Validation of Quality of Life Instruments for Cancer Patients – Colorectal Cancer (QLICP-CR) in patients with colorectal cancer in Northeast China

Liaoning Cancer Hospital & Institute is a provincial-level tertiary hospital (2330 beds) for cancer prevention, treatment, scientific research and teaching. It is the main hospital treating CRC in the north of China, because it hosts the Provincial Key Laboratory and Provincial Translational Medicine Center for CRC. Its diagnosis and treatment of CRC is advanced and high specification. In total, 98.55% of inpatients are from northeast China, suggesting that a study population sampled from this hospital could represent the CRC population in the region.

The inclusion criteria were (1) CRC patients who were hospitalized and received treatment; (2) patients with primary (not secondary) CRC, which had been pathologically diagnosed. The exclusion criteria were (1) patients who were under 18 years old; (2) patients with a mental illness or cognitive impairment; and (3) patients who also had another malignant tumor. Screening for mental illness or cognitive impairment used past history of the disease and the judgment of the doctor.

This study was conducted from November 2016 to January 2017. After obtaining CRC patients’ written consent to the survey, QLICP-CR (V1.0) and the Chinese version of FACT-C (V4.0) questionnaires were distributed to all the patients, who were surveyed three times. The surveys were performed on admission, 2–3 days later, and at discharge.

QLICP-CR (46 items) contains two modules, the Quality of Life Instruments for Cancer Patients – General Module (QLICP-GM) (32 items) and a specific domain (14 items, SCR1–SCR14). QLICP-GM is divided into four domains, physical (seven items, GPH1–GPH7), psychological (12 items, GPS1–GPS12), social (six items, GSO1–GSO6), and common symptoms and side-effects (seven items, GSS1–GSS7). Each item is rated on a five-point Likert-type scale (from “not at all” at 1, through “a little bit”, “somewhat”, and “quite a bit”, to “very much” at 5). Items that are positively stated are scored from 1 to 5 points, and negatively stated items from 5 to 1. The scoring guidelines for QLICP-CR state that the raw score for each domain is derived by summing the individual item scores, and the score for the total scale is the sum of the scores for the five domains, with higher scores indicating higher QOL. For comparison, raw scores for all domains were converted into standard scores by using the extreme difference method [11].

The Chinese version of FACT-C was used alongside QLICP-CR to assess the criterion-related validity. FACT-C (36 items) contains the Functional Assessment of Cancer Therapy – Generic Scale (FACT-G, 27 items) and a colorectal cancer subscale (CCS, nine items). FACT-G is divided into four domains, covering physical well-being (PWB, seven items), social/family well-being (SWB, seven items), emotional well-being (EWB, six items) and functional well-being (FWB, seven items). Each item is rated on a five-point Likert-type scale (i.e. “Not at all”, “A little bit”, “Somewhat”, “Quite a bit”, and “Very much”) with scores from 0 to 4. Again, items that are positively stated are scored from 0 to 4 points, and negatively stated items from 4 to 0. The scoring guidelines for FACT-C were used to calculate the score of the items, domains, FACT-G and FACT-C total scale [18, 33].

Demographic characteristics included age, sex, marital status and education. Only one patient (0.66%) was unmarried, three (1.97%) were divorced, five (3.29%) were widowed, and one was separated. They were therefore combined into a single “other” group.

Reliability was evaluated by internal consistency and test–retest reliability. Internal consistency was evaluated using Cronbach’s α coefficient. An α value of at least 0.7 was considered acceptable [34‐36]. Test–retest reliability was assessed using paired Student’s tests, Pearson correlation analysis and intra-class correlation (ICC) analysis to compare the difference between the first and second measurements. ICC with 95% confidence interval (CI) was calculated on the basis of absolute agreement with a single measure under the two-way mixed model. A Pearson correlation coefficient greater than 0.8 was considered to provide good reliability [37]. An ICC value of at least 0.7 was considered acceptable [38].

Validity was evaluated using Pearson correlation analysis and exploratory factor analysis (EFA). Pearson correlation analysis was used to calculate the item–domain correlation of QLICP-CR and correlation of domain scores between QLICP-CR and FACT-C. Convergent validity was defined as an item–domain correlation of 0.40 or higher. Discriminant validity was defined as an item having a higher correlation with its own domain than with other domains [21, 34]. Correlation coefficients of domain scores between QLICP-CR and FACT-C were calculated to evaluate the criterion-related validity. EFA with the principal component method and varimax rotation was used to assess QLICP-CR structure [11]. Bartlett’s test of sphericity and the Kaiser-Mayer-Olkin (KMO) test were used. An initial eigenvalue above 1 was set as the criterion for factor extractions. Factor loadings above 0.50 were considered indicative of item loading [39].

Responsiveness was evaluated using paired Student’s t tests and calculating the standardized response mean (SRM) to compare the difference between the first and third measurements. SRMs of 0.2, 0.5 and 0.8 were taken as poor, moderate and good responsiveness [40].

The study population included 152 patients with mean age of 57.5 ± 12.24 years. In total, 79 patients (52.0%) were male, and 73 (48.0%) were female, with 142 patients (93.4%) married or cohabitating. A total of 17 (11.2%) patients were illiterate or had completed primary school, 76 (50.0%) had completed junior middle school, 30 (19.7%) had completed senior middle school, and 29 (19.1%) had completed junior college/university or above.

Internal consistency and test–retest reliability of QLICP-CR are shown in Table 1. Cronbach’s α coefficients for QLICP-GM and each domain of QLICP-CR ranged from 0.62 to 0.93. In total, 150 patients completed the second survey and these data were used for the test–retest reliability analysis. Test–retest reliability coefficients (r) for the five domains were between 0.74 and 0.91, with QLICP-GM and the total scale being 0.90 and 0.87. The ICC for all domains ranged from 0.74 to 0.90, with QLICP-GM and the total scale being 0.90 and 0.86. There were no significant differences between the first and second measurements for the scores of the five domains, the total score for QLICP-CR and the score for QLICP-GM (P > 0.05).

The item–domain correlations for QLICP-CR are shown in Table 2. Pearson correlation analysis showed that correlation coefficients between items and their own domains ranged from 0.19 to 0.90. The correlation coefficients of items with their own domains were all greater than the coefficients with other domains.

Correlation coefficients for domain scores between FACT-C and QLICP-CR are shown in Table 3. FACT-C was used to evaluate the criterion-related validity. The Pearson correlation coefficients between the specific domain of QLICP-CR and the colorectal cancer subscale of FACT-C, QLICP-GM and FACT-G, and the two overall scores were 0.37, 0.77 and 0.70. Generally, the correlations between similar domains were higher than between different domains. For example, the highest correlation coefficient for the psychological domain of QLICP-CR was with the emotional well-being domain of FACT-C, at 0.75.

Principal components and the factor loadings of QLICP-GM are shown in Table 4. A scree plot was also provided, and the results are shown in Fig. 1. Based on Bartlett’s test of sphericity (χ² = 2633, P < 0.001) and the value of the Kaiser-Meyer-Olkin test (0.84), we used EFA of QLICP-GM with varimax rotation. Using the initial eigenvalues > 1, we extracted nine principal components. Using factor loadings above 0.50, the first principal component corresponded to most items of the psychological domain (nine items, 0.62–0.89), the second to most items of the physical domain (four items, 0.53–0.86), the third with most items of the social domain (four items, 0.55–0.81), and fourth with two items from the common symptoms and side-effects domain (0.68–0.79). The fifth principal component had only one item (0.84), from the physical domain, the sixth had three (0.56–0.82) from the common symptoms and side-effects domain, the seventh had two (0.63–0.80) from the social domain, the eighth had two (0.53–0.66) from the psychological domain, and the ninth had one item (0.75) from the common symptoms and side-effects domain. Overall, these principal components accounted for 70.21% of the cumulative variance and reflected four domains of QLICP-GM. A four-factor solution (breaking at the fourth factor) could be seen on the scree-plot diagram. The variance contribution rate of each factor gradually became smaller after the fourth factor. Three principal components were extracted from the 14 items in the specific domain of QLICP-CR, accounting for 72.26% of the cumulative variance.

Responsiveness of QLICP-CR was assessed by calculating the difference between the first and third measurements, shown in Table 5. In total, 145 patients completed the third survey, and the data were used to evaluate responsiveness. There were significant differences in all domains, QLICP-GM and the total scale (P < 0.05) between the first and third measurements. SRM values ranged from 0.60 to 0.81 for the common symptoms and side-effects domain, specific domain, QLICP-GM and the total scale, and from 0.30 to 0.48 for the physical, psychological and social domains.