OHDSI Home | Forums | Wiki | Github

Differences in definitions between a diagnosis, a sign, or a symptom?

(Kieran Mace) #1

Is there a well accepted set of definitions for diagnosis, sign, and symptom?

How do they differentiate themselves?

Are symptoms patient reported signs?

Would really like a concise, well accepted differentiation between them.

(Christian Reich) #2

There isn’t, @kmace. That is why we combine them all into the Condition domain. Take nausea. Can be a diagnosis or a symptom. Vomiting could be diagnosis (very crude, but the patient just arrived in the ER) or a sign. We treat anything that’s wrong with a patient as a continuum.

(Roy Mathew) #3

I would agree but there is still a clinical difference.
Sign is a provider elicited response from the patient - one of the most common would be tenderness is a sign during a physical exam and pain is the symptom that the patient is reporting.
Tachycardia is a sign (and a diagnosis) because the patietn may not know that the pulse is fast. Palpitation is a symptom but doesn’t suggest rate, it is the sensation of the heart rate - which can be normal yet strong like in anemia or possibly in a high output state.

If you look at Stedman’s Medical Dictionary - The Diagnosis is the final decision of what is ailing the patient.

However - this distinction in an EMR is not straight forward. Diagnoses would like be the principal condition under the condition domain in an inpatient setting. But if it is the admission diagnosis - it could still be a symptom or a sign.

i would think that conditions in an EMR should be considered diagnoses?

I think signs and symptoms would have to be extracted through NLP in free text or unstructured notes?

(Kristin Kostka, MPH) #4

In OMOP thinking, the answer is not always. We rely on our Standard Vocabulary here. Our good friend the DOMAIN_ID field in the CONCEPT table tells us where a concept belongs in OMOP. There are plenty of ICD10CM codes that don’t belong in the CONDITION_OCCURRENCE table because they’re not active conditions but rather clinical things we’re informed about during a patient encounter (e.g. family history of [something], nicotine dependence, etc).

It depends. Plenty of clinical encounters log symptoms. For instance, I fell on ice and bruised my rib last March while walking my dog. I have “rib pain” as a code in my EHR problem list that day. All depends on the source coding. Maybe your PCP is coding it. Maybe it’s in the clinical narrative.

Long way of saying: @Christian_Reich is right. Signs and symptoms come up in all sorts of ways but they’re all usually stored in the CONDITION_OCCURRENCE table if they’re coded at all. If they’re locked in a narrative, they’d be in the NOTE table (maybe – some sites don’t like to put things here because PII).

@agolozar did some work on signs and symptoms in COVID patients in her recent testing paper. She could weigh in on the fun of finding it in the data.