Evaluating the discriminatory power of EQ-5D, HUI2 and HUI3 in a US general population survey using Shannon’s indices

Public Health Epidemiology, Department of Social Medicine, Academic Medical Center, 22660, 1100 DD, Amsterdam, The Netherlands.
Quality of Life Research (Impact Factor: 2.49). 07/2007; 16(5):895-904. DOI: 10.1007/s11136-006-9160-6
Source: PubMed


To compare quantitatively the discriminatory power of the EQ-5D, HUI2 and HUI3 in terms of absolute and relative informativity, using Shannon's indices.
EQ-5D and HUI2/3 data completed by a sample of the general adult US population (N = 3,691) were used. Five dimensions allowed head-to-head comparison of informativity: Mobility/Ambulation; Anxiety/Depression/Emotion; Pain/Discomfort (EQ-5D; HUI2; HUI3); Self-Care (EQ-5D; HUI2); and Cognition (HUI2; HUI3). Shannon's index and Shannon's Evenness index were used to assess absolute and relative informativity, both by dimension and by instrument as a whole.
Absolute informativity was highest for HUI3, with the largest differences in Pain/Discomfort and Cognition. Relative informativity was highest for EQ-5D, with the largest differences in Mobility/Ambulation and Anxiety/Depression/Emotion. Absolute informativity by instrument was consistently highest for HUI3 and lowest for EQ-5D, and relative informativity was highest for EQ-5D and lowest for HUI3.
Performance in terms of absolute and relative informativity of the common dimensions of the three instruments varies over dimensions. Several dimensions are suboptimal: Pain/Discomfort (EQ-5D) seems too crude with only 3 levels, and the level descriptions of Ambulation (HUI3) and Self-Care (HUI2) could be improved. In absence of a formal measure, Shannon's indices provide useful measures for assessing discriminatory power of utility instruments.

Download full-text


Available from: M.F. Bas Janssen, Dec 17, 2013
  • Source
    • "The functional health state of the patient and the utility weights derived from the population are based on generic attributes and without regard to the underlying disease, disease-specific key symptoms, prognosis, or treatment. Evidence suggests that the sensitivity of the EQ-5D is low compared to the HUI, implying that the EQ-5D does not measure disability where the HUI does [13]. Additionally, Polinder et al. showed that among injury patients, the HUI is more sensitive for comorbid disease compared to the EQ-5D [14]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Three approaches exist to deal with the impact of comorbidity in burden of disease studies - the maximum limit approach, the additive approach, and the multiplicative approach. The aim of this study was to compare the three comorbidity approaches in patients with temporary injury consequences as well as comorbid chronic conditions with nontrivial health impacts. Disability weights were assessed using data from the EQ-5D instrument developed by the EuroQol Group and derived from a postal survey among 2,295 injury patients at 2.5 and 9 months after being treated at an emergency department. We compared the observed and predicted EQ-5D disability weights in comorbid cases using data from injury patients with and without comorbidity who were restored from their injuries at 9 months follow-up. The predicted disability weights were calculated using the maximum limit approach, additive approach, and multiplicative approach. The intraclass correlation coefficient (ICC) was used to test whether the values of the observed disability weights and the three model-predicted disability weights were correlated. The EQ-5D disability weight of injury patients increased significantly with the number of comorbid diseases. The ICCs of the additive, multiplicative, and maximum limit models were 0.817, 0.778, and 0.674, respectively. Although the 95% confidence intervals of the ICCs of the three models overlap, the maximum limit model seems to fit less well than the additive and multiplicative models. For mild to moderate chronic disease (disability weight below 0.21), the association between predicted and observed disability weights was low. Comorbidity has a high impact on disability measured with EQ-5D. Ignoring the effect of comorbidity restricts the use of the burden of disease concept in multimorbid populations. Gains from health care or interventions may be easily overestimated if a substantial number of patients suffer from additional conditions. The results of this study found that in accounting for comorbidity effects, all three models showed a strong association between the predicted and observed morbid disability weight, though the maximum limit model seems to fit less well than the additive and multiplicative models. The three models do not fit well in the case of mild to moderate pre-existing disease.
    Population Health Metrics 04/2011; 9(1):10. DOI:10.1186/1478-7954-9-10 · 2.11 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The population's health faces an array of diseases and injuries. Limited resources compel policy-makers everywhere to focus on threats that are regarded most relevant in terms of public health. The World Health Organization and Worldbank developed an innovative concept which expresses the burden of disease in Disability Adjusted Life Years (DALY). The DALY provides knowledge on the size of health problems and the potential benefit of proposed measures set against similar and comparable data of other problems. However, apart from the obvious advantages, the DALY-concept has been criticized concerning key aspects of the methodology. This thesis addressed four of these controversial aspects: namely (1) deriving disability weights for disorders with complex and heterogeneous recovery patterns, (2) the disregard of comorbid diseases, (3) the arbitrariness of the in- and exclusion of long-term sequelae, and (4) the absence of a criterion to identify cases that are relevant from a public health perspective. These controversial aspects were addressed with regards to two health (care) domains where these aspects are particularly problematic. Both health domains, injuries and infectious intestinal disease, are featured by heterogeneous health outcomes, including the extremes of the severity spectrum and duration, with all types of time-severity relations. This thesis demonstrated satisfactory alternative solutions to these four controversial issues of the DALY concept. As a result, the DALY-concept may be used for burden of disease studies of conditions in which these controversial issues emerge in particular. Application of alternative solutions proposed in this thesis may improve burden of disease estimates considerably.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To analyse international variation in clinical injury incidence, and explore the performance of different injury indicators in cross-country comparisons. Hospital discharge data of seven European countries (Austria, Denmark, Ireland, Netherlands, Norway, England and Wales) were analysed. We tested existing and newly developed indicators based on (a) health care use, (b) anatomical criteria, or (c) expected health outcome: admissions excluding day-cases (a), hospital stay 4+ (a) and 7+ days (a), (serious) long-bone fractures (b), selected radiological verifiable fractures 'SRVFs' (b), and indicators based on international (Global Burden of Disease) and Dutch disability weights). Assessment criteria were reduction in incidence variation and length of stay in hospital, and the association between incidence and mortality rates. Indicators based on health care use led to increased variation in incidence rates. Long bone fractures and SRVFs, and both indicators based on injuries with moderate to high disability showed similar variation in clinical incidence compared to the crude rates, smaller variation in median length of stay in hospital and a good association with mortality rates. No perfect or near perfect indicators of clinical injury incidence exist. For international comparisons, indicators based on disability weights, SRVFs and long bone fractures may be sensible indicators to use, in the absence of a direct measure of anatomical severity.
    Accident Analysis & Prevention 02/2008; 40(1):182-91. DOI:10.1016/j.aap.2007.05.001 · 1.87 Impact Factor
Show more