Table 1.
Performance of Each Query Method by Diagnosis Category
| Query Method | % Agreement | Kappa* (95% CI) | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|---|
| PJI | ||||||
| PS-PD | 89.3% | 0.51 (0.43–0.58) | 38.6% | 99.9% | 98.5% | 88.6% |
| AR-PD | 97.5% | 0.91 (0.87–0.94) | 86.0% | 99.9% | 99.3% | 97.1% |
| PS-AD | 90.4% | 0.58 (0.50–0.65) | 45.6% | 99.8% | 97.5% | 89.8% |
| AR-AD | 97.3% | 0.90 (0.87–0.94) | 88.3% | 99.1% | 95.6% | 97.6% |
| Loosening | ||||||
| PS-PD | 82.1% | 0.64 (0.59–0.69) | 69.7% | 94.8% | 93.2% | 75.4% |
| AR-PD | 82.1% | 0.64 (0.59–0.69) | 69.7% | 94.8% | 93.2% | 75.4% |
| PS-AD | 82.5% | 0.65 (0.60–0.70) | 70.9% | 94.3% | 92.7% | 76.1% |
| AR-AD | 82.4% | 0.65 (0.60–0.70) | 70.9% | 94.1% | 92.4% | 76.0% |
| Instability | ||||||
| PS-PD | 82.9% | 0.59 (0.54–0.65) | 56.8% | 97.5% | 92.8% | 80.1% |
| AR-PD | 82.7% | 0.59 (0.53–0.65) | 56.8% | 97.3% | 92.2% | 80.1% |
| PS-AD | 84.8% | 0.64 (0.59–0.70) | 62.2% | 97.5% | 93.3% | 82.2% |
| AR-AD | 84.7% | 0.64 (0.59–0.70) | 62.9% | 96.9% | 92.0% | 82.4% |
| Fracture | ||||||
| PS-PD | 97.1% | 0.24 (0.04–0.44) | 16.0% | 99.6% | 57.1% | 97.4% |
| AR-PD | 97.1% | 0.24 (0.04–0.44) | 16.0% | 99.6% | 57.1% | 97.4% |
| PS-AD | 97.1% | 0.36 (0.16–0.55) | 28.0% | 99.2% | 53.9% | 97.8% |
| AR-AD | 97.1% | 0.36 (0.16–0.55) | 28.0% | 99.2% | 53.9% | 97.8% |
PS-PD: Prosthetic joint-specific diagnosis codes in the principal diagnosis position; AR-PD: Any related code in the principal diagnosis position; PS-AD: Prosthetic joint-specific diagnosis codes in any of the 10 diagnostic positions; AR-AD: Any related code in any of the 10 diagnostic positions
*
Cohen’s kappa statistic was calculated and was assessed using grading criteria per precedence in the literature: κ = 0.20 to 0.39, fair; κ = 0.40 to 0.59, moderate; κ = 0.60 to 0.79, very good; κ greater than 0.80, excellent