Table 4.
Error analysis of incorrectly predicted CUIs by the best performing variant that selects the longest concept after filtering using EHR concepts
| (a) FHIRDATA | |||
|---|---|---|---|
| Boundary match | Semantic type match | Example | Count |
| ✓ | ✗ | Ques: What was the dose for amoxicillin? | 212 |
| Gold: Amoxicillin 200 MG Oral Tablet—C1298551 [Clinical Drug] | |||
| Pred: Amoxicillin—C0002645 [Antibiotic] | |||
| ✓ | ✓ | Ques: When was the onset of her diabetes? | 44 |
| Gold: Diabetes Mellitus, Non-Insulin-Dependent—C0011860 [Disease or Syndrome] | |||
| Pred: Diabetes—C0011847 [Disease or Syndrome] | |||
| ✗ | ✗ | Ques: How many times has he followed up on his asthma? | 311 |
| Gold: Asthma follow-up—C1273970 [Health Care Activity] | |||
| Pred: Asthma—C0004096 [Disease or Syndrome] | |||
| ✗ | ✓ | Ques: Was he ever admitted to an ER? | 12 |
| Gold: Emergency room admission—C0583237 [Health Care Activity] | |||
| Pred: Admitted to—C4482331 [Health Care Activity] | |||
| (b) ICUDATA | |||
|---|---|---|---|
| Boundary match | Semantic type match | Example | Count |
| ✓ | ✗ | Ques: When was the drainageapplied? | 8 |
| Gold: Drainage procedure—C0013103 [Therapeutic or Preventive Procedure] | |||
| Pred: Body Fluid Discharge—C0012621 [Body Substance] | |||
| ✓ | ✓ | Ques: How old is the line? | 4 |
| Gold: Intravenous Catheters—C0745442 [Medical Device] | |||
| Pred: Intravascular line—C0700221 [Medical Device] | |||
| ✗ | ✗ | Ques: Does she have a diaphragmatic tear? | 25 |
| Gold: Rupture of diaphragm—C0238088 [Injury or Poisoning] | |||
| Pred: Respiratory Diaphragm—C0011980 [Body Part, Organ, or Organ Component] | |||
| ✗ | ✓ | – | 0 |
Note: Gold concept boundaries are bolded while the predicted boundaries are underlined.
Ques: question; Gold/Pred: gold/predicted concept in the form of Display Name—CUI [Semantic Type].