Development of the multivariate administrative data cystectomy model and its impact on misclassification bias

Table 3 Cystectomy classification accuracy with codes and using Administrative Data Cystectomy Model (ADCM)

A. Cystectomy Status Using Administrative Database Codes
		Cystectomy-Incontinent Diversion Status
		+	-	Sens	97.1%
Coded With Cystectomy—Incontinent Diversion	+	270	191	Spec	100.0%
	-	8	428,228	+ PV	58.6%
				-PV	100.0%
				+ LR	2178.5
				-LR	0.0
		Cystectomy-Continent Diversion Status
		+	-	Sens	100.0%
Coded With Cystectomy—Continent Diversion	+	222	237	Spec	99.4%
	-	0	428,238	+ PV	48.4%
				-PV	100.0%
				+ LR	1807.9
				-LR	0.0
B. Cystectomy Status Using ADCM and Best Expected Probability Thresholds*
		Cystectomy-Incontinent Diversion Status
		+	-	Sens	99.6%
ADCM Cystectomy—Incontinent Diversion	+	277	1614	Spec	99.6%
	-	1	426,805	+ PV	14.6%
				-PV	100.0%
				+ LR	264.5
				-LR	0.0
		Cystectomy-Continent Diversion Status
		+	-	Sens	100.0%
ADCM Cystectomy—Continent Diversion	+	222	80	Spec	100.0%
	-	0	428,395	+ PV	73.5%
				-PV	100.0%
				+ LR	5355.9
				-LR	0.0

For cystectomy status classification by administrative database codes (Section A), patients were classified with cystectomy by diversion type if they met criteria listed in Appendix B. For cystectomy status classification using the ADCM (Section B), patients were classified with cystectomy-incontinent diversion if the expected probability from the ADCM was at least 0.0922%; patients were classified with cystectomy-continent diversion if the expected probability from the ADCM was at least 0.0626%. (Sens Sensitivity, Spec Specificity, + PV Positive predictive value, -PV negative predictive value, + LR Positive likelihood ratio, -LR Negative likelihood ratio). Values have been rounded to 1 decimal place. *As determined using Youden’s method

ISSN: 1471-2288