OHDSI Home | Forums | Wiki | Github

Acute Liver Failure

#1

Hi all,

At my first LAERTES work group meeting (which I thoroughly enjoyed) I noticed during one presentationa minimum of 8 representations or mappings involving acute liver failure.

I am curious to learn more about this. I can certainly see that many diseases may map to acute liver failure, but I wondering about the actual concepts/manifestations being used in this categorization if each mapoped relationship is not a disease. For example hepatic encephalopathy could be one, aterixis/live flap could be another, decreased hepatic synthetic capacity/coagulopathy could be another, portal hypertension/cirrhosis another etc etc.

Apologies if question is stupid, I have much learning to do.

Thanks

Manfred

#2

@Manfred:

I wasn’t in the meeting, so can’t tell you what was discussed. However, we have an older list of DILI definitions, which contain more or less of the sequelea of the actual liver failure you are alluding to. Because nobody really knows what works and what doesn’t work, so the easiest is to test and find out.

However, portal hypertension or cirrhosis - is that really acute? Sounds more like a chronic deterioration of liver function.

#3

Hi Christian,

Yes you are correct-that pair is chronic but I thought could conceivably map via acute-on-chronic somehow.

I’m just trying to understand the structure.

Thanks!

Manfred

#4

@hergchan:

Yes, that is correct. Again, the proof is in the pudding: Just test the hypothesis, and you will find out.

#5

@Christian_Reich, @ericaVoss, @lilipeng - can any of you provide a pointer or a copy of those definitions to this thread? thanks!

#6

@rkboyce:

They are the old OMOP definitions:

HOI Codes Acute Liver Failure Definition 1.xls (174 KB)
HOI Codes Acute Liver Failure Definition 2.xls (111.5 KB)
HOI Codes Acute Liver Failure Definition 3.xls (130.5 KB)
HOI Codes Acute Liver Failure Definition 4.xls (141 KB)
HOI Codes Acute Liver Failure Definition 5.xls (24 KB)
HOI Codes Acute Liver Failure Definition 6.xls (27.5 KB)
HOI Codes Acute Liver Failure Definition 7.xls (32.5 KB)

Querying LAERTES and data in OHDSI reference set
#7

Thanks both for providing and helping me get up to speed.

In examining each definition it appears that there are quite a lot of clinical components in them that are very nonspecific and in some instances nowhere close to qualifying as acute liver failure yet the appearance of one of these could meet the definition if I understand icorrectly.

Perhaps I need to understand better exactly how these terms were chosen and are being utilized and what is meant by the statement that these were the definitions that “worked”

Thanks

Manfred

#8

Manfred:

Yes, that’s correct. Some of the definitions attempt a very precise capture (high precision), others cast a very broad net (high recall). The idea was to find out what “worked”, meaning, what created the best discrimination in the OMOP experiment, testing known positive and negative drug-DILI cases. It turned out that the ideological debates about what to include or not to include in the definition are not justified and the potential effect of “misclassification” is widely overrated. At least in the cases we tested.

#9

But is the HOI DILI or acute liver failure?

In the accompanying tables there are some marked in one column according to whether they are mapped and another column labeled descendent.

What do those column labels mean exactly?

Thanks

#10

Well, LI, if you want, liver injury. Whether it is drug induced is something that we are trying to established.

The descendants are those concepts that are not part of the explicit definition, but implicitly included by using the hierarchical information in CONCEPT_ANCESTOR.

#11

But that is an important difference (liver injury versus acute liver failure) in the outcome definition.

Also if you use sufficiently broad terms in the case definition, and then use that definition to assign positive and negative control status to drugs, your observed performance is for liver injury, not just acute liver failure.

So maybe these definitions that are currently labeled/described as acute liver failure definitions are more aptly labeled as liver injury and should be changed in the results we were reviewing in past week.

Why are some terms that don’t imply liver injury, such as Reidel’s lobe, assessed as descendants? That is simple anatomical variant/accessory lobe whose major clinical importance is that it can be mistaken for a neoplasm/mass (although rarely it can undergo torsion).

Thanks

Manfred

#12

@hergchan:

Actually, what are you trying to do?

With respect to the liver injury, with failure or without: It would be great if we could figure out a way to establish the optimal phenotype for this and make it available to the public. Because there isn’t one. In the literature, everybody invents their own definition, from very broad to very specific diagnostic codes, or based on the sequelae of a liver function impairment, Hy’s law, or other lab tests. So, we recreated a suite of those different definitions to see whether it makes a difference to the actual question: Does the drug cause harm to the liver? And it turns out that the precision of the definition will not have a big effect on that ability to detect those effects.

I am fully aware that some of those descendant conditions no way could have been caused by the drug. But they were generated based on a bunch of ICD-9-CM codes, which were translated to SNOMED. Some of the ICD-9-CMs are so badly defined, that after translation and pulling the descendants you get weirdo concepts in there. But it turns out it doesn’t matter: if you get a few Reidel’s lobes in there they will be distributed equally in the exposed vs the non-exposed patients (since the drug has nothing to do with it), and in the worst case you will water down the effect a little. Since they are so rare, I doubt it it will make the slightest dent.

#13

Well I think i would want to do a few things including but not limited to the following: One is identify the most meaningful and/or useful phenotype. Next I would want to have a good label for that phenotype that made the most sense to the most people.

If one is including very non-specific events including any (even asymptomatic) elevations of transaminases or LDH or other non-specific abnormalities in the phenotype then I submit that acute liver failure is not the best label for the phenotype. It’s seems to me that the HOI / phenotype is more accurately labeled as liver injury or drug induced liver injury.

For Reidel’s lobe it is even more fundamental then drug versus non - drug induced. It is not even an injury but I guess when you are mapping and linking so many concepts and relationships they don’t always make sense.

Thanks

Manfred
Sent from my iPhone

#14

@hergchan:

That would be wonderful to have a well-curated phenotype library.

#15

DILI is a challenge in observational data. Here is an article by Rick Hanssen looking at a few different OMOP 1.0 definitions for Acute Liver Injury (as well as AKI and MI) and their performance on manual review. Overall the PPVs were not great, and a lot of chronic liver disease got pulled in as acute.

Improved phenotypes for acute liver injury would be a valuable contribution. I think the stronger exclusion criteria would be the key.

Jon

t