Dyadic Analysis of End-of-Life Treatment Preferences

Purpose: When terminally-ill patients become physically and cognitively unable to communicate their preference stop or continue life-prolonging treatment, surrogates often make such decisions in collaboration with health care providers. Spouses comprise about two-thirds of legally appointed surrogates of married persons. Therefore, it is important for partners to know one another's end-of-life treatment preferences. This study examined the extent to which spouses and cohabiting partners accurately report one another's treatment preferences, aiming to identify whether there are differences in degree and direction of discrepancy in reporting partners' treatment preferences by (1) gender, (2) relationship quality, and (3) socioeconomic differences between partners (e.g., age, race). Method: The study examined data from a national internet survey comprised of 1,075 married or cohabiting heterosexual American couples aged between 18 and 64. Participants were asked to rate the level of treatment that they would like to receive during a serious illness with very little chance of survival (0 = stop all life-prolonging treatment and 10 = continue all treatment). They were also asked to estimate the level of treatment that their partners would like to receive. The difference between a man's preference score and his partner's prediction of his preference score measured the degree of discrepancy. Direction of discrepancy was a three-category variable indicating that the woman's prediction would result in overtreatment, undertreatment, or treatment in line with the man's preference. Parallel measures were constructed for women. Results: The average preferences of men and women were 2.51 (SD = 2.64) and 2.59 (SD = 2.68), respectively, indicating an overall tendency to prefer low levels of life-prolonging treatment. Roughly 45% of partner predictions were accurate, 28-33% showed errors of overtreatment, and 22-25% showed errors of undertreatment. With regard to degree of discrepancy, ordinary least squares regressions revealed that an age gap of more than 2 years between partners (β= 0.57, p < 0.01) and legal marriage (β= 0.63, p < 0.05) are associated with a larger discrepancy between a man's preference for himself and his partner's estimation of his preference. If a woman is African American, a smaller discrepancy was observed between her preference for herself and her partner's estimation of her preference (β= -1.01, p < 0.01). With regard to direction of discrepancy, multinomial logistic regression revealed that women in a relationship with an age gap of more than 2 years are more likely to make an error of undertreatment regarding their partners' preference than to be accurate (OR = 1.67, p <0.05). When a woman feels happy about the relationship (OR = 1.34, p <0.05), has some college education (OR = 1.88, p <0.05), and has a higher income (OR = 1.06, p <0.05), her partner is more likely to make an error of undertreatment than to be accurate regarding her preferences. Implications: These findings suggest that some couples are more likely than others to know each other's end-of-life care preferences, and because there are several systematic predictors of inaccuracy, targeted interventions could foster conversations about end-of-life medical care within at-risk couples.