Results

Discussion

The measurement of any psychological concept such as QoL is inherently different from measuring a physical concept such as height, and it may therefore be inevitable that we must live with some limitation in any measure. However, this is not to say that we should give up on measuring QoL. For children, QoL is too important to be disregarded. Further development of measures depends crucially on experience gained in using the measures that are now available. This is relevant not only for refinement of currently available measures, but also to enable the development of more sophisticated measures in the future. For these reasons, it is important to recognise the limitations of currently available measures, while also acknowledging that improvements can only be made when we understand how current measures perform in practice.

Given the current state of the art, we draw on information about the performance characteristics of available measures summarised in tables2 and 3. Based on these data, we conclude that only three11 ,20 ,27 generic measures and two disease specific measures35 ,38 fulfil very basic psychometric criteria. Our own recommendations would be based on these measures and might involve the following.

For work evaluating clinical trials, whether in the context of high technology medicine such as childhood cancer, or in a community setting, there is a need for a brief measure of QoL that can be completed during a regular clinic visit. In order to recruit a large sample of patients, a measure is needed that is simple to administer with minimal training or expertise. The measure needs to include those aspects of functioning that are most likely to be compromised by the treatment protocol. Thus there is a need for measures that focus on physical symptoms and emotional wellbeing. Assessment of school or learning needs to be included especially for children (compared with adults), and if there is any concern about cognitive side effects of the protocol. Given the concern with physical symptoms, it is likely that disease specific measures might be more useful than generic. The Pediatric Quality of Life (PedsQL) and its associated modules for work in oncology, asthma, or diabetes20 is one of the more thoroughly developed measures currently available. In asthma, the measure by Juniper and colleagues35 also has much to recommend it.

The inclusion of QoL data in clinical trials creates new questions about statistical analyses which have not been resolved. The analysis of multivariate QoL data (and the inevitable missing data) poses a very different problem compared with analyses based on univariate outcomes such as survival. Strategies to manage missing data are important, as is the need for hypothesis driven trials.

The choice of measures for evaluation of psychosocial interventions is relatively similar. If the need is for a brief assessment, generic measures such as the PedsQL20 or HUI2 and HUI327 have some merit. However, it is unlikely that either of these will address the full range of functioning that might need to be assessed (and indeed they were not designed to do so). Additional measures will therefore need to be included, depending on the specific purpose of the intervention. Where the goal is to achieve greater school integration or improve family functioning, the Child Health Questionnaire11 may be more appropriate. However, advantages of this need to be set against the length of the currently available measure (though shorter forms are in the process of development).

There are also measures developed for specific purposes, such as the BASES36 for work involving children undergoing bone marrow transplantation. This does fulfil the basic criteria we identified, and has potential use in evaluating interventions involving children undergoing bone marrow transplantation. It is clear that there are many other specific contexts in paediatrics where QoL measures may be desirable (for example, palliative care), but no measure is currently available.

Our review highlights many inconsistencies and problems associated with measurement of QoL in children. These include the following.

(1) Confusion about the definition and measurement of QoL—This is reflected in the overlap between measures of QoL and health or functional status, and the variability in definition and number of domains assessed. This variability means that there may be little relation between QoL as assessed by different measures. There is an urgent need to determine how far currently available measures of QoL really assess the same underlying construct.

(2) Limited availability of disease specific measures—To the extent that generic measures are suitable to assess QoL regardless of the child's specific condition, such measures are assumed to be preferable when decisions need to be made regarding allocation of resources from public health perspectives. In contrast, disease specific measures are assumed to have merit when assessing the impact of a change in treatment, or when assessing outcomes in clinical trials. Among disease specific measures, asthma, cancer, and epilepsy have received most attention. For children with many other conditions it is only possible to rate QoL using a generic measure.

In practice, decisions about generic or disease specific measures may be less simple, given the limited number of measures available. Disease specific measures are inappropriate where a child has more than one condition. Furthermore, the low incidence of some conditions will preclude development of disease specific measures. There is also a need to understand the relation between generic and disease specific QoL. Development of a core generic instrument supplemented by disease specific modules may be one solution. This allows for direct comparison between illness samples, and additional information to be obtained concerning specific disease. Such an approach is central to the generic and module approach advocated by Varni and colleagues.20

(3) Discrepancies between child and parent ratings—We need to accept that both child and proxy ratings have value. The question is to clarify how differences in perception of QoL arise between child and proxy and the implications for the child's QoL. This applies as much to clinicians as parents, teachers, and other proxies. Parents may be influenced by the development of other children they know (their own or those of friends), their expectations and hopes for their child, additional life stresses, and their own mental health. It is important to clarify how parent mental health and their perceptions of the disease influence the child's QoL over time. This is relevant to issues concerning how parenting practices and family organisation can subsequently effect the child's QoL.

(4) Limited availability of measures for self completion by children—Measures are typically targeted at children across a broad age range, with very few measures available for those below 8 years. Based on findings that children and parents differ in their understanding of illness and treatment,7 there is a widely endorsed view that children should rate their own QoL wherever possible.38 They have different views about illness. Furthermore, parents' views about their child's QoL may be influenced by their own mental health and concerns about the child's illness. Despite this, many measures rely exclusively on parent report. A limited number of measures provide parallel forms for completion by both child and parent. These may be the measures of choice in situations where children are well and able to rate their own QoL.

Techniques need to be developed to enable self ratings to be obtained routinely from children, especially those below 8 years of age. In addition, given differences between children and parents, basic research is needed to identify situations where parents are able to respond for their children.

(5) Lack of precision regarding the content of domains of QoL—Most developers of scales define QoL as a multidimensional construct, and attempt to assess domains including physical, social, and emotional QoL. Other domains (for example, cognitive or spiritual) are less often assessed. In addition, the precise content of these domains varies considerably in emphasis and generality. In measuring physical QoL, the emphasis may be on physical symptoms, self care, participation in physical activities, or distress caused by limitations in physical activities. There is even greater variability in content of social domains.

(6) Cultural appropriateness of measures for use in the UK—Many measures have been developed outside the UK, which may prove unacceptable to British children, given cultural differences in the meaning of illness, relationships between parents and children, and organisation of health care services. Consideration also needs to be given to the language used. (Questions about “difficulties walking one block” mean little to children in the UK.)

Other issues may be even more critical. In the cancer specific QoL measure described by Varni and colleagues38 for example, a number of questions ask children to report their concerns about relapse. Inclusion of such direct questions (or even use of the term “cancer”) may be unacceptable to some paediatricians and families in the UK. Translating a QoL instrument for use in different countries may appear a cheap and satisfactory option, but in fact requires extensive work to establish true comparability.57

There is no doubt that much needs to be done to improve the quality of QoL measures, and hence the status of this work in clinical practice and research. However, the focus on QoL has done much to raise the profile of children's views about treatment and organisation of care. Recognition of the shortcomings of currently available measures must not be used as a reason to ignore QoL issues. At the least, attention to QoL has emphasised the need to consider the outcomes of paediatric medicine in terms of the whole child rather than focus on a narrow range of clinical indicators.