This study addresses the gap in reporting the association between EQ-5D utility and EQ-VAS score in a primary healthcare setting, which has not been widely reported or discussed elsewhere in Chinese population. From the analysis of the relationship between EQ-5D utility and EQ-VAS score, three key findings were extracted. Firstly, EQ-VAS score is more sensitive than utility in detecting minor variations in age-related health status. Secondly, a consistent and systematic relationship was found between EQ-5D utility and EQ-VAS score. Lastly, EQ-VAS score can provide important information regarding people's health status based on conceptual constructs other than utility.

Our study provides empirical evidence in primary care setting that using EQ-VAS with a single score might be easier for respondents to report their health status compared to using the EQ-5D descriptive system. This finding is consistent with a UK study, where 85% of National Health Service patients completed the EQ-VAS unambiguously [29]. Klimek also noted that the main benefit of using EQ-VAS from a subjective point of view is its ease of use and involved in decision-making process [30]. However, we also found that the EQ-VAS score was highly variable, even for the same health state reported by the EQ-5D descriptive system. These variations between individuals may be due to differences in socioeconomic, physical, or mental factors. Brazier reported that VAS techniques tend to generate meaningless results [4]. Feng, et al. also indicated that caution should be exercised when reporting the results of EQ-VAS, as it may lead to some unusable responses [5]. Nonetheless, in our study, most respondents (90%) clearly defined their health state on the EQ-VAS scale. A significantly positive utility–VAS correlation showed that people understood the meaning of the VAS scale and made reasonable choices based on their understanding. We acknowledge that expectations of HRQoL are closely tied to individuals' subjective experiences. Therefore, we anticipate the development of customized and localized guidelines for the EQ-VAS system, specifically tailored to different subpopulations. This will help mitigate the structural impact on HRQoL evaluation.

We discovered that the equilibrium point between utility and EQ-VAS score is continually shifting towards patients who are experiencing increasing physical or mental discomfort, which supported the argument of Carr et al.’s that experiences constantly shape individual’s health expectations [31]. This finding provides evidence that EQ-VAS and EQ-5D utility score have a dynamic construct for assessing HRQoL. The EQ-VAS seems to be more appropriate and flexible for measuring changes in HRQoL when respondents indicate that they have no health issues, as this is inherent in its design. However, when it comes to utility scores, even if respondents report no problems, a perfect health status is limited to 1.0 [5]. This limitation restricts its ability to accurately reflect changes in health status [32]. Robinson et al. also noted that people's health status is often influenced by contextual factors beyond the health intervention itself [33]. The construct of the EQ-VAS may determine its sensitivity in detecting even minor influences on HRQoL variations.

Previous research on EQ-5D has shown that older individuals are more likely to report a poor HRQoL [27, 34,35,36]. However, few studies have examined how the association between utility and EQ-VAS affects the magnitude of HRQoL variation across age groups. It is important to report and analyze substantial heterogeneity across different age groups before making any HTA decision. In this study, we observed that both utility and EQ-VAS scores decreased with increasing age. However, the EQ-VAS score decreased more rapidly than the utility score. Nevertheless, when we examined the trajectory of people's reported problems with EQ-5D (health state ≠ 11,111), we found some changes in the results. EQ-VAS score decreased more slowly than utility, which suggests that people with health problems tend to be more conservative, intentionally or unintentionally, in using the VAS scale to report their health status. Previous studies have shown culture [37], response shift [38], focusing illusion [39], and adaptation [31] could potentially explain these findings. This is because the EQ-VAS and utility are based on different conceptual frameworks [29]. However, the degree and direction of the effect that these frameworks have on quantifying people's variation in HRQoL in response to changing health conditions are not well understood.

Age, as another underexplored source of variation that consists of heterogeneity in reporting a population's health status, was investigated in this study. The EQ-5D utility and EQ-VAS score reflected considerable variations in self-reported health state across age groups. These findings are inconsistent with those of previous studies that emphasized the relationship between age and HRQoL, where age was a risk factor for harming HRQoL. Only Quintein reported finding as similar as ours that HRQoL exhibited poor physical HRQoL with increasing age but good outcome for social functioning among cancer patients [40]. Currently, cost-effectiveness analysis (CEA) often relies on a method that uses average preferences from a sample population to represent the preferences of the entire society [41]. However, this approach can be risky as factors like age can introduce biases in estimating the benefits of CEA [42]. This study provides empirical evidence that the influence of age on HRQoL must be considered in preference estimation, and the use of age-specific average preferences may be an optimal method for identifying and adopting heterogeneity and reporting a reasonable arrangement of scarce social care resources. However, considering the absence of evidence-based guidelines to justify which EQ-5D tool is more sensitive to detecting age heterogeneity when reporting HRQoL in different cases, our findings illustrate that reporting the results of utility and EQ-VAS score and providing an unambiguous comparison may be the best choice, particularly for the elderly.

This study demonstrated that even individuals who reported a ‘full health’ (utility = 1.0) status using the EQ-5D descriptive system had an extremely sharp decline in EQ-VAS score with increasing age. This result may suggest that EQ-5D utility has limitations in reflecting people’s real health status at the extreme end of the scale. On the other hand, EQ-VAS score appears to be sensitive to capturing the effect of the natural ageing process on HRQoL. Older people, even those with the same health state defined by EQ-5D descriptive system, tend to have poorer overall health status than young people. Cubi-Molla et al. observed that the health of older individuals tends to exhibit great variability, and the interpretation of the same underlying health state should differ depending on the individual's age [43]. Another study in Netherland emphasized the importance of considering the differences in health valuations between younger and older individuals when selecting or establishing outcome measures [44]. Our study suggests that although the effect of age may be small or uncertain, it should not be disregarded when assessing HRQoL burden changes. Therefore, the use of EQ-VAS is encouraged to estimate magnitude and size differences amongst different age groups in HRQoL.

Furthermore, in this study, the EQ-5D profile accurately predicted the EQ-VAS score. We observed that the regression coefficients for most dimension levels of the EQ-5D descriptive system were appropriately ordered. However, the dimensions of 'Usual activities' and 'Self-care' did not perfectly reflect the variation of the EQ-VAS score. This result differs slightly from the findings of a previous UK study, which indicated a significant relationship between the EQ-VAS score and all the EQ-5D-3L profiles [5]. No similar analysis has been reported for the EQ-5D-5L, therefore, we cannot determine whether the updated five-level version of EQ-5D makes EQ-VAS less sensitive to capturing the variation of health status, or if EQ-VAS cannot reflect the health status of HK people on these two dimensions.

Although EQ-VAS is an alternative method, it is not a true utility instrument [4]. Respondents are not required to trade off anything for a health status, and the value is not calculated based on preference. Therefore, we suggest that the EQ-VAS might be useful in the clinical practice for comparing the effectiveness of clinical interventions. However, for political decision-making, utility may be more appropriate for conducting HTA to allocate health and social resources [45]. The debate over which technique or variant is more appropriate for measuring and valuing health continues [4]. Current research is the first and an important step in exploring the meaning and mechanism of the EQ-VAS for populations with different socioeconomic characteristics. Researchers should exercise caution when reporting EQ-5D results, as the disparity between the utility and EQ-VAS data could potentially suggest methodological issues in data collection or analysis.

Although the discussion of this study mainly focuses on the methodology, the findings also have practical implications. Our findings support that EQ-VAS is a valid and simple instrument for rapidly assessing an individual's perceived health status. It can be widely used in primary care settings to assess the effectiveness of clinical interventions. Its simplicity and flexibility can be valuable in supporting doctor-patient communication and improving patient-centered care in HK. The performance of EQ-VAS in different patient groups should be further explored.

Several limitations of this study should be addressed. The primary limitation is that all participants were recruited from one GOPC, Although the difference of most key background characteristics between our sample and general population was statistically insignificant, selection bias might be existed. Given potential variations in patients' demographics, comorbidities, laboratory data, and medication usage across GOPCs, there may be a limited understanding of the Hong Kong population and insufficient support for robust statistical analysis. Future studies should consider collecting data from different local healthcare services. multiple GOPCs. By including a diverse range of GOPCs, we can obtain a more representative sample and enhance the overall validity and reliability of our results. Additionally, this was a cross-sectional design, thus, no causal relationship can be developed. Third, all participants in this study were primary care patients. Compared to the general population, our sample was slightly older, the comparisons between EQ-5D-5L utility score and EQ-VAS in local general population are unknown and need further analysis. Last, in this study, due to self-reported, no clinical data was collected by participants. The comparisons between EQ-5D utility score and EQ-VAS regarding different clinical conditions are not presented.