Table 2.

Percent of variation in gestational age explained by neonatal metabolic screening markers in the model-building dataset

Model description	No. model parametersa	No. observations	R2 (%)	RMSE
Final metabolite model	88	153,342	53.1	1.3
Final metabolite model excluding infants with abnormal neonatal screens	88	148,104	50.1	1.3
Final metabolite model excluding infants on total parenteral nutrition	88	151,293	47.9	1.3
Final metabolite model excluding statistical outliers as defined as those with residuals less than −1.96 and greater than 1.96	88	146,083	55.4	1.1
Final metabolite model + month and year of collection	104	153,342	53.4	1.3
Month and year of collection only	17	153,342	1.2	1.9
Final metabolite model + multiple gestation status (yes/no)	89	153,342	53.7	1.3
Multiple gestation status only	2	153,342	4.8	1.9
Final metabolite model including only singleton births	88	147,597	49.4	1.3
Final metabolite model + age at collection (hours)	91	153,322	53.1	1.3
Age at collection (hours) only	4	153,322	0.05	1.9
Final metabolite model in subset with nonmissing age at collection	88	153,322	53.1	1.3
Final metabolite model + infant sex	89	152,674	53.2	1.3
Infant sex only	2	152,674	0.02	1.9
Final metabolite model in subset with non-missing infant sex	88	152,674	53.1	1.3
Final metabolite model + infant weight, g	91	153,008	66.1	1.1
Infant weight, g only	4	153,008	54.5	1.3
Final metabolite model + weight for gestational age (AGA, LGA, and SGA)	90	153,008	53.1	1.3
Weight for gestational age (AGA, LGA, and SGA) only	3	153,008	0	1.9
Final metabolite model in subset with nonmissing weight	88	153,008	52.9	1.3

AGA, average-for-gestational age; LGA, large-for-gestational age; RMSE, root-mean-square error; SGA, small-for-gestational age.

Ryckman. Metabolic predictors of gestational age. Am J Obstet Gynecol 2016.

^aModel parameters include the intercept term.