3.1 Characteristics of the sample
A total of 498 patients were recruited at 15 centres in 13 countries representing 5 cultural areas. The mean age of these patients was 62 years. Of these patients, 54.9% were females. Various education levels, cancer sites and oncology treatment modalities were represented. Details of the patients are shown in Table 2 and supplementary table 2.
A total of 346 patients (69.5%) completed the QLQ-COMU-26 at the second assessment (201 of these patients assessed doctors while 145 assessed nurses). Reasons for non-completion were: declined (12 patients; 2.4%), deceased (28 patients; 5.6%), and administrative failure (46 patients; 9.2%). In some cases (66 patients; 13.2%), the reasons were not recorded. A total of 219 patients (44%) performed the re-test assessment (120 of these patients assessed doctors while 99 assessed nurses). All questionnaires had over 70% of their items answered at each assessment.
The total number of missing items from all the QLQ-COMU-26 questionnaires gathered at the first assessment was 86 (0.7%): the number of patients who failed to complete these missing items ranged from 0 to 13 (2.7%) per item.
3.2 Debriefing questionnaire on patient acceptability
Most patients (78%) completed the QLQ-COMU-26 in 15 minutes or less; 43.1% of patients were given help when doing so, with 26.9% being given practical help.
A total of 70 patients (14.3%), from 10 centres, found at least one item confusing. The highest frequencies were for item 17 (the professional listened when the patient expressed emotions), which was considered confusing by 15 patients (3.0% of the sample), and item 18 (help with managing emotions), which was confusing for 14 patients (2.8%). For both items, the patients were distributed among the various centres.
A total of 14 patients (2.9%) found at least one item upsetting. Three patients (0.6%) found item 16 (the professional tried to understand the patient’s situation) upsetting, while three patients found item 17 (the professional listened when the patient expressed emotions) upsetting. The other items were mentioned by fewer patients. See supplementary table 3.
3.3 Scale structure
Goodness of fit measures in confirmatory factor analysis confirmed the hypothesised scale structure of the questionnaire (CFI=1.00, TLI=1.00, RMSEA=0.025). All factor loadings were above 0.82 and all residual correlations but one were below 0.22. See Figure 2 for details on the investigated model. Scale structure was also supported by the results from multi-trait scaling analysis, which showed that all items had an item-own-scale correlation above 0.40 (corrected for overlap) and that all items but four had the highest correlations with their hypothesised scale (Table 3).
3.4 QLQ-COMU-26 descriptive statistics and reliability
Mean scores and standard deviations of the scales and individual items at first assessment are shown in Table 1.
The percentages for floor effects were low: item 19 showed the highest floor effect (11%). All scales except two had >50% of patients at ceiling (highest ceiling effect= 69.4%).
Cronbach’s alpha coefficients for all scales were between 0.85 and 0.91, while test-retest reliability was between 0.86 and 0.92 (see Table 1).
3.5 Convergent and divergent validity
Supplementary table 4 shows the correlations between the EORTC QLQ-COMU26 areas and the selected doctors’ items of the EORTC INPATSAT32 that evaluated convergent validity.
Correlations ranged from 0.50 to 0.70. The hypothesised relationships between the selected doctors’ items of the EORTC IN-PATSAT32 and the scales and items of the QLQ-COMU26 that were expected to be more conceptually related showed correlation coefficients of >0.60 for convergent validity. The correlations between QLQ-C30 and QLQ-COMU26 areas were <0.30 in all cases, which confirms the QLQ-COMU26 divergent validity (see supplementary table 1).
3.6 Known-groups validity
The results of known-groups comparisons are shown in table 4. Communication scores in all communication areas were higher in patients with higher levels of emotional functioning in the EORTCQLQ-C30 scales and higher satisfaction with communication in item 26 of EORTC QLQ-COMU26. No significant differences in communication scores were found between age-based groups.
Males showed higher communication scores than females in seven QLQ-COMU26 areas. Patients with lower education levels showed higher communication scores than those with higher education levels (nine QLQ-COMU26 areas). Better management of the patient’s emotions by professionals and greater satisfaction with communication were shown in patients who received treatment with palliative intention. Communication with nurses showed higher scores than communication with doctors in nine QLQ-COMU26 areas. Sometimes having a companion during the visit showed higher communication scores than having no companion in five QLQ-COMU26 areas, while sometimes having a companion showed higher sores than always having a companion in three QLQ-COMU26 areas.
3.7 Responsiveness to change
Mean changes between the two assessment points for the whole sample showed significant differences in three areas ranging from 2.5 to 3 points.
For a more in-depth analysis, three groups (from the whole sample) were created based on changes (or no changes) in the score of item 26 (satisfaction with communication improved, remained stable or worsened). These groups were studied independently. Patients whose score in satisfaction item 26 worsened showed significant worsening in eight QLQ-COMU26 areas (between 16.1 and 25.4 points). Patients whose score remained stable showed just two significant differences towards worsening (between 2.4 and 2.9 points). Patients whose score in satisfaction item 26 improved showed significant improvement in nine QLQ-COMU26 areas (between 14.7 and 24.3 points) (see Table 5).