Factors underlying surrogate medical decision-making in middle eastern and east Asian women: a Q-methodology study

This is a cross-sectional study on the relative importance of 28 surrogate decision-making factors to Middle-Eastern (ME) and East-Asian (EA) women from three perspectives: norm perception (N-perspective), preference as patient (P-perspective), and preference as surrogate decision-maker (S-perspective). A Q-methodology instrument was used because it is especially suited to examine relative importance. Briefly, Q-methodology involves construction/collection of a concourse of opinion statements related to the topic under study, sorting the statements into groups and collapsing them into a Q-set that adequately covers the various thematic domains, presenting the Q-set to respondents to model their point of view by rank-ordering the statements into piles (Q-sort) along a continuum defined by certain instructions, performing a special type of by-person exploratory factor analysis using the Q-sorts as variables, grouping of respondents who rank-ordered the statements into similar arrangements (i.e., loaded significantly on the same factor) into models, and interpreting the models [32, 33].

Eligibility criteria were having a ME or EA ancestry, age > 18 years, >high school education, and ability to understand study aim and procedures as subjectively determined by study coordinator. Volunteers were recruited through advertisement within Riyadh, Saudi Arabia. The study involved both men and women; however, because of the extent of data, only the results of women are reported here. A sample size of 120 ME and 120 EA women was based on convenience and limitation of the Q-methodology program (maximum of 120 Q-sorts).

The study was conducted in accordance with the Declaration of Helsinki after approval of the Research Ethics Committee of the King Faisal Specialist Hospital and Research Center (KFSH&RC) and obtaining respondents’ written informed consent.

Each respondent performed Q-sorting according to each of the three perspectives (3 Q-sorts), in three random same-day Q-sorting sessions. The sessions included writing comments on the 6 extreme selections and were separated by respondents’ completion of 1) respondents’ characteristics sheet and 2) a previously published social value scale questionnaire [23].

Each of the three Q-set consisted of 28 factors that may potentially underline surrogate medical decision-making (items), 16 of which were reported previously [5, 6] and the rest were identified from literature review [3, 13‐15]. The 28 items are listed in Additional file 1-Q-set items together with their abbreviated version that is used in the text. The items can be divided into eight domains, patient’s preference-centric (3 items plus one item shared with patient’s life-long narrative-centric), patient’s life-long narrative-centric (4 items plus one item shared with patient’s preference-centric), patient’s health-centric (4 items plus one item shared with surrogate’s emotion-centric), surrogate’s emotion-centric (2 items plus one item shared with patient’s health-centric), surrogate’s preference-centric (3 items), surrogate’s interest-centric (2 items plus one item shared with family-centric), family-centric (4 items plus one item shared with surrogate’s interest-centric), and society-centric (3 items). The wording of the 28 items (as well as the associated instructions) in the three Q-sets were modified to fit the three perspectives addressed in the study. The Q-sets were subjected to two cycles of pilot testing, validation, and revision as previously described [34]. Q-set items were randomly numbered.

Q-sets were given to respondents with sets of instructions and sorting sheets (Additional file 2-instructions and sorting sheets). Q-sorting requires respondents to arrange the items according to their subjective relative importance. In the current study, this was from 1 = least important to 9 = most important. Numbers of slots for each category were symmetrically distributed (Additional file 2-instructions and sorting sheets); for example, there was one slot under categories 1 and 9 and 2 slots under categories 2 and 8. The time spent in Q-sorting was recorded. Sorting sheets were checked for completion (i.e., each item number is used and only once) and were collected before respondents started the next Q-sorting. In case a mistake was found in copying items’ numbers onto the sorting sheet, respondents were notified to self-correct.

Respondents’ characteristics included age, nationality, employment status, self-rated religiosity, major religious affiliation, self-rated health, daily self-care status, pain in last month, pain interference with daily activities, life quality, life satisfaction, living arrangements, death in immediate family/close friends, and personal experience with surrogate decision-making.

The social value scale [23] consisted of three subscales: planning social change (subscale-1, 3 “conservative” and 3 “liberal” statements), free choice regarding medical ethical questions (subscale-2, 4 “conservative” and 3 “liberal” statements), and deciding good vs bad regarding personal welfare (subscale-3, 3 “conservative” and 3 “liberal” statements). Respondents were asked to score each statement from (1 = total disagreement to 5 = full agreement). To analyze responses, we multiplied “liberal” statement scores by − 1 and then calculated the average score for each subscale. Thus, score of subscale-1 could range from − 2 (extreme liberal) to 2 (extreme conservative), of subscale-2 from − 1.57 (extreme liberal) to 2.43 (extreme conservative), and of subscale-3 from − 2 (extreme liberal) to 2 (extreme conservative).

Q-sorts were analyzed separately for ME respondents and EA respondents and for each of the three perspectives, using PCQ for Windows (PCQ Software, Portland, OR, USA). Data analysis in Q-methodology requires a special program and involves exploring Q-sorts’ correlation, factor analysis (factor loading and rotation), factor scores computation, and factor interpretation. The PCQ program identified and used Q-sorts with a loading of ≥0.49 (p < 0.01) on a single factor to create model Q-sorts, which indicate how a respondent with 100% loading on a factor would have ordered the 28 items. Model Q-sorts were interpreted based on the 3 most important (i.e., items assigned to categories 8 and 9) and 3 least important items (i.e., items assigned to categories 1 and 2), respondents post-sorting comments, and relative importance of all items within and across models. Association between models and respondents’ characteristics was studied after grouping respondents who loaded significantly on one model and respondents with confounded loading who had higher loading on the same model. For averaging-analysis, items were considered “neutral” if their mean ranking score was >4 and <6. For Q-methodology, items were considered strongly-agreeable (most important), agreeable, disagreeable, or strongly-disagreeable (least important), if they received a model Q-set rank of 8–9, 7, 3, or 1–2, respectively. Bivariate analysis was performed using chi-square or Fisher Exact tests for categorical variables and ANOVA or t-test for continuous variables (IBM SPSS Statistics version 21 software). An unadjusted two-tailed p-value is reported and considered significant if <0.01.

Figure 1 shows mean ranking scores of the 28 items by ME and EA respondents from the N-perspective, P-perspective, and S-perspective. The items are grouped according to the underlying domains.

Overall, patient’s health was more important than other factors, including patient’s preference and life-long narrative; surrogate’s emotion, preference, and interest; family interests; and society interests. For ME respondents, of the items that received a mean ranking score >6, 5/5, 5/5, and 5/6 were related to patient’s health in the N-perspective, P-perspective, and S-perspective, respectively. For EA respondents, the corresponding ratios were 4/4, 3/4, and 3/5. The other items that received a mean score >6 were “26.Fear of loss of loved one” in ME and EA S-perspective and “9.Family needs” in EA P-perspective and S-perspective.

Notably: 1) In the patient’s preference domain, patients’ wishes were given about the same weights regardless of patients’ mental status. 2) In the patient’s life-long narrative domain, the weight given to “18.Life-long story of patient” was lower than the weight given to the other four related items, suggesting that it may have not been understood correctly by respondents. 3) In the patient’s health domain, “3.Trying everything possible to save patient” received more weight than “22.Medical facts.” 4) In the surrogate’s preference domain, there were no appreciable differences between the weights given to what surrogates want for themselves or wish for patients.

Using a Q-methodology specific program (PCQ program), we extracted 6 factors for each of the three perspectives of ME and EA respondents. The program calculates idealized item scores (model Q-sort) for each factor. A model Q-sort indicates how a respondent with 100% loading on a factor would have ordered the 28 items. Factor characteristics and item scores are presented in Tables 2S to 7S (Additional file 4-factor characteristics and item scores).

The final step in Q-methodology is to interpret and name program-identified model Q-sorts and explore their association with respondents’ characteristics. Tables 4 and 5, list model names in each perspective of ME and EA respondents, respectively. Details underlying factors interpretation are presented in Additional file 5-Q-methodology technical report. The following provides an overall description of models in each perspective.

According to the substituted judgment standard, surrogate decisions should be based on patient’s stated or predicted preferences rather than what the surrogate thinks is in the patient’s best interests, which is supported by the results of a national survey of US physicians [2]. However, in the current study, patient’s health was more important than other studied factors, including patient’s preference, for both ME and EA respondents and from all the three perspectives. Consistently, a previous study using simple rating and dichotomization showed that 78% of surrogates (mostly white educated US women) focused more on patient’s well-being than patient’s preferences [6]. A predominance of consequentialist attitude over autonomy-based attitude was also observed in lay people approach to medical use of placebos [34] and organ donation, [35] providing support to the concept that “good” may be more fundamental than “right” and to the importance of “harm/care” as one of the psychological foundations of morality [36].

On the other hand, placing more weight on patient’s health than patient’s preference may be emotion-driven. Indeed, for our respondents, trying everything possible to save patient was more important than other patient’s health-related items such as medical facts and patient’s best interests (Figure). Substituted judgement is known to be associated with emotional demands, [18] and about a third of surrogates had negative emotions lasting several months [37]. Alternatively, the emphasis on patient’s health over patient’s preference may be related to beliefs of life sacredness.

Overall, patient’s preference followed patient’s health in importance. Remarkably, current wishes (despite impaired mentation) were given similar weights to past wishes in most perspectives (Figure). The concept of prospective autonomy/ autonomy extension argues that desiring is only a basic notion, whereas valuing is a reflective appraisal that should be respected when previous and current desires conflict, [8] that autonomy is a right rather than necessarily a reflective capacity, that legitimate interests include not only experiential but also critical interests (such as interests in fulfilling religious commitments and interests in family welfare after death), and that personal identity/personhood is based on bodily continuity rather than psychological connectedness or continuity [7, 8]. In contrast, the current preference thesis takes into account only experiential interests, believes that preferences and interests do not survive loss of mental capacity, considers autonomy as reflective capacity/responsibility, and concludes that there is no moral authority for exercising control over one’s incompetent self and that one should instead promote the current interests of patients [7, 8]. The current preference thesis could provide yet another explanation for our respondents’ emphasis on patient’s health rather than patient’s preference.

Several of the items related to patient’s life-long narrative received similar weights as items related to patient’s preference. According to the narrative view, narratively well-structured lives have an aesthetic value, and life is to be concluded in a way that is best consistent with life-long themes [11, 14]. This view focuses on respect for persons rather than narrowly on respect for autonomy and on authentic decisions rather than autonomous decisions. In addition to taking into consideration patient’s prior wishes, it considers patient’s life-long dispositions, relationships, decisions, decision-making processes, and interests, balancing rather than prioritizing [9]. There are three closely-related versions of the life-story narrative model, authentic life, [10, 15] substituted interests, [13] and patient’s endorsed life [14]. The patient’s life-story is taken as evidence of what the patient would have wanted to continue, [15] of what the patient’s interests should be based upon, [13] or of what the patient have valued, respectively [14].

In most perspectives, our respondents put less weight on what continues the life the patient has led (authentic life) than on what promotes the life the patient has valued (endorsed life) or the way the patient used to make decisions (Figure). However, Q-methodology did identify both a relatively patient’s endorsed-life narrative-independent model (ME N-perspective) and a relatively patient’s authentic-life narrative-independent model (EA S-perspective). Finally, the fact that, in general, the importance of patient’s preference followed the importance of patient’s health appears to be most consistent with the substituted interests’ model, which integrates the substituted judgment and the best interests’ standards [13, 14]. Of note, “Life-long story of patient” was assigned much lower weight than the other items under the life-long narrative domain, suggesting that this item may have not been understood by our respondents as intended.

Consistent with the Golden Rule, there were no overall appreciable differences between the importance of “What surrogate would have wanted if in similar situation” and the importance of “Surrogate own wishes for patient.” It has been argued that the Golden Rule, a fundamental concept of Judo-Christian and Islamic ethics, should be considered as a more practical alternative to substituted judgment [5]. Interestingly, the Golden Rule was more important to ME than EA respondents, ME respondents put significantly more weight on surrogate’s wishes for the patient in the N-, P-, and S-perspectives and more weight on what the surrogate would have done in a similar situation in the S-perspective. Finally, the observation that in the ME S-perspective, these two items were as important as items related to patient’s preference provides support to previous results showing that surrogates may more project their wishes than simulate patient’s wishes [29].

In the current study, surrogate/family needs and burdens were given more weight in the P-perspective than in the S-perspective. Similarly, in a previous study, burden on family was the third most important end-of-life issue for patients but was not important for surrogates [38]. Several studies [26, 27, 29, 30, 38] and reviews [24, 25] have found that substituted judgment accuracy is low to moderate despite the fact that surrogates are more accurate than physicians in predicting patients’ preferences [25, 39]. Although this may be caused by inadequate knowledge of patient’s wishes and priorities, [29, 30] the current results suggest that it may be also role-dependent. This is supported by the fact that patient’ best interests and emotions (fear of loss of loved and feeling of guilt) were given more weight in the S-perspective than in the P-perspective.

There was also notable dissociation between the N-perspective and S-perspective. In the N-perspective, respondents put less weight on surrogate’s preference and family cohesion and more weight on patient’s best interest. Previous studies showed significant divergence between preference and perception of norm in regard to consenting for research on medical records and leftover tissue samples [40] but not in regard to disclosure of medical errors [41] or organ donation [42]. One can differentiate two kinds of norm perception. Perception of what is commonly approved/ disapproved (injunctive norm) and perception of what is commonly done (descriptive norm). Our instrument was designed to explore the perception of injunctive norm. The observed dissociation between the N- and S-perspectives suggests that social desirability bias did not have a substantial influence on our respondents.

End-of-life preferences may be culture and sex [21, 22, 43] as well as religiosity [23] dependent. We observed two notable differences between ME and EA respondents. First, EA respondents were more family-centric, whereas ME respondents were more patient-centric. In a previous study on attitude to organ donation, there was also stronger familism orientation in Christians/Philippines-educated than Muslims/Saudi Arabia-educated respondents. In familism, which is prevalent in East Asia, [16, 17, 44] the family makes decisions collectively based on the family’s interest [5]. Interestingly, familism orientation may be role-dependent. In our study, family needs and burden were given more weight in the P-perspective compared to the S-perspective. Similarly, a previous study showed that surrogates err more by attributing to the patient a preference for substituted judgment when the patient prefers patient-family shared decision-making or family-only decision-making than the reverse [27].

Second, in the N-perspective, ME respondents put significantly more weight on “What continues the life patient has led” and less weight on “What promotes the life patient has valued” and “Religious or spiritual believes of patient,” suggesting a differential subscription to the authentic- life narrative and endorsed-life narrative models.

Q-methodology, a special type of by-person exploratory factor analysis, produces grouping of respondents into thinking models [32, 33]. Comparing ME and EA models, we found substantial overlap but also notable differences. For example, in the N-perspective, relatively surrogate-independent, society’s interests-independent, and emotion-independent models were found in both groups; patient’s life-long narrative-independent, patient’s endorsed-life narrative-independent, family needs-independent, and Golden rule-independent models were found only among ME respondents; and religious/spiritual beliefs-independent and patient’s preference-independent models were found only among EA respondents. Similarly, in the N-perspective, there were relatively patient’s preference-dependent models in both groups; and emotion-dependent, medical facts-dependent, and patient’s religious/spiritual beliefs-depended models only among EA respondents.

Intriguingly, there was no association between Q-methodology models and characteristics of ME or EA respondents in the N-perspective. However, in the ME P-perspective, respondents who loaded on a relatively emotion-dependent model were younger and predominantly students; in the EA P-perspective, the relatively family needs-dependent model had the most “conservative” score while the relatively religious/spiritual beliefs-independent model had the most “liberal” score on social value subscale-3 (deciding good vs bad regarding personal welfare); in the ME S-perspective, the relatively patient’s preference-dependent model had the most “liberal” score and the relatively society’s interests-independent model had the most “conservative” score on social value subscale-1 (planning social change) and subscale-2 (free choice regarding medical ethical questions); and in the EA S-perspective, the relatively society’s interests-independent model had the most “conservative” scores on social value subscale-3.

The results suggest that guidelines for surrogate decision-making that are based solely on right of self-determination may, in general, poorly fit surrogate decision-making in Middle Eastern and East Asian cultures. However, there is no one-size-fits-all solution as patient’s preference-dependent models were identified in both groups. Further, the discrepancy between how one prefers to make surrogate decisions and how one prefers surrogate decisions be made on their behalf should be considered in clinical practice and in designing and interpreting related research studies. Finally, the results highlight the relative importance of familism orientation in surrogate decision-making, especially among East Asian women. It is suggested that clinicians develop measures for eliciting both patients’ and surrogates’ perspectives to include them into the care and education of patients at risk for requiring surrogate decision-making and their potential surrogates.

The study was unique in examining an expanded collection of potential factors that may underlie surrogate medical decision-making, in enrolling Middle Eastern and East Asian women, and in analyzing data by both averaging-analysis and Q-methodology. The use of shuffled items and forced-ranking permitted minimizing biases related to order effect and tendency to assign maximum importance to numerous items, respectively. The study design also allowed direct comparison of three perspectives, norm-perception, preference as patient, and preference as surrogate decision-maker. Finally, the three Q-sorts were performed in a random order and separated by completion of respondents’ characteristics and social value scale questionnaire to minimize potential carry-over effect.

The study has several limitations. First, generalizability of the results is limited due to convenience sampling, limiting enrollment to educated individuals, and limiting analysis to women respondents. Second, it is possible that some respondents may have not understood some items as intended (for example, “Life-long story of patient”). Third, it is likely that there are more than the models identified in the study, which is expected because only six factors were extracted and is consistent with the exploratory and none exhaustive nature of Q-methodology. Fourth, given the impressionistic nature of Q-methodology, there was overlap among the models in each perspective.