a review of disease-specific quality of life measurement ... · pdf filea review of...

A REVIEW OF

DISEASE-SPECIFIC

QUALITY OF LIFE

MEASUREMENT

SCALES

M E A S U R I N G D I S E A S E

Second Edition

Ann Bowling

Open University PressBuckingham · Philadelphia

Open University PressCeltic Court22 BallmoorBuckinghamMK18 1XW

email: [email protected] wide web: www.openup.co.uk

and325 Chestnut StreetPhiladelphia, PA 19106, USA

First published 1995Reprinted 1996, 1997, 1998

First published in this second edition 2001

Copyright © A. Bowling 2001

All rights reserved. Except for the quotation of shortpassages for the purpose of criticism and review, nopart of this publication may be reproduced, stored ina retrieval system, or transmitted, in any form or byany means, electronic, mechanical, photocopying,recording or otherwise, without the prior writtenpermission of the publisher or a licence from theCopyright Licensing Agency Limited. Details of suchlicences (for reprographic reproduction) may beobtained from the Copyright Licensing Agency Ltd of90 Tottenham Court Road, London, W1P 0LP.

A catalogue record of this book is available from theBritish Library

Library of Congress Cataloging-in-Publication DataBowling, Ann.

Measuring disease : a review of disease-specificquality of life measurement scales / Ann Bowling. –2nd ed. p. cm.

Includes bibliographical references and index.ISBN 0–335–20642–5 (hb) – ISBN 0–335–20641–7

(pbk.)1. Health status indicators 2. Quality of life.

I. Title.RA407.B68 2001613–dc21

00–060647

Typeset by Graphicraft Limited, Hong KongPrinted in Great Britain by St Edmundsbury Press Limited,Bury St Edmunds, Suffolk

CONTENTS

PREFACE xiii

ACKNOWLEDGEMENTS xvi

LIST OF ABBREVIATIONS xvii

1 HEALTH-RELATED QUALITY OF LIFE: CONCEPTUAL MEANING,

USE AND MEASUREMENT 1

Background: the ‘quality of life’ 1What is quality of life and health-related quality of life? 2Health-related quality of life and health 7Quality of life assessment and research on health care outcomes 9Utility assessments and quality of life assessment 11Who should rate quality of life? 13Patients’ preferences 14Disease-specific, domain-specific and generic measurement scales 16Methodological issues 19

2 CANCERS 23

The importance of measuring the quality of life of cancer patients 23What has been measured in studies of the quality of life of cancer patients? 23What should be measured? 24Recommended measurement scales 24Review articles on quality of life measurement and cancer 25The domains of measurement 25The measurement scales 30

Scales measuring physical functioning, pain and symptoms 30Functional status scales used with cancer patients 30The Karnofsky Performance Index (KPI) 30Other versions of the Karnofsky Scale 33The WHO Functional Scale 33The Zubrod Scale (or the Eastern Cooperative Oncology Group Performance Status

Rating Scale: ECOG PSR) 33Pain scales 34The McGill Pain Questionnaire (MPQ) 34Other symptom scales 35Symptom Distress Scale 35Lasry Sexual Functioning Scale for Breast Cancer Patients 36Summary of other symptom scales 36McGill Nausea Questionnaire (MNQ) 36The WHO Symptom Checklist 37Medical Research Council (MRC) UK Scale 37The Qualitator 37Disease-specific quality of life scales 37EORTC Modular Approach 37Rotterdam Symptom Checklist 42Functional Living Index – Cancer (FLIC) 44Functional Living Index – Emesis (FLIE) 46Cancer Inventory of Problem Situations (CIPS) and the Cancer Rehabilitation Evaluation

System (CARES) 47Visual Analogue Scales (VASs) 49Spitzer Quality of Life (QL) Index 49Linear Analogue Self-Assessment (LASA) Scale 52Ontario Cancer Institute/Royal Marsden LASA Scale 54Padilla Quality of Life (QL) Scale and variants, including the Multidimensional Quality

of Life Scale – Cancer (MQOLS-CA) 55Holmes and Dickerson 57Global Quality of Life Scale (Coates) 57Quality of Life Index 58VAS for bone marrow transplant patients 59Other scales 59Functional assessment of cancer therapy – Lung Quality of Life Instrument (FACT–L) 61Breast Cancer Chemotherapy Questionnaire (BCCQ) 60European Neuroblastoma Study Group Quality of Life Assessment Form – Children

(QLAF–C) 61Brief details of other scales 62Cancer Leukaemia Group B Studies (CALGB) 62Ability Index 62Burge Quality of Life Severity Scale 63Anamnestic Comparative Self-Assessment (ACSA) 63World Health Organization Quality of Life Assessment Instrument (WHOQOL) 63Time Without Symptoms and Toxicity (TWiST) 64Palliative care 64Support Team Assessment Schedule (STAS) 65McGill Quality of Life Questionnaire (MQOL) 66Conclusion 68

vi MEASURING DISEASE

3 PSYCHIATRIC CONDITIONS AND PSYCHOLOGICAL MORBIDITY 69

Quality of life research and psychiatric and psychological morbidity 69Measuring quality of life in institutions 69Community indicators of quality of life 70Measuring quality of life in individuals: the domains of measurement 71Subjective accounts and the value of patients’ ratings 74

THE MEASUREMENT SCALES 74

Part I Symptom scales 74Background 74Symptom questionnaires 78Hopkins Symptom Checklist (SCL) and the Symptom Checklist-90 78Brief Symptom Inventory (BSI) 79Symptom Rating Test (SRT) 80Center for Epidemiological Studies Depression Scale (CES-D) 81Rand Depression Screener 82Hospital Anxiety and Depression Scale (HADS) 84Goldberg’s General Health Questionnaire (GHQ) 85Beck Depression Inventory (BDI) 87Hamilton Rating Scale for Depression 89Zung Self-rating Depression Scale 91Quality of Life in Depression Scale (QLDS) 92State–Trait Anxiety Inventory (STAI) 93Profile of Mood States (POMS) 95Scales for children 97Conclusion to Part I 98

Part II Quality of life instruments 98Condition-specific quality of life scales 98Quality of Life Scale (QLS) 98Lehman Quality of Life Interview 100Variations 102Lancashire Quality of Life Profile 103SmithKline Beecham Quality of Life Scale (SBQOL) 105Oregon Quality of Life Self-Report Questionnaire and semi-structured interview

rating version 106Wisconsin Quality of Life Index for Mental health (W-QLI) 108Index of Health-Related Quality of Life 111Health of the Nation Outcome Scales for Severe Mental Illness (HONOS–SMI) 113Team for the Assessment of Psychiatric Services (TAPS) Measures 114General Well-Being Schedule (GWBS) 117Scales for assessing needs 118Key Informant Survey Scales 119MRC Needs for Care Assessment 120Camberwell Assessment of Need (CAN) 123Services Needed, Available, Planned, Offered, Rendered Instrument 125Patient attitude scales 126

CONTENTS vii

Social Problem Questionnaire 126Client Satisfaction Questionnaire (CSQ) 127Satisfaction with Life Domains Scale 129Quality of Life Enjoyment and Satisfaction Questionnaire (Q-LES-Q) 129Conclusion to Part II 130

Part III Role functioning and related instruments 131Role functioning, performance and social behaviour scales 131Social Functioning Schedule 132Social Role Performance Schedule (SRPS) 133Social Behaviour Schedule (SBS) 134Rehabilitation Evaluation (REHAB) 136Social Behaviour Assessment Schedule (SBAS) 137WHO Disability Assessment Schedules (WHO-DAS) – Psychiatric 138Global Assessment Scale (GAS) 140Global Assessment of Functioning (GAF) Scale 141Adjustment and adaptation scales for use with psychiatrically and/or physically ill people 141Social Dysfunction Rating Scale 142Structured and Scaled Interview to Assess Maladjustment (SSIAM) 142Vineland Adaptive Behaviour Scale 144Social Adjustment Scale-II (SAS-II) and Self-Report Version 146Katz Adjustment Scales 147Psychosocial Adjustment to Illness Scale (PAIS) 148Global Adjustment to Illness Scale (GAIS) 149Acceptance of Illness Scale 150Reintegration to Normal Living (RNL) Index 150Stress and coping scales for use with psychiatrically and/or physically ill people 151Ways of Coping Scale 153Health and Daily Living Form 155Stress in Life Coping Scale 156COPE 157Control over life 158Rotter’s Internal–External Locus of Control Scale 159Multidimensional Health Locus of Control Scales 160Generalized Self-Efficiency Scale (GSES) 161Conclusion to Part III 162

4 RESPIRATORY CONDITIONS 163

Quality of life in respiratory disease sufferers 163The domains of measurement 163Measurement scales 168Symptom-specific scales: dyspnoea 168The Fletcher Scale and the MRC Dyspnoea Grade and Respiratory Symptoms

Questionnaire 169American Thoracic Society Respiratory Questionnaire and Grade of Breathlessness

Scale 171

viii MEASURING DISEASE

Horsley Respiratory Symptoms Questionnaire 171American Lung Association Severity of Disability (COAD) 172Feinstein’s Index of Dyspnoea 172Severity of Symptoms Visual Analogue Scale (VAS) 173Oxygen-Cost Diagram 173The 6- and 12-Minute Walking Tests and Stair Climbing 174Borg Ratio of Perceived Exertion 174Mahler Baseline and Transition Dyspnoea Index 175Disease-specific quality of life measures 176Quality of life assessment in COPD/COAD patients: disease-specific scales 176Guyatt’s McMaster Chronic Respiratory Questionnaire (CRQ) 176St George’s Respiratory Questionnaire (SGRQ) 179Chronic Disease Assessment Tool (CDAT) 180Breathing Problems Questionnaire (BPQ) 180Quality of life assessment in asthma patients: disease-specific scales 182Living with Asthma Questionnaire 182Outcome Measures in Ambulatory Care (Asthma and Diabetes) (OMAC) 183Asthma Quality of Life Questionnaire 184Asthma Symptom Checklist 186Summary of other scales which await full psychometric testing 187Asthma Self-efficacy Scale 187Attitudes to Asthma Scale 187Life Activities Questionnaire for Adult Asthma (LAQAA) 188Conclusion 188

5 NEUROLOGICAL CONDITIONS 189

Measuring health-related quality of life in neurology 189Cerebrovascular accident (CVA/stroke) 189The domains of measurement 190Measurement scales 193Stroke scales 193Rankin Handicap Scale 193National Institute of Health (NIH) Stroke Scale 194Canadian Neurological Scale (CNS) 195Hemispheric Stroke Scale 195Disease-specific measures of quality of life: functional ability 196Barthel Index 196Modifications of the Barthel Index 198Granger’s Modified Barthel Index 198Functional Independence Measure (FIM) and Functional Assessment Measure (FAM) 198Gompertz’s Modified Barthel Index 199Nottingham Extended Activities of Daily Living Questionnaire 199Frenchay Activities Index 200Rivermead Mobility Index (RMI) 201Disease-specific measures of quality of life: stroke 202Kudos Battery for Stroke Patients 202London Handicap Scale 202

CONTENTS ix

Epilepsies 203The domains of measurement 204What instruments have been used? 204Measurement scales 207A severity measurement scale 207Seizure Severity Scale 207Disease-specific quality of life scales: epilepsy 208Washington Psychosocial Seizure Inventory (WPSI) 208Epilepsy Surgery Inventory (ESI-55) 209Katz Adjustment Scales (revised for use with epilepsy) 211Health-Related Quality of Life Model 212General tests of neurophysiological functioning 212Trail-Making Test 213Wechsler Scales 213Wechsler Memory Scale (WMS) 213Wechsler Adult Intelligence Scale (WAIS) 214Scales for children 215Tests for cognitive impairment in elderly people 216Abbreviated Mental Test Score (AMTS) 216Mini Mental State Examination (MMSE) 217Mental Status Questionnaire (MSQ) 218Older Americans’ Resources and Services Schedule (OARS) mental health measures 219Brief Cognitive Functioning Scale (SIP sub-scale) 220Geriatric Mental State (GMS) and Comprehensive Assessment and Referral Evaluation

(CARE) 221Conclusion 222

6 RHEUMATOLOGICAL CONDITIONS 224

Quality of life in people with joint disorders 224The domains of measurement 224The measurement scales 229Disease severity and functional classifications 229Disease-specific scales: physical functioning 230Stanford Arthritis Center Health Assessment Questionnaire (HAQ) 230Arthritis Impact Measurement Scales (AIMS) 233Functional Status Index (FSI) 235Western Ontario and McMaster Universities Arthritis Index (WOMAC) 237Office of Population Censuses and Surveys (OPCS) Disability Scale 238Rheumatoid Arthritis (RA) Impact on the Homemaker Questionnaire 240Broader disease-specific quality of life scales 241MACTAR (McMaster–Toronto Arthritis) Patient Function Preference Questionnaire 241Hörnquist’s Quality of Life Status and Change 242Rand Joint Problems Battery 243Summaries of other scales 245Arthritis Helplessness Index (AHI) 245EURIDISS (European Research on Incapacitating Diseases and Social Support) 246Conclusion 246

x MEASURING DISEASE

7 CARDIOVASCULAR DISEASES 247

Heart disease 247Trends in measuring quality of life in people with heart disease 247Most and least frequently used indicators 248What should be measured? 248The domains of measurement 249The measurement scales 255Categorizations and measures of physical activity 255New York Heart Association (NYHA) Functional Classification Scale 255Olsson Ranking Scale 256Canadian Cardiovascular Society (CCS) Functional Classification for Angina Pectoris 256Specific Activity Scale (SAS) 257Rankin Handicap Scale 258Symptom questionnaires 258Fletcher Questionnaire and the Rose (WHO) Questionnaires: The London School of

Hygiene Dyspnoea Questionnaire and Cardiovascular Questionnaire (angina of effort) 258Feinstein’s Index of Dyspnoea 261Ratio Property Scale 261Disease-specific quality of life scales: heart disease 261Rand Congestive Heart Failure (Shortness of Breath/Enlarged Heart/Heart Failure)

Battery and Rand Chest Pain (Angina) Battery 261Guyatt’s Chronic Heart Failure Questionnaire (CHQ) 263Seattle Angina Questionnaire (SAQ) 264Minnesota Living with Heart Failure (LIhFE) Questionnaire 266Quality of Life in heart failure Questionnaire (QLQ-CHF) 266Ferrans and Powers’ Quality of Life Index (CARDIAC) 267Rehabilitation Questionnaire 268Circulatory disease 268Trends in measuring quality of life in people with circulatory disease: hypertension 268What should be measured? 269Disease-specific quality of life scales: hypertension 270Palmer’s Symptom Checklist and Battery 270Bulpitt’s Hypertension Questionnaire and Batteries 270Quality of Life Impairment Scale – Hypertension 272Rand Health Insurance Study (HIS) Blood Pressure Battery 272Conclusion 273

8 OTHER DISEASE- AND CONDITION-SPECIFIC SCALES 275

Other developments in the measurement of disease-specific quality of life 275Popular generic measures used to supplement disease-specific scales 276Disease-specific measures of quality of life 277Diabetes 277Outcome Measures in Ambulatory Care (OMAC) (Asthma and Diabetes) 277Diabetes Impact Measurement Scales (DIMS) 278Quality of Life, Status and Change (QLSC) 279Rand Diabetes Mellitus Battery 280

CONTENTS xi

xii MEASURING DISEASE

Diabetes Quality of Life Measure (DQOL) 282Renal disease 283Functional classification 283The use of generic scales and domain-specific batteries 283Kidney Disease Quality of Life Questionnaire (KDQoLQ) 284Haemodialysis Quality of Life Questionnaire (HQLQ) 285Leicester Uraemic Symptom Scale (LUSS) and Battery 285Quality of Life Assessment 286Bowel diseases 287Inflammatory Bowel Disease Questionnaire (IBIQ) 287Rand Surgical Conditions Battery: haemorrhoids 288Human Immunodeficiency Virus (HIV) and Acquired Immune Deficiency Syndrome

(AIDS) 288Medical Outcomes Study HIV Health Survey 30-Item Form (MOS-HIV) 289HIV Overview of Problems–Evaluation System (HOPES) 290AIDS Health Assessment Questionnaire 291Varicose veins 291Varicose Veins Questionnaire 291Rand Surgical Conditions Battery: Varicose Veins 292Back pain 293Clinical Back Pain Questionnaire 294Children 295Older people 296A popular core (generic) measure for use in disease-specific studies of adults: the Short

Form-36 (SF-36) 298Individualized measures of quality of life 302Schedule for the Evaluation of Individual Quality of Life (SEIQoL) 303Patient Generated Index (PGI) 305

APPENDIX: A SELECTION OF USEFUL SCALE DISTRIBUTORS AND ADDRESSES 307

REFERENCES 313

INDEX 390

PREFACE

Whom would the just man fail to greet, in orderto stop an injustice?What medicine tastes too nasty to savea dying man?How much meanness would you not commit if theAim is to stamp out meanness?If you’d found a way to alter this planet, what wouldYouRefuse to do?

(Extract from ‘Alter the world. It needs it’, BertoltBrecht, 1930. Reprinted in Willett, J. (ed.) (1995)Bertolt Brecht: Bad Time for Poetry – 152 Poems

and Songs. London: Methuen. Reproducedwith permission.)

The aim of this volume is to introduce the keyliterature on the psychometric properties of meas-ures of disease-specific quality of life, includingthe symptom scales often used alongside them. Inaddition to updating the reviews published in thefirst edition of this book, the opportunity has alsobeen taken to include additional scales. These in-clude the Living with Heart Failure Questionnaire;the Heart Failure Questionnaire; the Seattle AnginaQuestionnaire; the Camberwell Assessment ofNeed, Quality of Life Enjoyment and SatisfactionQuestionnaire; the Quality of Life in DepressionScale; the London Handicap Scale; the BreathingProblems Questionnaire; the Medical OutcomesStudy HIV Health Survey; the McGill Quality of

Life Questionnaire; the McGill Nausea Question-naire; the Support Team Assessment Schedule; andthe Kidney Disease Quality of Life Questionnaire.Individualized measures are also included. Forexample, the Schedule for the Evaluation of Indi-vidual Quality of Life and the Patient GeneratedIndex. The final chapter on measurement issuesin the first edition of Measuring Disease has beenremoved from this edition, as these are coveredmore fully in the author’s Research Methods in Health(Open University Press 1997).

There has been a rapidly increasing interest inboth generic and disease-specific measures of qual-ity of life. Other current reviews of a broad rangeof measurement scales are those by Wilkin et al.(1992); McDowell and Newall (1996); Spilker etal. (1990); Spilker (1996) and Salek (1998). Jenkinsonand McGee (1998) have also described the prin-ciples for designing disease-specific measures.Databases of research on the related concept ofhappiness have also been compiled (Veenhoven1994). The rapid growth of an industry of stand-ardized scale development in relation to quality oflife reflects the international emphasis on the pro-vision of effective, evidence based health care, andthe measurement of the outcome of care in thebroadest sense. It has also led to a renewed con-cern about the content validity of the measuresproduced, and hence to interest in using individu-alized measures. This wide range of approaches isincluded here in the second edition of MeasuringDisease. The scale literature was updated by using

electronic database searches of the literature,including EMBASE, Ageline, Cancer, Medline,CINAHL and PsychINFO, manual searches of keyjournals and correspondence with scale develop-ers. The general rationale for the inclusion of scalesin this volume is the existence of a body of evid-ence to support their psychometric properties. Insome cases, where this body of evidence is notextensive or where testing is still in progress, scaleshave been included because of their potential orcontinuing popularity. It is, of course, impossibleto include every scale that has been developed,and decisions have had to be taken over includingsome and excluding others where their evidencebase is still emerging. The scales are presented inchapters which relate to some of the most com-mon conditions across all age groups. Areas wherescale development is less well developed, and/orthere is less of a tradition of outcome measure-ment, are grouped together in a final chapter alongwith brief reviews of commonly recommended orincreasingly popular core modules (standardizedand individualized).

THE STRUCTURE OF THE CHAPTERS

Each chapter follows a similar, but not identicalpattern, which is dependent upon the amount andtype of research on health-related quality of lifethat has been carried out in that field. Each chap-ter on condition- and disease-specific scales beginswith an introduction to the methods of measuringquality of life in that area, and also references thecommonly used generic and domain-specific (e.g.depression) health-related quality of life scales.Generic and several domain-specific scales havebeen reviewed by the author elsewhere, in Meas-uring Health (Open University Press 1997).

Disease- and condition-specific quality of lifescales often require further supplementation withdisease-specific symptom items, or other sensitiveindicators of the presence of perceptions of illhealth. These range from self-reporting of symp-toms and unwanted treatment effects, as in can-cer, to assessments of grip strength among peoplewith joint problems, to the reporting of symp-toms of anxiety and depression in the case of psy-chiatric conditions and psychological morbidity.Although they are not quality of life measures, it

is important to include perceived symptoms anddiscomfort in batteries of measures purporting tomeasure health-related quality of life, and thus theyare presented here before the reviews of quality oflife measures.

RECOMMENDATIONS

The strengths, weaknesses and coverage of eachscale are presented in the text. Readers need toacquaint themselves with the range of availablescales in their field of interest and select the oneswhich are most appropriate for the aims of theirinvestigation. All scales have their good and badpoints. Not all scales have been fully tested, andscale development is often ongoing. For this rea-son, no ‘quasi-scientific’ league table of ‘best buys’has been attempted, although readers will find textcomments, and a concluding section at the end ofeach chapter with some recommendations, whichshould help them to make up their own mindsabout whether a particular scale is appropriate fortheir study.

Finally, it is easier to criticize a scale than toconstruct one. Researchers are encouraged to useexisting scales, or adapt them where necessary,rather than design them from scratch. Where scaleshave not been fully tested, users are encouraged totest further the reliability and validity of selectedinstruments when carrying out their own research.In this way, a better, and continually updated, bodyof knowledge will be developed.

COPYRIGHT

Many scales are only available for purchase com-mercially. This is particularly true of those de-veloped in North America. In many cases, thepurchase price simply covers administration costsand the cost of a manual and questionnaire; inothers, the cost of the manual and scale can beseveral hundred pounds. Potential users are advisedto contact the authors or distributors of scales fordetails of permission of use and, where applicable,purchase.

The author would like to echo the plea of Wilkinet al. (1992): in view of the difficulty that can beexperienced in obtaining scales, authors should be

xiv MEASURING DISEASE

encouraged by publishers to either reproduce theirquestionnaire (if short) in a major article on thescale’s psychometric properties, or publish thename and address of the scale distributor. This

PREFACE xv

would save a great deal of unnecessary correspond-ence, searching electronically and manually forlikely references and scanning of the catalogues ofthe different distributors.

HEALTH-RELATED QUALITY OF LIFE 1

1HEALTH-RELATED

QUALITY OF LIFE:

CONCEPTUAL MEANING,

USE AND MEASUREMENT

Whenever we seemedTo have found the answer to a questionOne of us untied the string of the old rolled-upChinese scroll on the wall, so that it fell down andRevealed to us the man on the bench whoDoubted so much.

(‘The Doubter’, Bertolt Brecht, c. 1937. Reprintedin Willett, J. and Manheim, R. (eds) (1995)Bertolt Brecht, Poems 1913–1956. London:

Methuen. Reproduced with permission.)

BACKGROUND: THE ‘QUALITYOF LIFE’

In the developed world, with the emphasis onaffluence and with the increasing longevity ofpopulations, there is a general interest in how toachieve the ‘goodness’ of life, sometimes calledlife satisfaction or quality of life. But what isquality of life? Dictionary definitions of qualityindicate that it implies the degree of excellence ofa characteristic. But the quality of life may meandifferent things to different people. The theoret-ical definitions of related concepts of happiness,life satisfaction, well-being, the ‘good life’ and‘quality of life’ have attracted much conceptualconfusion, and have preoccupied a wide range ofdisciplines, dating far back to Aristotle (384–22

) and early Greek philosophy (see Ross 1947;Veenhoven 1991; Argyle et al. 1995).

The following extract is taken from a dialoguebetween Serafin and Summerchild in MichaelFrayn’s (1991) novel, A Landing on the Sun, inwhich the plot centres on a government investiga-tion of quality of life. It is used here, along withthe discussion that follows it, to illustrate the inef-fable nature and topicality of ‘quality of life’:

. . . Now . . . To recapitulate: ‘the quality oflife’, as you understand it, is some propertywhich is in one way or another promotedor enhanced by washing machines . . .[p. 81] . . . Does the class of things which toyour mind promote[s] . . . the quality of lifestretch far enough from the kitchen door toinclude the family television set? [p. 82] . . .Washing machines, and similar sorts ofdomestic machinery, are intended to saveour time and labour. Now no-one couldclaim that a television set saves time. On thecontrary – it uses up the time we’ve savedwith the dishwasher and the vacuumcleaner. And I think one might go on toargue that far from saving labour it createsit . . . [p. 83] . . . They enjoy watching it.Let’s leap from the television to the centralheating. Do they have central heatingbecause they enjoy it? [p. 83] . . . Do youfind the word ‘motor car’ on your list?[p. 85] . . . The list that you apparently keep

2 MEASURING DISEASE

inside your head of things that enhance thequality of life [p. 85].

(Frayn 1991)1

Walter and Shannon (1990) described the cur-rent interest in ‘quality of life’ in the developedworld, ranging from current concerns for the en-vironment to the marketing of the products webuy, and to the evaluation of the benefit–burdenratio involved in medical treatments. More glo-bally, quality of life as an outcome indicator hasbeen added to social, as well as health, service pro-gramme development (Department of Health andSocial Security 1989). It was added to the WorldHealth Organization’s (WHO) international HealthyCities Programme, now in its third phase (WorldHealth Organization 1998), and defined as span-ning the visual arts, recreation, employment, trans-port, housing, environmental and conservationissues, health and other indicators of what has beenlabelled as ‘the social temperature’. The Organiza-tion for Economic Cooperation and Development(OECD) long ago agreed on a list of quality of liferelated social concerns of member states, includ-ing health, command over goods and services,employment and quality of working life (Andrews1973), and the national press regularly publishescomparative annual statistics on indicators of qual-ity of life across the world. The salience of qualityof life across disciplines has resulted in the emer-gence of a generic Quality of Life Questionnaire(Evans and Cope 1989), and quality of life ques-tionnaires for use in health care evaluation (Chubon1987; Patrick et al. 1988; Siegrist et al. 1993).

WHAT IS QUALITY OF LIFE ANDHEALTH-RELATED QUALITY OF LIFE?

Quality of life

Quality of life research, then, spans a range of topics,from quality of life in the last year of life (Lawtonet al. 1990) to quality of life in urban environ-ments (Rogerson et al. 1989). As was illustratedearlier, quality of life is an amorphous concept thathas a usage across many disciplines – geography,

literature, philosophy, health economics, adver-tising, health promotion and the medical andsocial sciences (e.g. sociology and psychology).It is also a vague concept; it is multidimensionaland theoretically incorporates all aspects of anindividual’s life.

Social science and the ‘good life’

There are several meanings of the term ‘quality oflife’ in social research, which range from individualfulfilment and satisfaction with life – the satisfac-tion of basic human needs, the ability to lead a‘normal’ life – to the quality of the external envir-onment (Edlund and Tancredi 1985; Fowlie andBerkeley 1987; Rogerson et al. 1989). Quality oflife has been defined as the ‘output’ of the inputsof the physical and the spiritual (Liu 1974); as thedegree to which a person accomplishes life goals(Cella and Cherin 1987); and even quantifiedcrudely as a formula in which quality of life (QL)is a product of one’s natural endowment (NE) andthe effort made on one’s behalf by the family (H)and society (S), such that QL = NE × H × S (Shaw1977). Grant et al. (1990) defined quality of life as‘a personal statement of the positivity or negativ-ity of attributes that characterize one’s life’. Qual-ity of life is still taken to refer to equally amorphousconceptions of the ‘goodness of life’ (Zautra andGoodhart 1979).

Experiential social indicators research, whichdeveloped during the 1960s onwards, focused onthe importance of measuring subjective well-being (see Elster and Roemer 1993). The concernwith subjective indicators led to the first large sur-veys of life satisfaction, happiness, quality of lifeand the ‘good life’ among adults and among olderpeople (Gurin et al. 1960; Bradburn and Caplowitz1965; Bradburn 1969; Andrews 1973; Andrews andWithey 1976; Campbell et al. 1976; Lawton 1983).This research has increasingly indicated that, incontrast to subjective variables, objective, socialbackground variables account for relatively littleof the variance in happiness, life satisfaction andwell-being, thus leading to more emphasis on theimportance of subjective feelings of independence,control and autonomy as predictors of well-being(Larson 1978; Inglehart and Rabier 1986; Mroczekand Kolarz 1998). Issues of definition and meas-urement were the subject of critical investigation

1 Michael Frayn, A Landing on the Sun, Viking 1991, pp. 81,82, 83, 85. Copyright © Michael Frayn 1991. Reproduced bypermission of Penguin Books Ltd, Harmondsworth.


and debate, which continues today (Andrews andWithey 1976; Andrews and Crandall 1976; Larson1993).

Lawton (1983) first proposed a theoreticalmodel of quality of life as ‘the good life’, definedas psychological well-being, perceived quality oflife, behavioural competence and the ‘objective’environment. It was increasingly held that, in thedeveloped world where human needs have gener-ally been met, quality of life is the extent to whichpleasure and satisfaction have been obtained (An-drews 1974). Quality of life has been referred to asan affective response to one’s role situation andvalues (Andrews and Withey 1976), as the dis-crepancy between desired and actual circumstances(Krupinski 1980), and as the well-being or ‘ill-being’ of people and/or their environment (Bubolzet al. 1980). The gerontological literature on thetopics of ‘successful ageing’, ‘positive ageing’ and‘quality of older age’ makes a similar point,focusing largely on life satisfaction and morale(Neugarten et al. 1961; Havighurst 1963; Williamsand Wirths 1965; Bradburn 1969; Lawton 1975;Andrews and Withey 1976; Palmore 1979). Therehas been a rapid expansion of literature on cor-relates of life satisfaction as a proxy for quality oflife (Bowling et al. 1991; Shahtahmasebi et al. 1992);and life satisfaction has become a key variable inanalyses of health status and mortality in old age(Bowling and Browne 1991; Grundy et al. 1992;Bowling and Farquhar 1993; Bowling et al. 1994,1996a). The work of social gerontologists also hasroots in philosophy. As long ago as 44 , Cicero([44 ] 1979), argued that old age contains manyopportunities for positive change and productivefunctioning, and should not be confused with ill-ness. Quality of life research in gerontology hastaken the theoretical debate further with an em-phasis on the additional importance of feelings ofcognitive efficacy, social competence and produc-tivity, personal control and motivation (Baltes andBaltes 1990; Day 1991).

In non-experiential social indicators research,quality of life encompasses all external, or object-ive, circumstances of life – for example housing,leisure activities, work, the environment and soon (Campbell et al. 1976; Wingo and Evans 1978;Kaplan 1985). Environmental research has focusedon non-experiential objective background charac-teristics of neighbourhoods, while also attempting

to incorporate subjective public values and levelsof satisfaction and preferences. Rogerson et al.(1989) and Rogerson (1995) reviewed the conceptof quality of life, and pointed out that no definit-ive list of criteria of quality of life has yet beendeveloped, and neither has an acceptable weight-ing system for the incorporation of subjectiveand objective indicators. The research conductedby Rogerson et al. (1989) in human geographywas based on a national opinion survey of 1200respondents in Britain. They were asked to rate 20dimensions of quality of life in terms of their de-gree of importance in influencing their choice ofwhere to live, on a scale of 5 (very important) to1 (indicating minimal significance). The top fiveitems were violent crime, non-violent crime, healthprovision, pollution levels and cost of living. Thebottom five related to travel to work time, leisurefacilities, quality council housing, access to coun-cil housing and cost of private or rented housing.Thus, in research on quality of life in cities, healthcare ranks third in importance. This research isconsistent, and overlaps with housing surveys ofneighbourhood satisfaction and community needsassessment exercises (Percy-Smith and Sanderson1992; Burrows and Rhodes 1998).

It has been argued that human needs are thefoundations for quality of life and that quality oflife is the degree of satisfaction of those needs –for example, physical, psychological, social, ac-tivity, marital and structural (Hörnquist 1982).Needs satisfaction is the theoretical basis of manyof the quality of life scales developed for use inmental health. This theory is reminiscent ofMaslow’s (1954, 1962a, 1962b) hierarchy of need(physiological, safety and security, social and be-longing, ego, status and self-esteem, and self-actualization), and the argument that once basicbiological and survival needs have been met, emo-tional and social needs become more prominent.Gap (or relative deprivation) theorists argue thatin the developed world perceptions of quality oflife are less likely to be related to basic needs, wherethese are largely met, but more to one’s expecta-tions in life, and to social comparisons with pastachievements, and also to comparisons withothers (Michalos 1986). Quality of life could bedefined in relative terms of what one has lost, orlacks, rather than what one has. Calman (1984)proposed that quality of life is the difference, at a

4 MEASURING DISEASE

particular period in time, between the hopes andexpectations of the individual and their presentexperience. So quality of life is influenced by pastexperience, present circumstances and aspirationsfor the future. Aristotle (384–22 ) stated, in re-lation to man’s perception of happiness: ‘When hefalls ill he says that it is his health, and when heis hard up he has that it is money’. In relation tohealth, quality of life has also been defined in termsof the difference between reality, or the percep-tion of reality, and expectations (Calman 1984;Presant 1984). However, the relativity argumenthas been criticized on the grounds of conceptualconfusion (Veenhoven 1991) and inconsistent lit-erature in support of it (Headey and Wearing 1992).Also, gap theories are really explanations of per-ceptions of quality of life, and do not increaseknowledge about the constituents of quality of lifeas such. There has been little attempt to examinethese assumptions critically (Bauer 1966a; Ziller1974) although they provide the conceptual back-ground to many health-related quality of life scales.For example, gap theory (Calman 1984) under-pins the Patient Generated Index (PGI) (Ruta et al.1994b, 1999) and the needs satisfaction modelunderpins several mental health scales, in additionto the more subjective life satisfaction model (e.g.Lehman et al.’s (1982) Quality of Life Interview).

Philosophical approaches

Overlapping with this research is the history ofthe exploration of happiness dating from the workof early to present-day philosophers (e.g. Aristotle,referred to earlier) and social scientists (Morgan1934; Ross 1947; Barschak 1951; Tatarkiewicz1975; Gallup 1976; Mason and Faulkenberry 1978;Andrews 1981; Parducci 1984; Veenhoven 1988,1989, 1991). Bentham ([1834] 1983) introduced thedimension of psychological hedonism as the gov-erning principle of human conduct in the form ofwell-being and its measurement, which he definedas ‘the difference in value between the sum of pleas-ures of all sorts and the sum of pains of all sortswhich a man experienced in a given period of time’.He attempted to show how the moral theory ofutilitarianism could be developed into a calculusof pleasures whereby the effects of actions couldbe judged, and the right policy thus identified: theend of human conduct is the greatest happiness

of the greatest number. It was thus expected thatutilitarianism would provide a rational founda-tion for social and legal policy. This utilitarianfocus on pleasure and happiness was derived fromthe translation of Aristotle’s concept of eudaimoniaor ‘human flourishing’ as pleasure or happiness,rather than as excellence. Bentham’s argument, thatutility is identified with the amount of satisfactionof desires, allows that in some circumstances thehappiness of some people should be purchased atthe cost of the undeserved and uncompensatedmisery of others. Contemporary utilitarianismfocuses more on satisfaction than on happiness asthe relevant outcome. Current philosophical viewsof the ‘good life’ relate not only to hedonism, butto the satisfaction of desires or preferences, andthe realization of normative ideals (Brock 1993;Scanlon 1993). This influence can be seen in cur-rent measures of health-related quality of life,particularly in mental health with the predominantmodel of quality of life as needs satisfaction. Thefocus of Bentham’s utilitarianism on the greatesthappiness of the greatest number also underpinsattempts by policy makers and health economiststo justify health service rationing policies (par-ticularly age-based rationing policies) (A. Williams1997). The use of Quality Adjusted Life Years(QALYs) as indicators of health gain in healthservice resource allocation is frequently defendedon utilitarian grounds, in that they appear tofacilitate the greatest health gain, given scarceresources, for the greatest number. The ageisminherent in this formula, and the problems of clas-sic utilitarianism, are rejected by its defenders, whoargue that QALY values are derived from surveysof the public and therefore reflect the aggreg-ated preferences of the population (Williams 1994;Edgar et al. 1998). However, the basis of suchlogic was early on branded by John Stuart Mill(1861) as the worst tyranny – ‘the tyranny of themajority’, legitimized in democracy and expressedin legislation by elected tyrants. Mill argued there-fore that society has a moral duty to correct thismajoritarian bias and uphold the principle of equal-ity (he actually argued that our unelected legaljudges should have this duty). His own ‘Utilitari-anism’ (Mill 1861) was an attempt to elaborate onthis theory given the conflict in Bentham’s utilit-arianism between the demands of justice and thedemand to maximize general happiness.


Individual meaning

Phenomenologists argue that quality of life is de-pendent upon the interpretation, and perceptions,of the individual (Ziller 1974). Cohen (1982) hasalso pointed out that the simple listing of qualityof life domains is not a satisfactory way of meas-uring quality of life because it is unknown whetherall important domains have been included. Re-searchers who construct health-related quality oflife measurement scales are seldom philosophicallysophisticated or concerned with competing ac-counts of the good life. To some extent, the needto develop valid measures for use with large andvaried samples of people necessitates compromisesand the simplification of issues of philosophicalimportance (Brock 1993). Rosenberg (1992) hasargued that the psychometric translation of qual-ity of life into components such as emotionalstatus, social interaction, economic status, healthstatus and physical capacity, while incorporatingthe multidisciplinary nature of human beings, doesnot capture their subjectivity. He argued thathermeneutic thinking should be introduced intomodern medicine, so that a naturalistic conceptof mankind is presented along with a concept ofthe human being as a self-reflective individualresponsible for their own actions. Ziller (1974)has also argued that the approach to quality oflife is ‘through the eye of the experiencer; that is,a phenomenological approach’. The health-relatedquality of life scales designed by Guyatt et al.(1987a, 1987c, 1989a, 1989d), O’Boyle and col-leagues (O’Boyle et al. 1990, 1992, 1993; Hickeyet al. 1999), and Ruta and his colleagues (Ruta 1992;Ruta et al. 1994a, 1994b; Garratt and Ruta 1999),which take account of the individual’s perspec-tive, are exciting developments which are coun-teracting the trend towards the pre-definition ofquality of life by the researcher ( Joyce et al. 1999).O’Boyle’s work, for example, takes individualmeaning into account, and has involved the appli-cation of the techniques of human judgementanalysis (Brehmer and Joyce 1988) to assess health-related quality of life (O’Boyle et al. 1989; Hickeyet al. 1999). With this technique respondents areasked to list, rate and weight the five areas of life(‘cues’) that they judge to be the most importantto their overall quality of life (the techniqueis known as the Schedule for the Evaluation of

Individual Quality of Life or SEIQoL). This isdescribed more fully in Chapter 8.

Research on people’s own definitions of health-related quality of life and values, using a range ofmethods from Delphi techniques to semi-struc-tured individualized methods, includes work byGuyatt et al. (1987a, 1987c, 1989a, 1989d), Tugwellet al. (1987), O’Boyle and his colleagues (O’Boyleet al. 1989, 1992, 1994; Hickey et al. 1999), Par-tridge et al. (1991), Dalkey (1972), Berg et al.(1976), Farquhar (1995), Bowling (1995, 1996a,1996b) and Ruta (1992), Ruta et al. (1994a, 1994b).Ruta’s work is reviewed in Chapter 8.

The general conclusion is that perception andachievement of quality of life is dependent on anindividual’s preferences and priorities in life. Themeaning of the concept of quality of life is thusarguably dependent on the user of the term, theirunderstanding of it, and their position and agendain the social and political structure (Edlund andTancredi 1985): ‘Quality of life is a vague and ethe-real entity, something that many people talk about,but which nobody very clearly knows what to doabout’ (Campbell et al. 1976). Adequate measure-ment should therefore reflect these elements andbe preference weighted (Diamond and Becker1999). Individualized measures are a step in thisdirection.

Is health a domain of quality of life?

The empirical evidence appears to justify includ-ing health as a dimension of quality of life. Re-search on valued states of existence, from the earlydays of social indicators research to the presentday, has reported that health is among the mostvalued states, and among the most important areasof life and of quality of life nominated by people(Rokeach 1973; Kaplan 1985; Bowling 1995, 1996a,1996b; Farquhar 1995). It also increases in valueand priority with older age (Bowling 1996b).

What is health-related quality of life?

In relation to health, health status is increasinglyreferred to as quality of life, and, so as to narrowdown its operationalization in research studies,quality of life is referred to as health-related qual-ity of life. Quality of life in relation to health is

6 MEASURING DISEASE

rarely explicitly defined in published studies (Gilland Feinstein 1994), but often implicitly definedfrom a functionalist perspective of society, whichrelates to the ability to perform activities of dailyliving and fulfil role obligations (necessary for thefunctioning of society as a whole). It has tradi-tionally been viewed negatively in this context, interms of ‘dis’abilities, and measured accordingly(e.g. with activities of daily living and physicaland role functioning scales).

The use of the concept health-related quality oflife in scientific research requires more precise defi-nition of the concept itself. From a health (or dis-ease) perspective, quality of life has been said torefer to the social, emotional and physical well-being of patients following treatment (Greer 1984),mirroring WHO’s (World Health Organization1947, 1948, 1958) definition of health (see below),and to the impact of disease and treatment on dis-ability and daily functioning (Kaplan 1985). Itfocuses on the impact of perceived health status onthe ability to lead a fulfilling life (Bullinger et al.1993). It is a double-sided concept, incorporatingpositive as well as negative aspects of well-beingand life, and it is multidimensional, incorporatingsocial, psychological and physical health. It is also,ultimately, a personal and a dynamic concept for,as health status deteriorates, perspectives on life,roles, relationships and experiences change (Sher-wood et al. 1977; Morris et al. 1986).

Taking these definitions into account, health-related quality of life is defined here as optimumlevels of mental, physical, role (e.g. work, parent,carer, etc.) and social functioning, including rela-tionships, and perceptions of health, fitness, lifesatisfaction and well-being. It should also includesome assessment of the patient’s level of satisfac-tion with treatment, outcome and health statusand with future prospects. It is distinct from qual-ity of life as a whole, which would also includeadequacy of housing, income and perceptions ofimmediate environment. Quality of life as a wholehas also been divided into subjective and objectiveareas: it ‘encompasses what a person is capableof doing (functional status), access to resourcesand opportunities to use these abilities to pursueinterests, and sense of well-being. The formertwo dimensions are often referred to as objectivequality of life and the latter as subjective qualityof life’ (Lehman et al. 1995).

The theoretical framework of health-relatedquality of life, then, is largely based on a multi-dimensional perspective of health as physical, psy-chological and social functioning and well-being,along the lines of the WHO’s (World HealthOrganization 1947, 1948) definition of health: a‘state of complete physical, mental and social well-being and not merely the absence of diseaseor infirmity’. WHO (World Health Organization1984) has since added ‘autonomy’ to this list. Ware(1987) has argued that five health concepts areinherent in this definition: physical health, mentalhealth, social functioning, role functioning andgeneral well-being. He restricts his definitionbecause the goal of health care is to maximizethe health component of quality of life. Healthstatus may influence quality of life without deter-mining it.

WHO (World Health Organization 1991) has aworking party on quality of life under its umbrella:the World Health Organization Quality of LifeGroup (WHOQOL Group 1991, 1993a, 1993b,1993c, 1994a), in relation to its international inves-tigation into health-related quality of life (Sartorius1993), and has provided a definition of qualityof life which also takes individual perception andrelationship to the environment into account:

Quality of life is defined as an individual’sperception of their position in life in thecontext of the culture and value systems inwhich they live and in relation to theirgoals, expectations, standards and concerns.It is a broad ranging concept affected in acomplex way by the person’s physicalhealth, psychological state, level ofindependence, social relationships, and theirrelationships to salient features of theirenvironment.

(WHOQOL Group 1993b: 3)

This definition underpins the development of theWHOQOL instrument for measuring quality oflife that can be used in a variety of cultural settings(WHOQOL Group 1998b; Skevington 1999).

Unresolved questions

There are many vexing and unresolved questionsin quality of life research. For example, as Schipper(1983) has asked: How can the quality of life of,


say, a city and a country dweller be compared?Does their quality of life change in comparableways in response to illness? Are some parametersof quality of life weighted differently by differentgroups of people, and in relation to different ill-nesses? The issue is not straightforward, and amultitude of additional questions are raised whenattempting to consider the impact of an illness onan individual: What does the condition preventthe patient from doing? What do they miss doingas a result of the condition? What are their fearsfor the future? How do they cope with anger andfrustration related to their condition? How doesmood affect functioning and social relationships?Are there any positive consequences of the illness(e.g. bringing relatives closer together)? How dopeople spend their time, and how does this changewhen they are afflicted by certain illnesses (e.g.less time spent on leisure and domestic activitiesand social roles)? What is the impact of the condi-tion on other family members? How have theyadjusted, and have they been restricted?

Rarely does one health status measure encom-pass any of these items, and researchers are forcedto resort to lengthy, and sometimes impractical,batteries of several measurement scales.

HEALTH-RELATED QUALITY OF LIFEAND HEALTH

A concept of health

A concept of health-related quality of life mustrest on a concept of health as well as of quality oflife. A measure of health status should also be basedon a concept of health. A medical conception ofhealth is freedom from disease and abnormalities;a sociological view can be defined in terms of thepossession of acceptable levels of mental and physi-cal fitness in order to perform one’s social role insociety. As with the concept of quality of life,there is also a humanistic view of health in whichoptimal autonomy, self-mastery and a positiveperception of life are central components or influ-ences (Heyrman and van Hoeck 1993).

The concept of positive health has been describedelsewhere (Bowling 1997a). Briefly, most healthstatus instruments measure deviations away froma state of health and are really measuring ill health,

or the absence of illness and disease. They rarelyreflect the global definition of WHO (World HealthOrganization 1948), although this is the theo-retical underpinning, explicit or implicit, in morerecent attempts at scale development. There aremultiple influences upon patient outcome, andthese require a broad model of health. The non-biological factors which may influence or mediaterecovery include patient expectations, self-esteem,self-mastery, motivation, coping, adherence totherapy, socio-economic status, availability ofhealth care, social support networks, and individualcultural beliefs and health behaviours.

A negative conception and measure of healthis more appropriate when measuring severely illpopulations, but less appropriate in general popu-lation surveys. Negative definitions of, and meas-urements of, health status will tell us little aboutthe health of the population. Positive health istherefore an increasingly popular concept, en-compassing not just the absence of disease, butfeelings of mental and physical well-being, fullfunctioning, physical fitness, ability to cope, so-cial support, adjustment and efficiency of mindand body. Collectively, these positive states havebeen referred to in the literature variously as posi-tive health, social health and health-related qual-ity of life. Even in disease-specific studies, wherenegative measures of health are appropriate, amore balanced scale including positive measuresshould also be used in order to assess outcome inrelation to degrees of wellness as well as illness.

The public’s view of health

Dubos (1959) argued that health and disease can-not be defined merely in terms of anatomical,physiological or mental attributes, and that ‘theirreal measure is the ability of the individual to func-tion in a manner acceptable to himself and to thegroup of which he is part’. An absolute definitionof health is not possible to construct and even theWHO (World Health Organization 1947, 1948;WHOQOL Group 1993b) definitions can only beviewed as relative concepts (see above). Peopledefine their health variously, depending on socio-demographic factors and on their culture (D’Houtardand Field 1984; Currer and Stacey 1986). Their re-ported definitions include health as not being ill, asabsence of disease, as behaviour, as role functioning,

8 MEASURING DISEASE

as physical fitness, energy and vitality, as socialrelationships and as emotional well-being. Wright(1990) has summarized lay definitions of health ashealth as being, health as doing and health as hav-ing. Wright (1997) also pointed to the debate aboutwhether health is a ‘state’ or a ‘trait’ (constructswhich, he points out, are not mutually exclusive).He refers to the application of various constructswhich he summarizes as ‘dispositional resilience’,such as ‘hardiness’, ‘sense of coherence’ and ‘dis-positional optimism’ (Kobasa 1979; Scheier andCarver 1985; Antonovsky 1993). Wright arguedthat ‘dispositional resilience’ may influence healthoutcomes.

The concept of health is inevitably subject tocultural relativism (Heyrman and van Hoeck 1993),and while health may be a social goal common toall groups, the salience of health to individuals mustbe assessed relative to other goals. The place ofhealth in one’s value system will be reflected inone’s definition of health. This has been amplydemonstrated by the classic studies of Koos (1954)and Herzlich (1973), as well as by more recentresearch.

In a study of several thousand people attendinghealth check-ups in Nancy in France, D’Houtardet al. (1990) reported that the most common andconsistent definitions of health were ‘hygiene, liv-ing conditions, to feel well in one’s skin, to knowoneself well, work, luck, to be at the top of one’sform, personal unfolding and not to feel one’sbody’. These were comparable with findings fromsimilar research carried out in Nijmegen in theNetherlands. D’Houtard and Field (1984) reportedthe results of a study of 4000 people from a healthcentre in Lorraine, France in more detail. Whenasked an open-ended question about what healthmeant to them, the most common replies were‘not to be sick’, ‘to be at the top of one’s form’(more than 400 mentions), ‘good physical equilib-rium’, ‘good mental equilibrium’, and ‘joy of liv-ing’ (300–400 mentions).

In the British national Health and Lifestyle Sur-vey of 9003 adults living in randomly sampledhouseholds, Cox et al. (1987) reported that theconcepts of health most often used for describingwhat health is in ‘someone else’ were, amongfemale respondents in all age groups, ‘never ill, nodisease, never see a doctor’, ‘fit, strong, energetic,physically active’, ‘has healthy habits’ (e.g. not

smoking, taking exercise, taking care of health)and ‘able to do a lot, work, socially active’ (inorder of frequency mentioned). Among male re-spondents in all age groups, the most commondefinitions were ‘fit, strong, energetic, physicallyactive’, ‘never ill, no disease, never see a doctor’,‘has healthy habits’ (e.g. not smoking, taking exer-cise, taking care of health) and ‘able to do a lot,work, socially active’ (in order of frequency men-tioned). When asked about definitions of healthused for describing ‘what it is like to be healthyoneself ’, the most common response of more thanhalf the males and females in all age groups was‘feel psychologically fit’ (e.g. good, happy, ableto cope), followed in order of frequency by ‘fit,strong, energetic, physically active’, ‘able to do alot, work, get out and about’ and ‘never ill, nodisease, never see a doctor’. Respondents with Alevel education and above, particularly females,were more likely to mention ‘physically fit,strong’.

Others have defined health in the context ofinterference with the performance of normal socialroles and activities (Apple 1960), or of ability towork (Twaddle 1969). The various lay definitionsof health reported by Baumann (1961) pointed tothree main orientations: a general feeling of well-being, the absence of illnesses and the ability toperform social roles. Definitions vary according toage, sex, level of education (Cox et al. 1987), cul-tural group (Bowling 1994b) and socio-economicgroup, with those in the lower socio-economicgroups defining health more negatively (Blaxterand Patterson 1982) and more likely to perceivethe causes of health as being outside their control(Blaxter 1983; Pill and Stott 1985, 1988; Coulter1987; D’Houtard et al. 1990). On the other hand,Cox et al.’s (1987) national Health and LifestyleSurvey in Britain reported that those in poor eco-nomic circumstances were not significantly morelikely than those who were better off to associatepoverty with ill health. They also reported that,while poverty and prosperity were seen generallyas important determinants of health for society atlarge, it was rarely mentioned as a cause of illhealth in the context of respondents’ own lives.This pattern of responses was also found in rela-tion to individual behaviours (diet, smoking, exer-cise). These beliefs and definitions are pertinentto measuring subjective health status, and have


implications for the development of cross-culturalinstruments.

These examples from the research literature onhealth beliefs illustrate that some groups of peopledo include social roles and social functioning intheir definitions of health. However, for somedecades there was much criticism of the previ-ously described WHO (World Health Organiza-tion 1947, 1948) definition of health as a completestate of mental, physical and social well-being. Thiswas criticized as utopian and unrealistic, particu-larly given the many studies of community mor-bidity which imply that having unreported (tothe doctor) symptoms is a normal condition ofthe population (Dunnell and Cartwright 1972). Thelack of realism in the WHO definition, and the im-possibility – even undesirability – of achieving ‘per-fect health’ is emphasized by Dubos’ view of health:

Every manifestation of existence is aresponse to stimuli and challenges, each ofwhich constitutes a threat if not adequatelydealt with. The very process of living is acontinual interplay between the individualand his environment, often taking the formof a struggle resulting in injury or disease.The more creative the individual the less hecan hope to avoid danger, for the stuff ofcreation is made up of responses to theforces that impinge on his body and soul.Complete and lasting freedom from diseaseis but a dream remembered fromimaginings of a Garden of Eden designedfor the welfare of man.

(Dubos 1959: 1–3)

Indeed, ‘health is more than just a biomedical phe-nomenon; it involves a social human being func-tioning in a social environment with social roleshe must fulfill’ (Lerner 1973) – hence the import-ance of measuring health-related quality of life.

Gradually, definitions of health, like health-related quality of life, have moved away from atotal disease model to one which incorporateshealth and well-being. The most widely used meas-ures of broader health status, or health-related qual-ity of life, reflect this definition, and incorporatephysical functioning, psychological well-being andsocial support and activity items. Also, in recog-nition of the emphasis on the positive as well asthe negative evaluations of health and consequences

of illness, Hyland and Kenyon (1992) have devel-oped the Satisfaction with Illness Scale, and Argyleet al. (1989, 1995) have developed the Oxford Hap-piness Scale.

QUALITY OF LIFE ASSESSMENT ANDRESEARCH ON HEALTH CAREOUTCOMES

Why are we measuring health-related qualityof life?Life expectancy at birth in the developed worldhas increased over the past 150 years, althoughmost of the increase has taken place during thefirst half of this century. Expectation of life, andexpectations of a morbidity-free life at older ages,has also increased and has led to internationalattempts to measure health expectancy anddisability-free life expectancy (Bone 1992; Robineet al. 1992; Colvez 1996). Debate focuses on whe-ther the extra years of life are spent in good orpoor health quality, with one side arguing, withsome evidence, that chronic morbidity is beingcompressed into an increasingly shorter periodbefore death, and the other arguing that there isan expansion of morbidity in old age. This inter-est in health expectancy, as opposed to simplyanalysing mortality rates, coincides with the morepositive view of health measurement.

Purchasers of health care are increasingly ex-pected to allocate health care resources on thebasis of the evidence of the effectiveness of healthcare interventions – i.e. on the basis of health out-comes of intervention in the broadest sense. Healthoutcome indicators, defined as indicators of changein health status (Donabedian 1985) (a definitionthat does not take account of stabilizing andpalliative treatments), have traditionally includedinformation on avoidable mortality, survival rates,standardized mortality ratios, adverse reactions,complications, symptom relief, pain and physicaland biochemical markers of recovery. Purchasingdebates in health care now focus more on healthcare costs in relation to ‘health gain’ or broaderhealth benefits from the treatments and interven-tions that are contracted for (Øvretveit 1993; seealso overview by Normand and Bowling 1998).

Quality of life as a measure of outcome redirectsattention towards consideration of the impact of

10 MEASURING DISEASE

the condition, and the treatment, on the patient’semotional and physical functioning and lifestyle.Quality of life indicators help to answer the ques-tion of whether the treatment leads to a life worthliving, by providing a more patient-led baselineagainst which the effects of the intervention can beevaluated. The concept is not new to clinicians;many early health-related quality of life measureswere developed by clinicians some decades ago,although they were often crude and limited to func-tioning (Karnofsky et al. 1948; Steinbrocker et al.1949; Katz and Akpom 1976a, 1976b).

The fashion for health-related quality of lifeassessment

Quality of life assessment as a supplement to moreobjective clinical indicators is becoming moretopical in view of the increasing questioning of theeffectiveness and appropriateness of many exist-ing medical treatments and methods of organizinghealth services (Brook et al. 1983, 1990; Winslowet al. 1988; Enthoven 2000). It also represents a par-adigm shift in the approach to the operationalizationand measurement of health outcomes (O’Boyle1997). The measurement of the health outcome ofclinical interventions has become a cornerstone ofhealth services research; it is essential for the as-sessment of the effectiveness and appropriatenessof health care interventions (Brook 1990). Qualityof life assessment is increasingly popular amongpharmaceutical companies, with most reportingthat they have used some type of quality of lifeinstrument in their clinical trials of drugs (Luceet al. 1989). Information about broader patientoutcomes is also required in order to guide andempower patients about appropriate health care.

Health care outcome measures must, however,relate to the objectives and the known risks andbenefits of the intervention or care (Frater 1992).The Department of Health (1992) suggested thefollowing should be incorporated in outcomeassessment: survival rates, symptoms and com-plications, health status and quality of life, theexperiences of patients and their carers, and thecosts and use of resources. As their report con-tinued: ‘Many health technologies are intendedto improve general health and the quality of life,so it is important to measure patients’ subjectiveexperiences of illness and the care they receive’.

Quality of life assessment has become an indus-try in itself. The term was introduced by Medlineas a heading in 1975, and accepted as a conceptby Index Medicus in 1977; this was followed byacknowledgement and acceptance by various sci-entific bodies (Bech 1992). Since the 1970s, therehas been an explosion of interest in the subject,with an increasing number of citations of qualityof life in the medical literature (de Haes and vanKnippenberg 1985, 1987; Cella and Tulsky 1990;Aaronson et al. 1991b). There has also been a pro-liferation of study groups, conferences, bibliogra-phies and special journal issues (e.g. Advances inNursing Science in 1985; Journal of Chronic Diseasesin 1987; Psychotherapy and Psychosomatics in 1990;Medical Care in 1990; WHOQoL Group 1994a andSocial Science and Medicine in 1995). In 1991, a jour-nal entitled Quality of Life Research was publishedby Rapid Communications of Oxford, devoted tothe study of health-related quality of life. Bardelliand Saracci (1978) reviewed all the clinical studiespublished in six major cancer journals between1956 and 1976, and reported that in less than5 per cent an attempt had been made to measuresome aspect of quality of life. The situationchanged from the late 1970s and the number hassince exploded, although the methodology hasfrequently been relatively crude (Najman andLevine 1981; Fayers and Jones 1983; Aaronsonet al. 1988; Hollandsworth 1988; Falotico-Tayloret al. 1989; Padilla et al. 1992b). A search ofMedline by O’Boyle (1997) using the term ‘qual-ity of life’ yielded eight publications in which theterm was mentioned in 1974, but this had increasedto 1482 for 1996. Medline showed many morethan this for 2000. However, such a global, andfashionable, key search term is relatively useless asit is unclear what relevance most publications hadto quality of life.

Some investigators remain sceptical about usingpatient-based indicators in outcomes or descrip-tive research because of their subjective nature (e.g.self-reported health status and perceptions of well-being), quite apart from the poor quality of theiroperationalization and measurement (Muldoonet al. 1998). However, it is important to includepatient-based indicators in assessments of outcomeprecisely because they do not necessarily correlatewith objective measures of patients’ level of phys-ical functioning (Evans et al. 1985). There are many


examples of studies which have reported lowlevels of agreement among doctors in relation toobjective indicators (Wigton 1988), as well as be-tween doctors’ and patients’ assessment of outcome.Subjective indicators (based on self-ratings) of qual-ity of life, and health-related quality of life areincreasingly popular due to the recognition of theimportance of patient satisfaction and of how indi-viduals feel, rather than what statistics imply theyought to feel. Clinical indicators of outcome areno longer sufficient alone, particularly in view ofthe debate about quality of life in terminal condi-tions (see Jennett 1976). The argument in favourof measuring quality of life as an outcome of clin-ical interventions has been neatly summed up byEbbs et al. (1989). They argued that, in the case ofchronic conditions with only partial or temporaryamelioration of symptoms, a comprehensive evalu-ation is needed in order to determine the merits oftreatment. Sullivan (1992) also pointed out thatwith incurable conditions the realistic goal of careis to provide a life that is as comfortable, func-tional and satisfying as possible. This argumentcan be extended to intervention for more acuteconditions where the benefits of the treatment arecomplex and/or uncertain. The primary goals ofmedical treatment and care are to increase survivaland to add quality to the survival.

UTILITY ASSESSMENTS ANDQUALITY OF LIFE ASSESSMENT

Utility measurement by health economists dealswith valuations of health states – it is concernedwith the preference for (desirability of ) a particu-lar health state or outcome, and results in a sum-mary score which equals a single value. This singlevalue facilitates comparisons of treatments in cost–utility studies. It is conceptually and technicallydifferent from broader quality of life outcomeassessment. There is some blurring of the bound-aries between the measures used by health eco-nomists and those used by investigators of health-related quality of life treatment outcomes. Thishas occurred because the latter investigators oftenuse measures primarily designed to measure utilit-ies (and which were designed to produce a singlevalue of a health state to inform cost-effectivenessanalysis) in health research. Also, some investiga-

tors of existing broader, multi-dimensional healthstatus scales are attempting to adapt them toproduce a single value for use in utility studies(Brazier et al. 1996). The more recently developedEuroqol, designed to produce a single value, ispopularly used in broader outcome assessmentbecause of its relative brevity (Euroqol Group 1990;Kind 1996). Each of these approaches is presentedbriefly next in order to clarify the distinctionbetween them and the broader measures of healthand disease-specific related quality of life reviewedin this volume and in Measuring Health (Bowling1997a).

Utility assessments

In response to the interest of health policy makersin cost-benefit formulas for purchasing decisions,a main focus among health economists has beenon the development of utility assessment. Theresearchers initially responsible for applying utilitytheory to the health care field include Kaplan, Bushand their colleagues in the USA (Kaplan and Bush1982; Kaplan et al. 1984), Weinstein and his col-leagues also in the USA (Weinstein et al. 1980),Williams and colleagues in England (Kind et al.1982) and Torrance and his colleagues in Canada(Torrance et al. 1972, 1982; Torrance 1986, 1987).A generic utility questionnaire collects data onquality of life in a way that enables it to be appliedto a pre-scaled matrix of health state preferences(values).

Quality Adjusted Life Years (the QALY)

Cost–utility studies need a common measurementof health outcome. With the QALY the cost of anintervention is related to the number of QALYs, aconcept first introduced by Weinstein and Stason(1976). A QALY is a year of full life quality. Poorhealth may reduce the quality of a year (e.g. from1 to 0.5). In QALYs, improvements in the lengthand quality of life are amalgamated into one singleindex. Each life year is quality adjusted with autility value, where 1 = full health. The utility valueaims to reflect the health-related quality of life.Different types of medical interventions are thencompared by calculations of costs per gainedQALY (Williams 1985b; Office of Health Eco-nomics 1989). QALYs are not really measures of


quality of life but measures of units of benefit froma medical intervention, aiming to reflect the changein survival with a weighting factor for qualityof life.

QALYs have been severely criticized (Carr-Hill1989; Carr-Hill and Morris 1991) for their lack ofvalidation. There are inadequate data on their con-ceptual basis, and the data on reliability and valid-ity that do exist are based either on unrepresentative‘convenience’ samples or random samples of thepopulation with high rates of non-response, partlybecause of the complexity of the task. The appli-cation of QALYs to health policy making has ledto ethical concerns, particularly given their crudestate, over reliance on valuations by health profes-sionals and non-representative groups of people(Grimley Evans 1992). There is no evidence thatthe judgements determining the QALY for a par-ticular condition bear any relationship to realjudgements faced by patients suffering from thatcondition.

QALYS can be derived using several differentmethods (e.g. the Rosser Index of Disability(Rosser and Watts 1972), standard gamble, trade-off and rating scale techniques (Torrance et al. 1972,1982; Torrance 1986, 1987) ).

Rosser Index of Disability

In the UK the basic measurement tool underlyingthe calculation of the QALY is the Rosser Indexof Disability (Kind et al. 1982; Williams 1985a;Williams and Kind 1992), which was originallydeveloped as an indicator of hospital performance(Rosser and Watts 1972; Rosser 1992). The RosserIndex is based on the concept of a health index,and subjects are graded into one of eight areas ofdisability, from none to unconscious, with eachstate being graded on four levels of distress, fromnone to severe. States are scored on a scale rang-ing from 0 at death to 1 = healthy (with negativevalues for states valued to be worse than death,such as vegetative states). The scaling and valua-tion of health state techniques of the scale havebeen clearly described and reviewed by Wilkinet al. (1992). Kind and Gudex (1991) have alsodeveloped a survey questionnaire – the HealthMeasurement Questionnaire – designed to collectself-reported information capable of being pro-cessed to yield Rosser disability/distress categories.

The approach of Kaplan and Bush

Kaplan and Bush (1982) and Kaplan et al. (1984)adopted a slightly different approach. They placedindividuals with given health states into categoriesof mobility, physical activity and social activity,and then classified the symptoms and health prob-lems that they had on a given day. Four hundredcase descriptions were then compiled to illustratethe combinations of functional levels, symptomsor problems. The scale includes death. Randomsamples of the public gave preference ratings tothe descriptions, and weights were derived for eachlevel of mobility, physical activity, social activityand symptom or problem. It is based on a meas-ure of functional status. A utility value was as-signed to each functional level, and questionnaireresponses were used to assign subjects to one of anumber of discrete function states. Kaplan’s Indexof Well-being Scale is a single-score scale that hasbeen developed out of this methodology, and hasbeen used widely as a health status measure andmethodological tool for health care decision mak-ing (Kaplan et al. 1976; Bush 1984). The scale quan-tifies the health output of a treatment in terms ofyears of life, adjusted for the changes in qualitythat it is responsible for.

Standard gamble, trade-off and rating scaletechniques

In Canada, the McMaster group’s approach wasthat QALYs should capture the subject’s prefer-ences concerning the length and quality of life,and proposed three approaches to the derivationof utility values (Torrance et al. 1972, 1982;Torrance 1986, 1987). The first is the standardgamble technique, in which subjects are askedto choose between their own health state and agamble that they might die immediately or achievefull health status for the remainder of their lives(e.g. in relation to a specific treatment choice).The second is the time trade-off technique(Torrance et al. 1972), in which an individual isasked to consider a health state that is to last fora fixed period of time. A new procedure willgive the individual normal health for a shorterperiod of time, but they will probably die or beseverely disabled at the end of that time. The per-son is asked to ‘trade-off ’ the time with reduced


capacity for living with normal health for a shorterperiod of time. The time spent in normal health isvaried until the point of indifference is found. Sub-jects may also be asked to evaluate the equivalencesbetween the numbers of people helped by differenttreatments – how many people in state B must behelped to provide a benefit that is equivalent tohelping one person in state A (see review by Kaplanet al. 1993a). The time trade-off technique has beenreported to be more reliable and valid than thestandard gamble technique (Dolan et al. 1993).

Third, Torrance et al. (1982) developed a ratingscale approach, specifying six attributes that shouldbe included in a health state: physical function,emotional function, sensory function, cognitivefunction, self-care and pain. Each of these attributesis given several levels of gradation, and the char-acteristics of a given health state would includea description of functioning, self-care and painassociated with that state. The descriptions can beshown on video or in written vignettes. The rat-ings assigned to the descriptions are placed on avisual analogue scale ranging from 0 (death/leastdesirable) to 100 (healthy/most desirable). Multi-ple Attribute Theory is then used to determine thevalue for each level of the attributes and the utilityvalue of an associated health state.

The Euroqol

The aim of the Euroqol is to provide a stand-ardized non-disease-specific survey instrumentfor describing health-related quality of life, and togenerate a single index value for each health state(Euroqol Group 1990). It contains five questionsrelating to physical functioning, mental health andpain, and a self-rating of health on a ‘thermo-meter’ VAS. While its construct and convergentvalidity is adequate, it has suffered from moderateto low response rates in a number of populationsurveys, is highly skewed and has relatively poorsensitivity (Brazier et al. 1992, 1993a, 1993b; Bowl-ing 1998), particularly in relation to disease-basedoutcomes research (Casellas et al. 2000; Selai et al.2000). It has inherent methodological design faultsin the wording and range of its response scales,and it is possible that the length of the thermo-meter scale (0 to 100) is biasing. It is frequentlycriticized (Carr-Hill 1992; Jenkinson and McGee1998). However, it is widely used internationally,

and the development and refinement of the instru-ment is still in progress (Dolan et al. 1995).

Limitations of utility assessments of qualityof life states

The disadvantages of all these methods are theircost, the requirements for skilled interviewers andcomplexity (leading to reliance on non-random orunrepresentative samples of the public). Althoughthe Euroqol can be self-administered and has beenwidely used in postal approaches, inconsistenciesin reponses have been found suggesting that somemembers of the public find it too difficult (Carr-Hill 1992).

One of the main debates surrounding the use ofutility assessment techniques is who should pro-vide the utility values – the general public, healthcare providers and/or patients and their familiesthemselves? While patients’ assessments are un-doubtedly important, proponents of obtaining util-ity values of the public and health professionalsargue that patients’ values may change over thecourse of an illness and thus their utility valueswould not be stable. The ethical concern is thatthe judgements of quality of life and utility assess-ments are poorly understood (Cohen 1990). Areview by Kaplan et al. (1993a) has pointed toevidence from psychology experiments that sug-gest that methods commonly used for economicanalysis do not represent underlying true prefer-ences (Kahneman and Tversky 1983).

WHO SHOULD RATE QUALITYOF LIFE?

Discrepancies between doctors’ and patients’assessments

Patients own first-hand views are also essential inhealth-related quality of life outcome assessments.

Research suggests that there are wide discrepan-cies between patients’ and doctors’ ratings of out-come after specific therapies (Orth-Gomer et al.1979; Thomas and Lyttle 1980; Jachuck et al. 1982;Slevin et al. 1988). The case against observer rat-ings of another person’s quality of life, which isthen taken as a proxy indicator, has been made byseveral empirical studies comparing doctors’ and


patients’ ratings, and by the increasing literatureon discrepancies in treatment preferences. Thisliterature serves to indicate that the patient’sfeelings, values and opinions cannot be assumed.Jachuck et al. (1982) reported that, while all treat-ing doctors in their study rated their patients’ qual-ity of life as having improved after they startedanti-hypertensive treatment, three quarters of thepatients’ relatives thought that it was worse; 8 percent of the patients felt worse and 44 per cent feltthe same (the remainder felt they had improved).Poor correlations between professionals’ and pa-tients’ own self-assessments were also reported byPadilla et al. (1981). Large discrepancies in assess-ments between patients and doctors have been re-ported by Slevin et al. (1988), who compared theirassessments using the Karnofsky PerformanceScale, the Spitzer Quality of Life Index and theLinear Analogue Self-Assessment (LASA) Scale.

Relatively few investigators have based theiroperational definitions and measurement decisionson what the public say are the relevant domains ofquality of life. While some generic and disease-specific measures of health-related quality of lifehave included lay persons’ perceptions of theeffects of symptoms on their lives (e.g. Bergneret al. 1981; Hunt et al. 1986), investigators havemore commonly turned to the existing literature astheir starting point (Stewart and Ware 1992). Somescales, particularly in mental health and oncology,bypass the patient’s perspective altogether and baseratings on the perspective of a staff member or a‘significant other’ (e.g. relative or friend).

In health care, the debate sometimes revolvesaround whether the quality of life assessmentshould be made by the patient or by a health pro-fessional (e.g. a doctor). Objections to physicianratings include the argument that while patientsmay judge their quality of life to be low, theymay nevertheless value their lives as precious(Brock 1993). Some clinicians object to patients’ratings on the grounds of their subjectivity. Thissubjectivity, as it reflects the patient’s point ofview, should be viewed as their strength. There isnow general recognition among health servicesresearchers that measures of health outcome shouldincorporate the patient’s perspective, not simplyin terms of whether or not the treatment was asuccess, but more globally in relation to perceivedmental and physical well-being as a consequence

of an intervention. A person can feel ill withoutmedical science being able to detect any apparentdisease. A person’s ill health is indicated by feel-ings of pain and discomfort, or perceptions ofchange in usual functioning and feeling. It has beenknown for many years that the utilization of healthservices is more closely associated with the per-ception of symptoms and people’s feelings thanwith their actual medical condition (Mechanic 1962;Goldberg and Huxley 1980). Measures of healthoutcome need to take account of both the tradi-tional disease model (pathological abnormalityindicated by a set of symptoms and signs) andthe patient’s perspective.

The implication is that patients should completea questionnaire about their quality of life them-selves, or the questionnaire should be administeredto them by a trained interviewer. ‘Significantothers’ (e.g. relatives) and health care professionalsshould only complete ratings where their perspect-ive is also required, and where the patient is toofrail or ill to be questioned. However, a numberof scales require a health professional to completea questionnaire on behalf of the patient. These in-clude Spitzer et al.’s (1981) Quality of Life Index,the rating scales of dependency, work and schoolperformance, family and non-family relationshipsused by Horowitz and Cohen (1968), Taylor andFalconer (1968) and Rausch and Crandall (1982)in their studies of patients with temporal lobeseizures, and various other physical performanceindicators, such as the Barthel Index (Mahoney andBarthel 1965) and the Karnofsky Performance Scale(Karnofsky et al. 1948). On the other hand, al-though patient-based, self-report questionnaires arethe ideal to be aimed for, it is not denied that thereis also a practical need for supplementary, indirectmeasures of quality of life when the patient is toofrail or ill to respond. The criticisms of these havearisen because some investigators have selected thelatter rather than the former for primary routine use.

PATIENTS’ PREFERENCES

Patients’ preference assessments anddecision makingSome consideration should be given to patients’preferences for treatments in relation to their likelyeffect on their health and health-related quality of


life. Few outcome scales incorporate this aspectof the patient’s view, but it is arguably importantin understanding patients’ perceptions of, andsatisfaction with, their health outcome followingtreatment, and their preferences for different typesof invasive or non-invasive treatments.

A preference is an attempt to weigh up, con-sider and express a consistent value for alternativechoices of action (Till et al. 1992). A study ofpeople’s preferences involves the assessment oftheir attitudes (preferences) towards specificoptions, after informed deliberation on the risksand benefits, and in the light of people’s own valuesystems (Entwistle et al. 1998). The preferences ofpeople for alternative treatments need to be meas-ured in order to empower patients by taking theirviews into account, thereby improving opportun-ities for choice, and leading to a more rounded,patient-based body of knowledge on clinicalappropriateness. Such preference assessments aredistinct from the preference assessments of dif-ferent health states in the development of utilitymeasures (Blischke et al. 1975).

There have been few insightful attempts to iden-tify aspects of procedures that make them more(or less) acceptable to patients (Lippman et al.1985), and no attempt to identify the social andcultural variations in the views and preferencesexpressed, and the values underlying them.Research on preferences requires careful design toavoid complexity (Frogberg and Kane 1989), andquestion framing effects (McNeil et al. 1982), andto sample people for whom the options presentedare relevant, while in a hypothetical context. Onetechnique of eliciting people’s preferences, and theirintensity, is conjoint analysis, which has been usedmainly in market and transport research, and morerecently in health research on preferences for treat-ments, as well as in the development of healthstatus measures (Harwood et al. 1994; Ryan 1996;Ryan and Hughes 1997; Ryan et al. 1998). Withconjoint analysis, respondents are presented withhypothetical scenarios, and asked to express pref-erences between alternatives. Although it can becomplex and taxing for respondents, it has beenshown to be acceptable to people and achievesgood response rates in postal as well as interviewsurveys (Ryan 1996).

Understanding risk is likely to be importantfor assessing preferences and for both patient and

clinical decision making (Wennberg et al. 1988).People will vary in their understanding of per-sonal risks to health and in risk-taking attitudes,and this is likely to influence their preferences foreither conservative management or invasive inter-vention. A post-operative mortality risk of 5 percent, or even 25 per cent, may not be unaccept-ably high to the patient with risk-taking attitudes,especially when the prognosis without surgery ispoor in relation to quality of life and/or survival.Psychological research on risk locates this conceptwithin the individual and focuses on the conceptof perceptions of control (Conner and Norman1995), although the use of untested measures hasbeen widespread, and it has under-emphasized theeffects of social structure (e.g. family, friends,socio-economic status, workplace) on behaviour.Most health research on risk taking and perceivedcontrol relates to preventive behaviours and life-style among younger people, and has focused onlevels of ‘unrealistic optimism’ or ‘optimism–pessimism bias’ (Weinstein 1984). Little is knownabout most patients’ understanding and percep-tions of the risk to their quantity and quality oflife imposed by their condition and its treatment.Nothing is known about the effects of perceivedrisk, control, optimism bias, value attached tohealth, health status and social characteristics onpreferences for treatment. Such variables are rel-evant to patients’ perceptions of their treatmentand outcome, although it should also be notedthat not all patients may want to exercise a choice,and not all patients who do make choices opt forthe choice expected by clinicians (Wolberg et al.1987; Wilson et al. 1988; Fallowfield and Hall 1991;Wolberg 1991).

In relation to decision making about quality vsquantity of survival, attitudes diverge within pa-tient groups. Decision making is never clear-cut,as a brief review of some of the literature by Byrne(1992) emphasizes. While research on patientsacross disease conditions has reported that peoplegenerally agree about which health states are worsethan death (e.g. permanent coma, severe dementia,loss of essential functional ability such as beingable to feed oneself – Pearlman et al. 1993), thereis a great deal of variation among doctors in relationto treatment preferences (Mackillop et al. 1992), aswell as discrepancy between clinicians and patientson the treatment of choice (Mackillop et al. 1986,


1988, 1992; Palmer et al. 1990; Liddle et al. 1993;Schneiderman et al. 1993). There is also much re-ported discrepancy in relation to preference researchwith patients (McNeil et al. 1978, 1981; Slevinet al. 1990; Coates and Simes 1992).

DISEASE-SPECIFIC, DOMAIN-SPECIFICAND GENERIC MEASUREMENTSCALES

It is important to clarify the different types ofinstrument. A generic scale is useful when the rel-evant variables are covered and when investigatorswish to make comparisons of results between dif-ferent diseases and conditions. A domain-specificscale is required when the area covered is of par-ticular relevance to the study and its hypotheses,and where generic and disease-specific scales neg-lect that area. A disease-specific scale is used whendisease or condition-related attributes need to beassessed, and greater sensitivity to the clinical con-dition under consideration is required. It is oftenrecommended that a generic measure should beused alongside a disease-specific measure in clin-ical trials, in order to address both clinical andbroader policy questions, and to detect unexpectedpositive or negative effects of interventions (Bom-bardier et al. 1995; Fitzpatrick et al. 1998). Thisdoes carry the problems of respondent and researchburden, the need for an increased number of sta-tistical tests thus increasing the likelihood of ob-taining statistical significance by chance (Fitzpatricket al. 1998). Fitzpatrick et al. (1998) have describedthe requirements for selecting and judging theappropriateness of measurement scales, from psy-chometric criteria and their clinical significance tothe importance of ensuring that a body of evid-ence exists to justify the selection of measures thathave social significance (i.e. measures whichinclude domains and items that reflect areas ofimportance to patients).

Disease-specific scalesDisease-specific measurement scales have the aimof being more clinically and socially significant inrelation to specific conditions – of being able todiscriminate more finely between patients’ levelsof severity of condition, and of being more sens-itive to their clinical outcomes.

In choosing a measurement instrument, or set ofmeasures, key questions to consider are whether adisease-specific and/or a generic measure is needed,and whether either requires supplementation withsingle domain measures that are important to thestudy aims (e.g. depression). There is little pointin utilizing a health status measure alone if it isunlikely to detect the effects of the treatment inquestion, or the symptoms specific to the condi-tion. In addition, some measure of disease severitywill also be required. Several severity and co-mor-bidity indexes have been developed (Kellerman andHackman 1988; Parkerson et al. 1993a), althoughthese are often fairly crude (Linn et al. 1968) andclinical investigators prefer to rely on biomedicalindicators. Most health services researchers workclosely with clinicians to ensure that the appropri-ate disease-specific outcome and severity indica-tors have been included in their batteries.

Clearly, criteria for assessing outcome of carewill vary for different disease syndromes. A uni-versal questionnaire to elicit the relevant informa-tion for a number of conditions would require aquestionnaire of enormous length (Goligher 1987).Disease-specific quality of life scales are needednot simply for greater brevity, but to ensure sens-itivity to sometimes small, but clinically signific-ant, changes in health status and levels of diseaseseverity (for good illustrations of this point in re-lation to head and neck cancer, see Berg et al. 1976;Dhillon et al. 1982; Morris 1990). The domains offocus in relation to outcome assessment across thedifferent diseases and conditions vary considerably,and this diversity is reflected in the disease-specifichealth-related quality of life scales. For example,in psychiatry a main area of focus is on the ar-rangements for care, and consequently measure-ment scales reflect a needs-based approach (e.g.housing, money, shelter and so on, from the per-spective of both patient and carer). In this sensethe scale used will reflect the tradition of socialindicators research. In oncology, the toxicity oftreatment and the aggressive nature of the condi-tion are of importance; in relation to respiratoryand cardiovascular conditions, the symptoms andthe restrictions they impose on everyday life andactivities are relevant; in rheumatology it is lim-itations on activities and functioning, as well aspain, that are the focus for measurement. Thesevariations also reflect the evidence that different


areas of life are affected by different conditions(Bowling 1996a, 1996b).

However, while different procedures requireappropriate disease-specific quality of life out-come measures, there can still be a central core ofquality of life questions common to a wide rangeof disease areas. It is easy to assume that disease-specific measures per se will be more sensitiveindicators of patient outcome than generic meas-ures. However, a study of osteoarthritis patientsby Kantz et al. (1992) reported that while knee-specific function measures and pain scales weremore specific than the generic health status scale,the SF-36 (Ware et al. 1993), among patients withother co-morbid conditions, the SF-36 physicalfunctioning sub-scale, plus knee-specific adapta-tions of that scale were just as specific as thedisease-specific scales. Moreover, the disease-specific measures failed to distinguish betweentreated and untreated patients. A combined ap-proach to outcome assessment, using both disease-(or condition-) specific and generic measures ispreferable where a broad disease-specific qualityof life instrument has not been satisfactorily de-veloped. The generic measure of choice acrossmany diseases is increasingly the SF-36. As Ware(1993) concluded, because the SF-36 is short, welltested, and population norms exist, it may con-stitute a good generic core for use along withdisease-specific outcome measures. The potentialuse of this scale as a generic core will be discussedmore fully within the appropriate chapters of thisbook (see Chapter 8 for review).

Domain-specific scales

As the reviews of disease-specific indicators ofquality of life will show, they can often be criti-cized for being too narrow in focus, while neglect-ing the measurement of important outcome andmodifying variables (e.g. social support, adjustment,coping, life satisfaction, self-esteem, depression andother domains). Some investigators supplementtheir disease-specific scales with domain-specificmeasures. Generic and domain-specific measureshave been reviewed by several authors (e.g. Wade1992; Wilkin et al. 1992; McDowell and Newell1996; Bowling 1997a).

The domain-specific areas of interest will varyaccording to how the condition and its treatment

affect the patient. Thus measures of psychiatricstatus will, for some diseases and conditions,necessitate the inclusion of a memory test, as wellas a depression scale. The measurement of phys-ical functioning may be restricted to global categor-ies ranging from fully functioning to bed-boundfor more dramatic conditions and interventionswhere great changes are expected, but may needto be more refined and sensitive (i.e. at the ‘lessrestrictive’ end of the scale) in the case of moremoderate cases.

Other domains of potential relevance may in-clude occupational and social role functioning,including maintenance of social relationships andactivities (Kaplan 1985; Bowling 1994a, 1997a;Bowling et al. 1994). Psychological well-being canbe important, and includes happiness/satisfaction,self-mastery, expectations, personal control, adjust-ment and coping ability (influencing or modifyingvariables), well-being, emotional support and inter-action (outcome and modifying variables), self-concept and self-esteem, body image and somaticcomfort (Lewis 1982, 1989; Young and Longman1983; Schipper et al. 1984; Schipper and Levitt1985, 1986; Ferrans and Ferrell 1990).

Some argue that a definition of quality of lifeshould reflect the pre-illness situation of the pa-tient; others argue that it should reflect the attain-ment of an ideal quality (Calman 1984; Schipperand Levitt 1986). There is, of course, an individualjudgement involved and people differ in their val-ues. In medicine, the goal is to return patientsto normal lives, rather than attain the ideal, andmeasurement scales should reflect this (Selby andRobertson 1987). Measures of quality of life shouldtherefore reflect the range of normal activities thathave been potentially affected by the condition andtreatment. It is possible to obtain good responseand completion rates with a comprehensive ques-tionnaire (Sadura et al. 1992).

The specialized scales measuring specific do-mains are often long, and have to remain fairlylengthy in order to retain their sensitivity and psy-chometric properties. A caution must be made herein relation to the use of batteries of several scales,for example as a supplement to a brief disease-specific scale. Although disease- and condition-specific scales of quality of life are easy to criticizefor their brevity, and their often narrow scope, nostudy can hope to measure every relevant domain,


and most disease-specific scales are limited to themost essential. Domain-specific scales should beused when the area is of particular interest to theinvestigator, and the disease-specific (or generic)scale selected for use neglects that domain. Theyshould not be included for the sake of having acomprehensive battery, without any theoreticalor methodological justification. The danger is theproduction of a battery that contains too many –and also overlapping – items, which becomes tire-some to respondents, expensive to administer andanalyse, and produces an overwhelming amountof data which may not always be helpful.

Generic scales

Measures which implicitly or explicitly aim to taphealth-related quality of life are usually referred toas broader measures of health status. They shouldencompass the dimensions of physical, mental andsocial health. Investigators have tended to supple-ment generic health status measures with specificdisease items. They have used generic measures inorder to make comparisons with other conditions,to broaden their outcome indicators, and becauseof the slow development of disease-specific ques-tionnaires. On the other hand, generic instrumentshave an important constraint as they are unable toidentify the condition-specific aspects of a diseasethat are essential for the measurement of outcome(Hutchinson and Fowler 1992). Generic measureswill always require supplementation with disease-specific measures in order to detect important clin-ical changes (Guyatt et al. 1986). McKenna (1993)has argued that the role of generic measures inhealth services and clinical research will diminishas more disease-specific measures are producedwhich can focus directly on the research issues. AsMcKenna points out, the use of disease-specificmeasures avoids asking irrelevant questions of re-spondents and maximizes the chance of detectingclinically significant changes, which is essential inclinical and policy-orientated research. However,it is unlikely that they will obviate the need forgeneric measures, or at least a generic core, as longas comparisons across disease groups within andbetween specialities are required. The most popu-lar generic measures include the Rand batteries,in particular the increasingly used SF-36 (Carteret al. 1976; Deyo et al. 1982; Bergner 1984, 1988,

1993; Deyo 1984a; Stewart and Ware 1992; DeBruin et al. 1993; Ware et al. 1993), the SicknessImpact Profile (Bergner et al. 1981), the Notting-ham Health Profile (Hunt 1984; Hunt et al. 1986)and the McMaster Health Index Questionnaire(Chambers et al. 1976, 1987; Chambers 1993). Apopular and promising generic measure that isbeing developed for use in primary medical care isthe Dartmouth Coop Function Charts (Nelson etal. 1987, 1990; McHorney et al. 1992; see reviewby Wilkin et al. 1992). However, while health-related quality of life measures are increasingly usedin health services research, several reviews of clin-ical trials have indicated that they are still underusedby clinicians (e.g. Guyatt et al. 1989b).

Satisfaction with care and outcome

This domain is listed separately because it is animportant and neglected component of disease-specific and generic scales or batteries of scales. Aperson’s degree of satisfaction with their healthstatus and outcome, and degree of fulfilment ofexpectations of the treatment, should be included.It is arguably an outcome as well as a process indic-ator of the quality of care. It is usually neglectedbecause the detection of dissatisfaction has foiledall patient satisfaction questions and questionnaires.Reviews of the problems inherent in measuringpatient satisfaction can be found in Roberts andTugwell (1987), Ware and Hays (1988), Cartwright(1989), Fitzpatrick (1990), Wilkin et al. (1992) andvan Campen et al. (1992). Davies and Ware (1991)have developed a promising Consumer SatisfactionSurvey Questionnaire (including a ‘Visit-specificSatisfaction Questionnaire’) and user manual. The‘visit-specific’ questions were developed from theRand Medical Outcomes Study Questionnaires.Readers are referred to these sources where theirmeasurement scales of choice do not addresspatient satisfaction issues.

Computerized item banking

The issue of which domains to include in batteriesof measurement scales may be less problematicand taxing in the future if computerized item bank-ing is developed, as in educational research (Uttaroand Lehman 1999). With this, items measuring allpossible life areas of importance to people and also


those that are important to measure clinically areentered onto a computer database; the computer isprogrammed to select the relevant questions foreach respondent based on their responses to a setof initial screening questions about their condi-tion, its effects on their life, their circumstancesand their perceptions of the important areas of life.The aim will be to reduce the length and burden ofthe questionnaire for respondents, while retainingits sensitivity and a scoring system that reflectsindividuals’ concerns and enables comparisons be-tween people.

METHODOLOGICAL ISSUES

Other texts and specialized review articles havedescribed research methodologies, question anditem response design, scaling methods and a widerange of statistical issues – thus the principles ofresearch will not be repeated here. Readers arereferred elsewhere for information on thesemethodological issues (e.g. Webb et al. 1966;Oppenheim 1968; Blalock and Blalock 1971; Moserand Kalton 1971; Andrews and Crandall 1976;Pocock 1983; Sudman and Bradburn 1983; Barkerand Rose 1984; Kleinman 1986; Armitage and Berry1987; Cella and Cherin 1987; Selby and Robertson1987; Bellamy 1989; Spector and Thompson 1991;Fletcher et al. 1992b; Streiner and Norman 1995;McDowell and Jenkinson 1996; McDowell andNewell 1996; Bowling 1997b).

However, an introduction to some of the basicpsychometric concepts, as well as operationaliza-tion and level of data, is required for the inter-pretation of the descriptions of scale reliability andvalidity in the text.

Operationalization

There has always been considerable debate amongsocial scientists about ‘operationalism’, and theextent to which the gap between theory and em-pirical research can be bridged (for a discussion,see Blalock and Blalock 1971). However, for theconcept of quality of life to have any value in de-scriptive or health outcome research, it must bedecided what is to be measured and the agreed con-cepts need to be defined and translated into anobservable form. This is essential if quality of life

is to be researched in any coherent and scientificway – and necessary before a quality of life scalecan be said to have content validity.

Scepticism has often greeted attempts to meas-ure quality of life in clinical studies (Muldoonet al. 1998). This is unsurprising given that mostinvestigators in the past have not defined oroperationalized their terms (Kaplan and Anderson1990). Many assume that the concept refers tophysical functioning, and/or psychological andmental status or symptom levels. Many scales pur-porting to be, or used as, quality of life scales are,in fact, simply health status scales, and they arealso frequently single-domain scales (e.g. level ofphysical functioning or mental health). Becausemost so-called ‘quality of life’ instruments, andtheir proxy measures, do not attempt to measurethe philosophical, external or environmental as-pects of quality of life, but focus on health-relatedareas, Fitzpatrick et al. (1998) have consequentlylabelled the term ‘quality of life’ as misleading,and rejected its use in favour of the term ‘patient-based outcome measures’. Annas (1990) alsoargued that the term ‘quality of life’ is so misusedthat it should be banished from our lexicon.

Reliability and validity

Measuring the quality of life of patients is inevita-bly difficult. Questions about the sensitivity, reli-ability, validity and generalizability of data continueto be raised because of the complex nature of dis-eases, treatments (Smart and Yates 1987), expectedrecovery times (Bardsley and Coles 1992), and, ofcourse, the concept of quality of life itself.

Whichever measure, or battery of measures isselected, all instruments should satisfy the criteriafor reliability and validity. Unfortunately, manyinstruments have been developed ‘ad hoc’ andmany studies are based on instruments with weakpsychometric properties, or little evidence of theirreliability and validity (Coste et al. 1995). Validityis the extent to which an instrument measures whatis intended. Reliability is the extent to which meas-urements on the same respondent are similar on re-peated applications of the measure at different times.Another important property of quality of life scaleswhen used as outcome measures is their respon-siveness to change, particularly clinically import-ant changes. There is an unresolved debate about


whether responsiveness is an aspect of validity (Haysand Hadhorn 1992). Several texts describe these con-cepts in more detail (e.g. Carmines and Zeller 1979;Kline 1986; Streiner and Norman 1995; Bowling1997b). They are now described briefly here.

Reliability

Reliability refers to the ability to produce consistentresults, and consistent results on different occasions,when there is no evidence of change. This is testedby test–retest reliability (administration of the scaleon different occasions to the same population),internal consistency (measurement of the same con-cept by different scale items), and inter- and intra-rater reliability (consistency of a measure whenadministered by different interviewers or the sameinterviewer on different occasions). It is impor-tant that the setting of the administration is thesame on each occasion to prevent contamination(Ziebland and Fitzpatrick 1992).

A few of the appropriate statistical techniquesfor assessing reliability are referred to below, inorder to clarify the chapters in this volume wherevalues are presented. Readers are referred to thereferences and to appropriate methodological andstatistical texts for further details, and for detailsof other tests as well as appropriate levels of statis-tical significance in relation to sample size (e.g.Pocock 1983; Armitage and Berry 1987).

Internal consistency involves testing for homo-geneity. Tests assess the extent to which individualitems are inter-correlated and the extent to whichthey correlate with overall scale scores. Internalscale consistency is usually tested with inter-itemcorrelations and with Cronbach’s alpha (Cronbach1951), based on the average correlation among theitems and the number of items in the instrument(values range from 0 to 1). Although some regard0.70 as the minimally acceptable level for internalconsistency reliability, others accept the level >0.50as an indicator of good internal consistency, aswell as of test–retest reliability (Cronbach 1951;Ebel 1951; Helmstater 1964; Cronbach et al. 1972;Nunnally 1978). It is accepted practice to rejectscale items below r = 0.20 (Kline 1986). Streinerand Norman (1995) cite an alpha of 0.85 or aboveas indicating adequate internal consistency.

The items of the measure can also be split intotwo halves and the alphas for the alternative forms

compared (the Spearman–Brown formula uses thiscorrelation to estimate reliability after adjustingfor the number of scale items); or the scale can bedivided into two groups and coefficients can becomputed for each half and compared. Compar-able coefficients confirm the consistency of theresponses (Zeller and Carmines 1980; Williamsand Wood-Dauphinee 1989).

The interpretation of alpha is that, for example,if Cronbach’s alpha is high (e.g. 0.80), then theimplication is that the responses are consistent, andthe sum of the item responses yields a score forthe underlying dimensions that the item represents.If the items are adequately sampled from the qual-ity of life domain, the sum of the responses shouldgive a better indication of quality of life for therespondent than responses to any one item. A lowcoefficient alpha, on the other hand, indicates thatthe item does not come from the same conceptualdomain (Williams and Wood-Dauphinee 1989).

Other tests of reliability include inter-rater agree-ment (the concordance, or reliability, of scores bydifferent raters on a single occasion) and intra-rateragreement – the reliability of the rating by the samerater (of the same subjects) on different occasions(repeat testings, or test–retest). If the measure iscategorical, then the most appropriate statisticalmethod to employ is Cohen’s (1968) Kappa Testof Concordance. Kappa has a value of 0 if agree-ment is no better than chance, a negative value ifworse than chance, and a value of unity (1.0) ifagreement is perfect. Fleiss (1981a, 1981b) suggestedthat less than 0.40 is poor agreement, 0.40–0.59 isfair agreement, 0.60–0.74 is good agreement, and0.75–1.00 is excellent agreement. Spearman’s Rhoand Kendall’s coefficient of concordance, W, mayalso be considered (the latter in cases where morethan two sets of ranked data are involved). How-ever, simple correlation analysis is less appropriatefor studying concordance as it makes no allowancefor chance agreement. Pearson’s product–momentcorrelation is used for comparing continuous scores.Intra-class correlation coefficients compare the vari-ance between subjects, the variance between ratersand the variance between times with error variance(Shrout and Fleiss 1979; Fleiss 1986). An intra-classcorrelation coefficient of, for example, 0.80 ormore indicates that the scale is highly reliable.

The stability of a measure over a period of timeis measured by repeated administrations of the test


on subjects and domains not expected to changein their scores over a carefully defined time per-iod. Test–retest reliability is generally assessed bycorrelations of the measure administered on thedifferent occasions. A high level of consistency inresponse is desired. A test–retest correlation of0.85 or more indicates that the instrument has anacceptably low level of random measurement er-ror (McDowell and Newell 1996). However, theobserved associations can sometimes be difficult tointerpret – a low correlation may reflect a genuinechange in health status, rather than poor reliabil-ity. Some investigators also feel that correlationsare a weak measure of test–retest reliability, andrecommend the use of confidence intervals to as-sess the size of the difference between the scores(Bland and Altman 1986; Ruta et al. 1994a).

Validity

Validity refers to the extent to which the scalemeasures the underlying concept of interest. Scalesalso need to have predictive validity (to be sens-itive to real changes in condition) and to havesocial and clinical validity (relevance). That is, theyshould tap areas of social importance to the pa-tient and should be responsive to the different dis-ease stages and fluctuations due to the conditionand its treatment over time (Aaronson et al. 1993).Validity is measured by testing for face validity(at face value, does it appear to be measuring whatit is intended to measure; is it unambiguous andappropriate?), content validity (does the scale tapall relevant concepts of the attribute to be meas-ured?) and criterion validity (the extent to whichthe measure correlates with a ‘gold standard’).Validity is often difficult to assess because thereis no objective ‘gold standard’ measure of qualityof life. Criterion validity encompasses concurrentvalidity (correlations with an existing measure ofthe same construct) and predictive validity (correla-tions against other measures to assess predictivepowers). Quality of life scales are often reliant onpredictive validity to demonstrate their psycho-metric properties. There is also construct validitywhere hypotheses are generated and the scale testedagainst a measure central to the hypothesis.

Construct validity is usually divided into con-vergent validity (tests for correlations with otherindicators intending to measure the same concept)

and discriminant validity (lack of correlationswith unrelated indicators) (see Campbell and Fiske1959).

Hyland (1992a, 1993) has cautioned against reli-ance on claims for content validity by scale devel-opers. He pointed out that quality of life comprisesseveral connected constructs, and any question-naire may reflect one or more of these constructs.In the absence of an agreed theoretical descriptionof quality of life, validity can be difficult to assess.Moreover, standardized questionnaires, with stand-ardized scoring and weighting procedures, imposean external value system on respondents. Suchmeasures may often satisfy conventional tests ofreliability and validity, but may not be relevantto all individuals and thus lack content and socialvalidity (Hickey et al. 1996). Research using indi-vidualized measures has indicated that areas ofrelevance to people are not always represented instandardized scales (e.g. in relation to people withAIDS, important areas which emerged related tospirituality, sexuality and issues relating to death;in relation to older people nominated areas includedreligion, independence, finances and happiness)(Browne et al. 1994; Hickey et al. 1996).

Factor structure

Questions that deliberately tap different dimen-sions within a scale cannot be expected to neces-sarily have high item–total or full item–itemcorrelations. Therefore, factor analysis should beused to identify the separate factors within thescale. Factor analysis is a technique which definesa small number of underlying dimensions (factorsthat account for a high proportion of the com-mon variance of the items), and in so doingdemonstrates whether the items in the scales grouptogether in a consistent and coherent way. Ex-ploratory factor analysis is usually used in scaledevelopment in order to identify and discard itemsthat are not correlated with the items of interest.Factor analysis is also used to confirm that scaleitems principally load onto that factor and cor-relate weakly with other factors. While smallsamples may be used in analysis, ultimately a con-firmatory factor analysis should include a largersample for assessment of stability (Comrey 1978;Loo 1983). A factor is considered important, andits items worthy of retention in the scale, if its


eigenvalue (a measure of its power to explain vari-ation between subjects) exceeds a certain value.Jollife and Morgan (1992) state that this valueshould be 1.1, although 1.5 appears to be com-monly taken.

Translation and cultural equivalenceElaborate guidelines are available which stress theneed for rigorous assessment of cultural equiva-lence when questionnaires are translated, and inthe back translation (translation back to originallanguage) procedures (Bullinger et al. 1993). Thereremains concern that such procedures may still failto identify cultural values not initially anticipatedin the culture where the instrument was firstdeveloped, and therefore may have little contentvalidity (Fitzpatrick et al. 1998). The innovatoryapproach of the WHOQOL Group (1998a, 1998b)in their recent development of WHO’s Quality ofLife Assessment instrument, intended for interna-tional use, was therefore to develop the conceptsand items for inclusion in 15 study sites across theworld. The disadvantage of this approach is theconsequently long length of the scale (100 items).

Response formatsThere are several possible response formats for usewith closed questions. Most responses are usuallyin a categorical dichotomous format (e.g. yes/no),a categorical Likert scale (Likert 1952) (e.g. a ser-ies of ordered responses, such as ‘strongly agree,agree, neither agree nor disagree, disagree, stronglydisagree), or a visual analogue scale (VAS) (re-spondents are asked to place a mark on a num-bered or unnumbered line, usually 10cm in length,on which opposing statements or descriptionsare placed at each end). A range of other responsescales exist (Thurstone 1927a, 1927b; Guttman1944; Osgood et al. 1957; Cantril 1965; Andrewsand Withey 1976). They have been described morefully by Bowling (1997b). The choice of responseformat is important as this determines the type ofstatistical techniques that can be applied.

Measurement scales vs single item questions

The superiority of multi-item measurement scalesover single-item questions has been demonstratedby Davies and Ware (1981) and Manning et al.(1982). The quality of data collection is improvedby the use of standardized, well-tested scales, withgood psychometric properties. In addition, accountmust be taken of the type of scoring that the in-strument is based on and whether it yields:

• Nominal data: numbers are used simply for clas-sification (e.g. ‘died’, ‘survived’).

• Ordinal scale data: scale items stand in somekind of relation to each other (e.g. ‘very diffi-cult’ through to ‘not very difficult’).

• Interval scale data: the characteristics of an or-dinal scale but the distances between any twonumbers on the scale are of a known size (e.g.temperature).

• Ratio scale data: the characteristics of aninterval scale with the addition of a true – notarbitrary – zero point (e.g. weight).

The full definitions of these concepts can be foundin most research method textbooks (e.g. Blalockand Blalock 1971; Moser and Kalton 1971). Theirconsideration is crucial because different scales aresuitable for providing different types of data andfor answering different types of question. In addi-tion, interval and ratio scales are required ifpowerful statistical analysis is envisaged. Somemeasurement scales, which technically providecombined nominal and ordinal scale data, haveformulas for re-coding, summing and transform-ing response categories, using a scoring algorithm,into a scale ranging from 0 (worst) to 100 (best)(e.g. Ware et al. 1993, 1995). This allows the res-ults to be reported as means, rather than frequencydistributions, and for the use of more powerfulstatistical techniques. This is justified on thegrounds that the treatment of such data at intervallevel has minimal effects on most statistical pro-cedures, although this is not without controversy( Julious et al. 1995; Lamping et al. 1998a).

a review of disease-specific quality of life measurement ... · pdf filea review of...

Documents