Convergent and discriminative validity of the Frail-VIG index with the EQ-5D-3L in people cared for in primary health care

Validity is defined as “the degree to which an instrument truly measures the construct(s) it purports to measures” [1]. Validation is a continuous process and different forms of validation can be applied. Criterion validity and construct validity are two types of validity. Criterion validity is applicable when there is the gold standard for the construct that is measured, it refers to the degree to which the scores of a measurement instrument are an adequate reflection of a gold standard [1, 2]. By contrast, construct validity is applicable when there is no gold standard, it refers to the degree to which the scores of a measurement instrument are consistent with the available knowledge about the construct [1, 2].

On the other hand, frailty is defined as “a clinical state in which there is an increase in an individual’s vulnerability to develop negative health-related events (including disability, hospitalizations, institutionalizations, and death) when exposed to endogenous or exogenous stressors” [3]. It is a complex and multidimensional concept for which there are numerous and multiple operational definitions. This fact has contributed to the lack of an accepted gold standard [3, 4]. As a result, most frailty measurement instruments assess their validity by analysing the degree of consistency of their scores with different hypotheses about their relationship with other instruments or the differences between relevant groups. In other words, hypothesis testing for construct validity studies is carried out because of the difficulty in assessing criterion validity.

A new frailty index, the Frail-VIG index, has been developed recently [5, 6]. It is an instrument with a multidimensional approach and a pragmatic purpose that allows rapid and efficient assessment of the degree of frailty in the context of clinical practice. This measurement instrument has shown to have an optimal capacity to predict two-year mortality (Area Under Curve 0.85) [6]. The relationship of their scores with those of the Clinical Frailty Scale [7] has been evaluated in a cross-sectional study and a strong positive correlation (r = 0.706) has been established [8]. All of these studies have been conducted in an inpatient hospital setting and there have been no studies in primary health care (PHC) settings.

Previous research suggests an association between frailty and worse quality of life, but its findings are mixed and inconsistent. However, recent systematic reviews show a consistent negative association between frailty and quality of life among community-dwelling people [9, 10]. Besides, the validity of a measuring instrument does not reside in the instrument itself but in how it is used and hence depends on its appropriateness to the target population and the specific context of administration [2]. Therefore, a good approach to further developing evidence of the validity of the Frail-VIG index would mean to analyse the relationship of its scores with those of another instrument that measures the quality of life (both instruments administered in the context of PHC).

Consequently, we carried out this study in a PHC setting to investigate the convergent and discriminative validity of the Frail-VIG index regard to health-related quality of life (HRQoL) measured by the EQ-5D three-level version (EQ-5D-3L). Concerning convergent validity, we hypothesised that the relationship of scores between instruments was moderate to strong and negative. With regard to discriminative validity, we hypothesised that non-frail people would have higher scores on HRQoL than frail people.

Four hundred and twelve participants were included in this study. Table 1 shows their general characteristics and frailty status according to the Frail-VIG index scores.

Table 2 describes the 10 most frequent EQ-ED-3 L profiles according to frailty status. The worst of these profiles was most prevalent among people with severe frailty, while the best was more frequent among non-frailty people.

Pearson’s correlation coefficient between Frail-VIG index and EQ-5D-3L value was negative and moderate (r = −0.510; P < 0.001). After adjusting for age and sex variables, the multiple linear regression model revealed that Frail-VIG index independently correlated with EQ-5D-3L value (B = -0.945; 95% Confidence Interval, −1.098 to − 0.791; R² = 0.287). As you can see from Table 1, non-frail people reported a substantially better HRQoL than people with moderate and severe frailty people. Likewise, closer inspection of Table 3 shows that the EQ-5D-3L value declined significantly as the Frail-VIG index score increased.

In this study, involving people cared at-home by PHC professionals, we found that the Frail-VIG index demonstrated convergent validity with the EQ-5D-3L value. Furthermore, its discriminative validity was optimal, as their scores showed an excellent capacity to differentiate between people with better and worse HRQoL.

Consistent with the literature, this research found that HRQoL and frailty were negatively associated among community-dwelling older people [9, 10]. Studies on other cumulative deficit models of frailty have also found a negative relationship between the frailty state and the scores of instruments measuring HRQoL, well-being and life satisfaction [9, 10, 21‐23]. However, the results of this study show that not only is the frailty state associated with a worse HRQoL but also that there is a linear association between both indexes (Frail-VIG index and EQ-5D-3L value). Hypothesis testing is an ongoing process, so the more hypotheses are tested the more evidence is generated for construct validity [24]. Therefore, this research supports evidence from previous studies carried out in the hospital setting on the construct validity of the Frail-VIG index [6, 8].

This study has several strengths. Most validation studies of measurement instruments of frailty in community-dwelling people focus on demonstrating their predictive potential for adverse outcomes or resource use, such as disability, institutionalisation or hospital admissions [25‐27]. In contrast, studies that analyse their relationship to more positive outcomes such as HRQoL are less common [10, 28]. Likewise, the use of the Frail-VIG index in primary care has been poorly studied, and this is one of the first studies to analyse its construct validity in this care setting. However, some limitations exist. The representativeness of non-frailty people is very low (12 people), probably because the study population were people in a home-care programme. A greater representation of this population might have influenced the Frail-VIG index’s discriminative validity observed. In addition, we used the EQ-5D values for this assessment of the measurement properties of the Frail-VIG index instead of other values such as the EQ VAS. The EQ VAS can be considered a measure closer to the patient’s perspective than the EQ-5D values [29]. Nevertheless, as Devlin and Parkin point out [19], being able to summarise and represent a health profile with a single value has important advantages, including simplifying statistical analyses. On the other hand, both the EQ-5D values and the EQ VAS are able to discriminate between the quality of life of most groups of individuals with different socio-demographic factors, and those with or without clinical conditions [30]. Furthermore, some studies [31] and experts in the field of psychometrics [32] report that older people experience difficulties in understanding and completing direct estimation methods, such as the visual analogue scale. For all these reasons, we chose the EQ-5D values to carry out this psychometric study.

People living with frailty risk experiencing a decline in their quality of life [9, 10, 33]. The findings of this study suggest that the interventions aimed at decreasing frailty could have the added benefit of improving the HRQoL. PHC professionals are naturally positioned to identify frailty early and to implement interventions that prevent related adverse effects on the most vulnerable people [4]. Moreover, the assessment of frailty in PHC settings requires tools that are not time consuming as well as valid and reliable which is way the Frail-VIG index could provide a useful and appropriate tool for this care setting [34].

This study has identified a negative moderate correlation between the Frail-VIG index and the EQ-5D-3L values. It has also shown that the Frail-VIG index was able to discriminate significantly home-dwelling older people according to their HRQoL. These findings provide additional pieces of evidence for construct validity of the Frail-VIG index. Further research is needed on this new measurement instrument to determine its suitability for screening and preventing adverse effects of frailty in PHC settings.