Psychometric assessment of the Chinese version of the decisional conflict scale in Chinese women making decision for breast cancer surgery

Abstract

Objective

The decisional conflict scale (DCS) measures the perception of uncertainty in choosing options, factors contributing to decision conflict and effective decision making. This study examined the validity and reliability of the Chinese version of the DCS in Hong Kong Chinese women deciding breast cancer (BC) surgery.

Method

A Chinese version of the 16‐item DCS was administered to 471 women awaiting initial consultation for BC diagnosis. Confirmatory factor analysis (CFA) assessed the factor structure. Internal consistency, and convergent and discriminant validities of the factor structure were assessed.

Results

CFA revealed the original factor structure of the DCS showed poor fit to this sample. Exploratory factor analysis revealed an alternative three‐factor structure, Informed and Values Clarity, Uncertainty and Effective Decision and Support, was optimal. Cronbach's alpha ranged from 0.51 to 0.87. Correlations between decision‐making difficulties and satisfaction with medical consultation demonstrated acceptable convergent validity. Construct validity was supported by correlations between decision regret and psychological distress. Discriminant validity was supported by differentiation between delaying and non‐delaying decision‐makers.

Conclusions

The three‐factor DCS‐14 is a valid and practical measure for assessing decisional conflict in deciding BC surgery. It shows good potential for use in assessing decision satisfaction for women diagnosed with BC.

Introduction

Engaging patients in treatment decision making (TDM) has become increasingly important in medical care. Much work has focused on identifying and evaluating interventions such as decision aids that might facilitate patients' TDM.1 Several outcome measures have been developed to evaluate the effect of decision‐making support interventions. One variable extensively used for evaluating patients' decision‐making process in various medical contexts is decisional conflict. Decisional conflict is defined as a state of uncertainty about the course of action to be taken.2 Four conditions that are likely to result in such state have been identified: (i) when decisions involve risk or outcome uncertainty, (ii) when decisions involve significant potential gains and losses, (iii) when there is a need to make value trade‐offs in selecting a course of action and (iv) when anticipated regret over the positive aspects of rejected options.2 Manifestations of decisional conflict include verbalization of uncertainty, hesitation in choosing between choices, delayed decision making and questioning personal values and beliefs while attempting to make decisions.2

The decisional conflict scale (DCS)2 was developed based on the construct of decisional conflict. The DCS is a 16‐item measure of individual's perception of uncertainty in choosing options, perception of modifiable factors contributing to the uncertainty and perception of the effective decision making. The DCS has been widely used to assess decisions among individuals facing various medical decision making regarding hormone replacement therapy,3 prostate‐specific antigen testing,4 breast cancer surgery,4, 5 mammography screening,6 benign prostatic hypertrophy,7 end‐of‐life decision making8 and BRCA genetic testing.9, 10 The DCS was tested in North America populations and was found to yield good internal reliability and predictive validity.2, 8, 10 High decisional conflict as measured with the DCS has been found in association with delayed or perceived uncertainty to undergo vaccine immunization,2 breast cancer screening2 and genetic testing for hereditary breast and ovarian cancer.10 Studies that assessed the effect of decision–support interventions have found that individuals receiving decision–support intervention have lower decisional conflict than those without decision support.3, 7, 8 Dutch5 and French9 versions of the DCS have also demonstrated sufficient internal consistency and predictive validity when using the total score.

In contrast, the construct validity of the DCS is less clear. The original validation work reported the DCS composed of three subscales including Uncertainty (3 items), Factors contributing to the uncertainty (9 items) and Perceptions of effective decision making (4 items).2 Subsequently, the subscale of factors contributing to the Uncertainty was split into three subscales: Informed (3 items), Values clarity (3 items) and Support (3 items) leading to a total of five subscales for the DCS.11 Also, different factor structures for the DCS have been reported. Three previous studies examined the factor structure of the DCS, using an exploratory factor analytic (EFA) approach. Katapodi et al.10 examined the factor structure of the DCS English version in 342 women deciding genetic testing for hereditary breast and ovarian cancer, revealing three constructs accounting for 82% of the total variance. However, EFA failed to differentiate between the constructs of Uncertainty and Factors contributing to the uncertainty. Mancini et al.9 examined the factor structure of the French version of the DCS in 644 cancer patients facing decisions to undergo breast cancer genetic testing. EFA identified four factors, with Informed and Support subscales loading onto a single factor, accounting for 71% of the variance. Confirmatory factor analysis (CFA) showed the four‐factor model had a better fit than the original five‐factor model. The factor structure of the Dutch version of the DCS was examined in 170 patients with cancer making treatment decisions.5 They performed CFA to test the original three‐factor structure and demonstrated a poor fit. Using EFA, the study revealed four‐factor structure, accounting for 67.5% of the variance. The Dutch model failed to differentiate between the constructs of Uncertainty, Effective decision and Factors contributing to the uncertainty. The differences in factor structure identified across these studies may be due to clinical and cultural heterogeneity of the samples studied or to weak conceptual validity for decisional conflict. The DCS has been examined exclusively in Western countries, predominately using Caucasian populations. A Chinese version of the DCS has been developed, but has not been validated.

The main objective of this study was to examine the psychometric properties of the Chinese version of the five‐factor model of DCS. In particular, this study examines the factorial validity, construct validity and reliability of the DCS in a sample of Hong Kong Chinese women deciding breast cancer surgery.

Methods

Participants

Following the approval from the Ethics Committee of the University of Hong Kong and Hospital Authority, the study sample was recruited from two Hong Kong government‐funded breast cancer centres. Women were eligible for study participation if they met the following criteria: (i) newly diagnosed with breast cancer, (ii) awaiting their initial consultation wherein the surgeon discusses diagnosis and surgical options, (iii) were native Cantonese or Mandarin speakers and (iv) having no communication problems. All eligible women gave written informed consent and were asked to complete a baseline interview at their next follow‐up appointment scheduled between 4 and 7 days after the initial consultation and a follow‐up interview at 1 month post‐surgery. As the standard practice in these two breast cancer centres, women are asked not to make decisions for breast cancer surgery during their initial consultation and instead are asked to phone back a week later to indicate their decision, to minimize impulsive decisions that might be later regretted. Hence, any woman who had already made her surgical decision would have just confirmed her decision when she completed the baseline interview. All interviews were conducted by one of the trained research assistants. During the study period, 520 eligible women were identified and invited to participate, and 471 (91%) agreed to participate in the study.

Measures

Decisional conflict scale

The DCS comprises 16 items covering 5 domains: Informed (3 items), Values clarity (3 items), Support (3 items), Uncertainty (3 items) and Effective decision (4 items). The DCS has two full versions (statement format and question format).11 The full version with question format was used in this study as it allowed women with limited literacy to participate, therefore improving representativeness. Previous studies have shown no effective differences in data collected by these two methods. Patients were asked to rate each item using a five‐point Likert scale: 0, yes; 1, probably yes; 2, unsure; 3, probably no; 4, no. Domain scores are calculated by averaging the sum of individual domain item scores, then multiplying the product by 25. Hence, each domain score ranges from 0 to 100. Higher scores indicate more decisional conflict. The Chinese version of the DCS provided by the original author2 was adopted in this study. The DCS was translated into Chinese by one bilingual team member and then back translated into English by a second independent team member. Original and back‐translated versions were compared. Discrepancies were identified and resolved by discussion and translation procedure reiteration until a satisfactory semantic equivalence was found. The Chinese version of the DCS required approximately 5 min to complete.

Comparative measures

Decision regret scale

The decision regret scale consists of five items measuring distress or remorse after a health‐care decision.12, 13 Patients indicate the extent to which they agree with the statements on a five‐point Likert scale ranging from 1 ‘strongly agree’ to 5 ‘strongly disagree’. Higher scores indicate greater decision regret. The decision regret scale has good internal consistency (Cronbach's α = 0.81–0.92).13 We developed a Chinese version of the decision regret scale using the translation procedures described above. The Chinese version also has good internal consistency (Cronbach's α = 0.83) based on the present sample.

Perceived TDM difficulties scale

Perceived TDM difficulties were measured using the eight‐item Perceived TDM difficulties scale14 using a four‐point Likert response scale ranging from ‘strongly disagree’ to ‘strongly agree’. Higher scores indicate a greater level of decision‐making difficulties. Internal consistency (Cronbach's α) for the scale was 0.85, suggesting good internal reliability.

Medical interview satisfaction scale

Modified versions of the Cognitive and the Affective subscales of the Medical Interview Satisfaction Scale15, 16 measured women's satisfaction with their TDM medical consultation. The Cognitive subscale measures patients' satisfaction with their understanding of their doctor's explanation of the medical problem and its treatment; the Affective subscale measures patients' feelings of trust and confidence in their doctors' attention to their concerns.15 To minimize assessment load, only the four most relevant items from each subscale were used. Patients indicate the extent to which they agree with the statements on a five‐point Likert scale ranging from 1 ‘strongly agree’ to 5 ‘strongly disagree’. The scale has good internal consistency (Cronbach's α = 0.83).16 A total satisfaction score was obtained by summing the two subscales' scores.

Hospital anxiety and depression scale

Patient anxiety and depression were measured by the Chinese version of the 14‐item Hospital anxiety and depression scale that uses a four‐point categorical response format.17, 18 Respondents were asked to indicate how they had felt in the past week on a four‐point categorical response format. Both anxiety (Cronbach's α = 0.93) and depression (Cronbach's α = 0.90) scales demonstrate good internal consistency.17

All measures were gathered at baseline except for decision regret, which was measured at 1 month post‐surgery. Measures of anxiety and depression were assessed at both baseline and 1 month post‐surgery. Sociodemographic and medical data were also collected from patients and medical records, respectively.

Statistical analysis

To assess the factorial validity of the five‐factor model of the DCS, CFA was performed using EQS 6.1 for Windows.19, 20 In addition to the five‐factor subscores, a total score of the DCS was used to measure the extent of decisional conflict. Therefore, both hierarchical and correlated models of the DCS were compared. Several fit indices were used to assess how well the tested models fitted the data. While the chi‐squared statistic is the commonly used fit index, its sensitivity to sample size limits its applicability to evaluate model approximation.20 Therefore, other commonly used fit indices, the comparative fit index (CFI),21 the adjusted goodness‐of‐fit index (AGFI),22 the goodness‐of‐fit index (GFI),22 root mean square error of approximation (RMSEA)23 and the RMSEA 90% confidence interval24 were used to assess model fit. CFI and GFI values of ≥0.90, AGFI values of ≥0.80 and RMSEA values of ≤0.080 were taken as indicative of good fit.20 Mardia's normalized estimate was used to assess whether the data are normally distributed, with the estimate value of >5.00 suggesting that data are non‐normally distributed,20 when the Satorra–Bentler (S–B) chi‐squared, which incorporates a scaling correction for non‐normal sampling distributions is instead reported.

If the proposed five‐factor model of the DCS was unsupported, we adopted an exploratory factor analysis, followed by CFA to verify the revised factorial structure. Cronbach's alpha coefficient was calculated to reflect the scale's internal consistency.25 Minimal acceptable alpha was specified at 0.70.

Convergent validity, the extent to which theoretically related measures are correlated with each other, was evaluated by correlating the DCS with the perceived TDM Difficulties Scale as both measures assess the experience of TDM. Construct validity was evaluated by correlating the DCS with measures of decision regret, psychological distress and patient satisfaction with medical consultation. We hypothesized that the DCS would correlate with greater perceived TDM difficulties,26 decision regret,2 psychological distress27 and lower patient satisfaction with medical consultation.26 All correlation was performed using Pearson's correlation coefficient. We also assessed construct validity using a known‐group comparison approach (delaying vs. non‐delaying decision‐maker). Patient delay in making a decision was assessed at baseline interview. Participants were asked to indicate whether they had made a decision for breast cancer surgery. Delaying decision‐makers describes those who had not made a surgical decision by the time they completed the baseline interview, whereas non‐delaying decision‐makers describe those who had made the decision by that time point. We hypothesized that patients who delayed making their treatment decision3 would express greater decision conflict than those who made a prompt decision. We used Student's t‐test to examine this hypothesis. Apart from CFA, all analyses were performed using SPSS v. 18.0 (IBM SPSS statistics, Armonk, NY, USA).

Results

Participant characteristics

Of the 520 eligible patients, 471 (91%) provided full informed consent and completed the baseline questionnaire and 421 (89%) completed 1 month post‐surgery follow‐up assessment. Table 1 summarizes the demographic and clinical characteristics of the participants.

Table 1.

Sample characteristics (n = 471)

Characteristic	n (%)
Age (years)
Mean (SD)	54.4 (9.9)
Range	29–86
Marital status
Single	64 (13.6)
Married/cohabiting	313 (66.5)
Divorced/separated	42 (8.9)
Widowed	52 (11.0)
Education
No formal education	26 (5.5)
Primary	140 (29.7)
Secondary	258 (54.8)
Tertiary	47 (10.0)
Total monthly household income (HK$)a
Less than 10 000	171 (36.3)
10 001–20 000	151 (32.1)
20 001–30 000	67 (14.2)
30 001 or above	53 (11.9)
Missing data	29 (6.2)
Occupation
Full‐time employment	160 (34.0)
Part‐time employment	32 (6.8)
Retired	92 (19.5)
Housewife	144 (30.6)
Unemployed (before diagnosis of cancer)	34 (7.2)
Unemployed (after diagnosis of cancer)	6 (1.3)
Missing	3 (0.6)
Stage of disease
0	113 (24.0)
I	120 (25.5)
II	108 (22.9)
III/IV	39 (10.3)
Missing	91 (19.3)
Surgical decision made at baseline interview
Yes	412 (87.5)
Chosen surgical option reported at baseline interview (n = 412)
Breast conserving therapy	145 (35.1)
Modified radical mastectomy	225 (54.6)
Modified radical mastectomy and immediate breast reconstruction	40 (9.7)
Missing	2 (0.4)
Type of surgery reported at follow‐up interview
Breast conserving therapy	130 (27.6)
Modified radical mastectomy	230 (48.8)
Modified radical mastectomy and immediate breast reconstruction	59 (12.5)
Missing	52 (11.0)

SD, standard deviation.

^a1 US$ = HK$7.8.

Factorial validity

Both hierarchical and correlated five‐factor models of DCS were tested. Mardia's normalized estimates, the Z‐statistic of 54.64, suggested the data were non‐normally distributed. Therefore, we report the S–B chi‐square. Table 2 summarized the goodness‐of‐fit indices of the two models, indicating both models failed to meet the minimum fit criterion. Consequently, an exploratory factor analysis was performed.

Table 2.

Goodness‐of‐fit indices of confirmatory factor analyses of the DCS

	S–B χ2	d.f.	P‐value	RMSEA (90% CI)	CFI	GFI	AGFI
Original DCS
Five‐factor Hierarchical model	588.38	100	<0.001	0.106 (0.098–0.115)	0.677	0.812	0.744
Five‐factor Correlated model	296.97	94	<0.001	0.071 (0.062–0.080)	0.866	0.883	0.830
Revised DCS
Three‐factor Correlated model (14 items)	175.86	74	<0.001	0.056 (0.046–0.067)	0.917	0.935	0.907
Three‐factor Hierarchical model (14 items)	143.87	73	<0.001	0.047 (0.036–0.059)	0.943	0.935	0.906

DCS, decisional conflict scale; S–B χ², Satorra–Bentler scaled chi‐squared statistic; d.f., degree of freedom; CFI, comparative fit index; GFI, goodness‐of‐fit index; AGFI, adjusted goodness‐of‐fit index; RMSEA, root mean square error of approximation; CI, confidence interval.

Bartlett's test statistic for the significance of the correlation matrix was 2373.60 (P < 0.001), indicating appropriateness for factor analysis. Sampling adequacy was confirmed by the Kaiser–Meyer–Olin (KMO) statistic of 0.87 exceeding the minimum adequacy value of 0.50. Using latent root criteria with an eigenvalue >1 and scree plots to identify the optimal number of factors, exploratory factor analysis suggested a three‐factor model (Table 3), explaining 53% of variance. Items from the Informed subscale and items from the Values Clarity subscale loaded on to the same factor. Items from the Uncertainty subscale and items from the Effective Decision subscale seem to share the same dimension. The third factor comprised two of the three items from the Support subscale. The remaining item from the Support subscale ‘Are you choosing without pressure from others?' loaded on to the second factor. Two items (one from the Uncertainty subscale ‘Are you clear about the best choice for you?’ and another from the Effective Decision subscale ‘Do you feel you have made an informed choice?’) had cross‐loadings with similar factor loadings. These two cross‐loading items were therefore removed.

Table 3.

Exploratory factor analysis of decisional conflict scale (DCS; principal components analysis, oblimin rotation)

	Factor			Communality
	1	2	3
Know which options are available (I)	0.60			0.41
Know the benefits of each option (I)	0.79			0.62
Know the risks and side‐effects of each options (I)	0.79			0.63
Clear about which benefits are matter most (V)	0.78			0.60
Clear about which risks and side‐effects are matter most (V)	0.80			0.65
Clear about which is more important (V)	0.82			0.68
Have enough support from others to make a choice (S)			−0.71	0.51
Choosing without pressure from others (S)		0.49		0.28
Have enough advice to make a choice (S)			−0.80	0.67
Clear about the best choice (U)	0.62	0.51	−0.50	0.61
Feel sure about what to choose (U)		0.76		0.59
Decision is easy to make (U)		0.49		0.32
Have made an informed choice (E)	0.50		−0.56	0.48
Decision shows what is important (E)		0.58		0.36
Expect to stick with the decision (E)		0.77		0.60
Satisfied with decision (E)		0.68		0.51

I, Informed subscale; V, Values clarity subscale; S, Support subscale; U, Uncertainty subscale; E, Effective decision subscale (according to the original five‐factors scale).

CFA was performed to test the revised three‐factor DCS with omission of the two cross‐loading items. The 3‐factor model with 14 items had an excellent fit to the data (RMSEA <0.0.06, CFI, GFI and AGFI all >0.90; Table 2). We also tested the three‐factor hierarchical model with omission of the two items using CFA. The hierarchical model with 14 items also revealed an excellent fit to the data (RMSEA <0.05, CFI, GFI and AGFI all >0.90; Table 2). Because the three‐factor correlated model and the three‐factor hierarchical model with omission of the cross‐loading items are nested, we can statistically compare their adequacy using chi‐squared difference tests. The comparison of the three‐factor correlated model with the three‐factor hierarchical model reveals that the hierarchical model explains the data significantly better than the correlated model (Δχ²(1) = 31.99, P < 0.001).

Reliability

The overall internal consistency for the 14‐item version of the 3‐factor hierarchical DCS (DCS‐14) was high, with Cronbach's α = 0.81. While the Informed and Value Clarity subscale and Uncertainty and Effective Decision subscale demonstrated good internal consistency, with Cronbach's α = 0.87 and α = 0.71, respectively, the Support subscale showed poor internal consistency, with Cronbach's α = 0.51.

Convergent validity

All of the DCS‐14 subscales and the overall scale demonstrated an expected moderate correlation [correlation coefficients (r) ranged from 0.36 to 0.53] with the measure of perceived difficulties in TDM (Table 4).

Table 4.

Correlation matrix for convergent and construct validity

	DCS‐14 Informed and Values Clarity subscale	DCS‐14 Uncertainty and Effective Decision subscale	DCS‐14 Support subscale	DCS‐14 total
Perceived difficulties in treatment decision making	0.48a	0.35a	0.40a	0.53a
Decision regret	0.26a	0.10 (P = 0.035)	0.10 (P = 0.05)	0.21a
Patient satisfaction with medical consultation	−0.37a	−0.43a	−0.38a	−0.52a
Anxiety at baseline	0.42a	0.16a	0.21a	0.31a
Depression at baseline	0.41a	0.20a	0.22a	0.33a
Anxiety at 1 month post‐surgery	0.27a	0.08	0.20a	0.22a
Depression at 1 month post‐surgery	0.25a	0.14 (P = 0.006)	0.20a	0.25a

DCS, decisional conflict scale.

^a P < 0.001.

Construct validity

All of the DCS‐14 subscales and the overall scale correlated positively with the concurrent measure of anxiety (r ranging from 0.20 to 0.42), depression (r ranging from 0.20 to 0.41) as well as negatively with the measure of patient satisfaction with medical consultation (r ranged from −0.37 to −0.52; Table 4). All of the DCS‐14 subscales and the overall scale showed positive correlation (r ranged from 0.14 to 0.25) with the measure of depression and the measure of decision regret (r ranged from 0.10 to 0.26) assessed at 1 month postoperatively. With the exception of the Uncertainty and Effective Decision subscale, other subscales and the overall scale (r ranged from 0.20 to 0.27) correlated positively with anxiety scores measured at 1 month postoperatively (Table 4). Student's t‐tests were used to test the known‐group comparison (delaying vs. non‐delaying decision‐maker). Congruent with expectations, delaying decision‐makers reported significantly higher scores on all of the DCS‐14 subscales and the overall scale (Table 5).

Table 5.

Summary of known‐groups comparisons

	Delay vs. Non‐delay decision‐maker	Mean scores (SD)	n	P‐value
DCS‐14	Delay	36.85 (17.33)	32	<0.001
Informed and Values Clarity subscale	Non‐delay	14.33 (13.64)	407
DCS‐14	Delay	33.33 (31.93)	55	0.025
Uncertainty and Effective Decision subscale	Non‐delay	24.83 (25.40)	406
DCS‐14 Support subscale	Delay	34.95 (26.08)	54	<0.001
	Non‐delay	16.07 (20.59)	406
DCS‐14 Total	Delay	31.42 (17.59)	32	<0.001
	Non‐delay	19.07 (15.03)	401

DCS, decisional conflict scale; SD, standard deviation.

Discussion

The present study evaluated the factorial validity, construct validity and reliability of the Chinese version of the DCS in Chinese women making decisions for breast cancer surgery. Confirmatory factor analyses revealed that the original five‐factor structure of the DCS did not fit the data from the Chinese version of the DCS used in our sample of women choosing surgery for breast cancer. Subsequently, exploratory factor analysis was performed to identify the underlying dimension of the present data. Our EFA revealed a three‐factor structure: (i) Informed and Values Clarity, (ii) Uncertainty and Effective Decision and (iii) Support was optimal.

In contrast to the original proposed structure in which factors contributing to the uncertainty consist of three subscales (i.e. Informed, Values Clarity and Support subscales),11 our results failed to differentiate the Informed subscale and the Values Clarity subscale. Previous studies also failed to clearly differentiate items between these two subscales.5, 10 The informed subscale assesses the extent of which women felt informed about the treatment options, its benefits and risks, whereas the Values Clarity subscale measures women's personal values related to the treatment options. Both subscales assess the extent of women's understanding of the treatment options, and therefore, it is not surprising that they do not differentiate clearly. The Support subscale was differentiated from the Informed and Values Clarity subscale. However, one of the three items ‘Are you choosing without pressure from others?’ from the Support subscale failed to fit onto the expected factor, but loaded onto the factor measuring uncertainty and perceptions of effective decision making. Inspection of the items of the Support subscale indicates that the other two items primarily assessed the perceived support from others in making decision, whereas ‘choosing without pressure from others’ evaluated the extent of interference from others in making decision. This supports the work of Koedoot et al.,5 who evaluated the Dutch version of the DCS on cancer patients choosing breast cancer surgery or choosing palliative chemotherapy and found this item loaded onto the Uncertainty subscale.

In contrast to the previous studies,5, 9, 10 our analyses also failed to differentiate the Uncertainty subscale and the Effective Decision subscale. While the distinction between the Uncertainty and Effective Decision subscales allows to tap different aspects of the quality of the decision, our results suggest that in this sample of Hong Kong women making a decision for breast cancer surgery particularly may not involve intentionally differentiating the state of uncertainty related to the decision from the perceived effectiveness of the decision making. This needs further investigation. Moreover, two items, ‘Are you clear about the best choice for you?’ and ‘Do you feel you have made an informed choice?’ were eliminated as they loaded on more than one factor. Our CFA suggests that the alternative three‐factor DCS with 14 items is the best fitting measurement model in this patient setting. Furthermore, the better fit of the hierarchical model over the correlated model supports the utility of using the total score of DCS to measure decisional conflict rather than subscores which may be less reliable. The finding that various alternative factor structures have been identified across studies is important, reflecting that the original proposed factor structure measuring decisional conflict may not be applicable across different populations. The factor structure differences across studies may be due to the heterogeneity of the sample in terms of type of medical decision making and ethnicity. For example, the original five‐factor structure was validated based on healthy Canadians deciding about influenza immunization or breast cancer screening, whereas the French and the Dutch four‐factor versions were validated based on cancer patients deciding cancer treatment. A three‐factor structure was identified based on American women deciding about genetic testing for hereditary breast and ovarian cancer. DCS is often used as a primary outcome measure in evaluating the effectiveness of decision‐making intervention.1 While DCS was developed based on the construct of decisional conflict and is currently a widely used measure of patient decision‐making across studies, the variability of the factor structure of the DCS across studies suggests that we should not assume the DCS score items have the same meaning for different patient populations. Further research is needed to confirm the factor structure of the DCS in other medical decision contexts and in other cultures. A second possibility is that different cultural construction of experience interacting with different decision‐making tasks may lead to different compartmentalizing of the process within different cultures. This might help to explain why the different dimensions of the decision‐making process cluster differently in different groups. For some, the emotive aspects of the process may be more prominent for others separating the cognitive tasks of decision making from its influences may be imperceptible. Hence, studies using DCS in assessing the effectiveness of decision‐making intervention should first validate the factor structure of the DCS to identify its appropriate scoring method.

The Chinese version of the three‐factor DCS 14 (Ch‐DCS‐14) demonstrated good internal consistency for the overall scale, as well as most of the subscales. While the Support subscale demonstrated weak internal consistency, this is probably due to this subscale having only two items. The Ch‐DCS 14 also showed good convergent validity demonstrated by positive correlation with the measure of perceived difficulties in TDM. Construct validity is good, as indicated by positive correlation with measures of decision regret, anxiety and depression, as well as by negative correlation with the measure of patient satisfaction with medical consultation.

Clinical validity was good. The DCS 14 demonstrated expected patterns for known‐group comparisons in terms of decision status. Our data show that women who delayed in making their decision reported greater levels of decision conflict, compared to non‐delay decision‐makers.

The strength of the present study was the adoption of longitudinal design, which allows assessment of validity from prediction of subsequent outcomes in the form of decision regret and psychological distress. On the other hand, this study is limited to Hong Kong Chinese women with first diagnosis of breast cancer. Future studies should examine the instruments' validity for use among Chinese patients with other types of cancer or medical conditions where deciding medical treatment is involved.

In summary, the present study found that the alternative three‐factor structure for 14 items of the Chinese version of the DCS best fitted the data for Chinese women deciding for breast cancer surgery. Moreover, this three‐factor Ch‐DCS 14 is a valid and practical measure for assessing decisional conflict in making decision for breast cancer surgery. Hence, this instrument shows good potential for use in assessing decision satisfaction for women diagnosed with breast cancer. Lastly, the clinical validity of the DCS is consistently demonstrated across studies, the factorial validity is less clear. Therefore, caution should be taken using DCS to assess patient decision‐making process in other health or cultural contexts.

Sources of funding

This work was supported by a grant #07080651 from Health and Health Services Research Fund, The Government of Hong Kong, Hong Kong SAR.

Conflicts of interest

None.