essential in many industries but more so in healthcare.

Who's writing the software ? Humans

1 human error in the software = millions of people affected

1 human error by a pharmacist = 1 person impacted

And yet, who's writing calculator software?

Beside overflows, it doesn't error.

And yet, who's writing chess software?

And yet - there's no human in the world that can ever beat a chess program.

And yet, who's writing Google?

And yet - there's no human in the world that can recommend webpages like Google.

The list goes on.

Humans are bad at memorizing a GIGANTIC lists and not making mistakes.

Humans are EXCELLENT where error is okay and it's not really clear what the goals are.

I don't really know anything about pharmacy - but it does not appear to fall into category 2.

> And yet, who's writing calculator software?

Push calculators far enough and they'll all give you different results

We're talking potential human life here.

If you don't see the difference between a dumb game like chess and the infinitely complex world of medication and facing patients I honestly don't know what to tell you. Every customer facing people we replaced by computers degraded the experience for the customers

The pharmacists still have to consult per existing law. The robots would not hand the prescriptions directly to the patient.

Right now large chain pharmacists basically have to talk and fill at the same time to keep up. That’s not good for patients either.

How many people use compounding pharmacies outside of hospitals nowadays? I'm personally happy to see the possible automation of low-level work in the pharmacy world.

Reply on news.ycombinator.com