Sunday, September 16, 2012

Here Is The Sort Of Thing That Makes Medicare Dispensing Information Pretty Useless Clinically.

Here is a little story which has both good news and the opposite.
First the good news is that when I made my weekly visit to the NEHRS the system was actually up and apparently working. It all seemed to working just ‘tickety boo’ but is, as always pretty slow and tedious.
However there are two issues I noticed, one of which dramatically decreases the value of the system from a clinical perspective.
Looking at my medication record I found these two entries (out of a total of 9 entries from 3 prescriptions - one of which is wrongly allocated to myself rather than my wife.)
First we have this:
Generic Name: OMEPRAZOLE
Prescribed:  30-Jun-2012
Supplied:  01-Jul-2012
Form & Strength:  TABLET 20MG
Code:  08333N
Second we have this record.
Generic Name:  OMEPRAZOLE
Prescribed:  05-May-2012
Supplied:  06-May-2012
Form & Strength: TABLET 20MG (AS MAGNESIUM)
Code:  09110L
(Note I have left out the quantity and repeats fields as they are identical)
The point to observe here is that an identical (generically substitutable) medication finds itself with two different codes which I can find no relationship between. These are the same generic medicine and should have exactly the same code if there was any clinical common sense applied to the coding. Indeed it is quite probable the medications are simply different packages of tablets manufactured by the same pharmaceutical major (Astra Zeneca).
What this means is that the coding system used by Medicare is essentially useless for clinical research and clinical decision support as it is not apparent that each is the same medicine. If any medication coding and history is to be useful the functionally and chemically identical same medications need to have the same code - it is as simple as that.
The problem is of course, that the designers of the Medicare system were concerned with bean counting and not clinical utility. Using information which might be fit for one purpose for another purpose is always fraught with risk.
Less important is that the system is clearly the designed for those with 20/20 vision. While the print button is hard to see - the search and restriction functionality on the top left is even more obscure. Hard to know what the tiny symbols mean until you click them. To me the system needs much larger controls on the screen and it needs a clear warning to users to read the ‘help’ carefully. I have not seen such a warning and it would be really sensible to have it up in big clear type - or to send users here to browse:
It all seems to be there - it just needs highlighting.
Extremely disappointingly the search function does not look into document content - just the document type so you can’t search for all instances of ‘omeprazole’ for example. As the record acquires more records I suspect this search will be less than very useful.
As far as the drug coding is concerned this really is a shame that in even in the tiny sample I can see such silliness has been allowed.


Anonymous said...

Wasn't the AMT meant to address the issue you are talking about? The PBS code, as you have suggested is built for a different purpose. But it might have been nicer if the Nehrs mapped and sorted it against the AMT- that would definitely be a value add that nehrs could provide over just viewing it in your Medicare Australia record. That must be coming in phase 2 ;).

Cris Kerr said...

' ... If any medication coding and history is to be useful the functionally and chemically identical same medications need to have the same code - it is as simple as that. ... '

I agree David. In my feedback on the PCEHR, in my template example I included the active ingredient/s.

Patto said...

The AMT (or mims etc) can work out the clinical equivalence here - as has been pointed out the PBS codes are really billing codes and not suited for this. The new PharmCIS data that PBS is trialling (this is the data that is used to publish the PBS schedule) has AMT mappings for each item and so could be used to facilitate finding the AMT code.

But the PharmCIS stuff is all pretty new so I wouldn't expect it to be in the pcehr module yet, if ever. My limited understanding of the plan (*cough*) is that medication lists are meant to come from GP systems, and as is well known these aren't happening yet.

I think the PBS data feed is just there because they had the data and some spare millions to spend - surely no one really believes it is of particular clinical usefulness - as you have pointed out the data appears 3 months after the event and can be error prone..

On the more interesting question as to whether GP health summary medication lists will be coded in a common format (which would be a useful result IMHO) I fear they also are not.

Anonymous said...

AMT is, as always, the direction. But it is not here yet, it needs takeup at the clinician end, in GPs, pharmacists and hospitals, and as already noted here, is coming for PBS in the not too distant future.

Given that, the question is whether PBS data has usefulness. Most of the clinicians I've spoken to think it does, as long as you understand the limitations.

The upside is that the PBS medications list is pretty complete for the areas that it covers. If a pharmacist dispenses a PBS item they almost always claim for it - otherwise they don't get paid. In that sense it is a very rich data source, as opposed to a data source that relies on clinicians and jurisdictions connecting to PCEHR - which inherently means the data would be incomplete until every clinician is connected.

On the downside, the data is coded with PBS codes, these have some limitations. The data doesn't cover every medication - there are a set of non-PBS community pharmacy medications, there are a set of medications dispensed in hospitals. Those are not captured from this data source.

And as with all Medicare data, we know that getting it to the right Medicare card is reasonably reliable, getting it on the right person's record on that card is occasionally not occurring. In a claiming sense that isn't the end of the world, in a clinical sense it's clearly not ideal.

Within those constraints, the question is whether there is value. If you ask a clinician (particularly one in a hospital) whether a list of drugs that a person was most likely prescribed is a useful thing, the overwhelming answer is yes. It never absolves them from asking the patient "did you actually take all these drugs" and "is there anything else not on this list that you need to tell me about", but it is an improvement upon asking an 85 year old "could you please give me a list of the medications you're currently on" when they turn up in an ED.

Andrew McIntyre said...

Use of the AMT is not going to help much as its brain dead and cannot do the things that are needed. Its probably been designed to allow pharmacy stock checks rather than clinical use. AMT is SNOMED-CT in name only, its virtually devoid of the semantics needed for clinical use. Someone will probably say its coming, but its a long way off judging by my exposure to the organisation. Another failed Nehta initiative in reality.

Patto said...

And yet the AMT can handle the above use-case perfectly well.. so you know, baby steps and all that..

Andrew McIntyre said...

You mean Billion dollar baby steps?? AMT can distinuish between different brands on exactly the same drug, this is true. However it defines no relationship between different PPIs, or eg Penicillin and Amoxil. It seems to be aimed at Pharmacy ordering, but nothing else from what I can see.

Its a question of fitness for purpose and value for money and on both fronts it totally fails. Its not fit for purpose, and the concern is the lack of concern about its capabilities.

Patto said...


The AMT is a perfectly valid clinically sound coding system that meets many use-cases in many different clinical settings - GP, hospital, pharmacy. But you already know that.

It doesn't meet ONE of YOUR use-cases of providing a SNOMED compliant substance grouping. And I don't doubt it is a super important use-case, but in keeping with my baby themed sayings, I think there is something about babies and bath water etc.

They didn't spend a billion dollars on it (the AMT), and I'm not sure how trashing the only medicines terminology we have saves us money, or gets us any closer to having a medicines terminology that does what you want.

(as opposed to a PCEHR which I would be happy if they trashed - because I don't think it's much use - but given everyone seems to agree in principle with a common medication coding system I don't see exactly where the alternative comes from?)

Patto said...

btw Andrew, I actually understand your concerns, and get frustrated by the pace of fixing problems, implementation etc.

I just think your spin is perhaps overly negative - and the risk of the overwhelming pessimism is that we go from a flawed but slowly improving FUNDED medicines terminolgy


absolutely nothing

And I'm not sure how nothing is any better?? Do you think there is scope here for a private player to make an impact? I'm genuinely interested in what you think might be a viable replacement.

Andrew McIntyre said...

The issue is that in reality its not a terminology, but a dumb flat list of drugs. We would have been far better licensing MIMS, if all we wanted was common codes. (And MIMS provides much more that that) If you can't get drug groups, substances or do allergy checking with AMT they it's a waste of time and money. I don't think the people running it are really aware of what a terminology should do, or more likely are to scared to put any smarts into it in case they get it wrong. We would have been far better off extending the existing SNOMED-CT medication hierarchy with local brand names. That was not good enough, so they opted for a safe alternative, that appears complex, but is actually so dumbed down its useless. Currently there is no value in using AMT, and surprise surprise its uptake is poor. In its current form its useless for decision support. Whats the point in sharing drug codes if you can't do any reasoning with them (Beyond translating between exact equivalence between brands). For what has been spent its a disgrace, but mirrors the productivity and utility of the general government eHealth spend, which is well over a Billion dollars. I am appropriately negative about what my tax dollars has been spent on.

krowwolf said...

AMT clearly is a terminology, when treated as such.

This can be seen in the latest version of Minnow (free download which includes a combined SNOMED CT-AU and AMT index.

The basics are as follows:

1. NEHTA has built a map between SNOMED CT substances and AMT substances.

2. This map is distributed as one of the Reference Sets in AMT (since about version 2.30)

3. We (AEHRC at CSIRO) have used this map and processed it with out terminology tools (Snorocket and Ontoserver) to build a combined version of SNOMED CT-AU (May 2011) and AMT (2.35).

What does this all mean? It means that certain AMT substances will now appear as equivalent to SNOMED CT-AU substances (in this combined version) and, as a consequence of this, belong (as descendants) to appropriate SNOMED CT-AU substance classes.

For example, the AMT substance 2714011000036109 | warfarin | Is a 59545008 | Pesticide | and also Isa 372862008 | Anticoagulant | which Is a 373708006 | Hematologic agent | and so on.