Table 3.

Endorsement Rates by Mismatch Count in the Four Clinical Scenarios

			Endorsement Rate at Each Mismatch Point*
Scenario	Endorsement Rate	Model	0	1	2	3	4	Wald Chi Square	QICC‡	Sig pairs	Sig linear effects
High gain	81% ± 5	5 point	88% ± 6	93% ± 3	77% ± 5	73% ± 7	56% ± 11	χ2=22.26, p<.001	322.43	0–4,1–2, 1–3,1–4	p<.001
		3 point	93%± 4	79% ± 4	67% ± 6			χ2=20.1, p<.001	322.61	0–1,0–2, 1–2	p<.001
Low gain	8% ± 2	5 point†	3% ± .5	7% ± 2	12% ± 3	10% ± 4	17% ± 8	χ2=17.26, p=.002	215.88	0–2
		3 point	3% ± 2	9% ± 2	19% ± 4			χ2=13.3, p<.001	205.18	0–2,1–2	p<.001
Ineffective	37%± 7	5 point	28% ± 9	32% ± 8	32% ± 5	37% ± 8	58% ± 11	χ2=9.59, p=.048	444.31		p=.03
		3 point	29% ± 8	32% ± 6	46% ± 8			χ2=5.16, ns	437.96
High risk	44% ± 7	5 point	46% ± 12	41% ± 7	49% ± 7	39% ± 9	45% ± 11	χ2=3.9, ns	468.15
		3 point	50% ± 9	42% ± 6	41% ± 7			χ2=1.9, ns	464.09

^*Number of vignettes at each mismatch point: 5 point model, 1, 4, 6, 4, 1; 3 point model: 4, 8, 4

^†The 5-point model in the low gain condition did not converge with a logit link function. An identity link function was used for this model.

^‡QICC: Pan’s Corrected Quasi-Likelihood under Independence Criterion. Lower numbers indicate a better model fit