A single note mentions the same condition many ways — "DM2," "type 2 diabetes,"
"diabetes mellitus" — across PMH, HPI, and A&P, some negated, some historical.
A usable problem list collapses those mentions into one concept per problem,
drops what the patient does not have, and assigns a clinical status (active /
resolved / historical). This skill turns OpenMed's per-mention entity stream plus
ConText axes into that reconciled, de-duplicated list, shaped for USCDI "Problem"
exchange.
extracting-clinical-entities and resolving-clinical-context, when theuser wants a clean problem list, condition reconciliation, or dedup of repeated
diagnosis mentions.
mentions.
one entry per concept.
import openmed
from openmed.clinical import resolve_span_context, NEGATED, HISTORICAL, HYPOTHETICAL
note = ("PMH: type 2 diabetes, prior MI 2019 (resolved). "
"A&P: poorly controlled DM2; denies chest pain.")
ents = openmed.analyze_text(note, model_name="disease_detection_superclinical",
output_format="dict")
def normalize(surface: str) -> str:
# Cheap synonym folding; replace with SNOMED grounding (out-of-process).
s = surface.lower().strip()
return {"dm2": "type 2 diabetes", "diabetes mellitus": "type 2 diabetes"}.get(s, s)
problems = {} # concept -> reconciled record
for e in ents:
surface = e["word"]
ctx = resolve_span_context(surface, note)
if ctx.negation == NEGATED:
continue # patient does NOT have it -> exclude
concept = normalize(surface)
status = ("resolved" if ctx.temporality == HISTORICAL else
"active")
if ctx.temporality == HYPOTHETICAL:
continue # not asserted as present
rec = problems.setdefault(concept, {"concept": concept, "status": status,
"mentions": 0})
rec["mentions"] += 1
# Active anywhere wins over a historical mention of the same concept.
if status == "active":
rec["status"] = "active"
problem_list = list(problems.values())
# -> [{"concept": "type 2 diabetes", "status": "active", "mentions": 2}, ...]
# "chest pain" excluded (negated); "MI" -> historical/resolved.
analyze_text across the wholenote (or per section if you ran segmenting-clinical-sections).
resolve_span_context (or theaxes from resolving-clinical-context): negation, temporality, uncertainty.
NEGATED mentions (patient denies /no evidence of) and HYPOTHETICAL mentions (conditional, not asserted). These
must never land on the active list.
(abbreviations, word order, lexical synonyms) to a single canonical key.
Cheap normalization gets you started; SNOMED CT concept grounding is the
robust path — run it out-of-process with the user's own license and key on the
concept code, not the surface string.
RECENT/activeanywhere (typically A&P) is active; one seen only as HISTORICAL
("history of," "resolved," PMH-only) is resolved/historical. Active wins
over historical when the same concept appears both ways.
count, and provenance offsets — shaped for USCDI Problem / FHIR Condition.
extracting-clinical-entities: consumes analyze_text Diseaseentities. Run on a sectioned note (segmenting-clinical-sections) for best
active-vs-historical signal.
resolving-clinical-context: this skill depends on the negation /temporality / uncertainty axes — reconciliation without them would put "denies
chest pain" on the active list.
from openmed import analyze_text and`from openmed.clinical import resolvespancontext, NEGATED, HISTORICAL,
HYPOTHETICAL`.
clinicalStatus active/resolved (from temporality) and verificationStatus
refuted/provisional (from negation/uncertainty). SNOMED CT codes are
user-supplied and grounded out-of-process — OpenMed produces the dedup'd
concept and status, not the terminology binding.
normalizer knows the synonym. Lexical folding handles the easy cases; lean on
SNOMED CT grounding for real reconciliation, and never bundle SNOMED — call it
out-of-process with the user's credentials.
plus "asthma exacerbation" in A&P is one active problem, not two entries.
Aggregate before assigning status.
HISTORICAL /"resolved" stays resolved; don't promote it to active just because it appears.
problem-list entries. Keep them out entirely.
can trace each entry back to the note text.
support for clinician review, not an autonomous diagnosis.
https://www.healthit.gov/isa/united-states-core-data-interoperability-uscdi
clinicalStatus (active/resolved) andverificationStatus: https://hl7.org/fhir/R4/condition.html
https://www.snomed.org/