Population-level norm values by EQ-5D-3L in Hungary - a comparison of survey results from 2022 with those from 2000

Over the past 22 years the HRQoL of women aged 25–74 and men aged 35–64 improved considerably. Interestingly, our study could not replicate the relatively high index values observed in the 2000 national health survey for the age 85+ (for both men and women). However, in that survey, only 1% of respondents had this age which meant that some of the outliers may have had a potentially larger impact. Since 2000, the health status of elderly people has improved (life expectancy at age 65 improved from 17.3 to 18.7 for women and 13.4 to 14.6 for men between 2004 and 2016) [35], which (if better HRQoL is also assumed) may contradict our results showing lower index values. It seems that the elderly population with a better HRQoL may have been overrepresented in the 2000 national health survey, and the small number of participants of this age provided less robust estimates for this age band.

Overall, beyond using a more relevant value set for EQ-5D-3L, our study also offers larger sample size, thus enhanced statistical power, narrow age bands for more precise economic evaluations, inclusion of adolescents and potentially improved sampling, compared to the 2000 national health survey. We strongly believe that these factors contribute to more credible population norm estimates.

Higher index values derived using the Hungarian value set compared to the UK value set (shown in Fig. 2) can be explained by two key factors. Firstly, the Hungarian value set uses a 0.020 constant (decrement, to be used for health states other than 11111) instead of 0.081 used by UK value set. Secondly and more importantly, the new Hungarian value set does not apply the constant N3 (an additional  -0.269 decrement for L3 responses in any dimension used by UK value set, beyond the respective dimension-specific L3 decrement), as its impact has been considered in larger L3 decrements in the Hungarian value set compared to the UK one. On the other hand, for L2 responses which were reported much more frequently (Supplementary Table 2), the new Hungarian value set tend use smaller decrements compared to the UK one.

Nikl et al. published population norms for Hungarian population on a sample of 2000 adults, reported to be broadly representative in terms of sex, age groups, highest level of education, geographical region, and settlement type, using the 15D instrument [36]. The mean 15D index value was 0.810 using the Norwegian 15D value set. In that study, with advancing age categories, the 15D index values showed an inverse U-shaped curve with highest index values of 0.82 for both age bands of 25–34 and 45–54; and numerical results could be considered somewhat consistent with index values (derived by the UK value set) of this research. However, different HRQoL instruments, sample sizes, value sets, recruitment (i.e., voluntary registration from online panel) make the more detailed comparison of the two studies difficult. In Poland, Golicki and Niewada published population norms on a sample of 3963 adults, representative of the Polish population in terms of age, sex, geographical region, education, and socio-professional group, by using a self-administered EQ-5D-5L instrument [37]. To calculate index values, an interim EQ-5D-5L value set for Poland was used based on a crosswalk methodology. Index values (0.96 for those aged 18–24, 0.94 for 35–44, 0.9 for 45–54 and 0.81 for 65–74, respectively) were broadly consistent yet somewhat lower than the Hungarian ones (derived by the Hungarian value set). Again, differences in the applied instruments, sample sizes and administration make further direct comparison of the norms between the two countries difficult. In both studies, similarly to our results, men tend to have higher values in almost all age bands, especially above 35 years. Finally, Zrubka et al. compared EQ-5D-3L population studies from Hungary, Slovenia and Poland and reported issues in terms of comparability due different national characteristics, different data collection methodologies and times [38]. Importantly, data gaps for age 65+ were reported to be a general concern, confirming our findings in the oldest age category in the 2000 national health survey.

To minimize the impact of adult value set applied also for adolescents, our regression analysis included adults only. The lower EQ-5D-3L index in elderly individuals may be explained by the fact that elderly people tend to suffer from more diseases, including multi-morbid conditions [39], which potentially have a major impact on HRQoL. Better education is shown to be associated with healthier lifestyles [40‐42], higher participation in prevention programs, and appreciation of being healthy in general. Active occupation (i.e., employee, student) may lead to more physical activities and/or social contact, which may contribute to higher index values. On the contrary, inactive people may lack these, while those involved in childcare may feel isolated, sleepless or experience maternal depression, potentially associated with poorer HRQoL. Finally, interpersonal relationships are likely to be stronger in households with two or more people. Moreover, in larger families, the average household index value seems to be increasing significantly with household size even after controlling for the other factors listed in Table 3. Our findings on the association of age and education with HRQoL were also confirmed by similar conclusions from the 2000 national health survey [14] and the EuroVaQ study [16]. Interestingly, these prior studies also found that sex had a significant impact on the EQ-5D-3L index, as did household income (however, this latter variable was not included in the 2022 population survey). This is in line with our observation that there was an interaction between sex and age in our study.

This research has significant policy implications. The new population norm values for those under 18 introduced by this research will have a significant impact both on the accuracy of health burden estimates and the economic evaluation of health technologies for adolescents. Similarly, the application of the new Hungarian value set that truly reflects the preferences of Hungarian people for different health states are expected to improve the accuracy of health burden estimates and economic evaluations for adults. Moreover, as a Hungarian value set for the EQ-5D-Y-3L has also been available since 2022 [43], future research could compare the impact of using the EQ-3D-3L (with the adult value set) and the future use of the EQ-5D-Y-3L (with the new value set for adolescents) to conclude on the applicability of the adult EQ-5D-3L for adolescents aged 12–15 years, as considered to be acceptable by the EuroQoL Group [44]. Finally, this study together with the research conducted in 2000, with some limitations (difference in sampling methodology, mixing EQ-5D-3L administration methods, various factors influencing population change over 20 years etc.), allows researchers to estimate changes in HRQoL of the population over a 20-year time horizon.

The large representative random sample and the wide age range are the main strengths of this study. Compared to some other large-scale surveys, our study had more robust outreach for those above 65, a cohort especially relevant from public health and health economic point-of-view. However, it has some limitations, too. First, as the intention with the population-level health survey was to establish norm values for previous and ongoing disease-specific research in Hungary, and also to ensure comparability with the 2000 national health survey, the EQ-5D-5L was not considered for this research. Second, different administration methods of the questionnaire may have introduced bias even in a homogeneous sample. Third, the analytical weights had a relatively large range and, as noted by Potter and Zeng [45], ‘extreme variation in the sampling weights can result in excessively large sampling variances when the data and the selection probabilities are not positively correlated’. Finally, using the adult EQ-5D-3L also for 12-15-year-old adolescents may also have had an impact on the results.