In total, 500 individuals completed the online survey, including 140 healthy individuals, 122 individuals with hepatitis B, 107 with depression, 90 individuals with GAD and 101 with HIV/AIDS. Some respondents reported multiple conditions, e.g. 68 individuals reported both depression and GAD. In general, the whole study sample was young (mean age: 35.8, SD: 8.64) and well educated. The gender proportions of the five groups were generally balanced except for the group of HIV/AIDS, in which, about 87.1% of individuals were female. In terms of the age distribution, the healthy group was mostly young; the chronic hepatitis B group has more participants aged between 40 and 49; the depression and GAD groups had individuals from all four age groups and the HIV/AIDS group aged mainly from 30 to 49. Individuals with tertiary education accounted for over 80% for all four disease groups and the healthy group had more individuals with secondary education. Table 1 shows the demographic information by condition [see Additional file 1].
There were 13, 34, 28, 26 and 26 unique states reported for the healthy, chronic hepatitis B, depression, GAD and HIV/AIDS group respectively for the 3L. The corresponding numbers were 18, 43, 46, 42 and 35 for the 5L.
It is evident from table 2 that the ceiling effects of the 5L was smaller than the 3L for all dimensions and for all condition groups [see Additional file 2]. When all dimensions are considered, the number of health profiles 11111 decreased by almost 40% when reporting using the 5L. The reduction of ceiling effect is more salient in the pain/discomfort and anxiety/depression dimensions. The most prominent difference in ceiling effects between 3L and 5L was observed in the HIV/AIDS group, i.e. all dimensions had a relative reduction of over 30%.
Table 3 shows the median responses to 3L and 5L dimensions for the 5 groups [see Additional file 3]. When measured by the 3L, the median responses of healthy and HIV/AIDS groups were ‘no problems’ across the five dimensions. In comparison, when measured by the 5L, while the median responses for the healthy group remained to be ‘no problems’, the median responses of HIV/AIDS group were all ‘slight problems’. Similarly, the median responses for the chronic hepatitis B group were ‘slight problems’ for the last three dimensions when measured by the 5L, while the only pain/discomfort dimension had ‘moderate problem’ measured by the 3L. The median responses for the GAD and depression group were similar. The Mann-Whitney results were all significant at 0.01 level suggesting all 4 condition groups had a different distribution of responses against the healthy group, for both 3L and 5L.
For the 3L, the healthy group had a mean utility of 0.948 (SD: 0.104), followed by HIV/AIDS of 0.791 (SD: 0.116), chronic hepatitis B of 0.781 (SD: 0.166), depression of 0.721 (SD: 0.180) and GAD of 0.711 (SD: 0.178). In comparison, the mean utilities of the 5L were 0.947 (SD: 0.080) for the healthy group, 0.785 (SD: 0.180) for the chronic hepatitis B group, 0.779 (SD: 0.115) for the HIV/AIDS group, 0.746 (SD: 0.202) for the depression group and 0.718 (SD: 0.217) for the GAD group. As shown in Figure 1, the 5L index value had a smoother distributions, less clustering and wider range (-0.305 to 1) compared with the 3L (ranged from -0.03 to 1). An exception is the 5L utility score clustered at 0.734 for the HIV/AIDS group. This is the utility score of health state 22222, which had been reported most frequently in the HIV/AIDS group.
Table 4 shows the mean index values of each condition groups and shows the relative efficiency of the 3L and 5L between groups [see Additional file 4]. The index values of the 5L ranged from 0.711 of the GAD group to 0.948 of the healthy group. In comparison, the index values of the 3L ranged from 0.718 of the GAD group to 0.947 of the healthy group. It is clear that two versions of EQ-5D produced comparable index values for each sub-group and both versions demonstrated good known-group validity. Compared with the 5L index, the 3L index did not show a difference of statistically important in the comparison of the healthy and HIV/AIDS groups. The relative efficiency of the 5L index was higher in the comparison of the healthy group with the condition groups.