29.06.2013 Views

Evaluating Patient-Based Outcome Measures - NIHR Health ...

Evaluating Patient-Based Outcome Measures - NIHR Health ...

Evaluating Patient-Based Outcome Measures - NIHR Health ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

cases questionnaire items may ask for reports<br />

of very specific behaviours, for example, ability to<br />

walk a certain distance, or use of physical aids, that<br />

observers such as carers or therapists can in principle<br />

readily verify from observation. Even with such<br />

behavioural items, the questionnaire still largely<br />

elicits perceptual information. It is this reporting of<br />

the personal and the subjective by the patient that<br />

uniquely identifies patient-based outcome measures<br />

from other health information used as outcomes,<br />

such as laboratory data. Clinical scores and scales are<br />

a different kind of subjective perceptual evidence;<br />

they are the perceptual judgement of doctors or<br />

of other health professionals. It is the inherently<br />

subjective source of patient-based material that<br />

leaves grounds for anxiety in some minds about the<br />

‘hardness’, robustness and ultimately scientific value<br />

of such evidence (Fries, 1983; Deyo, 1991). Such<br />

concerns are addressed when we consider desirable<br />

measurement properties of patient-based measures<br />

in chapter 3.<br />

Dimensions such as ‘QoL’ and ‘subjective health<br />

status’ can be assessed by what may be considered a<br />

continuum of methods. At one extreme, health<br />

professionals or others make judgements with<br />

minimal input from the patient, and, at the other<br />

extreme, assessments are largely determined by the<br />

patient with minimal influence from other observers.<br />

This review is largely concerned with instruments<br />

of the latter kind because there is a prima facie<br />

case that such measures more directly elicit the<br />

respondent’s perspective rather than the observer’s<br />

(O’Brien and Francis, 1988; Rothman et al., 1991;<br />

Berkanovic et al., 1995). However there is a continuum<br />

of approaches and much of what is discussed<br />

in this review may be relevant to assessments such as<br />

disability scales or standardised psychiatric assessments<br />

which are completed by observers on the<br />

basis of evidence from a patient, but without the<br />

patient himself or herself literally selecting the<br />

items or description that most fit their view.<br />

There are circumstances where patients are unable<br />

to provide their unique report of their perceptions,<br />

due to ill-health, physical or cognitive problems, or<br />

some other incapacity. In these cases, proxy reports<br />

may be necessary because of the need for some<br />

assessment to inform a clinical trial. Because there<br />

is consistent evidence of lack of agreement with<br />

patients’ judgements of their QoL by observers<br />

such as health professionals, informal carers, and<br />

other so-called ‘proxy’ judges, this is increasingly<br />

considered a second-best solution to be used only<br />

when the patient cannot contribute (Mosteller et<br />

al., 1989; Clarke and Fries, 1992; Sprangers and<br />

Aaronson, 1992). However, there<br />

<strong>Health</strong> Technology Assessment 1998; Vol. 2: No. 14<br />

is also substantial evidence that patients with<br />

poorer health are less likely to complete patientbased<br />

outcome measures (Bleehen et al., 1993).<br />

Since such patients are an important group in<br />

relation to assessment of outcomes in trials and<br />

their omission may result in bias, effort is required<br />

to examine the extent to which proxy ratings of<br />

outcome are valid. Whilst there is clear evidence<br />

of discrepancies in judgements between patients’<br />

and proxy reports from others, it is important to<br />

examine closely the scope for obtaining proxy<br />

reports when patients’ are unable to contribute.<br />

Sneeuw and colleagues (1997) used a simple QoL<br />

instrument with relatively few distinctions between<br />

levels of QoL with patients with a range of cancer<br />

diagnoses, and their informal carers and physicians<br />

also rated the sample of patients on the same<br />

instrument. For five out of six dimensions, there<br />

was broad agreement between patient and physician<br />

or carer in 85% of patients, and 75% agreement<br />

on a sixth dimension (‘social activities’).<br />

Agreement also increased for some dimensions<br />

at a follow-up assessment. Thus there is some<br />

support for using evidence from proxies at least<br />

when relatively simple judgements are required.<br />

Theories and concepts<br />

It is sometimes argued that this field lacks a<br />

rigorous underpinning theory and clear and<br />

precise definitions that flow from theory (Schipper<br />

and Clinch, 1988; Ventegodt, 1996). There is some<br />

basis for this criticism; much of the work stems<br />

from very applied and pragmatic problem solving,<br />

rather than deriving from an explicit theoretical<br />

framework. However it is not entirely true that the<br />

field of patient-based outcome measures lacks<br />

theories of the phenomena that investigators wish<br />

to measure. Psychometric theory provides a well<br />

established foundation for most patient-based<br />

outcome measures (Nunnally and Bernstein, 1994).<br />

This scientifically rigorous field is concerned with<br />

the science of assessing the measurement characteristics<br />

of scales and involves such properties as<br />

validity, reliability and responsiveness (Hays et al.,<br />

1993). More recently developed is the field of<br />

‘clinimetrics’ (Feinstein, 1987; Feinstein, 1992;<br />

Wright and Feinstein, 1992). Closely associated<br />

with psychometric theory, this field focuses on the<br />

clinical challenges of constructing scales that<br />

clinicians use for measuring health status of<br />

patients (Fava, 1990). Similarly, economic contributions<br />

to this field have also a broad range of<br />

theoretical literature on which to draw from<br />

evidence such as decision-theory (Torrance, 1986).<br />

Thus strictly speaking, a far greater difficulty<br />

than the dearth of theory is that there are a large<br />

5

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!