Cancer is the leading cause of death in men and women worldwide. Cancer incidence was at 224,177 with the five-year survival rate reported 69.4% in 2013 in Korea [1], while in the United States, it was 69.0% in 2011 [2]. With cancer patient survival rates increasing, national policy and clinical researchers have begun to focus on the management of psychological distress and quality of life of cancer patients [3].

More than 30.0% of cancer survivors have a negative attitude toward cancer, and the cancer patients who experienced a stigma toward cancer showed more than 2.5 times higher rates of depression compared to cancer patients with positive attitude toward cancer without a stigma [4]. Stigma was linked to guilt and shame, self-blame, blame attribution, and depression, and in cancer research, these concepts were associated with one another [5, 6]. Stigma may have properties contrary to social identity and, therefore, those with cancer may feel undervalued or tagged, and may experience negative stereotyping and discrimination, which together lead to social rejection, social isolation, lack of social support, and low social status [7]. In particular, health-related stigma (HRS) represents experiencing rejection, blame, or devaluation due to his or her illness [8, 9]. HRS has been associated with illness-induced stress and contributing to psychological, physical, and social morbidity [9, 10]; such that a negative psychosocial impact results in depression, receiving limited social support, decreased treatment adherence, and adverse treatment effects [11]. Because HRS can be an obstacle to seeking professional help and health promotion activities [8], it is important for both clinical research and practice professionals to use a valid and reliable tool to periodically evaluate cancer patients for HRS and effectively manage HRS.

As cancer can be a fatal disease, it is very important to find interventions to improve psychosocial and physical health for cancer survivors to help them avoid long-term ill-effects and possible death. There are increasing empirical studies on how stigma affects cancer survivors [12]. Although researchers found the fact that cancer stigma affects individuals with cancer, most of the studies focus on patients with lung cancer [8, 11, 13, 14]. Van Brakel [9] reported that many instruments have been developed to assess the intensity and qualities of stigma, but these are often condition-specific including cancer type-specific. Stigma has an effect on individuals and their families, as well as on the effectiveness of health programs. The similarity in the consequences of stigma in many different cancer's types [12] suggests that the development of generic instruments to assess HRS of all types of cancer patients may be possible [9].

Through a literature review of 63 research papers that addressed the issues related to measuring stigma or stigma-related constructs, and that contained a sample of the instrument or items used, aspects of HRS can be grouped into five categories. First, the experience of actual discrimination and/or participation restrictions on the part of the person affected; second, attitudes towards the people affected; third, perceived or felt stigma; fourth, self or internalized stigma; and fifth, discriminatory and stigmatizing practices in health services, legislation, media and educational materials [9]. While Cataldo Lung Cancer Stigma Scale (CLCSS) developed by Cataldo et al. [13] is a multidimensional measurement tool that can measure all aspects of HRS, it was developed to measure the perceived stigma of lung cancer patients. It consists of four subscales, stigma and shame domain, social isolation domain, discrimination domain, and smoking status domain and it has been reported to have a satisfactory level of validity and high reliability at the time of development. In a study of cancer-related stigma by Else-Quest et al. [12], it was reported that cancer patients including patients with lung cancer have internal causal attributions where they believe cancer occurred due to their harmful lifestyle habits. So, most cancers can conjure a similar attribution of blame as is found with lung cancer because most cancers are frequently associated with individual's behaviors. Because of the lack of a valid and reliable tool, empirical evidence of cancer-related stigma is limited. Thus, to develop a generic cancer stigma scale, the existing instruments should be further validated, developed, or adapted for generic use, where possible.

In particular, the measurement tools for cancer stigma were as follow: (a) a single item for perceived stigma, "People judge me for my type of cancer," was rated 1 (strongly disagree) to 5 (strongly agree) [12]; and (b) CLCSS [13]. The CLCSS has been used in Korea and China [8, 15]. The CLCSS was adapted to 123 lung cancer patients without testing the validity in the Korean patient population [8]. Therefore, we modified the CLCSS to form the Korean Cancer Stigma Scale (KCSS) and conducted a study to test the validity of the KCSS in patients with one of the five major types of cancer in Korea (breast, colon, lung, stomach, and uterine cervix cancer). This paper is to report the validity and reliability of the KCSS.

KCSS was developed to measure the cancer-related stigma that can be experienced in the illness trajectory process of cancer patients. The findings in this study provide a primary basis for the reliability and validity that the tool, through verifications of Korean version KCSS such as content validity, construct validity, item convergent and discriminant validity, concurrent validity, known-group validity, and reliability. As a result, the six factors that emerged in this analysis were reflected in the six subscales: social isolation, distancing or avoiding, discrimination, guilt, attribution, and lack of medical support.

The reliability of four subscales of the original version of the CLCSS for lung cancer as reported was that the Cronbach's alpha coefficient was .96 for 31-item instrument and ranged from .75 to .97 [13]. In this study, the Cronbach's alpha coefficient was .89 for the overall instrument and ranged from .62 to .86 for each subscale. Thus, the reliability of the measured variable satisfied the standard of ≥.70 [29] except for attribution (.64) and lack of medical support (.62) subscales. These results were similar to those of Chinese version of the CLCSS [15] which reported Cronbach's alpha coefficients between .60 and .88 for lung cancer patients. As Cronbach's alpha coefficients for the overall instrument were shown to be relatively high and above the standard of .70 [29], this result ensures the reliability of the instrument. However, the attribution subscale and the lack of medical support subscale among the KCSS sub-factors should be reaffirmed through a test-retest reliability verification study in the future.

In the two studies using CLCSS that were intended for lung cancer patients, the same four factors were extracted [13, 15]. In the original measure for lung cancer, the four factors were stigma and shame, social isolation, discrimination, and smoking status by EFA, explaining 57.0% of the total variance; the final version comprised 31 items [13]. The findings of Yang et al. [15] study with Chinese lung cancer patients were the same as the results reported by Cataldo et al. [13] and the amount of total variance explained was 58.6% from the four factors extracted by EFA. In this study, the total variance explained was 65.7% so that it was higher than the total variance of the original CLCSS and Chinese version CLCSS. Therefore, KCSS is determined as a very useful measure of stigma for Korean cancer patients. However, as the KCSS was made with modified terms of the items to apply it to five types of cancer patients whereas the CLCSS targets lung cancer patients only, there may be partly limitation of validity.

The item distribution by subscale of 24-item KCSS and item distribution of original 31-item CLCSS are to be reviewed. Guilt subscale and lack of medical support subscale items of KCSS were stigma and shame subscale in the original CLCSS, social isolation subscale and distancing or avoiding subscale items of KCSS were social isolation subscale in the original CLCSS, discrimination subscale items of KCSS were discrimination subscale in the original CLCSS and attribution subscale items of KCSS were smoking status subscales in the original CLCSS so that the similar item distributions were seen. That is, the original CLCSS has four subscales and the KCSS has six subscales so that the subscale might be further subdivided and the item distribution was similar with item distribution by subscales in original CLCSS. However, for the lack of medical support which is the sixth subscale of KCSS, item 9 was located in the smoking status of the original CLCSS and item 30 and 8 were located in the stigma and shame of the original CLCSS. This means that the factor structure can be changed for each survey that varies by target group and sample number and especially, as "lung cancer" is changed to 'cancer' for all sentences in the study, it could be because "smoking status" subscale of original CLCSS disappeared and several wordings were modified to use with the five types of cancer.

For this study construct validity, item convergent and discriminant validity, concurrent validity, and known-group validity were used as methods of testing validity. To test construct validity, χ², Q (χ²/df), GFI, AGFI, NFI, TLI, RMR, and RMSEA were calculated through CFA. The χ² assesses whether the actual data corresponds to the model in CFA. However, when the sample size increases, it generally becomes significant. Thus, many other suitability indicators in addition to χ² were examined. Q, RMR and RMSEA met the recommended level among the fitness index, but the GFI, AGFI, NFI, and TLI didn't meet .90, the recommended level of the best-fit model in this study. It is not desirable in the evaluation of CFA result model for construct validity to judge the model by relying only on a single fit index and several fit indexes should be considered at the same time [21]. Therefore, in this study, a measurement model consisting of 24 items and 6 subscales was accepted in the end. However, there was the result seen that the validity of the CFA result model for construct validity did not meet some criteria in this study. Therefore, in order to establish stability of construct validity, re-test should be done through follow-up study.

In this study, the results of verifying item convergent and discriminant validity by multi-trait multi-items matrix analysis have showed 100% convergent validity and 97.5% discriminant validity of KCSS items. Items 29 and 8 were within the critical value of other subscales' items. This may mean that items 29 and 8 could be affected or confounded by other subscales. However, Cronbach's alpha coefficient of each subscale was higher than the correlation coefficients among the other subscales in this study, so that each subscale properties were discriminated [28]. Therefore, it was shown that the measurement items consistently measure construct concept and independent among subscales was maintained.

In order to verify the concurrent validity, this study utilized the QLQ-C30 questionnaire to assess the 'cancer-related quality of life" variables. In this study, significant correlations were found between the 24-item KCSS and subscales of the QLQ-C30. However, the range of correlation coefficient for the functional subscale of the QLQ-C30 and for 24-item KCSS was -.19, falling short of the recommended range for correlation coefficients (r=.40 to r=.80) [30] to establish concurrent validity. Such a low correlation is presumed to be because the participants of this study were those who were in aggressive treatment and about 69% of them are currently on anticancer chemotherapy so that it seems that there could be some limit for the stigma caused by cancer to influence the functional level of the participant compared to the global health status and treatment-related symptoms.

In order to test the known-group comparisons, we used two methods: comparison methods of the KCSS scores between (1) five types of cancer sites, and (2) the distress and non-distress groups of the PSI scores. The KCSS score of patients with colorectal cancer patients was the highest; lung, cervix, breast, and gastric cancer followed in order. However, there were statistically no differences from the KCSS scores between the five types of cancer patients. Therefore, the 24-item KCSS could potentially be applied to Korean cancer patients. As HRS that the cancer patients experience in the process of cancer treatment and its recovery means rejection, blame, or devaluation due to his or her illness [9, 10], HRS related to the illness may increase the degree of psychological distress in cancer patients [5], it is considered that psychological distress is proper research variable for the known-group validity verification of KCSS. Hence, a significant difference in the stigma scores of the groups with different psychological distress levels establishes the known-group validity of the KCSS. This research analyzed the differences in the KCSS mean scores for each group according to the distress and non-distress groups of PSI scores. In this study, the KCSS score had a statistically significant difference between the distress and non-distress group for the PSI scores; the distress groups of insomnia, anxiety, and depression domains had a significantly higher perceived cancer stigma than the non-distress groups. Thus, the known-group validity of the KCSS instrument was established.

However, the limitation of this research is that it did not test responsiveness, which evaluates the change in degree of patientreported cancer stigma over time. Therefore, there is a need to test the responsiveness through a future longitudinal study. And also, as this study investigated cancer patients in only one city, it is suggested that further studies should be conducted on cancer patients in various regions in order to generalize KCSS in clinical practice. Lastly, as KCSS is a tool to be applied to the five types of cancer patients, we suggest the necessity of cancer-specific or sensitive stigma tool development in the future.

Cancer stigma is a very important psychological concept that cancer patients can experience during diagnosis and treatment and it affects patients' daily lives so that it is very important to objectively evaluate the cancer stigma perceived by patients in the clinical fields. Therefore, this study was conducted to test the reliability and validity of KCSS. The study result has shown that the 24-item KCSS has a relatively acceptable reliability and validity. However, in the CFA for construct validity, the fit of the model did not meet some criteria and there were two subscales (attribution subscale and lack of medical support subscale) which had relatively low Cronbach's alphas. Therefore, if these are complemented by further studies, this tool will provide a useful instrument in clinical trials investigating stigma, as well as its impact on the quality of life and psychosocial distress in Korean cancer patients.