New Feature: LABEL PRIORITY: allow for custom rules for scoring the chart review

[comorbidity]

Providing two examples 1) current suicide study https://github.com/smart-on-fhir/cumulus-library-suicidality-icd10 2) hypothetical example of a heart disease study

---

1) suicide study prioritizes the severity of taking action of suicide over ideation of suicide. Action, including self-harming behavior or attempting suicide, has label priority. Present tense also has higher priority than past tense. In this example, there are 4 labels "ideation" or "action" in the past or present tense.

2) heart disease. There could be a label for “any heart disease” and a more specific label for “heart attack” or "heart failure" The precent rule there might be “heart attack” is more specific and implies “any heart disease”

[comorbidity]

label-X greater than label-Y

Fully expanded example

action-present > action-past action-present > ideation-present ideation-present > ideation-past action-past > ideation-past

[comorbidity]

possibly another ticket is better, but continuing this thread 🧵

some labels indicate something WRONG about the NOTE and that note should NOT be included in the chart-review analysis. case in point : when we rarely tag a note as suicidality-NOS that means exclude.

These are the steps I perform manually, the science is more reproducible if these steps aren't manually done by me anymore.

[mikix]

From talking to Andy in slack, just some extra clarity for non-medical folks like me:

Mike: If Truth says “any heart disease” and the reviewer says “heart attack”, is that a True Positive? Andy: False positive

Mike: If Truth says “heart attack” and the reviewer says “any heart disease”, is that a True Positive? Andy: True positive

So sounds like when scoring agreement, the reviewer can be more generic than Truth, but not more specific.

i.e. if Truth says "action-present', both "action-present" and "action-past" are TP. If Truth says "action-past", only "action-past" is TP.

[mikix]

some labels indicate something WRONG about the NOTE and that note should NOT be included in the chart-review analysis. case in point : when we rarely tag a note as suicidality-NOS that means exclude.

This feature request has now been landed in the ignore config key.

The main label priority request has still not been addressed yet.

[mikix]

Will be fixed by #12 when that lands.