Table 4.

Association Between Communication and Surrogate-Rated Outcomes: Standardized Coefficients from Linear Regression Models

	Multivariable linear regression with Maximum Likelihood Robust to Skewness (MLR) estimator*
	Post-traumatic stress		Satisfaction		Decision quality†
	Univariate models‡	Adjusted model‡	Univariate models‡	Adjusted model‡	Univariate models‡	Adjusted model‡
Information	−0.04; p = 0.4411	0.23; p = 0.0221	0.58; p < 0.0001	0.61; p < 0.0001	−0.44; p < 0.0001	−0.14; p = 0.1461
Emotional support	−0.13; p = 0.0137	−0.30; p = 0.0034	0.47; p < 0.0001	−0.01; p = 0.9182	−0.56; p < 0.0001	−0.44; p < 0.0001
White		−0.15; p = 0.0154		−0.03; p = 0.6000		0.08; p = 0.1861
Age (patient)		−0.13; p = 0.0329		0.02; p = 0.7699		−0.06; p = 0.3561
Age (surrogate)		−0.02; p = 0.6988		0.12; p = 0.0703		−0.00; p = 0.9437
Education (patient)		−0.02; p = 0.7860		0.02; p = 0.8163		0.02; p = 0.7816
Education (surrogate)		0.02; p = 0.7350		−0.20; p = 0.0029		−0.03; p = 0.7035
SES (comfortable vs. not enough)		−0.02; p = 0.8658		0.11; p = 0.2870		0.06; p = 0.5621
SES (just enough vs. not enough)		−0.06; p = 0.5545		0.04; p = 0.6828		0.09; p = 0.3286
Decision effectiveness		0.07; p = 0.2964		0.00; p = 0.9715		–
Death		0.25; p < 0.0001		−0.04; p = 0.5143		−0.00; p = 0.9345
Stress at baseline§		0.41; p < 0.0001		–		–

*Slope coefficients for information and emotional support are per the minimal important difference, which is a three-point change in informational and a two-point change in emotional support

^†For each surrogate, we selected the decision with the highest score

^‡Univariate models consist of separate analyses, with each subscale entered individually as the independent variables. Adjusted models include covariates, with parameter estimates given. High scores on the Impact of Event Scale-Revised (IES-R) are 22 and above

^§Stress at baseline measured by the Kessler-6