OHDSI Home | Forums | Wiki | Github

Phenotype Phebruary Day 15 - Acute Myocardial Infarction (STEMI/NSTEMI/UA/Chronic Angina)

Part of the Valentines week

Prep work:

  1. Start with clinical description here - key insights
  1. Review clinical vignettes to understand the phenotype from an individual patient or treating clinician perspective.

Key insight:

  1. There are significant difference in experiences based on whether it is UA/NSTEMI/STEMI from patients perspective.
  2. People may be hospitalized with suspected cases of Myocardial Infarction.
  1. Literature review - understand prior work
    Key insights:
  • Most publications document good validity of cohort definitions. There has been significant prior work
  • We reviewed the codes used in the definitions from prior studies. Most have removed Old MI related codes from the entry event criteria - there is debate about the use of “subsequent” MI codes here and here and is currently unresolved. How would we generate empirical data to support this decision? Should we keep or not keep subsequent MI codes.


  1. A starting set of cohorts (3 each) for Angina pectoris, unstable angina, NSTEMI, STEMI and Acute MI were built - and are in https://data.ohdsi.org/phenotypePhebruary/ and ATLAS . Evaluation is being done using the framework described here

The paper attempts to structure the evaluation to detect three types of errors - and the evaluation below will follow that framework.

Sensitivity error - we may say that a given cohort has sensitivity errors, if we observe evidence that there are persons in the data source who may have the phenotype of interest but are not identified by the PA.

Specificity error - we may say that a given cohort has specificity errors, if we observe evidence that subjects identified by the PA do not have the phenotype of interest.

Index-date misclassification - we may say that a given cohort has index-date misclassification error, if we observe evidence that subjects identified by the PA have incorrect designation of event start date.

  1. Evaluation of phenotypes using Cohort Diagnostics at https://data.ohdsi.org/phenotypePhebruary/

Key points:

  • I like to start with cohort definitions that capture the occurrence of the cohort for the first time in persons history, evaluate those cohorts and then decide if the definitions can be changed to allow multiple records per person.
  • Since this phenotype is an acute event that is expected to be managed (per clinical description) in a hospital - datasources that do not have inpatient records will not be considered in this evaluation (because these datasource maybe expected to have sensitivity errors by not capturing persons experience while in hospital)

As described in the hypertension post I start with my notes - sources from medical text books

The phrase Acute Myocardial Infarction - is ambiguous and i read three related chapters in the text book and summarized them as my notes below

Chronic Stable Angina, Unstable Angina, Non ST Segment Elevation of Myocardial Infraction, ST Elevation Myocardial Infarction

Chronic Stable Angina, Unstable Angina, Non ST Segment Elevation of Myocardial Infraction, ST Elevation Myocardial Infarction

Chronic Stable Angina:
most common clinical manifestation of coronary artery disease (CAD) - imabalnce between myocardial metabolic O2 demand vs supply commonly because of atherosclerotic coronary artery obstruction.
Associated with exertion or emotional upset
Relieved quickly by rest or nitroglycerin

History of established risk factors: Cigarette smoking, hypertension, hypercholesterolemia, diabetes, obesity
EKG - maybe normal
Stress testing - Exercise vs IV dipyridamole - radionuclide/EKG/MRI vs.
Coronary arteriography - in many
Outpatient management
Identify and manage risk factors (Cigarette smoking, hypertension, hypercholesterolemia, diabetes, obesity)
Aspirin 81-162mg/d +/- ACE inhibitor
Acute: sublingual nitroglyerin q5min
Long term: long-acting nitrates, beta blockers, calcium anntagonists, ranolazine
Mechanical revascularization - percutaneous coronar intervention with stent if stenosis, coronary artery bypass graft
Maybe self limited or progress to unstable angina/myocardial infarction if not managed.

Unstable Angina and Non ST Elevation Myocardial Infarction
UA and NSTEMI are similar in mechanism, presentation and management - and considered part of spectrum
NSTEMI is UA with evidence of myocardial tissue injury (necrosis/death e.g. elevated cardiac biomarkers)
New onset severe angina (more than usual), angina at rest/minimal exertion, recent change in angina pattern.
Different from STEMI by EKG - no ST elevation, no Q wave changes
All: nitoglycerin, beta blockers
Hospitalize - except in very low risk
Low risk of NSTEMI: serial EKR and biomarkers +/- anti thrombotic therapy + stress testing once chest pain resolves
High risk or NSTEMI:
Antithrombotic therapy (ASA, platelet receptor antagonist, anticoagulant)
Reperfusion therapy - immediate, early, delayed or ischemia-guided
Depends on risk factor management.

ST Elevation Myocardial Infarction
Characteristic EKG, symptoms and signs. Intense acute chest pain.
ST elevation, Q-wave development
Elevated cardiac biomarkers
Hospitalize, emergency
Reperfusion therapy
Primary PCI > fibrinolysis
Aspirin, Platelet receptor antagonist
Short term outcomes depends on time between symptom and revascularization
Degree of cardiac necrosis + complications (arrythmia, heart failure, shock, hypotension, mechanical complications, aneurysm, pericarditis)

My take based on the review is

  • a person may NOT have more than one of the clinical ideas at one point in time - and STEMI > NSTEMI > UA > chronic angina.
  • if you have STEMI, you cannot have NSTEMI, UA . Also - you can have history of chronic angina before, during or after STEMI/NSTEMI/UA – and chronic angina is NOT myocardial infarction,
  • We could debate if UA is a MI – because a person may not have cardiac tissue injury/necrosis - but the presentation and treatment is almost identical - and it is probable that we cannot differentiate between the phenotypes in observational data. The “debate” is an indication to me that phenotyping them as separate clinical ideas maybe challenging - and there may be cross over.

Next step - clinical vignette

This vignette describes chronic (stable) angina pretty well - and it helps me understand the clinical profile of the persons belonging to this category


This appears to be a case of myocardial infarction (most probably a STEMI but the presentation appears to be late)

Here is a case report - where a person evolved into STEMI while in direct observation (with serial EKG)

A case report - where the chest pain has already resolved

And another (again not MI)

Next step literature review - there appears to be many publications that have evaluated the validity of cohort definitions (or atleast code sets) in observational data. From common attributes that i was able to find

Thank you again Ms. Gayle Murray for leading the effort on this systematic literature search. Her contribution is also described here

  1. Many have limited to age > 18, some have limited to age > 65
  2. Many have provided PPV, but only two have provided other measures such as sensitivity and specificity
  3. Some limited the definitions to inpatient/hospitalizations only.
  4. Some have used the primary diagnosis to limited eligible population.

Some key materials are

sensitivity specificity PPV NPV
98% (94–100%) 91% ( 83–97%) 95% (89–98%) 97% (91–100%)

BMC Health Serv Res. 2018 Nov 26;18(1):895.

Based on the literature search - the codes used by various literature sources were collated (presenting only ICD and DRG for simplicity)

Code Vocabulary
121 DRG
122 DRG
123 DRG
I21 ICD10
I21.x ICD10
I22 ICD10
I23 ICD10
I21.x ICD-10-CM
I22.x ICD-10-CM
410.0 ICD9
410.01 ICD9
410.02 ICD9
410.1 ICD9
410.11 ICD9
410.12 ICD9
410.2 ICD9
410.21 ICD9
410.22 ICD9
410.3 ICD9
410.31 ICD9
410.32 ICD9
410.4 ICD9
410.41 ICD9
410.42 ICD9
410.5 ICD9
410.51 ICD9
410.52 ICD9
410.6 ICD9
410.61 ICD9
410.62 ICD9
410.7 ICD9
410.71 ICD9
410.72 ICD9
410.7x ICD9
410.8 ICD9
410.81 ICD9
410.82 ICD9
410.9 ICD9
410.91 ICD9
410.91; ICD9
410.92 ICD9
410.x ICD9
410.x0 ICD9
410.x1 ICD9
410.xx ICD9
410 ICD-9
410 ICD9CM
410.x ICD9CM
410.X0 ICD9CM
410.X1 ICD9CM

AHRQ CCS DXCCSR ICD-10-CM code list for Acute Myocardial Infarction
as described here

These lists of ICD-9 and 10 codes are very similar to those I’ve used to identify acute myocardial infarction in studies I’ve worked on. However, when our goal has been to identify an actual AMI event date, we don’t include codes for old MI (ICD-9 412 and ICD-10 I25.2), or any codes indicating a “subsequent episode of care”, such as 410.x2 (where ‘x’ can be 0 through 9).

1 Like

Agreed, @ershanno. I know in past OHDSI trainings (and in the cohort definition in Atlas Phenotype), we’ve simply excluded old MI from the MI standard concept, but it does look like concept 45766114 (Subsequent ST segment elevation myocardial infarction) could be a candidate for exclusion as well. I am trying to see if this concept truly acts a “subsequent” event in our data.

The ICD9CM range for old subsequent MIs (410.x2) do not seem to map to anything about the event being subsequent. Perhaps a mapping fix is needed?

So @Ajit_Londhe @ershanno what do you think of the clinical description written here

I agree with you the old MI does not fit well here - do you think we need to clarify that in our clinical description?

I know it’s a little off topic but I think this is where your use case maters. If you need to know that today is the day that you first had a heart attack this is very different from knowing if you have had an AMI or CAD or UA in the past for trial inclusion or covariate adjustment. We validated and fairly easily do you meet a clinical condition for outreach for enrollment in a clinical trial: Validation of a claims-based algorithm identifying eligible study subjects in the ADAPTABLE pragmatic clinical trial
A dental (they make me smile!) note indicating past medical history and medication reconciliation is enough to aid in confirmation of trial eligibility. Versus I need to know that today is the day some period post exposure to a COX-2 inhibitor that you first had an AMI.
Our clinical descriptions may change across research settings necessitating alternate phenotypes.

Absolutely - that why we want make writing of clinical description upfront a best practice.

So i think we agree - every cohort definition should accompany a clinical description that is based on the use case.

Just built the cohorts and am executing cohort diagnostics - lets review results tomorrow

I am anxious to learn from Cohort Diagnostics - can we truly differentiate between STEMI/NSTEMI/UA/Chronic Angina. In previous OHDSI work - i believe we have not differentiated between them.