TABLE 3.

Performance matrix of the prediction models.

	Sensitivity	Specificity	Accuracy	AUC	95% CI of AUC
					Lower limit	Upper limit
No-odor modela	0.893	0.791	0.799	0.901	0.864	0.933
Single-odor modelb
Orange	0.880	0.791	0.798	0.901	0.867	0.933
Leather	0.880	0.804	0.810	0.905	0.870	0.935
Cinnamon	0.893	0.804	0.811	0.904	0.867	0.936
Peppermint	0.893	0.781	0.790	0.904	0.866	0.936
Banana	0.920	0.768	0.780	0.906	0.873	0.936
Lemon	0.920	0.749	0.762	0.902	0.869	0.933
Licorice	0.867	0.814	0.818	0.902	0.869	0.932
Coffee	0.893	0.786	0.794	0.900	0.867	0.932
Cloves	0.920	0.749	0.762	0.902	0.866	0.932
Pineapple	0.893	0.796	0.804	0.903	0.866	0.932
Rose	0.893	0.786	0.794	0.904	0.869	0.936
Fish	0.893	0.791	0.799	0.902	0.864	0.933
OI	0.933	0.751	0.766	0.904	0.869	0.935
Full modelc	0.907	0.828	0.834	0.916	0.882	0.945
Stepwise modeld	0.880	0.831	0.835	0.914	0.881	0.943
Simple modele	0.880	0.781	0.790	0.901	0.859	0.931

^aPredictors: sex, age, body mass index (BMI), height, education, smoking, drinking, coronary artery disease (CAD), hypertension, diabetes, depression, stroke, apolipoprotein (APOE)-ε4, and Mini-mental State Examination (MMSE). ^bOther predictors are the same as in the no-odor model. ^cPredictors: sex, age, BMI, height, education, smoking, drinking, CAD, hypertension, diabetes, depression, stroke, APOE-ε4, MMSE, and the 12 odors. ^dPredictors: age, weight, education, depression, stroke, APOE-ε4, leather, peppermint, banana, lemon, pineapple, rose, and MMSE. ^ePredictors: MMSE, age, peppermint, stroke, and education.