EQ-5D-5L reference values for the German general elderly population

The measurement and valuation of health-related quality of life (HRQoL) have become two of the most important components for the assessment of health and social care as well as public health interventions [1, 2], however, with limited applicability within the latter two sectors, since benefits of interventions may also be non-health-related [3, 4]. Nonetheless, using measures of HRQoL such as the EQ-5D to describe the benefits of treatments in all three sectors is still common and even recommended by the National Institute for Health and Care Excellence [5, 6]. The EQ-5D is a well-established and extensively used measure of HRQoL developed by the EuroQol Group [7, 8]. It can be characterised as a succinct and generic instrument, intended to measure health and deviations from it [9]. The EQ-5D has evolved to become one of the most widely used instrument to operationalise utilities for use in economic evaluation [10, 11]. However, informing resource allocation decisions is not the sole use of the instrument. It is further used for clinical appraisal, in epidemiological studies, in population health surveys and as a routine outcome measure in health care [9, 12‐14]. The EQ-5D consists of a classification system with five dimensions (mobility, self-care, usual activities, pain or discomfort and anxiety or depression) and a subjectively rated visual analogue scale (EQ VAS) [15, 16]. The instrument provides a variety of information in terms of a descriptive health profile, the individual EQ VAS rating and an index score [17].

Health information based on these three EQ-5D components can be analysed and interpreted on an individual level. However, it can also be summarised and described at a population level. Such a set of aggregate comparator data is referred to as reference values or population norms and is useful for clinicians and health economists to compare sampled values against a specified reference group, e.g. the general population, to determine deviations in health [17]. Similarly, reference values may be used to populate health economic models with health state values [2]. Reference values generally refer to a defined population. Primarily, definition is based on regional aspects, where reference values are usually provided for the general population of that area, e.g. for the USA, Germany, Indonesia or South Australia [18‐21]. Similarly, the reference population may be further stratified to represent diseases, i.e. patient groups, or socio-demographic groups [22, 23]. Furthermore, norms data may be derived based on different instruments such as the EQ-5D-3L and 5L, since those differ with regard to the provided information [24‐26].

Several studies found that higher age is significantly associated with lower HRQoL as measured by the EQ-5D-5L [14, 19, 27]. Moreover, there is further evidence to suggest that variation in HRQoL is large in the elderly, which is even more pronounced in the oldest old [28, 29]. With regard to Germany, two sets of preference-based reference values were identified. Grochtdreis et al. [19] published EQ-5D-5L norms data for the German general population including elderly respondents, but using only two broad age bands for those aged 65 years and above. Secondly, König et al. [23] published reference values for the oldest-old (85+) in Germany based on EQ-5D-3L, which are, however, practically restricted by the underlying value set used to derive utilities. Given the expected increase in the proportion of elderly in Germany [30], we would like to argue that additional reference values based on EQ-5D-5L for elderly men and women of narrower age groups may be helpful to interpret changes in reported health more adequately, in particular, if returning to perfect health does not seem feasible [31]. Hence, the aim of this study is to provide reference values for Germany for the general elderly population, which are based on EQ-5D-5L using smaller age bands as compared to conventional population norms.

The aim of this study was to provide population norms for the German population aged 65 years and above. To the best of our knowledge this is the first study to provide reference values for the elderly population in Germany, which is based on the recently developed tariff for the German version of the EQ-5D-5L [32].

The overall mean utility score in our study was slightly lower than the reported overall value for the German general population (0.84 vs. 0.88), which included all age groups [19]. However, the mean index reported here is considerably higher than that of the oldest-old in Germany as reported by König et al. [23] (0.84 vs. 0.77(GER)/ 0.68(UK)). This finding replicates the negative association of age with mean utility scores, which was found in other population norm studies as well [14, 21, 27, 36‐38]. Another source for the deviation may be the choice of the instrument and value set.

Conventional population health studies only use two age categories to represent the elderly population, which we further split into two smaller age bands to facilitate a more detailed comparison based on age groups. When comparing the computed mean utilities with the values reported for the German elderly by Grochtdreis et al. [19], we find that the two middle categories (70–74 and 75–79 years) largely agree with the broader values reported for Germany. On the contrary, mean index values of those in the age group 65–69 years are considerably higher, whereas the opposite is the case for the oldest old [19]. This may indicate that the decline in HRQoL that is associated with high age is even more pronounced than may be assumed based on the broader age categories from conventional population norm studies. Evidence of a considerable decline beyond the age of 80 can be found in other studies, as well [23, 27, 29, 39].

Furthermore, men reported similar or higher mean utility than women. This difference was largest in the oldest age category. Again, this pattern can be confirmed for Germany [19, 23, 38]. However, Hinz et al. [38] calculated sum scores based on the severity levels of the descriptive system rather than utilities, which limits the comparability. The observed negative association of increasing age and female sex with HRQoL was also found when comparing mean EQ VAS values, which was also found internationally [17]. On average, men reported higher EQ VAS values than women (74.2 vs. 72.3) and this was consistent across age groups. The only exception were women aged 70–74 years reporting higher EQ VAS values than their male counterparts. Similar findings were described by Huber et al. [28] for Germany.

Considering the health profile level, 21.4% of this elderly sample reported to have no problems in all dimensions, which is considerably less than the proportions found in other German studies for the general population [19, 28, 38] and comparable to the share reported by König et al. [23] for the German oldest-old. Generally, the observed response distribution in this study largely adheres to that observed for similar age categories in another study for Germany [19] and internationally [14, 21, 27, 36]. Overall, we observed an increasing prevalence of reported health problems with higher age. In this sample, problems were most frequently reported in the dimensions of pain or discomfort and mobility, whereas only 15.5% of the total sample reported problems with self-care. Such a pattern, where age-related health impairments seem to predominantly manifest as problems with pain or discomfort and mobility, was also observed elsewhere [19, 23, 29, 40]. The diminished proportion of respondents at the ceiling in our sample of the elderly seems to indicate that the EQ-5D-5L is sensitive towards age-related health problems. This aligns well with the argument made by Konnopka and König [41] that ceiling effects decrease with higher levels of morbidity, which are regularly observed in samples of the elderly [19, 25, 27, 29, 36, 42].

There is a wide body of population norms indicating that problems with anxiety or depression remain at a similar level independent of age [14, 19, 21, 25, 29, 36, 40, 43]. However, this was only the case for men in our sample; women, on the other hand, reported more problems with an increasing rate at older age. Overall, the rate of reported problems with anxiety and depression increased with increasing age similar to the other dimensions, but we found this trend to be driven by the higher proportion of women in the higher age groups who tend to report more problems with anxiety or depression [23, 27, 29]. Interestingly, Jiang et al. [18] and others found evidence of a decreasing prevalence of problems with anxiety or depression in older respondents [20, 44], which they explained with social stigma being attached to mental health problems and, hence, may lead to a lower frequency of acknowledged mental health problems. A similar response pattern was also described by König et al. [23] for the oldest-old (85 +) using the EQ-5D-3L.

One strength of this study is the use of data from a sample of German general population, which is representative in terms of age, gender, education, employment status and area of residence. Further, this study provides reference data on all three information components of the EQ-5D-5L. Importantly, the index values were derived using the recommended tariff by Ludwig et al. [32] providing additional information compared to earlier studies, which described health of the German population based on the EQ-5D-5L, but using unweighted sum scores [38] or EQ VAS values [28]. Furthermore, we provide references values for smaller age bands, which may enable a more detailed comparison when using these references. However, some limitations of this study must be considered. While information on respondents’ experience with severe illness is available, the data lacks detailed information on prevalent long-term conditions or comorbidities. Since the data was collected as part of the German EQ-5D-5L valuation study, where respondents engage in a cognitively demanding task, it can be assumed that this sample may be cognitively and physically healthier than older people who are not participating in valuation interviews. Similarly, due to the primary purpose of the underlying data set, respondents were not sampled to represent the German elderly population per se, but to represent the general population; by this, individuals living in institutions, such as residential aged care facilities, may be underrepresented. Both of these limitations may have introduced a selection bias, which potentially led to an overestimation of the elderly population’s health by this sub-sample. Moreover, the sample size can be considered small for an EQ-5D-5L reference value study. Therefore, the precision in the confidence intervals for the means is relatively low, leading to some overlap in CIs between adjacent age groups, which limits the certainty of detecting true differences between age groups in mean EQ VAS and EQ-5D-5L index values. Secondly, reference values for the oldest age group are based on very few observations (i.e. n = 38). Thus, the robustness and generalisability of the reference value for this age group are limited and should be used cautiously.

Due to the secondary nature of this data set, the sample size results as a consequence to the EQ-5D-5L valuation protocol with a target of N = 1000 respondents [45, 46], where this sub-sample represents the proportionate share of the elderly population from the original data. While we believe this to be an efficient use of existing data to generate benchmark values, future research may take the special characteristics of the older population into account and improve in sample size to increase generalisability and precision of the results.

These values were derived from a representative sample of the German elderly using the recommended tariff for the EQ-5D-5L. The findings may enable empirical comparisons of EQ-5D-5L based HRQoL with other samples, to assess change in health over time or burden of disease. Generally, the observed mean utilities and VAS values as well as the dimension-level response distribution correspond well to earlier findings from a large German population norm study. However, a more detailed provision of reference values for the elderly population seems helpful given that HRQoL in the oldest old is less consistent and falls off in comparison to the young elderly, while the sought age categorisation of conventional population norms studies seems to mask these differences. Hence, our findings may facilitate a more precise comparison across elderly age groups. Nevertheless, we suggest conducting further research to explore aspects and determinants of HRQoL for the age group above 80 years.