[clinical-curation] Thyroid module assumes every thyroid flag is hypothyroidism

[aerenkaradag]

Topic

src/syntha/modules/thyroid.py treats every patient with the Tiroid source flag as hypothyroid and prescribes levothyroxine (50 mcg/day). Hyperthyroidism / thyrotoxicosis / goitre are never modelled.

This assumption is currently silent (no tracking issue) but was surfaced during code review and looks like the highest-risk of the code-derived clinical assumptions.

What the system does now

# modules/thyroid.py
REASON = ("40930008", "Hypothyroidism (disorder)")   # hardcoded hypothyroid
TSH    = ("3055005",  "Thyroid stimulating hormone level test (procedure)")
...
for drug in LEVOTHYROXINE:   # always levothyroxine 50 mcg
    out.add(R.medication_request_resource(...))

The generic source flag Tiroid (SNOMED 14304000, "Disorder of thyroid gland") is mapped down to hypothyroidism + levothyroxine for all such patients.

What we need from a clinician

• Is it acceptable to model all Tiroid-flagged patients as hypothyroid in a Turkish cohort, or should hyper- vs hypothyroid be split?
• If a split is needed, is there a data-driven proxy (e.g. TSH direction) we could use, or should prevalence weighting be applied?
• Is levothyroxine 50 mcg/day the right default first-line for the hypothyroid subset in Turkish primary care?

Affected files

• src/syntha/modules/thyroid.py
• src/syntha/fhir/rxnorm.py (LEVOTHYROXINE)
• src/syntha/fhir/codes.py (if a hyperthyroid SNOMED/ICD-10 is added)

Context

This question is consolidated, together with the other Turkish-clinician validation needs, in the aggregate review document opened in #53.