Table 10.

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

Sociodemographic conditions		Optimal threshold	Recall	Specificity	Accuracy	Differencea
Sex		66.60				1.85 b
	Male		50.00	97.88	97.68
	Female		51.85	97.88	97.49
Marital status		18.60				0.79
	Never been married		88.24	88.36	88.36
	Other groups of marital status		87.50	88.31	88.30
Working condition		33.80				9.92
	Working ≥35 hours		77.78	93.33	93.24
	Other groups of working condition		68.75	94.22	94.06
Race		19.90				25.92
	White		89.66	88.27	88.28
	Black		66.67	91.20	91.07
Income		68.00				3.19
	An income of <US $20,000		50.00	97.00	96.79
	Other groups of income		51.61	98.58	98.25

^aDifference between recall and specificity after bias mitigation.

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