Coding samples from stillborn neonates

Using SNOMED CT
1

I received the following question from the NL scientific association for medical microbiology. They need a way to code samples, e.g. cultures, from stillborn neonates, with SNOMED.

These neonates generally do not get a social security number of civil identification number of their own; nor a patient identification number of the hospital. Cultures are typically put on their mother’s ID. But if they then use the SNOMED code for ‘throat swab culture’, it will appear as the throat swab of the mother - which is not correct.

One solution could be to always give stillborn babies a patient identification number of their own in the specific organization (hospital or laboratory). However, this will not adequately facilitate communication with other institutions or, for instance, insurance companies, because those will be unable to identify this person in the national registries. A practical solution for this specific use-case (cultures) is to culture for instance placenta instead of the baby, which is of the same compartment and should give the same result, but this implies a change in behavior from the clinicians - and would perhaps not work for every use case.

So: we are looking for a way to specify with SNOMED that the specimen came from the (stillborn) child of the patient.

We have discussed the use of the concept “neonate” (which in our opinion would also be appropriate for deceased individuals), but this is in SNOMED a person and not a specimen or body structure. Concepts which have been made for intra-uterine use, e.g. 309502007 |Fetus specimen| or 258428005 |Products of conception tissue specimen|, are not appropriate either: the child is full-term or nearly so. Moreover, the results from a culture in the sterile uterus (for instance from a amnion fluid aspirate) are very different from those of a culture after rupture of the membranes and birth.

Our final thought is to create a new concept for “sample from stillborn child”. This excludes the use of this concept in the context of a newborn that is alive but not has an ID yet. Another option is “sample from newborn child of patient”. This of course would include all newborns.

Does anyone have this or a similar use case? Thoughts, suggestions?

2

As you indicate, your problem is principally the result of longstanding failings in information models and their allied infrastructure: local or national identifiers with which to open a new record belonging to the neonate is typically either actively prohibited by policy or not enabled by actual practice, whilst individual patient record information models typically don’t provide a standard (or, often, any) means by which to unambiguously hold statements about third parties.

As a general rule, my personal advice would be that its always a bad idea to attempt to compensate for such failings by co-opting the terminology as the cheapest bit of the software stack to change. Such quick fixes simply add to the overall technical debt mountain, and will eventually bite you later. A large part of the current interoperability problem is down to the kind of blurring of the boundary between information and terminology models that results.

Your specific problem has been a feature of both maternity and transplant surgery records for years, but is also seen in the context of documenting abusers and abusees, and in certain aspects of care guardianship. In maternity especially, there is pervasive confusion or absent differentiation between the different patient actors involved such as when recording the antenatal or birth history of a now adult patient, and the antenatal or birthing history of a currently pregnant but multigravid mother. As a result, if you see any of these codes in the record of a young woman:

394698008 Birth history (situation)
736060002 History of exposure to diethylstilbestrol in utero (situation)
128161000119105 History of low birth weight status, less than 500 grams (situation)
407613009 Born by breech delivery (situation)
161769009 History of being infant bottle fed (situation)
138091000119101 History of prematurity (situation)
609424003 Suspected fetal damage from maternal alcohol addiction (situation)
77931000119108 History of fetal growth retardation (situation)

…then, without looking at the date of entry, how do you know whether these relate to the patient or to her offspring? Is a patient whose lifetime EPR contains this code:

200133007|Delivery by combination of forceps and vacuum extractor (finding)|

…more at risk of cervical spondylosis or uterine prolapse in later life?

If you were to make the neonate simply another specimen type, then you’re still effectively saying that the neonate is as much a biologic and genetic part of the mother as would be a neoplastic tumour. Whilst administratively convenient, this seems biologically incorrect: the neonate may not have a separate legal identity, but it does critically have a different genetic one.

It is therefore perhaps both significant, and symptomatic of the unresolved confusion in the middle of all this, that you could also already use SNOMED’s context model to create, for example, the following as SNOMED’s built-in solution to the question of how to identify statements about 3rd parties:

Neonatal throat swab taken ==
{
363589002|Associated procedure|=312880001|Taking throat swab|,
408730004|Procedure context|=385658003|Done|,
408732007|Subject relationship context|=133933007|Neonate|,
408731000|Temporal context|=410512000|Current or specified|
}

(Noting in passing the lack of a Natural child neonate (person) concept such as would be subsumed by both 133931009 Infant (person) AND 75226009 Natural child (person))

Meanwhile, the examples below all demonstrate that authoring is historically confused on the whole topic:

609415007 Suspected fetal disorder (situation)
77931000119108 History of fetal growth retardation (situation)
289446001 Fetal heart sounds absent (situation)

All three are fully defined, but all three currently specify 410604004 Subject of record (person) as the value of 408732007 Subject relationship context (attribute). This modelling therefore expects that they will only ever be recorded in the (non-existent) record of the foetus whereas in reality they are of course far more commonly added directly to the maternal record.

3

I don’t recall a requirement like this, but ‘subject’ and ‘subject relationship’ requirements are one of the more common “we know this isn’t a great way to do it” patterns we have encountered over time. What has actually been done in each case will depend on a whole load of other variables.

A couple of thoughts:

1. Since it hasn’t been mentioned yet it is probably worth pointing out the long-standing approved attribute 18170007|Specimen source identity|, defined thus:

Specimen source identity specifies the type of individual, group, or physical location from which a specimen is collected

…and takes a range of:

<< 125676002 | Person (person) | OR << 133928008 | Community (social concept) | OR << 260787004 | Physical object (physical object) | OR << 276339004 | Environment (environment) | OR << 35359004 | Family (social concept) |

As such it is apparently well suited to contribute to the formal definition of content of the form “X specimen from newborn” (and is already used in 122562003|Blood specimen from newborn|) and offer a glimmer of (formal) separation from a corresponding “X specimen” but is almost certainly as challenging as ‘subject relationship context’, ‘inheres in’ and ‘recipient category’ in terms of distinguishing whether ‘newborn’ refers to a ‘newborn subject of the record in which the statement is made’ or refers to a ‘newborn child of the subject of the record in which the statement is made’.

2. Do you know what happens if a more extensive set of post-mortem investigations were to be done - how are they recorded and distinguished from (but also possibly linked to) a maternal record? If there is an established convention for this can it not be co-opted for use by the medical microbiology team?

Ed