FormalPara Key Points for Decision Makers

Only 16% of patients reported discordance in perceived and preferred roles, suggesting that current clinical practice in Singapore seems to provide an environment for most patients to participate in decision making at their desired level.

Patients who were involved at a greater level than they preferred reported outcomes that did not differ from those who reported concordance. This finding is very reassuring for the advocates of shared decision making in Asian countries, where patients, especially elderly patients, tend to take a backseat in decision making.

Being less involved in decision making than desired was significantly associated with poorer quality of physician communication and care coordination compared with concordance. Our findings suggest that encouraging patient participation is a safer approach to minimizing poor outcomes.

1 Introduction

Shared decision making (SDM) is now widely accepted as the gold standard to promote patient empowerment and care pathways that align with a patient’s values and preferences [1]. Previous studies have shown that SDM, particularly active patient participation, is associated with better psychological health outcomes and higher satisfaction with care. However, SDM or an active role in decision making is not desired by all patients with cancer [2,3,4,5]. Some patients may not want to take the responsibility of making important decisions at this vulnerable time in their lives [6]. They may lean on their family caregivers [7, 8] and/or physicians [2, 9] to make the decisions for them. Conversely, other patients want to be the ones making the final call on their care plans [10]. If patients are involved in decision making more [11] or less [9, 12] than they desire, they may experience anxiety and stress and perceive quality of care (QoC) to be poor. In particular, if they do not participate as much as they desire, they may lose their confidence or be dissatisfied with care as they may end up with a treatment that they do not desire [13].

Previous studies show evidence of discordance between patients’ perceived and preferred roles in decision making [14]. Age [15, 16] and education [16] were found to be significant predictors of experiencing discordance. Yet, only a limited number of studies investigated the association between discordance, psychological distress, and satisfaction with consultation or care [13, 17,18,19]. The findings from these studies have been mixed: only one study found a significant association between discordance and anxiety [13], and one other found a significant association between discordance, depression, and satisfaction with the decision-making process [18]. Also, these studies have focused exclusively on the patient–physician dyad in decision making. More research is needed to understand the relationship between discordance and patient outcomes when decision making is also influenced by family caregivers [20,21,22]. This is especially the case in Asian societies where adult children often consider caregiving as part of their duty to their parents, and the elderly expect their family members to take care of them, including making medical decisions on their behalf [23].

The objectives of this study were twofold: (1) to examine discordance between patients’ perceived and preferred roles in decision making in the management of advanced cancer (2) to investigate the associations of discordance with psychological distress and perceived QoC. We hypothesized that patients with discordance in decision making will have greater anxiety, more depressive symptoms, and poorer perceived QoC than those who experience concordance between preferred and perceived roles in decision making. We also hypothesized that the type of discordance matters, in that participating in treatment decision making at a lesser level than preferred will be associated with worse outcomes than being involved at a greater level than preferred. The findings from this study can inform development of interventions to implement SDM models where caregivers are important stakeholders and to match patients’ perceived roles with their preferred roles in decision making.

2 Methods

2.1 Participants and Study Setting

This cross-sectional study employed baseline data from an ongoing cohort study described elsewhere [24]. From July 2016 to March 2018, patients were recruited from outpatient oncology clinics at National Cancer Centre Singapore and National University Hospital in Singapore. Patients who were Singapore citizens or permanent residents, aged at least 21 years, and diagnosed with stage IV solid cancer were identified from medical records and approached by trained interviewers. They were screened for adequate functional status for participation by scoring an Eastern Cooperative Oncology Group performance status ≤ 2. The Abbreviated Mental Test was also administered for participants aged ≥ 60 years. The study team approached 1042 eligible patients. Of these, 600 (57%) consented to participate and completed the survey. The main reasons for declining to participate in the study were not being interested in the study, feeling tired or ill, not having enough time for the survey, and caregiver declining to participate on behalf of the patient. One patient who did not answer questions regarding his/her perceived role in decision making was excluded, resulting in an analysis of 599 patients (Fig. 1). The study was approved by the SingHealth Centralized Institutional Review Board (2015/2781).

Fig. 1
figure 1

Participant recruitment flow diagram

Full size image

2.2 Survey Development

The survey instrument was first developed in English. Through cognitive interviews, the survey was tested with ten participants for readability, and framing of the questions was finalized based on the feedback from these interviews. The English version was then translated to Mandarin and Malay, and then back translated to English by professional translators to cover the three most spoken languages in Singapore. The Mandarin and Malay versions were also reviewed by the bilingual study team members for readability. The survey was administered by trained interviewers who could speak the language of the patient’s choice. The following measures were used in the study.

Perceived and preferred roles in decision making Participants were first asked who had been responsible for making the most important treatment decisions and presented with the options of myself, my family, and my doctor (electronic supplementary material [ESM], A). Participants could select as many of the options as they felt were applicable. A second question elicited the role and capacity to which the individual(s) selected in the first question were involved in making decisions. These questions were used to identify the perceived roles in decision making. A similar two-step question format was used to assess patient’s preferred roles for these parties in treatment decisions. These questions were adopted from the Cancer Care Outcomes Research and Surveillance (CanCORS) study [25] but modified to incorporate family involvement in addition to that of patients and physicians. Decision-making options were subsequently grouped into six categories: (1) Patient-alone decision making, (2) patient-led decision making (i.e., patient decides after considering their family’s and/or physicians’ opinion), (3) collaborative decision making (i.e., patient decides with their family and/or physicians), (4) family-led decision making (i.e., patient’s family decides after considering patient’s and/or physicians’ opinion), (5) physician-led decision making (i.e., patient’s physicians decide after considering patient’s and/or family’s opinion), and (6) no patient involvement (i.e., family and/or physicians decide) (ESM B).

Discordance in decision making Discordance was defined as any discrepancy between preferred and perceived roles in decision making. Discordance may occur within the same category or across categories. We classified discordance into three categories: (1) involvement in decision making at a lesser level than preferred, (2) involvement in decision making at a greater level than they preferred, and (3) no change in patient involvement while there is discordance on the role of family and/or physician (e.g., patient decides with their family but prefers making decisions with their physicians).

Psychological distress outcomes Participants completed the Hospital Anxiety and Depression Scale (HADS) that measures anxiety (HADS Anxiety) and depression (HADS Depression). Each subscale has seven items, resulting in total score of 0–21 per subscale. A higher score indicates higher anxiety and depressive symptoms.

Perceived QoC outcomes Patients’ experience with physicians and care setting was assessed using questions previously used by the CanCORS study. Items are grouped into two domains: physician communication (QoC-PC) (five items), and healthcare coordination and responsiveness (QoC-CR) (seven items). Items related to the quality of nursing care was not used in the analysis of this paper since it was deemed to be unrelated to the decision-making process. Scores for each domain ranged from 0 to 100, with a higher score indicating better perceived QoC.

Patient characteristics Participants reported their age, ethnicity, marital status, years of education, current working status, and monthly household income. Cancer type was identified through medical records. Symptom burden was measured as the sum of patient-reported symptom severities on a Likert scale from 0 to 4 (not at all, a little bit, somewhat, quite a bit, very much). The list of symptoms, taken from FACIT-PAL-14 [26], were constipation, dryness of mouth, lack of energy, nausea, shortness of breath, weight loss, swelling in parts of body, dry throat, and vomiting.

2.3 Statistical Analysis

We first presented descriptive statistics on patient characteristics and outcomes then on perceived and preferred roles in decision making. We used the Stuart-Maxwell test, which is the appropriate test for comparing multiple correlated (or within-subject) proportions, to compare patients’ perceived and preferred roles in decision making.

We also investigated the predictors of discordance between perceived and preferred roles in decision making using a multinomial logit model. The dependent variable was the types of discordance: (1) involved at a lesser level than preferred, (2) involved at a greater level than preferred, (3) no change in patient involvement (reference category: concordance in perceived and preferred levels of involvement in decision making). The independent variables were medical characteristics (symptom burden and cancer type [respiratory cancer, colorectal and gastrointestinal cancer, breast cancer, genitourinary cancer, gynecologic cancer, and other {reference}]) and sociodemographic characteristics (age, ethnicity [Chinese = 1, Malay/Indian/others = 0], marital status [married = 1, otherwise = 0], education [post-high school or higher education = 1, high school and below = 0], work status [have part-time or full-time job = 1, otherwise 0], and income [Singapore dollar {S$}4000 per month or higher = 1, otherwise 0]). Gender was not included as it was correlated with types of cancer.

In univariable and multivariable regression analyses, we investigated the association of discordance with psychological distress and perceived QoC. For psychological distress, we employed anxiety and depression (as measured by HADS) as dependent variables. For perceived QoC, we used physician communication and healthcare coordination and responsiveness (as measured by QoC) as dependent variables. The independent variables of interest were types of discordance as described above. We first ran univariable analyses with only variables on the type of discordance. We then ran multivariable analyses controlling for patient medical and sociodemographic characteristics as described above. Assuming linear relationships between the dependent and independent variables, ordinary least squares regressions were used for all univariable and multivariable models. All analyses were conducted using Stata version 14.2.

3 Results

3.1 Patient Characteristics

Table 1 presents patient characteristics. The mean ± standard deviation (SD) age of patients was 60.4 ± 10.6 years. Less than half of patients were male (46%), had a post-high school education (32%), had part- or full-time jobs (36%), or had a total monthly household income of more than S$5000 ($US3606) (26%). The majority of patients were married (73%) and ethnically Chinese (79%). The mean ± SD symptom burden was 5.05 ± 5.05. Respiratory cancer (28%) was the most commonly reported, followed by colorectal and gastrointestinal (26%) and breast (16%) cancers. The mean ± SD scores were quite low for anxiety (2.34 ± 2.77) and depression (2.95 ± 2.92) and quite high for perceived quality of physician communication (90.78 ± 19.12) and care coordination (91.93 ± 17.91).

Table 1 Patient medical and sociodemographic characteristics (N = 599)
Full size table

3.2 Patients’ Perceived and Preferred Role in Decision Making

Table 2 shows descriptive statistics on patients’ perceived and preferred roles in decision making and Stuart-Maxwell test results for discordance in decision making. The most common perceived role in decision making was collaborative, where patients made decisions together with their family and/or physicians (32%), followed by patient-led decision making (29%). About 15% of patients reported no involvement in decision making, whereas 13% reported making decisions alone. Only 8% of the patients reported that decisions were led by their physicians. Although 4% of the patients reported that decisions were led by their family, 65% of all patients reported family involvement (to some extent). A similar pattern was observed for patients’ preferred roles in decision making, where 32% of patients preferred collaborative decision making and 26% preferred patient-led decision making.

Table 2 Descriptive statistics and Stuart-Maxwell test findings on perceived and preferred roles in decision making (N = 599)
Full size table

The proportion of patients reporting patient-led decisions was larger than the proportion of patients who preferred this style (29 vs. 26% for perceived and preferred, respectively; p = 0.009), and the proportion of patients reporting no involvement in decision making was smaller than the proportion of those who preferred this style of decision making (15 vs. 17% for perceived and preferred roles, respectively; p = 0.040).

3.3 Discordance in Perceived and Preferred Roles in Decision Making

Overall, 93 (16%) patients reported discordance in decision making (Table 3). Among patients with discordance, 31 (33%) reported being involved in decision making at a lesser level than they preferred, 44 (47%) reported being involved at a greater level than they preferred, and 18 (19%) reported discordance where patient involvement did not change. The multinomial logit regression results showed that, compared with experiencing concordance, older patients were less likely to be involved in decision making at a lesser level than they preferred (β = − 0.041 [95% confidence interval {CI} − 0.079 to − 0.003]), and patients with higher education (β = 0.735 [95% CI 0.005–1.466]) were more likely to be involved at a greater level than they preferred (ESM C). Ethnicity, marital status, work status, income, and symptom burden were not significant predictors of discordance.

Table 3 Discordance and types of discordance (N = 599)
Full size table

3.4 Association of Discordance with Psychological Distress and Perceived Quality of Care

Table 4 presents the results from eight univariable and multivariable regressions, respectively. Univariable regressions revealed that patients who were involved in decision making at a lesser level than they preferred reported greater anxiety (β = 0.906 [95% CI −0.010 to 1.911]), poorer quality of physician communication (β = −10.251 [95% CI − 17.146 to − 3.356]), and poorer QoC coordination (β = − 14.107 [95% CI − 20.509 to − 7.706]) compared with those who experienced decision-making concordance. Univariable regressions also revealed that patients who reported discordance with no change in patient role reported poorer quality of physician communication (β = − 9.319 [95% CI − 18.258 to − 0.381]) and poorer QoC coordination (β = − 9.610 [95% CI − 17.909 to − 1.311]) compared with those who experienced concordance.

Table 4 Univariable and multivariable estimates for discordance in decision making and its association with psychological well-being and perceived quality of care – OLS estimates
Full size table

Multivariable analyses that controlled for patient characteristics found that involvement at a lesser level than preferred and discordance with no change in patient role were associated with poorer quality of physician communication (β = − 9.478 [95% CI − 16.303 to − 2.653] and β = − 9.184 [95% CI − 18.066 to − 0.301]) and poorer care coordination (β = −11.658 [95% CI − 17.718 to − 5.597] and β = − 8.856 [95% CI − 16.744 to − 0.968]) compared with concordance. Compared with patients who experienced concordance, patients who were involved less than they preferred reported 9.48 (about half of SD) and 11.66 (about two-thirds of SD) lower scores for the quality of physician communication and care coordination, respectively. Similarly, patients who reported discordance with no change in patient role reported 9.18 and 8.86 (about half of SD) lower scores for the quality of physician communication and care coordination, respectively. Full multivariable regression model outputs are shown in ESM D.

4 Discussion

Our findings suggest that the perceived and preferred degree of involvement in decision making varies across patients with advanced cancer. The most common experience in decision making reported by patients was a collaborative approach, followed by a patient-led approach. We found a similar pattern for preferred roles in decision making. Passive (i.e., family-led, physician-led, or no involvement) decision making was reported and preferred by 27% and 28% of patients, respectively. Families were important stakeholders in making decisions regarding cancer care; yet, the extent to which patients reported their family’s involvement in decision making varied substantially. Overall, 65% of patients in our sample reported their families being involved in decision making to a certain extent, including 4% reporting that decisions were led by their families with the patients’ and physicians’ involvement and a further 4% reporting that decisions were solely made by their families. These figures were higher than those from the CanCORS study, in which family-led decision making was reported by 1.3% and 2.3% for Caucasian patients with cancer and English-speaking Asians, respectively. However, they were one-third of the figures reported for non-English-speaking Asians (12.8%) in the CanCORS study [20].

Overall, 16% of patients reported discordance between perceived and preferred roles in decision making. This is similar to the 20% discordance reported by women with breast cancer in Hong Kong [19] but lower than the 66% reported by patients with cancer in Australia [13]. However, when comparing studies, differences in the questions assessing roles in decision making should be considered. The low level of discordance in our study may be due to the extent of family involvement in patient’s decision making, a factor that has not been investigated in previous studies. Future research should investigate discordance in decision making in other countries and the underlying factors for differences across different cultures.

We also found that, among patients with discordance in decision making, more patients reported being involved at a greater level than they preferred compared with those who reported being involved in decision making at a lesser level (47 vs. 33%). This finding should be interpreted through a cultural lens. As stated in the introduction, patients in Asian societies generally expect their family members to take care of them and see this as the duty of the family. This may set patients with an expectation of being passive instead of taking charge of their care management. However, this expectation may not be realized for all those who desire it, as our findings showed.

We also investigated the predictors for types of discordance. Consistent with the previous studies [15, 16], we found that age and education were significant predictors of discordance. Specifically, our findings show that older patients were less likely to be involved at a lesser level than they preferred. This may be due to the authority older individuals have in Asian societies, such that they are not likely to be left out of the decision-making process [27]. We also found that patients with a higher level of education were more likely to be involved at a greater level than they preferred. Individuals with a higher level of education may end up being involved more than desired as they discuss their treatment with their physicians and family caregivers.

Our findings indicate that the type of discordance matters. Being involved in decision making less than desired was significantly associated with poorer quality of physician communication and care coordination compared with concordance. It is possible that patients who were involved less than they desired may have had unanswered concerns or questions or may have felt that their preferences were disregarded by their physicians. In addition, they may have been placed on a treatment regimen that was not consistent with their preferences and values. Conversely, patients who participated as much as or more than they desired were more likely to have raised their concerns and questions to their physicians, received a treatment consistent with their preferences and ultimately were more likely to be satisfied with physician communication and care coordination. Interestingly, a small number of patients reported that the roles of the physicians and/or families were not at the preferred levels, and they reported poorer quality of physician communication and care coordination compared with concordance.

The strengths of this study include its large sample size, inclusion of patients with different types of cancer, and the use of customized questions on decision-making roles to assess the involvement of both families and physicians. Most studies in this area were conducted in Western countries and focus on the patient–physician dyad in decision making. Since families are particularly important stakeholders in decision making in Asian countries, it was necessary that we focused on the triad of patients, physicians, and family members. This study is also one of the few to investigate the association between discordance in perceived and preferred roles, and patient outcomes.

Our study has several limitations. First, our data can only indicate associations and cannot investigate causality between discordance and patient outcomes. Second, although we recognize that patients’ perceived and preferred role in decision making may vary based on the type of decision (e.g., chemotherapy, palliative care), we only asked about their role in general decision making. Asking for specific types of decisions would have been cumbersome and tiring for the patients and possibly more susceptible to recall bias than asking for treatment decisions in general. Third, reverse causality between discordance, psychological distress, and perceived QoC outcomes may exist. Finally, patients reported their preferred role in decision making immediately after they reported their perceived role in decision making. This format and timing of survey questions might have led patients to report more concordance in decision making, resulting in the underestimation of the number of patients who may have experienced discordance.

5 Conclusions

Our findings show that current clinical practice in Singapore seems to provide an environment for most patients to participate in decision making at their desired level. Our findings also suggest that patients who were involved at a greater level than they preferred reported no difference in outcomes from those who reported concordance. This finding is very reassuring for the advocates of SDM in Asian countries, where patients, especially those who are elderly, tend to take a backseat in decision making. Our findings suggest that involving patients at a lesser level than they prefer can lead to poorer perceived quality of physician communication and care coordination and that encouraging patient participation is a safer approach to minimizing poor outcomes. Future research should develop tool(s) and strategies that aid healthcare providers to involve patients in decision making.