Proceedings of the 4th National Big Data Health Science Conference

Table 1 (abstract O28). Impaired cognition score for cognition performance based on the development and prediction datasets

	Tertiles of impaired cognition scores *
	Tertile 1	Tertile 2	Tertile 3
NHANES III (training dataset)
Impaired cognition score based on logistic model
Cases/Controls †	27/1186	91/1123	217/996
Model 1, OR (95% CI) ‡	1.00 (reference)	3.56 (2.30-5.51)	9.57 (6.36-14.4)
Model 2, OR (95% CI) §	1.00 (reference)	3.18 (2.05-4.94)	8.15 (5.36-12.4)
Impaired cognition score based on SVM model
Cases/Controls	25/1188	82/1132	228/985
Model 1, OR (95% CI)	1.00 (reference)	3.44 (2.18-5.43)	11.0 (7.22-16.8)
Model 2, OR (95% CI)	1.00 (reference)	3.18 (2.01-5.04)	8.77 (5.57-13.8)
NHANES 2011-2014 (validation dataset)
Impaired cognition score based on logistic model
Cases/Controls	25/811	47/790	113/724
Model 1, OR (95% CI)	1.00 (reference)	1.93 (1.18-3.17)	5.06 (3.25-7.90)
Model 2, OR (95% CI)	1.00 (reference)	1.66 (0.99-2.76)	4.31 (2.62-7.08)
Impaired cognition score based on SVM model
Cases/Controls	21/815	44/793	120/717
Model 1, OR (95% CI)	1.00 (reference)	2.15 (1.27-3.65)	6.50 (4.04-10.4)
Model 2, OR (95% CI)	1.00 (reference)	1.59 (0.92-2.74)	3.95 (2.34-6.66)

NHANES, National Health and Nutrition Examination Survey; SVM, support vector machine; OR, odds ratio; CI, confidence interval
* The tertiles was based on the distribution of all participants
† Poor cognition performance was defined as the lowest 10% of the distribution of the MMSE score (NHANES III) or a composite score based on the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD), the Animal Fluency (AF), and the Digit Symbol Substitution Test (DSST) tests (NHANES 2011-2014)
‡ Model 1 is the crude model without any adjustments
§ Model 2 adjusted for age, sex, and race

ISSN: 1753-6561