Health-related quality of life

Health-related quality of life is the part of quality of life that concerns health: a subjective, multidimensional construct that the patient reports, and one of the patient-reported outcomes. Its measures rest on models of how health relates to quality of life and turn answers into indices or profiles on ordinal or continuous scales. They fall into generic, disease-specific, dimension-specific, individualised and preference-based families, and their scoring rules and conditions of use decide how their numbers can be reported.

Health within quality of life

Humans have always desired to live a good life, a high quality of life, although what that means differs for each person. ¹ Quality of life is complex and encompasses many aspects of life. ¹ The quality-of-life index developed by The Economist Intelligence Unit, for example, includes material well-being, health, political stability and security, family life, community life, climate and geography, job security, political freedom and gender equality. ¹ Health is one of these determinants, and quality of life applied to medicine in this way is often called health-related quality of life. ¹

The International Society for Pharmacoeconomics and Outcomes Research defines health-related quality of life as a broad theoretical construct developed to explain and organise measures concerned with the evaluation of health status, attitudes, values, and perceived levels of satisfaction and general well-being, with respect either to specific health conditions or to life as a whole, from the individual perspective. ¹ One account of quality of life describes its more specific aspects, its dimensions, as general health with a focus on functional status both physical and emotional; checklists of symptoms; daily living activities; personal well-being; sexual functioning; and the impact of illness on social, emotional and family functioning. ^{1, 2, 3} Its authors conclude that health-related quality of life is a multidimensional construct, and they emphasise that subjectivity is a basic and unquestionable feature of it, so that it can be evaluated only by patients themselves. ³ The only exceptions they allow are patients incapable of providing the information, such as small children, patients with communication problems, or patients intellectually incapable of responding effectively; for them, proxy measures are acceptable. ³

The World Health Organization defines health as a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity; the definition does not mention health-related quality of life, but it clearly makes aspects of well-being principal attributes of full health. ^{3, 4} In the context of that definition, health-related quality of life and health status are closely interrelated and should not be considered separate. ⁴ Health-related quality of life belongs to the family of patient-reported outcomes, and its measurements attempt to turn subjective reports into objective data. ⁵ That requires properly developed and well-validated measures, built following strict and well-defined rules such as those in the guidance on patient-reported outcome measures for medical product development to support labelling claims that the US FDA published in 2009. ⁵ Developing such a measure is laborious, time-consuming, and needs a highly skilled and knowledgeable team of researchers. ⁵

Models behind the instruments

A variety of models of health-related quality of life have been proposed, and they are the basis for the various measures. ⁶

• Ware's model, one of the first, specifies three generic health concepts, physical health, mental health and general health, and places them on a continuum. ³ Physical conditions are closely linked to physical symptoms, which lead to physical limitations and reduced well-being; mental conditions relate to mental symptoms and so cause psychological distress and poor well-being; and both can severely impair perceived general health. ³ This model underlies the SF-36. ³
• Wilson and Cleary's model is also continuum-based and highlights increasing complexity as one moves from biological and physiological factors at the lowest level, through symptoms, functioning and general health perceptions, to overall quality of life at the highest. ⁶ Across those levels it also incorporates relationships and other factors, such as the characteristics of the individual and of the environment and non-medical factors. ⁶
• Nagi's disablement model, introduced in 1965, was the first of its family and comprises active pathology, impairment, functional limitation and disability. ⁶ Impairment is a structural abnormality at the anatomical level of cells, tissues or organs; functional limitation is difficulty performing activities; disability is categorised as physical, mental, social or emotional and covers a person's ability to fulfil a role in life. ⁶ Nagi's main contribution to patient-centred medicine was to move disability away from pure physical dysfunction towards interactions between patients and their environment, and more recent disablement models, such as the National Center for Medical Rehabilitation Research model and the WHO International Classification of Functioning model, are rooted in his concept. ⁶
• Calman's expectation model assumes that quality of life reflects the distance between an individual's present experience and their expectations. ⁶ A smaller gap corresponds to better quality of life, and quality of life can be improved either by improving the current situation, for example by treating disease, or by modifying expectations. ⁶ The Schedule for the Evaluation of Individual Quality of Life and the Patient Generated Index are based on it. ^{6, 7}
• The need-based model applies a similar rationale to human needs: quality of life depends on the personal capacity to satisfy them, poor health interferes with satisfying them and so lowers quality of life, but the model assumes that as long as the primary needs are fulfilled — by compensation mechanisms, for instance — quality of life remains unchanged. ⁷ The Quality of Life Assessment of Growth Hormone Deficiency in Adults and the Quality of Life in Depression Scale are examples. ⁷

One patient-reported outcomes researcher describes Wilson and Cleary's as the conceptual model well known within the German healthcare system, basic but very effective at classifying different measures of health outcome, and holds that it might be useful for medical writers new to patient-reported outcomes. ⁸

Items, indices and profiles

Instruments are built from items — questions or statements — and can contain one or several. ⁹ A single-item measure generates an index, a single number. ⁹ A multi-item measure may be unidimensional or multidimensional, and generates either an index or a profile, a set of scores for each measured dimension or subscale. ⁹

Profiles provide more detailed information about the characteristics of health-related quality of life and enable a better understanding of the problems respondents face, which makes them suitable for clinical practice — but they may not capture an overall change in magnitude and direction. ⁹ The Nottingham Health Profile is a good example of a multidimensional measure, covering sleep, pain, emotional reactions, social isolation, physical mobility and energy level. ⁹ For some profiles a simple sum of dimension scores is accepted, although its accuracy is questionable, because the calculation assumes each dimension is of equal importance, which is often not the case. ⁹ Applying weights — relative values for each dimension or even each item — overcomes the problem, and a single aggregated score derived that way is believed to be robust and appropriate; for the Nottingham Health Profile an index can be computed from weighted or unweighted dimension scores. ⁹

Scales and scores

Information is collected in different ways and many, but not all, measures are based on an ordinal scale. ¹⁰ The simplest are dichotomous variables describing health status, such as non-diseased or diseased, or a yes-or-no answer about a specific problem; this is often how need-based measures are constructed. ¹⁰ Their score can be generated by summing the number of yes answers — the number of recognised problems — so a higher numerical score denotes poorer quality of life and a decrease in the score indicates improvement. ^{10, 11} Such a descriptive classification distinguishes between categories and orders them hierarchically. ¹¹

The most frequently used scales for measuring quality of life are rating scales such as the Likert scale and the visual analogue scale. ¹² Likert introduced his in 1932 to measure social attitudes in the US; the items he included, particularly in the "Negro scale", are now considered shocking, but the way he proposed to collect information is widely used. ¹² A Likert scale contains one or more statements, each linked to several choices usually ordered from lowest to highest — worst to best, not important at all to extremely important, never to always, completely disagree to fully agree — with intermediate choices between. ¹² Five options are usual for most items, the number can vary from three to nine, and an odd number is recommended so that a neutral choice is available. ^{12, 11} For analysis the choices are coded as sequential numbers, for example 1 for worst to 5 for best, and those numbers are summed to give a single index. ¹¹

The raw score, called the working score, is sometimes standardised to a 0-to-100 scale to make comparison between different measures easier. ¹¹ The standardised score is called the scale score, and standardisation to a 100-point scale is called the standard scoring method. ¹¹

The visual analogue scale, first used by Priestman and Baum to measure quality of life in patients with breast cancer under the name linear analogue self-assessment, is a 100-mm horizontal or vertical line on which a respondent places a mark in response to a question. ¹² It is anchored at one end by the lowest choice and at the other by the highest, and the score is the measured distance from the left end to the respondent's mark, so it is a continuous scale that generates a single score. ¹²

Uses and the requirements they set

Patient-centred medicine has come out of the increasing importance of patients' voices in disease management, and as part of this health-related quality of life has become an important part of assessing treatment outcome and the quality of patient management. ^{2, 13} Health-related quality of life is widely used in daily patient management, clinical trials, health economics and medical decision-making, and each application imposes different requirements on the measure. ¹⁴ Clinical use usually requires one that captures specific changes within a certain disease, within a patient population in a clinical trial, or for an individual patient in daily practice. ¹⁵ Pharmacoeconomic evaluation often requires health status expressed as a single summary index capable of identifying and quantifying differences across diseases and aggregating changes in patient health status over time. ¹⁵ That divergence of requirements explains why so many measures have been developed. ¹⁵

Families of measure

Measures are classified by the purpose of measurement and the target population as generic, disease- or population-specific, dimension-specific, individualised, or preference-based; preference-based measures are sometimes called utility measures because they primarily serve to generate utilities, though utilities can also be derived from certain generic and disease-specific measures. ¹⁶ Some measures fall into two categories at once. ¹⁶ Each type serves a different purpose and should be applied to different populations: generic measures can be used in general populations and across various diseases, whereas disease-specific ones address specific diseased populations. ¹⁷

Generic measures are designed for use in any population irrespective of disease status. ¹⁸ Many focus on physical function and measure impairment, disability or handicap; others cover psychological issues. ¹⁸ They are often considered insufficiently sensitive to detect changes specific to certain diseases, but they allow comparisons across different conditions and with general populations. ¹⁸ The most widely used are the EQ-5D, the SF-36 and the Nottingham Health Profile. ¹⁸

• The EQ-5D defines five dimensions of health: mobility, self-care, usual activities, pain or discomfort, and anxiety or depression. ¹⁹ In the original version, now called the EQ-5D-3L, each dimension is categorised into three levels of burden — no problem, a moderate problem, an extreme problem. ¹⁹ Respondents first indicate the level that applies to them and then record their perception of their general health state on the EQ visual analogue score. ¹⁹ It is available in more than 170 languages and is used in clinical and economic studies and population surveys worldwide. ¹⁹ Two further versions exist: the EQ-5D-5L, introduced in 2009 to increase sensitivity and reduce the ceiling effect of the three-level version by adding the intermediate categories slight and severe, and the EQ-5D-Y, which targets children and adolescents aged 8 to 15 and is also available as a proxy measure. ¹⁹ All can be administered on paper or electronically. ¹⁹ Using any of them requires the study or project to be registered with the EuroQol Research Foundation and the conditions of use to be agreed. ¹⁹
• The SF-36 measures physical and mental health and provides an assessment of general health. ²⁰ Physical health comprises physical functioning (10 items), physical role functioning (4), bodily pain (2) and general health (6); mental health comprises vitality (4), emotional role functioning (3), social role functioning (2) and mental health (5). ²⁰ Most items use a Likert scale. ^{20, 19} Shorter versions exist — the SF-6D, SF-12 and SF-20 — of which the SF-6D is used primarily in health economic evaluations. ¹⁹ These measures are in the public domain and free of charge, subject to legal conditions such as proper acknowledgement, and can be downloaded from the RAND Corporation. ¹⁹ Following the scoring instructions is crucial, because the items for physical and mental health are constructed in opposite directions. ¹⁹ Raw scores can be standardised to a 100-point scale assuming equal weighting for each item, and for some countries, Germany among them, country-specific weights are available and should be used for national data. ¹⁹ Overall, a lower score denotes poorer quality of life. ¹⁹
• The Nottingham Health Profile focuses on feelings and emotions rather than physical performance, and comprises 38 statements in six dimensions. ²¹ The respondent selects yes or no according to whether a problem applies, and the score is the number of yes answers, so a higher score denotes poorer quality of life. ¹⁹ Galen Research holds the copyright and should be contacted for permission to use it. ¹⁹

Disease-specific measures are developed to monitor patients with increased accuracy and to give enough sensitivity to detect the features of specific conditions, and many are available across patient populations. ²²

• The EORTC QLQ-C30, developed by the European Organisation for Research and Treatment of Cancer, was one of the first disease-specific measures. ^{23, 24} It is multidimensional, with five functional scales (physical, role, cognitive, emotional and social), three symptom scales (fatigue, pain, and nausea and vomiting), a global health status and quality-of-life scale, single items such as dyspnoea, loss of appetite, insomnia, constipation and diarrhoea, and an assessment of the economic impact of the disease. ²⁴ Responses use Likert scales with a different number of choices for different items. ²⁴ Scores for all scales and single items generally range from 0 to 100, with a higher score corresponding to a greater response — which means a higher score on a functional scale or on global health status is better, while a higher symptom score means more or more severe symptoms or problems. ²⁴ It is modular: the core addresses issues generally encountered by patients with any cancer and specific modules cover different cancers or their treatments. ²⁵ EORTC holds the copyright and provides a manual for computing and standardising the scores. ²⁴
• The AcroQoL contains 22 items describing problems experienced by patients with acromegaly. ²⁶ The items cover three dimensions: physical, psychological and appearance, and psychological and personal relations. ²⁶ Responses use a Likert scale coded 1 to 5 for frequency or degree of agreement depending on the item. ²⁶ The raw item scores are summed and standardised to a scale of 22 to 110, with a higher score corresponding to better quality of life. ²⁶
• For children and adolescents, KINDL, originally developed by Monika Bullinger in 1994, assesses quality of life in healthy or diseased children and adolescents aged 3 to 17. ²⁷ Beside its core generic module there are disease-specific modules for paediatric patients with asthma, epilepsy, cancer, diabetes or obesity. ^{27, 25} The core measure contains 24 items in three versions for the age groups 4 to 6, 7 to 13 and 14 to 17, each completable by a child or adolescent and their caregiver. ²⁵ It exists in a paper-and-pencil version and an electronic one called CAT-SCREEN. ²⁵ All versions are copyrighted. ²⁵ Interest in paediatric measures is growing, and many have been or are being developed. ²⁷

Dimension-specific measures focus on one domain such as pain, fatigue, or anxiety and depression; examples are the Hospital Anxiety and Depression Scale, the McGill Pain Questionnaire and the Multidimensional Fatigue Inventory. ²⁸ Their structure and principles are similar to those of disease-specific measures, and depending on the nature of the items they can be used in any diseased population or even in a general one, so they could also be considered generic. ^{28, 16}

Individualised measures evaluate quality of life from the respondent's own perspective, letting them either choose their own items and weight them or weight predefined ones. ²⁹ Where respondents choose, they first select the most important issues relating to their quality of life, then self-rate the level of problems they face, then allocate weights; where the items are predefined, only the last two steps apply. ²⁹ The Schedule for the Evaluation of Individual Quality of Life and the Patient Generated Index use all three steps and laid the groundwork for the family, and the administration manual for the former, published in 1993 by O'Boyle and colleagues, describes the whole process in detail. ²⁹ In principle, each item's score is the self-rating multiplied by the allocated weight, and the sum of the item scores is the final index. ²⁹ The Questions on Life Satisfaction Modules-Hypopituitarism, developed for adult patients with growth hormone deficiency, is an example of a disease-specific individualised measure with predefined items. ²⁹

Preference-based measures emerged from decision-making theory and are used mainly in pharmacoeconomic evaluations, also known as cost-utility analyses. ¹⁵ Their basic requirement is to incorporate patient or general population weights — utilities — for different health states assigned under uncertainty. ¹⁵

A number of techniques are used to generate utilities. ¹⁵ Time trade-off, used in the EQ-5D under the name EQ-VT, asks respondents how many years of life in a described condition they would give up in order to live in full health — whether they prefer a shorter life in full health to a certain number of years longer in the given health state. ¹⁵ Standard gamble, used in the SF-6D, presents alternative treatments with probabilities of better and poorer outcomes against life in the given health state, and respondents give the highest acceptable risk of treatment failure, such as death. ¹⁵ A visual analogue scale with relevant anchors is a third route. ¹⁵ Standard gamble and time trade-off are the gold standards for measuring health utilities, but utilities can also be generated from a combination of standard gamble and multi-attribute scaling analysis, as in the Health Utilities Index 2 and 3, or derived from the SF-6D and the EQ-5D. ¹⁵

Reporting health-related quality-of-life data

Medical writers working with health-related quality-of-life data and writing documents about them should keep the following points in mind. ¹⁵

1. Most scales used in health-related quality-of-life measures are ordinal, and because an ordinal scale has no well-defined unit of measurement it can indicate the direction of a change but not its magnitude: the distance from "not important at all" to "little important" is not necessarily the same as the distance from "little important" to "important", so a change from the first category to the second need not equal a change from the second to the third.
2. Responses and therefore scores are subjective, and the values behind them differ between respondents according to personality, health, life experience and cultural norms.
3. Understanding how the measure is constructed and how the choices are coded is important when writing about them: whether a higher score is better or worse, and whether an increase indicates improvement or deterioration. Explain how to interpret the scores.
4. When comparing results from different instruments, check whether they are working scores or scale scores.
5. Confirm that the researchers used a legal version of the measure and that proper acknowledgement is included. Where a measure is publicly available and needs no licence, say so and acknowledge the source, and give the version number and the mode of administration.
6. Where a translation was used, confirm that it was properly translated and validated, and give a few lines about it in the manuscript.
7. For manuscripts presenting trial data that include patient-reported outcome measures, follow the 2013 CONSORT-PRO extension.

^{15, 14, 11, 10, 17}

Presenting a measure also means describing its type and specifics, how it was used and how its scores were computed. ¹⁷

Understanding the theoretical background and the basic rules governing health-related quality-of-life research is essential for interpreting and presenting the data correctly: one must understand what the numbers mean and remember that for health-related quality of life, two plus two is not always four, and two is sometimes more — better — than three. ¹³