Table 8.

The details of implementing bias mitigation for the neural network model using Adam.

Sociodemographic conditions		Optimal threshold	Recall	Specificity	Accuracy	Differencea
Sex		22.70				4.35 b
	Male		78.5	95.81	95.73
	Female		81.48	94.37	94.27
Marital status		0.60				1.53
	Never been married		100	65.43	65.64
	Other groups of marital status		100	66.96	67.17
Working condition		35.40				1.26
	Working >35 hours		77.78	96.75	96.68
	Other groups of working condition		78.13	95.84	95.70
Race		5.90				4.95
	White		96.55	90.11	90.15
	Black		100	88.61	88.66
Income		45.60				15.18
	An income of <US $20,000		80	94.48	94.34
	Other groups of income		67.74	97.40	97.24

^aDifference between recall and specificity after bias mitigation.

^bItalicized values indicate an improvement compared to the initial values (50% threshold).