Table 2. Sensitivity, secificity, PPV, and NPV of treatment identified with SAS-based text mining for stage IV non-small cell lung cancer patients (n = 17,310), 2012–2014, California.

	Sensitivity		Specificity		PPV		NPV
Treatment Group	%	95% CI	%	95% CI	%	95% CI	%	95% CI
Platinum doublets	90.0	89.7, 91.9	99.4	99.2, 99.5	96.4	95.6, 97.1	98.3	98.1, 98.5
Pemetrexed-based regimens	93.4	92.2, 94.4	98.9	98.7, 99.1	92.5	91.4, 93.6	99.1	98.9, 99.2
Bevacizumab-based regimens	88.1	85.1, 90.7	99.8	99.7, 99.9	93.0	90.5, 94.9	99.6	99.5, 99.7
Pemetrexed and bevacizumab regimens	97.3	95.7, 98.4	99.3	99.1, 99.4	84.4	81.8, 86.6	99.9	99.8, 99.9
Single agents	88.6	84.6, 91.8	98.9	98.7, 99.0	60.4	56.8, 63.8	99.8	99.7, 99.8
Tyrosine kinase inhibitors	93.2	91.9, 94.3	98.2	97.9, 98.4	84.8	83.2, 86.2	99.2	99.1, 99.4
No systemic treatment	88.3	87.4, 89.1	92.4	91.9, 92.9	84.4	83.6, 85.2	94.5	94.1, 94.8
Unknown systemic treatment	74.3	72.9, 75.7	96.5	96.2, 96.8	85.7	84.6, 86.8	92.9	92.6, 93.3

Abbreviations: CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value