Symptom-based outcome measures for asthma: the use of discrete choice methods to assess patient preferences
Introduction
This paper considers the use of discrete choice modelling as a preference-based outcome measurement technique, reporting on an application to assess patient weights over different asthma symptoms.
Valid and reliable outcome measures are required for assessing the extent and severity of illness, as well as judging the effectiveness of care or an intervention. The search for such measures is difficult due not least to contrasting and sometimes conflicting requirements of clinicians and health economists. Clinical outcome measures are needed to judge treatment success and establish good clinical practice and to assist judgements in individual patient clinical decision making. In asthma, the effectiveness of an intervention is often judged using a measure such as expiratory flow rates or airway responsiveness, or a measure of health service utilisation such as attendance at hospital or general practice.
The usefulness of clinical measures of effectiveness such as measures of lung function (FEV, FVC) and peak flow rates is limited not least because they tend to be associated only with the acute episodes of asthma and do not reflect the chronic consequences for the patient. Except during an acute phase, for many patients, and particularly mild asthmatics and those whose asthma is fairly well controlled, the impact on day-to-day quality of life and clinical status from their illness can be fairly small. However, mild asthma is still an important resource consideration since it affects a large group of patients who, as a whole, have high prescription costs.
There is a need to be able to measure patient outcomes in such a way as to incorporate the episodic nature of the illness. Measures reflecting the impact of asthma on the patient's day-to-day life, such as, days lost from work or school, have been used in some studies. Composite, clinically-based, measures which could be useful as single effectiveness indices for cost effectiveness studies and which reflect both acute and chronic disease aspects are the ‘episode free day’ (EFD) and the ‘symptom free day’ (SFD) [1], [2], [3], [4]. However, a potential weakness of existing SFD measures is that equal weights are assigned to each symptom in the SFD calculations. Perhaps the weights are not equal. Some patients may feel that certain symptoms are more important at any given level of severity than others. In this paper we explore the use of discrete choice techniques as a method of identifying preference-based symptom weights. The role of preference-based asthma symptom outcome measures was explored in another study to develop a preference-based outcome measure that reflected patient's views about their symptoms [5]. However, their study used a combination of standard gamble (SG) and visual analogue score (VAS) techniques to measure preferences, methods that have been the subject of some concern. For example, the VAS can be difficult to administer since people can find it difficult to assign numerical values to health while standard gamble techniques, suffer from the sorts of cognitive limitations associated with choice under uncertainty [6], [7].
The study reported in this paper uses discrete choice modelling, based on random utility theory, to test whether or not patients weight symptoms equally. There has been rapid growth in the use of discrete choice techniques, one form of conjoint analysis, as a method of deriving preference-based values in health economics [8], [9], [10]. Much of the earlier work considered process or intermediate outcomes, such as the method of service delivery, although more recently there have been attempts to apply these techniques to health state outcomes. [11], [12].
The data also allowed two important methodological issues in discrete choice modelling to be explored, namely the impact of attribute order and the linearity of the utility function.
Section snippets
Methods
Patient preferences over asthma symptoms were elicited using discrete choice methods. The sample chosen for the study was taken from a group of moderate asthmatics, aged 16 years or over, who were either currently, or had recently been, included in an integrated care scheme between the out patient respiratory clinic in Grampian and the patient's general practitioner.
The method of discrete choice is one form of conjoint analysis, a technique used for establishing the relative importance of
Data
Questionnaires were posted to 272 moderate to severe asthma patients who had attended an asthma out patient clinic in the region. One hundred and seventy four questionnaires were returned with all eight choice questions answered. The 12 respondents who violated the assumption that individuals will not prefer more of a symptom to less were excluded from further analysis. This reduced the sample to 162 respondents and 1296 responses, a usable response rate of 59.6%.
Where responses to the two
Results
The results of the ordered probit regression analysis, using the linear additive model, are reported in Table 5. All of the coefficients had the expected negative sign, and all were statistically significant. A negative coefficient implies that an increase in the asthma symptom would lead to a decrease in the utility difference between week B and week A, other things equal. The β values for cough and breathlessness are considerably higher (in absolute terms) than the coefficients of the other
Discussion
The aim of this paper was to consider the usefulness of discrete choice modelling as a method of preference-based outcome measurement in health care, focussing on a study of patient preferences over asthma symptoms, or the consequences of these symptoms perhaps in terms of activity limitations. The results have implications for the measurement of outcomes in the evaluation of asthma interventions. Firstly, the results make it possible to identify what the most important symptoms are, a
Acknowledgements
HERU is funded by the Chief Scientist Office of the Scottish Executive Health Department. This study received financial support from the NHS R&D Delivery of Care in Asthma Programme. The views expressed in this paper are those of the authors and not necessarily those of the Scottish Executive Health Department. We would like to express our thanks to the patients who took part in this study and to Jackie Fiddes for her help with the questionnaire survey.
References (16)
- et al.
Integrating patient preferences into health outcome assessment
Chest
(1998) - et al.
The episode-free day as a composite measure of effectiveness: an illustrative economic evaluation of formeterol versus salbutamol in asthmatic therapy
PharmacoEconomics
(1993) - et al.
Cost effectiveness of inhaled corticosteroid plus bronchodilator therapy versus bronchodialtor monotherapy in children with asthma
PharmacoEconomics
(1993) - et al.
Costs and effects of inhaled corticosteroids and bronchodilators in asthma and COPD
American Journal of Respiratory Critical Care Medicine
(1995) - et al.
The cost-effectiveness of Budesonide in severe asthmatics aged one to three years
British Journal of Medical Economics
(1993) - et al.
Measuring preferences for health states worse than death
Medical Decision Making
(1994) - et al.
Choices, values and frames
American Journal of Psychology
(1983) Using consumer preferences in health care decision-making: the application of conjoint analysis
Office of Health Economics
(1996)
Cited by (33)
Magnitude of effect of asthma treatments on Asthma Quality of Life Questionnaire and Asthma Control Questionnaire scores: Systematic review and network meta-analysis
2015, Journal of Allergy and Clinical ImmunologyCitation Excerpt :Considerations of adverse effects and cost to patients and society might also be relevant. For subjective outcomes, discrete-choice modeling techniques might assist in this decision.25,26 Our study highlights the importance of clinical study design if the difference between treatments is under study.
Asthma outcomes: Symptoms
2012, Journal of Allergy and Clinical ImmunologyCitation Excerpt :The scale is also ideal as a cost-effective analysis that requires a condition-specific utility index. The ASUI has been used in multiple clinical trials19-39 and appears to be the symptom scale most frequently quoted and used in the literature. The ASUI score is calculated as the product of 5 utility functions, 4 representing symptoms and 1 representing medication side effects.
Understanding medication compliance and persistence from an economics perspective
2008, Value in HealthCapturing the value of complementary and alternative medicine: Including patient preferences in economic evaluation
2008, Complementary Therapies in Medicine