Table 4.
Best fitting generalized estimating equations Type III model effects and parameter estimates, predicting diagnostic accuracy.
| Effect | B | SE | Odds Ratio Exp(B) |
95% CI |
Wald χ2(df) |
p- value |
|---|---|---|---|---|---|---|
| Type III Model Effects | ||||||
| Case Consensus Diagnosis | 20.69(3) | <.001 | ||||
| Case Breast Density | 8.17(1) | .004 | ||||
| Case Difficulty Rating | 20.06(1) | <.001 | ||||
| Pathologist Experience | 18.58(1) | <.001 | ||||
| Fixation Durations | 6.84(1) | .009 | ||||
| Fixations within versus outside ROIs |
11.24(1) | .001 | ||||
| Zoom Behavior | 3.08(1) | .079 | ||||
| Number of Fixations | 2.13(1) | .144 | ||||
| Pathologist Experience × Consensus Diagnosis |
9.02(3) | .029 | ||||
| Pathologist Experience × Zoom Behavior |
3.49(1) | .062 | ||||
| Model Parameter Estimates | ||||||
| Case Consensus Diagnosis (reference: benign) |
||||||
| Atypia | −.629 | .357 | .533 | (.265, 1.07) | 3.10(1) | .078 |
| DCIS | −.588 | .384 | .555 | (.262, 1.18) | 2.35(1) | .125 |
| Invasive | 1.91 | .614 | 6.72 | (2.02, 22.4) | 9.64(1) | .002 |
| Case Breast Density (reference: BI-RADS 1–2) |
||||||
| BI-RADS 3–4 | .813 | .285 | 2.26 | (1.29, 3.94) | 8.17(1) | .004 |
| Case Difficulty Rating | −.695 | .155 | .499 | (.368, .677) | 20.06(1) | <.001 |
| Pathologist Experience | −.054 | .279 | .947 | (.547, 1.64) | .038(1) | .846 |
| Fixation Durations | .201 | .077 | 1.22 | (1.05, 1.42) | 6.84(1) | .009 |
| Fixations in versus outside ROIs |
.420 | .125 | 1.52 | (1.19, 1.94) | 11.24(1) | .001 |
| Zoom Behavior | −.207 | .118 | .813 | (.645, 1.03) | 3.08(1) | .079 |
| Number of Fixations | −.166 | .114 | .847 | (.677, 1.06) | 2.13(1) | .144 |
| Pathologist Experience × Consensus Diagnosis |
||||||
| Atypia | .590 | .335 | 1.80 | (.936, 3.48) | 3.11(1) | .078 |
| DCIS | .910 | .328 | 2.49 | (1.31, 4.72) | 7.72(1) | .005 |
| Invasive | 1.21 | .540 | 3.37 | (1.17, 9.71) | 5.05(1) | .025 |
| Pathologist Experience × Zoom Behavior |
−.273 | .146 | .761 | (.572, 1.01) | 3.49(1) | .062 |