Skip to main content
. Author manuscript; available in PMC: 2015 Jan 1.
Published in final edited form as: JAMA Intern Med. 2014 Jan;174(1):114–121. doi: 10.1001/jamainternmed.2013.11958

Table 3.

Discrimination ability of the clinical indication prediction model to predict the four breast cancer specific clinical indications for breast MRI in the validation data set from the HPHC-Atrius Health subset.

Abstracted Indication
Model Predicted Indication Screening Diagnostic Staging/Treatment Surveillance Total
    Screening 131 21 1 0 153
    Diagnostic 26 263 17 38 344
    Staging/Treatment 0 15 111 10 136
    Surveillance 1 17 8 188 214
    Total 158 316 137 236 847
Prediction Accuracy
    “Sensitivity”a 0.83 0.83 0.81 0.80 0.82
    PPVb 0.86 0.76 0.82 0.88

MRI = magnetic resonance imaging; HPHC = Harvard Pilgrim Health Care

a

Sensitivity = number correctly predicted by the model/total number of MRIs assigned after medical record abstraction, for each indication and overall.

b

Positive Predictive Value = number of MRIs correctly predicted by model /total number assigned after medical record abstraction, for each indication and overall.