Reference Data and Known-Groups Validity of the EQ-5D-5L for Vietnam


 PurposeThis study aims to provide EQ-5D-5L reference data of the general Vietnamese population and to test the EQ-5D-5L’ known-groups validity among people living with hypertension in Vietnam.MethodsThe EQ-5D-5L population norms were obtained via a representative sample from a general population of 1200 adults. Outcomes of the population norms were presented through five dimensions and five levels, EQ-VAS, and EQ-5D-5L indexes. Descriptive statistics of these outcomes were categorised by gender and age groups. Evidence of the known-groups validity was based on a sample of 577 non-hypertensive adults, 242 individuals with undiagnosed and 477 individuals with diagnosed hypertension. A hypothesis was formulated that people with worse health status would have lower EQ-VAS and EQ-5D-5L indexes. ResultsFor the EQ-5D-5L population norms, 54.4% of the respondents reported having full health. The mean EQ-VAS and EQ-5D-5L indexes were 81.10 and 0.94, respectively. The EQ-VAS and EQ-5D-5L indexes were higher among males, people at younger ages, those with more education, a paid job, and single. The mean EQ-VAS and EQ-5D-5L indexes of people in the diagnosed hypertension group were statistically significantly lower than they were in the other two groups. ConclusionThe EQ-5D-5L population norms were derived for the Vietnamese general population. The EQ-5D-5L can distinguish the quality-of-life differences among Vietnamese with hypertension.


Introduction
Vietnam's life expectancy at birth has been increasing signi cantly in recent decades [1], yet simultaneously so has the prevalence of chronic illnesses [2]. For instance, the prevalence of people with hypertension among the general population aged 25-64 years was reported to increase from 15.3-20.3% in 2010 and 2015, respectively [3]. The characteristics of people living with such diseases varied among living conditions, sex, and economic status, which were clearly shown in the 2015 national survey on non-communicable disease factors [4]. Vietnam's Ministry of Health (MOH) is upgrading the healthcare system to provide su cient healthcare services for people living with long-term illnesses [5]. The MOH has enacted several targeted programmes to prevent non-communicable diseases; nevertheless, e cient measures to monitor and evaluate these programmes are still in need [2]. Intermediate outcomes (e.g. levels of systolic and diastolic blood pressure) and natural measures (e.g. number of deaths / or averted cases) are sometimes not adequate enough to evaluate the effectiveness of a healthcare intervention for people living with chronic illness [6]. For example, health interventions on hypertension may simultaneously exert effects on people with cardiovascular diseases [7]. Hence, multi-dimensional health outcomes are needed to identify the additional health bene ts offered by such interventions [8]. A multi-dimensional health outcome, the health-related quality of life (HRQOL), is now getting more attention [6,9]. The HRQOL, which has been commonly used to investigate the impact of health status on quality of life [10][11][12], can be described by many different health dimensions and at different levels achieved within each dimension [8]. Several measurement instruments attempt to describe HRQOL, but overall, it can be grouped into either disease speci c measures -designed to measure the HRQOL of people with some speci c health problems (e.g. EORTC QLC-C30 or EORTC-8D for cancer patients); or generic measures to describe HRQOL of any health status that can be represented by different health dimensions with their respective levels of achievement/severity (e.g. a pro le-based measure: 36-Item Short Form Survey -SF-36; index-based measures: EQ-5D, SF-6D, Health Utility Index -HUI) [8]. The disease speci c measures are customised according to disease' characteristics, whereas the generic measures are responsive to overall HRQOL, and are more exible in their use for comparisons across disease areas [8]. Therefore, utilising a generic measure for HRQOL estimations support consistent policy-making processes, and enable comparisons between the amount of health gained and lost within society [8].
Vietnam's MOH has taken the rst steps in making use of a generic measure for HRQOL in healthcare service's evaluations by enacting the national health technology assessment (HTA) guidelines and upgrading the health insurance bene t package with cost-effective drugs based on HTA evidence [13]. Those facts implied a demand to promote evidence-informed policymaking in the national healthcare system, initially in health insurance. According to the national guideline on HTA submissions, the measure of generic quality adjusted life years (QALY) is a suggested index in Vietnam. Health outcomes as HRQOL and/or QALY have also been requested by other countries when it comes to HTA [14][15][16][17].
The number of QALY(s) can be estimated by multiplying the number of years of life with the quality of life during those years [8].
Whilst years of life is an obvious indicator, quality of life can be represented by a generic index-based measure of HRQOL, under the term "utility". This utility conventionally runs on a scale from 0 -representing "death" to 1 -which is "full health". Several discussions on negative value of utility (for worse than death states of health) still generate controversy [8,9]. Utilities are commonly measured by three instruments to quantify HRQOL, including EQ-5D; SF-6D, and HUI [8]. In Vietnam, EQ-5D-5L is currently the only instrument that can produce utility that is based on preferences of the general Vietnamese population [18].
Therefore, EQ-5D-5L has been suggested in the Vietnam's HTA guidelines. The EQ-5D-5L instrument includes ve dimensional ve-level questions, a visual analogue scale (EQ-VAS), and a value set which was tailored for the Vietnamese people [18]. The application of EQ-5D-5L can be included but not limited in QALY estimations. The instrument itself can re ect people's quality of life via their health status being reported by either the ve dimension ve-level questions, or the EQ VAS scores, or the health state' values.
The EQ-5D-5L has already been suggested in the national HTA guidelines, yet there are still two big concerns regarding the instrument. The rst one is that Vietnam needs reference data allowing utility comparisons between people with certain health conditions and the general population of the same age/gender. The reference data deriving from EQ-5D-5L is one of such utility reference data, and it is also referred to the name as "EQ-5D-5L population norm data" or "EQ-5D-5L population norms" [19]. The EQ-5D-5L population norms typically provide three outcomes, including the reference data of descriptive ve dimensional ve levels, EQ-VAS, and EQ-5D indexes. The EQ-5D-5L descriptive refence data is presented in percentages of the general population at each of ve levels (from "no problems" to "unable to do/extreme problems") on each of ve health-related dimensions. The EQ-VAS reference data can often yield a table of mean(s)/median(s) EQ-VAS scores by age-sex groups, while the mean(s)/median(s) EQ-5D-5L indexes were derived from the national value set for the EQ-5D-5L indexes reference data [20]. The population norms using the EQ-5D-5L were developed globally, from Western countries [21][22][23][24][25][26][27][28][29][30] to Asian countries [31][32][33][34][35][36][37]. A hint of population norms using the EQ-5D-5L for the Vietnamese has been done elsewhere, but the study included an urban population only, and furthermore, used Thai preferences [37]. Since Vietnam has now had a country-speci c value set, this is timely to develop the country-speci c EQ-5D-5L population norms.
The second concern is whether EQ-5D-5L can be justi ed for use in Vietnam. Psychometric properties of the EQ-5D-5L have been proven in several countries and for several disease areas [38][39][40][41][42][43][44][45][46][47][48]. The instrument's reliability, which concerns stability in measuring people's quality of life, can often be proven by the consistency of HRQOL results obtained through different measurement instruments [38][39][40][41][42][43], and/or by the repeatability by each time using the instrument [44][45][46]. Validation of the EQ-5D-5L was also tested via different construction validations, commonly including convergent and known-groups validation. The convergent validation was for strong correlations between postulated dimensions and other dimensions that should, in theory, be considered relevant [49]. Thus, correlations of the ve dimensions (mobility, usual activities, self-care, pain/discomfort, and anxiety/depression) with other respective dimensions of other different instrument(s) are often evaluated for convergent validation purposes [46][47][48]. Meanwhile, the known-groups validation to evaluate the sensitiveness of the instrument is expected to yield distinctive results among different groups of patients [44,47]. A study on the reliability and convergent validation of the EQ-5D-5L in Vietnam was conducted among HIV/AIDS patients [40], yet the construct validation among people with chronic diseases is still limited. To ll the research gaps, this study aims to provide (1) HRQOL reference data using EQ-5D-5L among the general population and (2) validity tests for the EQ-5D-5L instrument among people living with hypertension in Vietnam.

Methodology
Data was pooled from two separated studies. Data to derive the EQ-5D-5L population norms was taken from the Vietnam EQ-5D-5L valuation study which had been conducted in the general population in 2017 [18]. Due to the shortage in HRQOL data measured from different measurement instruments, the reliability and convergent validity tests were not included. The knowngroups validity test was conducted using data from a hypertensive sample of the end-line survey of the "Evaluation of the Ho Chi Minh City Communities for Healthy Hearts" -CH2 project [50].

Samples
The sample for developing EQ-5D-5L reference data From the EQ-5D-5L valuation study [18], a general population sample of 1200 adults was used to develop population norms. A multi-stage strati ed cluster probabilistic quota-based sampling method was applied. The rst stage was to determine an urban and a rural cluster from six provinces of six different geographical regions.
The next stage was to determine quotas for each cluster. The probabilistic quotas were developed based on the fractions of the population's regions, residency, age groups (18-29 years, 30-44 years, 45-59 years, and 60 + years) and sex (male and female).
Details of the sampling have been published elsewhere [18].
The sample for validity testing of the EQ-5D-5L instrument.
From the CH2 cohort study [50], a sample from the post-evaluation community survey was used for the known-groups validity of the EQ-5D-5L. The survey was conducted in eight districts of Ho Chi Minh City during 2019. A combination of multistage cluster random sampling techniques was employed to recruit 1296 CH2 participants whose age was 40 years old or above for the survey. The World Health Organization's classi cation for blood pressure levels to identify people living with hypertension are those with an average systolic blood pressure (SBP) of ≥ 140 mmHg, and/or an average diastolic blood pressure (DBP) of ≥ 90 mmHg [51]. Three groups were created based on participants' hypertension-related status, which included 577 non-hypertensive individuals (SBP|DBP < 140|90 mmHg); 477 who had been clinically diagnosed with hypertension, and 242 individuals with undiagnosed hypertension (SBP|DBP ≥ 140|90 mmHg).

The EQ-5D-5L instrument
The EQ-5D-5L instrument included ve questions and the EQ-VAS. The ve questions represented ve health dimensions on mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Respondents were asked to describe their current health status for each dimension by ve levels of severity: (1) no problems, (2) slight problems, (3) moderate problems, (4) severe problems, and (5) unable to do/extreme problems. The EQ-VAS is a hash-marked scale ranging from 0 to 100, in which "0" indicates the worst imaginable health and "100" means the best imaginable health. The EQ-5D-5L indexes were calculated from the Vietnamese EQ-5D-5L value set [18].

Data collection
In both surveys, trained public health staff carried out face-to-face interviews. Participants were recruited using the door-to-door approach. Data used for developing the population norms were participant's demographic characteristics and self-reported health statuses using the EQ-5D-5L. For the validation study, data of participants' background; medical history on hypertension (including whether there were clinical diagnoses and comorbidities); health statuses reported via the EQ-5D-5L, and physical measurement outcomes such as height, weight, blood pressure were collected. Physical measurements were carried out by trained public health staff on all participants, with procedures adhering to guidelines of Vietnam's MOH on general health checkup, the process of height/ weight/ blood pressure measurements, follow-up work after measurements, safety management, etc Analysis Generally, differences on distributions of the ve dimensions ve levels among sub-groups were tested using Pearson chi-square tests. Even though the EQ-VAS and EQ-5D-5L indexes were not normally distributed for both samples, these sample sizes were, however, large enough to assume normal distributions according to the central limit theorem [53]. Therefore, the differences of the EQ-VAS and EQ-5D-5L indexes among sub-groups were tested using either T-tests (applying to two-group categorical variables), or ANOVA tests (applying to homogenous variance and more-than-two-group categorical variables), or Kruscal Wallis H-Tests (applying to non-homogenous variance and more-than-two-group categorical variables). The Poc-hoc analysis to examine differences among multiple pairwise comparisons was performed using Bonferroni tests. A signi cance level of 0.05 was used for all statistical tests. Data was analysed using STATA version 14 software [54].
The EQ-5D-5L population norms The EQ-5D-5L population norms w derived from the data given by the general population sample. Of which, the EQ-VAS and the ve dimension questions were self-reported, and the EQ-5D-5L indexes were calculated using the Vietnam value set [18]. The analysis on EQ-5D-5L population norms followed the standardised method recommended by the EuroQol Group [20]. Descriptive statistics of the ve dimensions ve levels, EQ-VAS and EQ-5D-5L indexes were categorised into gender and age groups. Among these, percentages of answers for the ve dimensions ve levels were presented; EQ-VAS and EQ-5D-5L indexes were reported in means, standard deviations, ranges of min -max and interquartile, respectively. Differences of the EQ-5D-5L' outcomes were statistically tested.

Known-groups validation
As data was given from the CH2 project using only the EQ-5D-5L to measure people's HRQOL, the known-groups validation was tested. The EQ-5D-5L indexes were computed based on the Vietnam value set [18]. Literature suggested that the presence of hypertension and comorbidities was associated with lower HRQOL [55][56][57]. Also, patients who were aware of their hypertension status reported a poorer quality of life [58,59]. Hence, a hypothesis was formulated that, as the EQ-VAS and EQ-5D-5L indexes would be higher among non-hypertensive people, indicating better HRQOL, such indexes would be lower among those with hypertension. Correspondingly, people with undiagnosed hypertension would have better HRQOL than those of the hypertensiondiagnosed group. In addition, HRQOL of hypertensive people was often suggested to be lower among females, people with more comorbidities, less education, older ages, and higher body mass indexes (BMI) [60,61]. Statistical tests were performed to test the hypothesis and differences of EQ-VAS and EQ-5D-5L indexes regarding gender; age; education; marital status; BMI classi cations, and number of comorbidities. The associations of these characteristics with the EQ-VAS and the EQ-5D-5L indexes were tested using a multivariate linear regression model among people living with clinical diagnoses for hypertension.

Results
The EQ-5D-5L population norms Figure 1 shows characteristics of the general population sample. Overall, the sample distribution was similar among sub-groups of genders; age groups; geographic regions, and education levels. The present sample was most attributed to Kinh (major ethnicity) people, groups of individuals living in rural areas, married individuals, and those having a paid job. Socio-demographic characteristics of the present sample were inline with those of the national adult population. Nevertheless, the education level of participants and proportion of younger people in this sample were higher than the national average. Table 1 shows the percentages self-reporting the ve dimensions ve levels by age groups of the general population sample. The percentage of participants that reported having full health was 54.4% and such indexes decreased by age. Respondents reported fewer problems in self-care and usual activities than the other dimensions. The number of individuals that reported having problems at higher levels increased for the subsequent age groups. People of the youngest age group (18-24 years) were shown to have slight problems in all ve dimensions; and moderate problems with anxiety/depression (3.2%) and pain/discomfort (2.3%). People aged 25-64 years reported having from "slight" to "extreme problems" in all dimensions, while the worst problems were mainly in mobility, pain/discomfort, and anxiety/depression. Females seemed to have more problems than males in all age groups. The proportion of females reporting "no problems" in self-care (about 98%) and usual activities (about 95%) were similar for males, yet their reporting of "no problems" was slightly lower in mobility (88.9%; a 2.9% difference), and lower in pain/discomfort (61.3%; a 8.8% difference), as well as anxiety/depression (77.7%; a 5.6% difference), respectively. Patterns of the ve dimensions ve levels for females and males in all age groups are in online resource 1.  Table 2 shows results of EQ-VAS and EQ-5D-5L indexes by sub-groups of the general population sample. Overall, the mean EQ-VAS and EQ-5D-5L indexes were 81.10 and 0.94, respectively. The mean EQ-VAS and EQ-5D-5L indexes were found to be 1.46 and 0.02 higher among males than among females (p-value 0.06 and 0.00), respectively. A signi cant reduction trend was observed in EQ-VAS and EQ-5D-5L indexes across people aged from 25 to 65 + years (p-value < 0.01). The EQ-VAS and EQ-5D-5L indexes, among people having an education level at high school or higher, or not being unemployed, or being single, were signi cantly higher than the other counterparts (p-value < 0.05). By geographical region, the EQ-VAS was shown to be statistically lower among people living in the Central Coast areas, in comparison with individuals in the Red River, South East, and Mekong River areas. Results of statistic tests on differences of EQ-VAS and EQ-5D-5L indexes of the general sample from the Poc-hoc analysis are presented in online resource 2. Known-groups validation Table 3 shows CH2' sample characteristics and its results on the EQ VAS, EQ-5D-5L indexes. Overall, demographic characteristics were similar among the non-hypertensive; diagnosed hypertension and undiagnosed hypertension groups. Nevertheless, characteristics of older age, higher rates of unemployment, and higher numbers of comorbidities, were more frequent among individuals living with hypertension both with and without a diagnosis, than those who were non-hypertensive. The percentage of "full health" self-reported by people living with hypertension (62.70%), was smaller than among those who were in the nonhypertensive and undiagnosed hypertension group (both at 71.90%). The mean EQ-VAS was found to be the highest in the undiagnosed hypertension group (76.95), followed by the non-hypertensive (76.65), and the diagnosed hypertension group (71.48). The mean EQ-5D-5L value showed a slight downward trend with the non-hypertensive (0.97), undiagnosed hypertension (0.96), and diagnosed hypertension groups (0.94), respectively. The mean EQ-VAS and EQ-5D-5L indexes of people in the diagnosed hypertension group were statistically signi cantly smaller than they were in the other two groups (p-value < 0.05).
Results of EQ-VAS and EQ-5D-5L indexes were reported comparably between people from the non-hypertensive and undiagnosed hypertension group. Results of statistic tests on differences of EQ-VAS and EQ-5D-5L indexes of the CH2 sample from the Pochoc analysis are presented in online resource 3. Notes: *Results from T-Tests. ** Results from ANOVA. Table 4 shows associations between demographic factors and, respectively, the EQ-VAS and EQ-5D-5L indexes among the diagnosed hypertension group. A total of 472 people with a diagnosis for hypertension self-reported their health status via the EQ VAS. Statistics shows decrements in the EQ-VAS by higher age and number of comorbidities, whereas the factor of having completed high school and above was associated with a higher EQ-VAS. The EQ-5D-5L indexes were derived from answers from all 477 respondents in the diagnosed hypertension group. Associations for lower EQ-5D-5L indexes were found in people of older ages (β= -0.001; p-value = 0.019), females (β= -0.024; p-value = 0.048), people suffering from an incremental comorbidity (β= -0.013; p-value = 0.028), and measured obese people (β= -0.081; p-value = 0.007). An education level of having completed high school and above was associated with higher EQ-5D-5L indexes (β = 0.027; p-value = 0.049).

Discussion
This study has provided the EQ-5D-5L reference data in Vietnam, which was presented with regards to age and gender for the descriptive part of the ve dimensions ve levels, EQ-VAS and EQ-5D-5L indexes. Additionally, this study demonstrated the validity of the EQ-5D-5L instrument among people living with hypertension. The EQ-5D-5L was shown to be responsive to changes in HRQOL among participants with less desirable health statuses.
A strength of this study's EQ-5D-5L population norms was the neutral context sample. Responses were pooled across the country by geographical regions, gender, age, and residence settings. In a previous EQ-5D-5L population norms study, results were derived from the data of an urban population and EQ-5D-5L indexes were calculated using Thai value set [37]. However, EQ-5D-5L indexes in the present study were estimated using the Vietnamese preference-based value set. Additionally, the percentage reporting full health in the previous study was about 67.4%, which was 13% higher than the present study. Findings here of EQ-5D-5L population norms, therefore, could be perceived as more neutral context HRQOL reference data. The mean EQ-5D-5L value for Vietnamese adults was about 0.94, which was in line with the range of indexes across countries, from 0.89 in Poland [21] to 0.96 in China [31]. The present study shows the same pattern of the EQ-5D-5L reference data compared to the previous studies. For example, the EQ-5D-5L indexes were reported to be lower for females than males; or higher for people having an education from high school and higher. Such results have been similarly found in the previous Vietnamese EQ-5D-5L population norms study [37]; China [31]; Hong Kong [33]; Indonesia [34] and Spain [24]. The EQ-5D-5L indexes in this study showed a linear relationship with age for both genders. Nevertheless, the linear relationship was inconsistent for females, i.e., the mean EQ-5D-5L value was slightly lower among younger females aged 18-24 years than those in the age group of 25-34 years. The EQ-5D-5L population norms in Australia [23] and Hong Kong [33] also reported similar linear relationships between the EQ-5D-5L value and age. Moreover, results showed statistically signi cant differences of EQ-VAS across the six geographical regions, seemingly that