There are limited reports on the detailed examination of steroid profiles for setting algorithms for 21-hydroxylase deficiency (21OHD) screening by liquid chromatography-tandem mass spectrometry (LC-MS/MS). We aimed to define an algorithm for newborn screening of 21OHD by LC-MS/MS, measuring a total of 2077 dried blood spot samples in Tokyo. Five steroids (17α-hydroxyprogesterone [17αOHP], 21-deoxycortisol [21DOF], 11-deoxycortisol [11DOF], androstenedione [4AD], and cortisol [F]) were included in the panel of LC-MS/MS. Samples from 2 cohorts were assayed: Cohort A, 63 "screening positive" neonates who were referred to an endocrinologist (n = 26 with 21OHD; n = 37 false-positive; obtained from 2015 to 2020); and Cohort B, samples (n = 2014) with 17αOHP values in the 97th percentile or above, in the first-tier test with 17αOHP ELISA from 2020 to 2021. Analysis of Cohort A revealed that the 3 indexes 21DOF, 11DOF/17αOHP, and (4AD + 17αOHP)/F had higher area under the curve (AUC) values (0.999, 0.997, 0.989, respectively), while the 17αOHP AUC was lower (0.970). Accordingly, in addition to 17αOHP, the 3 markers were included for defining the screening algorithm. The assay of Cohort B revealed that the new algorithm gave 92% of predicted positive predictive value without false-negative cases. We also determined the reference values for the 5 steroids at 4 to 7 days after birth, according to sex and gestational age (GA), revealing extremely low levels of 21DOF at any GA irrespective of sex differences. Our study demonstrated the high relevance of 21DOF, (4AD + 17αOHP)/F, and 11DOF/17αOHP, rather than 17αOHP, for 21OHD screening.