Adaptation and scale of reference bias in self-assessments of quality of life

https://doi.org/10.1016/S0167-6296(99)00037-5Get rights and content

Abstract

Adaptation behaviour and different scales of reference can bias self-assessments of well-being by individuals. In this paper, we analyse the impact of these biases on a subjective measure of the quality of health and on the QALY weights derived from this health measure. It is found that the scale of reference of the subjective health measure changes with age. Accounting for adaptation and scale of reference bias lowers most of the QALY weights for health problems and disabilities.

Introduction

It is frequently observed that people have a remarkable ability to adapt to discomfort and illness. Chronically ill patients generally report levels of quality of life that are much higher than one would expect given their condition. For example, in Adang (1997), type I diabetes and end-stage renal disease patients are asked to rate their quality of life on a 10-point scale before and after they receive a combined pancreas–kidney transplantation. These are people who are seriously ill, suffer a great deal of discomfort and are highly restricted in their activities. Before the transplantation, the average quality of life is rated at 5.5. After a successful transplantation this increases to 7. If, however, after the transplantation these patients are asked retrospectively about their pre-transplant quality of life, this is rated at 3.3. What this indicates is that patients highly adapt to their situation. This is just one example that shows that patients adapt to their situation (for other examples, see Adang, 1997and Heyink, 1993).

Adaptation is defined by Heyink (1993)as “…an intrapsychic process in which past, present, and future situations and circumstances are given such cognitive and emotional meaning that an acceptable level of well-being is achieved” (Heyink, 1993, p. 1332).1 Adaptation is used by people to recover psychologically from a setback. Adaptation is related to coping behaviour — i.e., behaviour aimed at reducing or eliminating psychological distress. The difference between coping and adaptation is that coping theory does not predict whether the outcome of the process will be positive or negative, while adaptation implies recovery and thus a positive outcome of the process (Heyink, 1993). In the adaptation model of well-being of Chamberlain and Zika (1992)it is argued that well-being is stable in the absence of situational change (such as the occurrence of an illness or disability), sensitive to change when situational change occurs, and adaptive to occurrence of change.

Adaptation may not be the only explanation for the frequent finding that people with health problems or handicaps report higher levels of well-being than expected. An alternative explanation might be that questions on well-being are answered relative to a certain reference group. If you are asked about your well-being you do so by comparing yourself to people in a similar situation as yourself (friends, relative, people of your age, people with similar education, etc.), or by comparing the sources of your well-being with those of others (people with more health problems than yourself, people without a job, people with a low income, etc.). The finding that patients with diabetes report relatively high levels of well-being may be because they compare themselves to other patients in a similar physical condition, or they may compare themselves to the situation they had expected themselves to be in at that stage of their disease. In short, questions on subjective well-being do not have the same meaning to everyone. Subjective measure of well-being do not have a natural reference point. Rather, the reference point of well-being is determined by individual specific situations and characteristics.

Empirically, it may not be possible to distinguish between adaptation and differences in the scale of reference. In a sense, adaptation may be a specific form of changes in the scale of reference. Once you have health problems or a disability your point of reference changes to adapt to the new situation.

As shown by the example of type I diabetes and end-stage renal disease patients, adaptation behaviour and different scales of reference between individuals may bias the answers to survey questions on subjective well-being or subjective quality of life. Subjective measures of the quality of life are an essential part of Quality Adjusted Life Years Studies (QALYs). Subjective measures of the quality of life are used as QALY weights to diseases. Three different ways to measure QALY weights can be distinguished.

The first is to ask patients to judge their quality of life. Either patients can be asked to evaluate their quality of life before and after a medical intervention (as was done in Adang, 1997), or the quality of life of patients can be compared to the quality of life of healthy people, or patients can be asked to judge their quality of life with the disease and in the hypothetical situation that they are healthy.

The second method is to ask informed experts — for example doctors — about the quality of life of people with a certain disease.

The third method is based on questioning a random sample of the entire population. They are either asked about the quality of their own life and about their health status, and then the quality of life of people with health impairments can be compared with that of healthy people. Or, they are asked hypothetically about the quality of their life with a certain disease or handicap. A difference with the other methods is that in this method QALY are usually defined in terms of diseases reported by the population, rather than in terms of diseases diagnosed by a medical expert.

Of these three methods, the second one can be thought to be free from adaptation, as it is unlikely that informed experts adapt to the diseases of the people they are asked to evaluate. However, almost by definition, this method suffers from problems in the scale of reference. The scale of reference of an informed expert is not identical to that of the patient. Further, if medical experts base their evaluation of the quality of life of patients on their own experience with these patients, adaptation may bias the responses of medical experts through the adaptation of patients to their situation.

There are other objections to this method also that make that this method is unsatisfactory. Little is known, for example, about how the opinions of medical experts differ from those of the patients and from non-medical experts more generally. Also, medical experts may have interests of their own that can bias the results. If medical experts, for example, have an interest in a positive outcome of the QALY analysis of a medical treatment (because they want insurance companies to fund the treatment), they may have an incentive to underestimate the quality of life of patients before the treatment and to overestimate the quality of life after the medical intervention.

The US Panel on Cost-Effectiveness in Health and Medicine that was set up to provide guidelines for QALY research favours using the third method (see Gold et al., 1996, Cutler and Richardson, 1998). The Panel recommends using community based weights for the quality of life that should be collected from a representative sample of the general population.

Recently, Cutler and Richardson (1997), Cutler and Richardson (1998)have used information on self-reported quality of health measure from a representative sample of the US population to calculate QALY weights of a wide range of different diseases.2 Cutler and Richardson (1997), Cutler and Richardson (1998)use these QALY weights to determine the increase in the value of health between 1970 and 1990. They calculate that health improved by US$100,000–US$200,000 per person between 1970 and 1990. This increase in the value of health is the outcome of two opposing trends. On the one hand longevity has improved. On the other hand, the prevalence of most diseases and handicaps is increasing. The reason for these opposing trends is that more and more diseases that in the past were fatal are not so anymore. This has increased life expectancy. Instead, many of these formerly fatal diseases have become chronic conditions. This is why the prevalence of these diseases has increased. Because of the increase in the prevalence of these diseases and handicaps, more people now have to adapt to limitations on their health condition. Adaptation has become more widespread as the prevalence of the diseases has increased. The wider prevalence of adaptation among people with diseases or handicaps will overestimate the increase in the value of health. This is noted by Cutler and Richardson (1998)when they discuss that the quality of life for each disease or handicap has improved over time and that people report themselves in less worse health than they did in the past (Cutler and Richardson, 1998, Cutler and Richardson, 1998, p. 98).

The results in Cutler and Richardson (1997), Cutler and Richardson (1998)probably also suffer from scale of reference bias. This is clear from their finding that the QALY weight for women in 1980 increases at very old age. They ascribe this finding to age-norming. Respondents are explicitly asked to take people of their own age into account when answering the subjective health question (“How would you rate your health as compared with individuals your age?”). In this way, older respondents are invited to use a different scale of reference than younger people.

Unfortunately, the psychological literature does not provide a formal treatment of the notions of adaptation and scale of reference bias. Here, in this paper, we assume that adaptation and scale of reference bias occur in the transformation of the `true' health state into the `reported' health state. We assume that people who are in same health states can perceive their situation differently — depending on the reference group with which they compare themselves — and that these differences in perception may lead to differences in the response to a question on the assessment of their health status. So, a health condition that is assessed as `good' by one person, can be perceived as only fair by another. If we observe that the same health status gives rise to different answers on a quality of health question, we may also observe the opposite, i.e., that people in different health states give similar assessments of their health situation. For example, someone with an objective handicap or disease may have adapted to the situation and evaluate his/her quality of life at the same rate as someone without this handicap or disease.

So the first condition under which the procedure proposed in this paper is appropriate is that scale of reference bias or adaptation occurs in the translation of the true health status into the response to a question on the evaluation of this health status. The second condition is that the discrepancy between the `true' and the reported health status is related to observable characteristics such as age, gender and the prevalence of health problems or disabilities. A limitation of this procedure is that it does not correct for adaptation and scale of reference bias generated by unobservable or individual specific characteristics. To correct for this would call for a longitudinal approach.

In this paper, we use these notions to formulate a method to purge self-reported quality of life measures from adaptation and scale of reference bias. In this method, we allow for the underlying distribution of health to depend on the health status and on other individual characteristics of individuals.

This method of correcting for adaptation and scale of reference bias is particularly useful for studies in which QALY weights are derived from a cross-section of people with and without health problems and disabilities, i.e., from national samples on the quality of life and health status of the population. In this paper, we use data from a large longitudinal sample of the British Population, the British Household Panel Survey 1995. QALY weights are derived from the following self-assessment of the overall-health quality of life question that was included in the survey: “Compared to people of your own age, would you say that your health has on the whole been: excellent, good, fair, poor or very poor?”. This quality of life variable is identical to the one used by Cutler and Richardson, 1997, Cutler and Richardson, 1998in their analysis of QALY. Like them, we use a representative sample of the population to derive QALY weights from the effects of the actual health status of an individual on his/her subjective health.

One further advantage of using self-assessments of health for calculating QALY weights is that the cognitive burden on respondents is lower than with other techniques, such as the standard gamble and the time trade-off methods. In Fryback et al. (1993), it is shown that the scores on this self-assessed overall quality of health correlate highly with the scores of other quality of life indicators that are frequently used in QALY analysis, such as the time trade-off assessment, the quality of well-being index and the outcomes of a general health perception questionnaire.

The outline of this paper is as follows. In Section 2, we describe the empirical model to derive QALY weights from self-assessments of health status. We further show how we can correct QALY weights for adaptation and scale of reference bias. The data used in the empirical analysis are described in Section 3. The results are presented in Section 4. Section 5concludes.

Section snippets

The quality of health model

In modelling the quality of life, we follow the same procedure as Cutler and Richardson, 1997, Cutler and Richardson, 1998. We build on their model by allowing for preference drift and adaptation to the health condition in the self-reported quality of health measure.

The starting point of the model is the concept of“health capital”as introduced by Grossman (1972). The value of Health Capital (HC) at year t is defined as:HCt=Vk=0Et[Ht+k](1+r)kwhere V is the value of a year in perfect health, r

Data and descriptive analysis

The data are taken from the 1995 wave of the British Household Panel Survey (BHPS, 1995). Details about this survey can be found in Taylor (1992). The sample includes all individuals over the age of 15. After eliminating a small number of observations with missing values on the self-reported health status and on the health condition variables, 9462 observations could be used in the analysis. As was already mentioned, the subjective health measure Hs is defined by the response to the survey

Estimation results

The parameter estimates of the ordered probit model are found in Table 2. As the subjective health measure runs from excellent to very poor a positive sign of the coefficient indicates that the variable lowers the health status, while a negative coefficient indicates that an increase in that variable will improve health. All objective health problems and disabilities lower the health status significantly. If we look at the effects of the other control variable, then it is found that years of

Conclusion

Through their effects on QALY weights, adaptation and scale of reference bias may affect conclusions about the cost-effectiveness of medical interventions (see Adang, 1997for an example). In this paper, we have shown how we can correct for adaptation and scale of reference bias in self-assessments of well-being of individuals. It was found that if people are asked about the quality of their health status compared to others of their age, the response suffers from age-norming: older people appear

Acknowledgements

I would like to thank two anonymous referees for helpful comments on a previous draft of this paper.

References (21)

  • T Chirikos et al.

    Economic determinants and consequences of self-reported work disability

    Journal of Health Economics

    (1984)
  • V Douzdjian et al.

    Treatment strategies for insulin-dependent diabetics with ESRD: a cost-effectiveness decision analysis model

    Am. J. Kidney Dis.

    (1998)
  • R Haveman et al.

    Market work, wages and mens health

    Journal of Health Economics

    (1994)
  • Adang, E., 1997. Medical technology assessment in surgery: costs and effects of dynamic graciloplasty and combined...
  • R Alessie et al.

    Habit formation, interdependent preferences and demographic effects in the almost ideal demand system

    Economic Journal

    (1991)
  • K Anderson et al.

    The retirement-health nexus: a new measure of an old puzzle

    Journal of Human Resources

    (1985)
  • Bazzoli, G., 1985. The early retirement decision: a comparison of expectations and realizations. In: Wise, D. (Ed.),...
  • J Bound

    Self-reported versus objective measures of health in retirement models

    Journal of Human Resources

    (1991)
  • K Chamberlain et al.

    Stability and change in subjective well-being over short time periods

    Social Indicators Research

    (1992)
  • Cutler, D., Richardson, E., 1997. Measuring the health of the United States Population. Brookings Papers on Economic...
There are more references available in the full text version of this article.

Cited by (232)

  • Inequalities in utilization of institutional care among older people in Estonia

    2022, Health Policy
    Citation Excerpt :

    To account for health deprivation adequately, we include into our models self-assessed general health. Although this indicator may be prone to various biases, including non-random measurement error [35,36] and reporting heterogeneity [37–40], it is widely used as a predictor of health care service utilization [41–44], because it captures elements of health that more guided and objective measures cannot [45]. A memory variable included in the models serves as a proxy for cognitive abilities: we use the delayed recall indicator that takes a value from 0 to 10 depending on the number of words a person recalled out of 10 listed by an interviewer several minutes ago.

  • Health Technology Assessment With Diminishing Returns to Health: The Generalized Risk-Adjusted Cost-Effectiveness (GRACE) Approach

    2021, Value in Health
    Citation Excerpt :

    These issues are exacerbated by the phenomenon of adaptation. People with a particular disability often rate it as less costly than would nondisabled persons (cf. 24). Thus, health ratings based on general population surveys may understate QoL for disabled people.

View all citing articles on Scopus
View full text