OHDSI Home | Forums | Wiki | Github

ICD10CM concepts representing death

There are a few codes in ICD10CM that may represent death.
For instance,
S06.0X7 Concussion with loss of consciousness of any duration with death due to brain injury prior to regaining consciousness
(the whole list’s attached examples_death.xlsx (12.5 KB))
We are thinking of implementing some guidelines on deciding whether or not those S-group condition concepts have actually resulted in the death of a patient or not. The assumption is that if no events/some suspicious events happen after the appearance of these codes then we can infer the patient is dead; as brain injuries require stay in ICU and drug administration, some records must persist after S-codes occur.
We’ve looked into patient data to find out what concepts actually occur after the alleged death.
From the medical perspective, the following concepts when appearing after death do not revoke it:

  1. Cardiopulmonary resuscitation (eg, in cardiac arrest) 92950
  2. Cardiac arrest 427.5
  3. Accident caused by unspecified electric current E925.9
  4. Acute respiratory failure 518.81
  5. Atropine, adrenalin administration

So, the idea is to use ICD10CM S codes to determine the death when we see no events or events from the list above after the death.

Please share your thoughts.

Did anybody test these in real data? If the patient really died?

Hm, rather lonely chat with 291 views and no answers.
@Christian_Reich Let’s do that; at least for the sake of closing the hot debate here.
Can be tested using wash-out period of, say, 90 days.

Agreed. Test against death record (not often), and cessation of data with the distribution.

I am also curios if subsumed death in a diagnostic code is also in terminologies used outside US.

E.g., in UK or France, are there billing diagnostic codes that subsume death as well.

And keeping within the thread - there is also ICD9CM and same problem. (as in this related thread) Data quality metric around death (discussion; location of death)

Curious if this study was ever run? What are the results? Do the ICD “death” codes really represent a Person who is no longer alive? Or are these ICD death codes erroneous data?

EHR data I have access to don’t have these codes, but as far as I remember the original post was related to a claims data instance.

Ok, thanks for the info!