OHDSI Home | Forums | Wiki | Github

Differences between Cohort and Phenotype


(Tanvir Ahmed) #1

Hi, while trying to understand about Phenotyping, I got a bit confused about the differences between cohort and phenotype. Can someone help me by explaining their differences?

(Jon Duke) #2

@tanvir Understandable! Quite easily confused.

The way I like to think about it is:

  • A phenotype = cohort definition
  • A cohort is the group of people who match that definition.

So a cohort is a list of people and dates, while a phenotype is a set of criteria.

Others may have a more extended perspective on the topic.



ps. One other subtlety perhaps is that a cohort definition sort of implies you are defining “people with X” whereas a phenotype can be defining “people with X” or simply “X” in the abstract, whether or not you are applying it to cohort creation.

(Gregory Klebanov) #3

great definition - thank you for sharing this, Jon

(Chris Knoll) #4

@jon_duke has a great summary. This topic was covered at the cohort definition tutorial, which you can see the slides here: https://docs.google.com/presentation/d/1gPDPJALdb3pYAPi7rDN4Nl9N_VB-6X9bJfzjbAbXgG4/edit?usp=sharing

To put it simply: phenotype is the ‘what’, cohort is the ‘when’. A phenotype can be identified via a set of rules (or criteria as Jon says) but also probabilistic (as a measure from 0 to 100%). To visualize how that might look in a patient record, consider this diagram:

From a rule perspective (bottom half) we see the person jumps from ‘no’ to ‘yes’ over time, representing the episodes which this person met the phenotype. From the probabilistic perspective, we take the same person’s probability of matching the phenotype, and established 2 different durations, one with a higher threshold of probability, the other with lower probability threshold, and made 2 distinct cohort eras from this person based on the threshold.


(Tanvir Ahmed) #5

@jon_duke awesome answer! I got it. Thanks a lot.
@Chris_Knoll thanks for the details and links!