Acceptability
Of the 247 participants at Time 1 (baseline), 87.9% (217 persons) answered all 14 questions of the GP-CORE, 8.5% (21 persons) did not responded to one question, 2.8% (7 persons) to two questions, 0.4% (1 person) to four questions and 0.4% (1 person) to five questions. The non-response rates for the fourteen questions are as follows: Q1 (0.4%, 1 person); Q2 (0.8%, 2 persons); Q3 (1.6%, 4 persons); Q4 (5.7%, 14 persons); Q5 (0%, 0 persons); Q6 (1.2%, 3 persons); Q7 (0%, 0 persons); Q8 (1.2%, 3 subjects); Q9 (0.4%, 1 subject); Q10 (0.4%, 1 subject); Q11 (0%, 0 subjects); Q12 (1.2%, 3 persons); Q13 (2.4%, 6 persons); Q14 (2.4%, 6 persons). All participants answered the following; “I have been troubled by aches, pains or other physical symptoms.` (Q5), “I have had difficulty getting to sleep or staying asleep.` (Q7), and “I have felt unhappy.` (Q11).
A univariate statistical description
Table 3 shows that for the main group, the average value on GP-CORE was 1.36, with a standard deviation of 0.61. The differences between men and women was not significant (t 245 = -0.98, p > .05, Cohen´s d = 0.12). The positive values for the skewness statistic implies that the distributions are positively skewed but the skewness statistic divided by its SE was less than two which indicates that the scale did not significantly deviate from a symmetric distribution (see SPSS) [21,22]. Similarly, the kurtosis statistic implies that the distributions deviated from normality, but the kurtosis statistic divided by its SE was greater than two only for women, which indicates that the scale was significantly platykurtic only for women (see SPSS) [21,22].
Reliability
Internal consistency
The Cronbach alpha coefficient for the total group was 0.79 (95% CI: 0.74 to 0.83). For the women, the alpha was 0.80 (95% CI: 0.75 to 0.84). For the men, the alpha was 0.74 (95% CI: 0.62 to 0.83).
Test-retest stability
Intra-class correlation coefficient (ICC) – mixed model, absolute type, average measures – was calculated for the test-retest group of 29 participants and gave a value of 0.88 (95% CI: 0.72 to 0.94).
Validity Table 4 shows the Spearman correlations between the GP-CORE and the other variables of the present study. The variables diagnosed illness and medicate regularly were excluded from these analyses because these two variables had very highly skewed distributions (see Table 1). For the total group, the correlations between the GP-CORE and the four background variables (gender, age, education, and cohabitant) were all not significant. However, women with a cohabitant had higher values on GP-CORE than women without a cohabitant (r = 0.17, p < .05). For men, the relation was the opposite but not significant (r = -0.18, p > .05). The Fisher´s z-test indicated a significant gender difference (z = 2.37, p < .05) and was followed up by a 2 (gender) x 2 (cohabitant) ANOVA for independent measures with GP-CORE as the dependent variable. The ANOVA indicated a significant antagonistic interaction (F 1, 243 = 5.48, p < .05, η Partial 2 = 0.02) and showed that women with a cohabitant had lower well-being than men with a cohabitant but women without a cohabitant had a higher well-being than men without a cohabitant.
The correlations between GP-CORE and the six health-related quality of life questions were all in the expected direction and all, except one, significant. No gender differences regarding correlation between GP-CORE and the six health-related quality of life questions were observed. In summary, higher values on GP-CORE (lower psychological well-being) was associated with greater problems with mobility, greater problems connected with self-care, greater problems performing usual activities (e.g. housework, family, leisure activities), greater levels of pain/discomfort, greater levels of anxiety/depression and lower self-rated health.
Cut-off value and sensitivity/specificity
The participants were divided based on their answers to the general health question in EQ-5D-5L. The median value of the answers for this question was calculated to be 50 points. Those respondents who rated below the median (0-49 points) were defined as having bad health (n=84) and those who rated equal to or above the median (50-100 points) were defined as having good health (n=163). Participants who were classified as having bad health had M=1.60 (SD=0.56) and participants who were classified as having good health had M=1.24 (SD=0.59) on the GP-CORE at Time 1 (t 245 = 4.68, p < 0.001, Cohens d = 0.63). To calculate the cut-off value on the GP-CORE that could be used to separate those with bad health from those with good health, the following formula was used (Jacobson and Traux, 1991):
The cut-off value was calculated to 1.43. Using this cut-off value, the sensitivity and specificity for the GP-CORE at Time 1 was estimated. The sensitivity (the percentage of those participants that rated their health as bad (0-49 points) falling over the cut-off value) was 59.5%. The specificity (the percentage of the participants that rated their health as good (50-100 points) falling on or below the cut-off value) was 62.6%.
Average changes
To investigate sensitivity to change for the GP-CORE, the means and standard deviations on GP-CORE for men and women at the three measurement occasions are shown in Table 5. A 2 (Gender) x 3 (Time) ANCOVA with cohabitant as the covariate and GP-CORE as the dependent variable was performed. The analyses showed significant time differences, F 2, 398 = 3.08, p < .05, ηPartial2 = 0.02, where pairwise comparisons (Bonferroni post-hoc tests) showed significant differences between Time 1 and Time 2 (M Time 1 – M Time 2 = 0.39, p < .001, Cohen´s d = 0.60), significant differences between Time 1 and Time 3 (M Time 1 – M Time 3 = 0.40, p < .001, Cohen´s d = 0.59) but no significant difference between Time 2 and Time 3 (M Time 1 – M Time 2 = 0.01, p > .05, Cohen´s d = 0.02). No significant gender differences were observed, F 1, 199 = 3.86, p > .05, ηPartial2 = 0.02. No significant interaction between gender and time was observed, F 2, 398 = 0.859, p > .05, ηPartial2 = 0.01. In summary, the results showed that at Time 2 well-being (as measured by GP-CORE) improved on average by 0.39 points relative to Time 1, with an associated Cohen´s d = 0.60 which can be interpreted as a medium effect size (Cohen, 1988), and at Time 3 well-being was still on average 0.40 points better than at Time 1, with an associated Cohen´s d = 0.59 which can be interpreted as medium effect sized [23] but that there were no significant differences in well-being between Time 2 and Time 3.
Individual changes
To further investigate the sensitivity to change of GP-CORE, it was analysed for every participant if the participant´s change between two measurement occasions is beyond what could be attributed to measurement error or in other words if it is a reliable change. The formula to calculate reliable change (see [24]) is as follows:
Changes that fall outside the interval of formula 1 are unlikely to occur more than 5% of the time by unreliability of the measurement alone and are considered as reliable changes. Using formula 1, it was analysed how many participants had reliable changes from Time 1 to Time 2, from Time 1 to Time 3, and from Time 2 to Time 3. The results are presented in Table 6. As can be seen in Table 6: after three months of intervention or homecare (at Time 2), 52 of the 227 participants (22.9%) perceived that their well-being (as measured by GP-CORE) had improved; after six months of intervention or homecare (at Time 3), 45 of the 202 participants (22.3%) perceived that their well-being had improved, and finally; the improvement between Time 2 and Time 3 is negligible, where only 7 of 202 participants (3.5%) perceived to have improved. In summary, 22.9% of the participants had better well-being after three months of intervention or homecare compared to baseline, and 22.3% of the participants had better well-being after six months of intervention or homecare compared to baseline, but the improvement plateaued and was negligible between Time 2 and Time 3.