Accounting for biases in survey-based estimates of population attributable fractions

Table 1 Hypothetical sample data

	Non-smoker	Smoker	Total	Counterfactual	pe	pd	RR
Non-Hispanic black
Survived	500	200	700	769
Died	150	150	300	231
Total	650	350	1000	1000	0.35	0.50	1.86
q_x	0.231	0.429	0.30
Non-Hispanic white
Survived	2700	500	3200	3375
Died	500	300	800	625
Total	3200	800	4000	4000	0.20	0.375	2.40
q_x	0.156	0.375	0.20
Combined
Survived	3200	700	3900	4144
Died	650	450	1100	856
Total	3850	1150	5000	5000	0.23	0.409	2.32
q_x	0.169	0.391	0.22

pe proportion smoker in entire sample, pd proportion smoker among deceased, q_x probability of death, RR risk ratio

ISSN: 1478-7954