Table 4

Factors affecting AI model influence on report, patient management, or imaging recommendation

Predictor Change ORs (adjusted CI)P value Benjamini-Adjusted threshold Significance 
No of critical findings  Report  1.306 (1.132 to 1.507)00.0042Yes
No of critical findings  Patient management  1.267 (1.056 to 1.521)0.0010.0083Yes
No of critical findings  Imaging recommendation  1.319 (1.035 to 1.681)0.0040.0125Yes
Lateral CXR  Imaging recommendation  6.495 (1.297 to 32.530)0.0050.0167Yes
Lateral CXR  Patient management  2.158 (0.837 to 5.565)0.0610.0208No
Lateral CXR  Report  1.542 (0.848 to 2.805)0.1050.025No
Radiologist experience  Report  0–5 years: Baseline
6–10 years: 0.255 (0.043 to 1.521)
>10 years: 0.305 (0.065 to 1.439)
0.1200.0292No
Radiologist experiencePatient management  0–5 years: Baseline
6–10 years: 0.165 (0.009 to 3.214)
>10 years: 0.378 (0.054 to 2.654)
0.2620.0333No
Radiologist experienceImaging recommendation  0–5 years: Baseline
6–10 years: 0.357 (0.034 to 3.783)
>10 years: 0.380 (0.044 to 3.287)
0.5160.0458No
Inpatient/outpatientImaging recommendation  1.550 (0.613 to 3.919)0.3260.0375No
Inpatient/outpatientReport  0.794 (0.476 to 1.323)0.3580.0417No
Inpatient/outpatientPatient management  0.818 (0.408 to 1.640)0.5720.0500No
  • Significance testing by the Benjamini-Hochberg algorithm to account for multiple hypotheses. ORs derived from stepwise logistic regression coefficients with CIs calculated with Benjamini-adjusted thresholds. Radiologist experience analysed as a categorical variable with derived from stepwise logistic regression coefficients with CIs calculated with Benjamini-adjusted thresholds. Radiologist experience analysed as a categorical variable with ORs representing effect of changing experience levels from the baseline (0–5 years) to a different level.

  • AI, artificial intelligence; CXR, chest X-ray.