Monday, September 19, 2011

"Why don't these systems talk to each other?"

Another frequent question I get asked in my job is, "These systems are all plenty expensive - Why don't they talk to each other?"

What this is referring to, of course, is the common phenomenon that the EMR at one hospital may not seamlessly transfer a patient's record to another EMR down the street.

There are actually a few reasons why this is so, but one of the most interesting ones is a phenomenon called the "patient identifier problem."


Here's what it boils down to : It's much harder to identify a human being than you might imagine.

Allow me to explain. (Names below are purely fictional, just for teaching purposes.) :)

So at first glance, it should be easy to identify a human being. After all, we have names, right? When we see our neighbor John mowing the lawn, riding his mower - His name is John - We recognize him - Yep, that's him. Easy, right?

Well the problem is what happens when we actually try to identify someone on paper - That is, have a record that we can match to an actual human being.

At first, we might try to label a chart "John's Chart". The problem with this approach is that there may be lots of Johns, so in a small town (even on a small street), you might have two "John's Charts".

So you might add the last name : "John Smith's Chart". This might work in a small town, but when you expand to collect charts for your whole state, or the whole country, you might find over 700 "John Smith's".

So names are generally a bad way to label a chart for a few reasons :

  1. There might be over 700 "John Smiths" across the country - How will you know which chart is the right one to look for?
  2. Your neighbor, John Smith, might register at Clinic A as "John Smith", at Clinic B as "Johnathan Smith", and at Clinic C as "Jon Smith". This could potentially make three records. How will you know which is the proper record to search for?
  3. Names may also be misspelled by registration staff - If a "Karen" registers in a clinic, will the registration staff write "Karen", "Caryn", "Karin", or "Karyn"?
  4. Ethnic names, over a large country, also may suffer from the poor understanding of the host country. How exactly does one spell Dimitry? Dimitri? Dimytri? Moroch? Morocz?
So one might try to straighten this out with some simple recipe - One I often hear first is, "Why not use the first three letters of the first name, first three letters of the last name, and the date of birth?"

The problem with this approach, again, is that someone might register with a different name in a different clinic. Is it going to be "JOHSMI01011970" (John) or "JONSMI01011970" (Jon)?

Then the suggestions usually continue...

Q : "Dirk, what about by the Medical Record Number?"

The medical record number for this patient at your hospital (123456) may not be the same as the medical record number for the office down the street (654321).

Local medical record numbers might work for a hospital, or a small regional group (if you have centralized registration), but they generally don't work across different healthcare systems.

Q : "Hmmm... Why not use the social security number to identify people?

The social security number suffers from a few problems too :
  1. There is no check-digit in the social security number. A check-digit is a number (or series of numbers) that are mathematically linked to the other numbers, so you can figure out if the number has been falsified. The social security number was invented back in 1935, before things like "identity theft" were around. As a result, the social security number is probably one of the most abused identifiers, often used for fraud by criminals. 
  2. The social security number is a 9 digit number - So in total, we should be able to issue about 999,999,999 of them, BUT... because of certain restrictions (e.g. no numbers that start with 666, no numbers with -13- in them, no numbers with all of the digits the same), there is really only a pool of about 820 million to draw from. Currently the U.S. population is about 350,000,000. Which sounds OK, except that we maintain that number by having some people die every year, and some new babies added every year. In total, about 620 million numbers have already been handed out, so we could potentially run out of social security numbers sometime around 2100. Yes, that will be some time from now, and hopefully we will be able to fix that before it happens - but in our current political climate, will the government ever be able to assign a personal identifier again?
It's funny - I've spoken to informatics people around the globe, and they usually ask me "Dirk, why are you guys in America having so much trouble getting a national health record? In our country it's very simple - Either :
  • "...our national government maintains our national health record.
  • ... or ...
  • "...our national government assigns a health identifier for all citizens."
Well, the problem is that we're Americans. Authors like George Orwell and Ayn Rand have left a significant impression on our national consciousness. We just don't like the idea of the government assigning a number to track all of our health information. In fact, in 1998 Congress forbade the HHS, by law through HIPAA, from creating a health information identifier - Despite many groups asking for an identifier, and an estimated $77 to $154 billion savings in healthcare that a national patient identifier could provide. And perhaps (just to be fair), this is for good reason - see this letter opposing government-issued medical identifiers and this document summarizing the potential abuses. (Please note : I'm not taking sides, just presenting both sides of the argument.) 

Q : "So Dirk, how does the VA (Veteran's Administration) do it? I heard they saved lots of money through their VISTA/CPRS medical record, and their record is a major source of data for reasearch."

The VA essentially has a national patient record because, well, most veterans have a different opinion. When you ask most vets, "Do you care if the government has a number to track you?", they say things like "No, the government has been keeping a file on me since the day I enlisted!" right before they rattle off their rank and military ID number from memory. In reality, the VA has also been using Social Security numbers, but I understand there is currently a movement underfoot to move away from those identifiers to another number - I'm not an expert on the VA architecture, but this might explain why they divide the VA record up into different VISN systems. (Any VA Informatics people reading this willing to help explain the architecture?)

In short - 
  • The culture at the VA supports a nation-wide medical record number.
  • The culture of private and teaching hospitals (the "rest of America") does not.
This is why, when I get asked :

Q : "Dirk - The VA has free EMR software - Written by the government, so it's public domain - Why don't private hospitals use it?"

I usually answer, "Private hospitals *could* use it, but because of these culture differences they probably wouldn't see the cost and efficiency benefits that the VA did."

(In reality, there are also other support reasons why a private hospital might not implement CPRS/Vista - But that might be changing some with cool open-source projects like OpenVISTA.)

Q : "So Dirk, is there any hope for a national EMR? Will patient data ever be truly portable?"

Well, currently there is the NHIN/Direct project (see and which seeks to allow physicians to transmit patient data, securely, between different offices - But without a common patient identifier, this may not have the workflow some patients and most physicians ideally want. Still, it would allow a maximum of privacy and patient control, and it's at least a step in the right direction.

There are also a number of regional Health Information Exchanges currently in use, and new ones being built - But without a common patient identifier, nobody seems to be sure about how this is going to work on a bigger, national level.

So yes, if you're traveling from Texas to NYC for vacation - You had probably better bring your medication list and medical history written on a piece of paper, just in case you need medical care.

Finally - I think there is actually some hope for a solution to this that could fly politically in America.  I've tested the idea with both republicans and democrats and oddly, both seem to like it. It's called the voluntary patient identifier. Unfortunately, I think so far this effort suffers from poor understanding, poor marketing, and quite frankly, poor patient interest. 

But I think there is a way to change that - I'll describe it in my next post.

(Ooh - Cliffhanger ending!) :)

Always glad to share - Feel free to leave comments, thoughts, and questions! :)

Thursday, September 15, 2011

Where Exactly Do My Med Orders Go?

Where Exactly Do My Med Orders Go?

Ever wonder where your orders go? One of the things I do when training a doc on CPOE is explain to them the basic medication delivery workflow in a standard hospital. When a doc "goes CPOE", he/she is suddenly confronted with some hard realities of the ordering process - Exactly what they order, and exactly how they order it, will largely determine their success in getting a drug at the right time in the right place.

(PLEASE NOTE WITH THIS DISCUSSION : YOUR MILEAGE MAY VARY GREATLY - This workflow is a general summary, but lots of other hospitals may do things differently, and for a very good reason - I'm just summarizing some general themes, but for specific information, ask your directors and informatics staff.)


So I usually start to frame the discussion with the basic unit of care - A doctor, and a patient. (You'll notice these slides borrow from my piece on "What is Med Reconciliation, anyway?", where the physician and patient lovingly spend time with each other in an area I call the "patient care cubicle") -
  • Inside a hospital, this "patient care cubicle" is an "inpatient care cubicle".
  • Outside a hospital (or in an ambulatory setting like an ED or day surgery), it is an "outpatient care cubicle".
Let's focus first on the inpatient care cubicles, where a lot of medications are ordered as patients are admitted into a hospital :

It's first important to consider how exactly an order is created. If the doctor and patient are sitting next to eachother, then the basic unit of care (from the physician perspective) is the physician order - On this slide, it's labeled the "MD ORDER".

First, let's look at this "MD ORDER" for a medication in the INPATIENT setting :


There are then basically three different ways a med order can come into existence :
  1. As a written order - This is an acceptable way of making an order - Even the VA, during downtimes, allows written orders. Most hospitals with EMRs still allow some form of written orders for downtimes or other complicated orders. The problem : This type of order requires a nurse or pharmacist to transcribe it, so if the doctor has poor handwriting, this can result in a small percentage of error. 
  2. As a telephone / verbal order - This is also an acceptable way of making an order, especially if the physician's hands are tied up doing a procedure or surgery. The problem : This type of order also requires a nurse or pharmacist to transcribe it, so if the doctor doesn't speak clearly, this too can result in a small percentage of error. ("Did you say Metoprolol 15 or 50mg?") Another problem : This legally requires a physician to go back and sign this order afterwards - This is very complicated and time-consuming for most physicians.
  3. As a CPOE order (Computerized Physician/Provider Order Entry) - This is the entry method preferred by most hospitals, regulatory bodies, and the government. I suspect this is primarily because there's no intermediary who needs to transcribe the orders, so there is the common belief that this is less error-prone. I haven't seen really good data about this yet, but I do believe that good CPOE requires good training. There is also some data to suggest total medication turnaround time decreases with CPOE use - See this HHS piece on Medication Turnaround Time.)
So no matter how a physician creates the order on the floor, the order gets made. Let's say, for our little example, that the order is "Ativan 2mg IV x1 dose STAT".

Any idea where the order goes next? If you guessed "Pharmacy", you're right!

Whether you're in an electronic hospital, or a paper hospital, Joint Commission requires all inpatient medication orders for acute care hospitals to be "verified" by pharmacy. ("Verify" is basically a fancy word for "double check".)

So a pharmacist suddenly sees the medication order - "Ativan 2mg IV x1 dose STAT". And there are a few things a pharmacist can do to try to help ensure the safety of this order :
  • They can check the allergy profile of the patient.
  • They can check the dose of the drug.
  • They can check for drug-drug interactions.
  • In some hospitals, they can even sometimes do fancier checks, like check the renal dosing of the drug, weight-based dosing of the drug, etc. (this sometimes varies considerably, depending on the types of services offered)
What pharmacists can't really do well is verify the need for the drug - They usually aren't sitting in front of the patient, with the patient's chart - So if you order heparin on a patient with a bleeding ulcer, a pharmacist is probably not going to be able to prevent that type of error. (Nurses, usually right in front of the patient, are generally much better at finding that sort of error.)

Anyway, after a pharmacist does his/her best to verify the safety and dosing of the order, or adjust the order, they generally click a button to "verify" the order, and then the order travels back to the floor where it does two things :
  1. It unlocks the Pyxis drawer (in this example the Ativan drawer)
  2. It creates a blank entry on the eMAR (electronic Med Administration Record, aka Cardex, aka Codex)
This then allows a nurse to take the drug out of the Pyxis drawer, give the drug, and chart it on the eMAR. Once it's charted, that's generally when a hospital gets to generate a bill for having given the drug. 

So that's generally the way it works in the inpatient world.

Any problem with this pharmacy verification workflow? Having a pharmacist double-check the orders helps reduce errors, so ... Is there any drawback a doc should be concerned about?

Well, imagine if you had a patient seizing in front of you, and you had to give them the "Ativan 2mg IV x1 dose STAT". How long exactly does it take a pharmacist to verify those medication orders? 

The interesting thing is that there are actual guidelines about this (these are approximate - Your state/region may vary on this) :
  • Priority = STAT : In many places regulations allow up to 30 minutes (in reality, most inpatient pharmacies verify STAT orders in about 5-10 minutes)
  • Priority = ROUTINE : In many places regulations allow up to 90 minutes (in reality, most inpatient pharmacies verify ROUTINE orders in about 10-15 minutes)
So if the regulations are typically around 30 minutes, and even if your inpatient pharmacy can do it in 5 minutes, can you wait 5 minutes to give a seizing patient ativan?

The answer, of course, is obviously no. So what can you do? You generally have two choices :
  1. OPTION 1 : Call the pharmacy and say "I just put in an order for Ativan - Can you verify it ASAP so the nurse can take it out?"
  2. OPTION 2 : BYPASS the system. In most hospitals, a "Code Blue" (or "Rapid Response") is a perfectly acceptable reason for a nurse to hit the "Emergency Bypass" button on most Pyxis machines - This allows a nurse to get the drug and give it in an emergency. 
Note that most Pyxis machines actually TRACK the number of times they have had an emergency bypass - There should be a valid reason to bypass this important safety mechanism, so a unit where there are more emergency bypasses than needed/expected may be a cause for concern and investigation.

So that's generally the way medication orders get created, verified, and followed in an inpatient setting.


Here's where it gets interesting - That pharmacy verification thing? Joint Commission, on seeing the success of this in reducing errors, thought 'Wouldn't it be nice to have pharmacist verification in the OUTPATIENT setting (e.g. Emergency Department, Surgical Daycare, etc?)'

And do you know what happened? Most emergency departments, for good reason, argued "We can't handle the time delay of verification!" - And so, as of yet, it's not a mandate to have pharmacy verification in the Outpatient setting. (Perhaps for good reason - The delays it could cause could create a whole different set of problems.)

So how exactly *does* it work in the outpatient setting?

ORDERING : Generally the same process - A doctor can use either :

  1. Written order (same as above)
  2. Telephone / Verbal order (same as above)
  3. CPOE order (same as above)
The difference is, however, that without pharmacy verification, where exactly does the order go?

In most hospitals with electronic systems, then, this order then will simply go to the eMAR. (It makes a space in the software for charting the administration of the drug.)

So then the typical workflow is this :

  1. Nurse sees order in EMR or eMAR
  2. Nurse opens Pyxis (in most EDs, the Pyxis will open up automatically)
  3. Nurse gives the drug
  4. Nurse charts the administration on the eMAR
Again, after charting the administration, this usually lets a hospital generate a bill for having given the drug.

So in most Emergency Departments (mostly for reason of avoiding delays), there is no pharmacist between the doctor and the nurse part of the workflow. There are still two people double-checking every order - The physician, and the nurse.


The reason you, as a physician, might care about these workflows is that often, inpatient doctors are asked to admit patients from the Emergency Department.

And how exactly does a computer know which workflow to follow - The inpatient or outpatient workflow?

Let's recall - In both settings :
  • The physician usually places the order via written, telephone/verbal, or CPOE methods.
  • The nurse usually gives the drug to the patient
  • The nurse usually charts the administration on the eMAR.
So how does the computer know whether to follow the inpatient or outpatient workflow for a medication order?

Q : "Dirk, is it by physician type? Like, orders from ED docs follow the outpatient workflow, and orders from Hospitalists or other inpatient docs follow the inpatient workflow?"

A : Usually not. If systems were built to do that, then a Hospitalist would have significant challenge in running a code on a patient physically located in the ED.

The answer is usually, as we say in New York, "Location, location, location." That is, the location of the patient determines which workflow most EMRs will use on the order.

So for example - A patient in a location, and the workflow :

  • ED Room 1 = Outpatient
  • ED Room 2 = Outpatient
  • ED Room 14 = Outpatient
  • ED Holding Unit = Inpatient *
  • Room 125B = Inpatient
  • Room 202A = Inpatient
  • ICU Bed 3 = Inpatient
  • ICU Bed 4 = Inpatient
* - Remember - Most holding unit beds are actually an inpatient level of care, and so they follow the inpatient workflow.

There are a few reasons why I teach this to the docs I'm training on CPOE. They are :

  1. You are going to be a working doc - You need to know how to troubleshoot the system, just in case of emergencies.
  2. You might admit someone from the ED to an inpatient bed.

Reason #2 is especially important for Hospitalists and other inpatient docs to understand. Why? Because if you are admitting a patient, and the computer still shows the patient to be in an ED bed (e.g. ED Room 1), then those orders entered into your EMR may not be verified by pharmacy - Orders not verified by pharmacy means the Pyxis machine won't open up when the patient gets up to the floor. This is when you get that call from the nurses : "Dr. ____________, remember those orders you put in while the patient was in the ED? For some reason, the Pyxis machine isn't opening up now - Can you re-enter those orders in the computer?"

This, of course, leads to delays and frustration. So what's the best thing you can do as an inpatient doctor, if you get called by the ED to admit a patient to your service?

1. First, go down to the ED and see and evaluate your patient
2. Look on the computer at their location -
     - If it says "ED Room 1" - Ask your registration to change their bed to an inpatient location.
     - If it says "ED Holding Unit" or "125B" or "204A" or "ICU 3" - Go ahead and start your orders.

Again, please remember - Your hospital's workflows may vary - This is a gross generalization for teaching purposes only. Please ask your local regulatory agencies, your local administrators and directors, and your local informatics people for more details about the workflows in your hospital.

Hope you enjoyed this post - Feel free to send questions, I love the feedback and I'm always glad to create posts to answer them!