OHDSI Home | Forums | Wiki | Github

What makes observations distinct from measurements or disorders?

General
#1

I’d like to get a better understanding of what the observation table can “say” vs what the conditions and measurements tables can say.

My team at the University of Pennsylvania strives to make a distinction between data that has been gathered about a patient and the processes the patient participated in, like health care encounters, but also disease processes themselves.

The data I’m working with right now comes from the Synthea project if that makes any difference. I’m using their OMOP ETL with very few modifications. I have my top 20 most frequent conditions, measurements
and observations and at the end of this post.

It seems like each conditions row could be a diagnosis, using SNOMED codes in this case (instead of ICD codes). Or am I supposed to interpret this as meaning that the patients were definitively suffering from the mentioned conditions?

It seems like many of the measurements are objective, continuous values that one would obtain with an instrument, like a meter stick, a sphygmomanometer, or a clinical chemistry analyzer. But I also see some more subjective, categorical (or self-reported) items in there, like pain severity and smoking status.

I see histories and allergies in observations… can’t many of those be expressed with SNOMED or ICD-10 Z codes, like you might find in disorders? I also see BMI-based obesity findings and smoking behavior, like you might find in measurements.

So what’s observations “unique selling point”? Where can I read more about the intention or semantics of these tables?

thanks for reading,
Mark

select
	cond.code,
	cept.concept_name,
	count(1) as condcount
from
	cdm_synthea10.conditions cond
join cdm_synthea10.concept cept on
	cond.code = cept.concept_code
group by
	cond.code,
	cept.concept_name
order by
	count(1) desc
limit 20

my joining of the concept table is probably a little under constrained here

code concept_name condcount
444814009 Viral sinusitis 1200
195662009 Acute viral pharyngitis 721
72892002 Normal pregnancy 568
10509002 Acute bronchitis 531
162864005 Body mass index 30+ - obesity 468
38341003 Hypertensive disorder 294
15777000 Prediabetes 289
40055000 Chronic sinusitis 234
65363002 Otitis media 194
19169002 Miscarriage in first trimester 191
43878008 Streptococcal sore throat 152
44465007 Sprain of ankle 125
408512008 Body mass index 40+ - severely obese 109
55822004 Hyperlipidemia 92
196416002 Impacted molars 86
82423001 Chronic pain 83
55680006 Drug overdose 78
68496003 Polyp of colon 76
124171000119105 Chronic intractable migraine without aura 76
75498004 Acute bacterial sinusitis 74 
select
	meas.measurement_concept_id,
	m_cept.concept_name,
	u_cept.concept_name,
	count(1) as obscount
from
	cdm_synthea10.measurement meas
join cdm_synthea10.concept m_cept on
	meas.measurement_concept_id = m_cept.concept_id
join cdm_synthea10.concept u_cept on
	meas.unit_concept_id = u_cept.concept_id
group by
	meas.measurement_concept_id,
	m_cept.concept_name,
	u_cept.concept_name
order by
	count(1) desc
limit 20
measurement_concept_id concept_name concept_name obscount
43055141 Pain severity - 0-10 verbal numeric rating [Score] - Reported No matching concept 12701
43054909 Tobacco smoking status NHIS No matching concept 10484
3036277 Body height centimeter 10344
3012888 BP diastolic millimeter mercury column 10344
3004249 BP systolic millimeter mercury column 10344
3025315 Body weight kilogram 10344
3038553 Body mass index kilogram per square meter 8997
3000483 Glucose [Mass/volume] in Blood milligram per deciliter 4577
3051825 Creatinine [Mass/volume] in Blood milligram per deciliter 4577
3032503 Calcium [Mass/volume] in Blood milligram per deciliter 4577
3018572 Chloride [Moles/volume] in Blood millimole per liter 4577
3014094 Carbon dioxide, total [Moles/volume] in Blood millimole per liter 4577
3004295 Urea nitrogen [Mass/volume] in Blood milligram per deciliter 4577
3000285 Sodium [Moles/volume] in Blood millimole per liter 4577
3005456 Potassium [Moles/volume] in Blood millimole per liter 4577
3022192 Triglyceride [Mass/volume] in Serum or Plasma milligram per deciliter 3809
3027114 Cholesterol [Mass/volume] in Serum or Plasma milligram per deciliter 3809
3009966 Cholesterol in LDL [Mass/volume] in Serum or Plasma by Direct assay milligram per deciliter 3809
3007070 Cholesterol in HDL [Mass/volume] in Serum or Plasma milligram per deciliter 3809
3004410 Hemoglobin A1c (Glycated) percent 3395 
select
	obs.observation_concept_id,
	cept.concept_name, count(1) as obscount
from
	cdm_synthea10.observation obs
join cdm_synthea10.concept cept on
	obs.observation_concept_id = cept.concept_id
group by
	obs.observation_concept_id,
	cept.concept_name
order by
	count(1) desc
observation_concept_id concept_name obscount
4060985 Body mass index 30+ - obesity 468
4256640 Body mass index 40+ - severely obese 109
4304110 Allergy to mold 86
439406 Allergy to animal dander 83
4302207 Allergy to grass pollen 73
4048169 Allergy to house dust mite 66
4306014 Allergy to tree pollen 62
433644 Shellfish allergy 50
4323208 History of appendectomy 41
4324181 History of cardiac arrest 41
45766064 History of single seizure 40
4174876 Allergy to bee venom 34
4240902 Allergy to peanuts 34
4219399 Allergy to fish 28
4169137 Allergy to wheat 26
438614 Allergy to nut 26
4163874 History of myocardial infarction 23
442116 Allergy to eggs 22
4102123 Latex allergy 21
4139681 Allergy to dairy product 19
42709996 Smokes tobacco daily 13
4038238 Suspected lung cancer 9
36684378 Allergy to soy protein 9
4058850 H/O: lower limb amputation 3
4168004 Burn injury 1
#2

I can’t speak to what the founders intended (does this make me a strict constructionist? :smiley:), but it what I observe looking at the values that you’ve drawn is that some of the observations are things that the physician may not have observed directly (like the history of a disease, that the patient would tell you but you might not otherwise know). It does seem that allergies may be a diagnosis or condition rather than an observation.

I would expect pain to be a measurement even if it’s subjective because the physician cannot observe or judge the pain; only the patient can judge the pain. So, I would say that measurement is the appropriate place for it.

1 Like
#3

You are definitely not the only one who gets confused :slight_smile:
Long answer short: Observation doesn’t have a selling point as for now, it’s rather a vague domain for everything that won’t fit other domains.

Histories and allergies are more or less simple - History of goes to Observation according to existing conventions. Same with allergy, as we assume that these are usually self-reported unless a specific presentation (e.g. rash or rhinitis) is stated. Latter goes to Condition. As you pointed out, some ‘subjective’ concepts can also be found in Measurement, which is justified by the assumption that they are scales or questionnaires to be filled at the point of care.

I’m not particularly sure that Body mass index XXX - obesity/severely obese etc. should belong to Observation. Look like valid conditions to me.

1 Like
t