Feature/FilterSurgeries

[nriedman]

Load robust Surgery data from FHIR Resource and filter procedures using manual filters and LLM reasoning

♻️ Current situation & Problem

At first pass, the SurgeryItem had the following footprint:

struct SurgeryItem: Identifiable {
    var id = UUID()
    var surgeryName: String = ""
    var date: String?
}

To load the surgery from FHIR data into this struct, we had a shallow query that only took the display name and date from the FHIRResource datatype instead of loading the robust data from the Procedures datatype specifically. This looked something like this:

func getProcedures() {
    let procedures = fhirstore.procedures

    for pro in procedures {
        var newEntry = SurgeryItem(surgeryName: pro.displayName)
        
        if let date = pro.date?.formatted() {
            newEntry.date = date
        }
        data.surgeries.append(newEntry)
    }
}

Furthermore, there was no filtering done on the procedures, meaning that the datapoints we displayed included items such as "Medication Reconciliation". Technically, these count as procedures, but we want to filter them out and only keep relevant surgeries.

⚙️ Release Notes

Now, we load data from the Procedure itself, and have access to richer information: notes, complications, status, location, and body sites, as well as more specific date information that allows for operations that span more than one day. See the Documentation section for more information about the specific changes to the SurgeryItem footprint and the new loading function.

Additionally, the function that loads in the surgeries now filters them using a manual filter and LLM reasoning. First, the manual filter excludes all procedures whose display name includes words in a manually curated stopWords list (see inline documentation). Then, the remaining display names are passed to an LLM, which removes extraneous labels such as "(procedure)" and selects which surgeries seem most relevant. This is done with few-shot prompting, and works reasonably well (see in-line documentation). More work is required to fine tune this process for better results.

📚 Documentation

The SurgeryItem now includes more robust information:

struct SurgeryItem: Identifiable {
    var id = UUID()
    var surgeryName: String = ""
    var startDate: String = ""
    var endDate: String = ""
    var status: String = ""
    var location: String = ""
    var notes: [String] = []
    var bodySites: [String] = []
    var complications: [String] = []
}

Note that there will only ever be an endDate if there is also a startDate. Both are optional.
See inline documentation for details about the loading function and filtering.

✅ Testing

This section will be thoroughly tested along with the rest of the code.

📝 Code of Conduct & Contributing Guidelines

By submitting creating this pull request, you agree to follow our Code of Conduct and Contributing Guidelines:

• I agree to follow the Code of Conduct and Contributing Guidelines.

[codecov]

Codecov Report

Attention: Patch coverage is 0.89686% with 221 lines in your changes are missing coverage. Please review.

Project coverage is 13.08%. Comparing base (897bd71) to head (451eefb).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #59      +/-   ##
==========================================
- Coverage   13.86%   13.08%   -0.77%     
==========================================
  Files          55       55              
  Lines        2671     2845     +174     
==========================================
+ Hits          370      372       +2     
- Misses       2301     2473     +172     

Files	Coverage Δ
Intake/Home.swift	90.59% <100.00%> (+0.12%)	⬆️
Intake/Intake.swift	100.00% <100.00%> (ø)
Intake/Surgery/SurgeryView.swift	0.00% <0.00%> (ø)

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 897bd71...451eefb. Read the comment docs.

[zoyagarg]

FIYA