Table 5.

Proportions of unique patients with current SDOH found only in structured data, only by NLP extraction, or in both

	All patients (N = 618K)				Patients with social history text (N = 225K)
		SDOH source				SDOH source
SDOH	# Patients with SDOH info	Struct.	NLP	Both	# Patients with SDOH info	Struct.	NLP	Both
Alcohol (current)	148 221	87%	5%	8%	106 658	82%	7%	11%
Tobacco (current)	34 871	69%	19%	12%	25 675	56%	26%	17%
Drug (current)	42 309	86%	10%	5%	31 222	81%	13%	6%
Employment (any status)	575 278	88%	1%	11%	200 926	66%	4%	30%
Living Status (homeless)	11 567	64%	32%	4%	9390	55%	39%	5%

Note: For purposes of comparison, in the right-most three columns, we further limit structured data to only patients who had social history narrative text as well.