Utilities Measured by Rating Scale, Time Trade-off, and Standard Gamble: Review and Reference for Health Care Professionals

Utility is a simple expression of health-related quality of life in individuals with different states of health. A number of studies on utility measurements were conducted and published in the past. We retrieved 164 English-language articles which appeared in 1966 through 1999 for a systematic review.The number of reports has been increasing at an accelerating pace, especially during the past decade. The most widely used method of utility measurement was time trade-off, TTO (40%), followed by rating scale, RS (31%) and standard gamble, SG (29%).The utility of chronic health status was more frequently reported as compared with acute health status (907 vs 86). Accordingly, frequently explored clinical categories were cardiology, neurology, nephrology, and gastroenterology. Specifically, coronary heart disease (52 utilities), physical disability due to neurological diseases (45 utilities), chronic renal failure (74 utilities), and colorectal cancer (29 utilities) were subject to utility measurement. Mental or social dysfunctioning accounted for only a small proportion (48 utilities). There is a strong tendency for RS to yield the lowest and SG to yield the highest values.We compiled an extensive list of the results of studies on utility as a reference for health care professionals in this field.


INTRODUCTION
Health-related quality of life (HRQOL) is now widely recognized as one of the major health outcome indices in clinical studies, health services research, and health policy formation, thus making HRQOL itself a truly important research subject.There are many ways to measure HRQOL as related to a specific disease or individual health status 1), all of which have both strength and weakness.Among these, utility is one of the simplest ways to represent HRQOL and many studies have evaluated the utilities of various states of health or disease.The concept of health state utility was developed in the early 1970s and has now a solid body of theory and a set of compelling axioms 2,3).The utility of any health state or disease is assessed as a number between zero and one.Typically, Zero means death and one perfect health.Since the utility approach provides a single cardinal measure of HRQOL and its reliability had been verified in many studies 4), it is suitable for quantita-tive and statistical analysis 5).Accordingly, many researchers have used utilities for decision analysis and cost-effectiveness analysis.In fact, utility is the only way to make use of HRQOL in cost-effectiveness analysis 2,3).
A number of methods have been proposed to measure utility, of which rating scale (RS), standard gamble (SG) and time trade-off (TTO) have been used most frequently.The RS consists of a line on a page with zero at one end, equivalent to death, and one at the other, equivalent to perfect health.The subjective health status, i.e., preference, is placed on the line between zero and one in such a way that the distance from zero to the placement corresponds to the degree of preference as perceived by the subject.The SG was developed by von Neumann and Morgenstern 6).It is based on a paired comparison in which the subjects choose one of the two strategies.One strategy has two possible outcomes: perfect health with probability p and death with probability 1-p.The other strategy leads to a certain health status which is intermediate in desirability between perfect health and death.Probability p is varied until the respondent registers indifference to the two strategies, at which point the utility value is p.The TTO was developed specifically for use in health care by Torrance et al. to directly assess how long a period in a state of perfect health is equivalent to a given period of ill health 7).By converting a year in a given health status to its equivalent in a state of perfect health, the value of utility can be assessed.Unlike the RS in which the subject provides explicit preference values, the SG and the TTO methods derive preference values implicitly by basing them on the subject's responses to decision situations.Concurrently, several studies tried to convert RS to SG or TTO and showed satisfactory results 8,9).In spite of a growing accumulation of utility studies, we still have to elicit utility data on a given health status on our own.The reasons seem to include inherent variability among individuals, a uniqueness of health status dealt with in each study, a lack of systematic review and modification for clinical use of already accumulated utility data, and a difficulty searching for relevant studies in computerized database.In fact, the word "utility" is used in many senses in the medical literature, i.e., as an efficacy of examination, a therapeutic intervention, and a representation of HRQOL.Furthermore, utility is not registered as a Medical Subject Headings (MeSH) in the MED0-LINE database, which makes it difficult to search for articles on utility.Recently, Brazier et al. published an omnibus review on the use of health status measures 10).However, their report focused on technical aspects and did not deal with utility values for diseases as elicited from a variety of subject populations.
We therefore conducted a systematic review of published utility data and listed the results as a reference for researchers in this field.

METHODS
The MEDLINE database was searched by using the following terms: rating scal*, analog scal*, linear scal*, categorical scal*, time trade-off, time trade off, time tradeoff, standard gamble, and standard reference gamble.Since "quality adjusted life years" is registered in the MeSH of MEDLINE, we also used this term.The MEDLINE database covers articles published between January 1966 and December 1999.After the first tier of screening of retrieved abstracts for the appropriate contents, full texts were collected and reviewed.Listed reference in the relevant articles and personal files were also searched.
The inclusion criteria for our review were as follows.1) utilities on a particular health status or disease were obtained from patients with that health status or disease, or from patients with other disease, or from healthy volunteers, 2) target health status or disease was clearly stated, 3) studies were written in English.The exclusion criteria, on the other hand, were as fol-lows.1) utilities were not represented by definite values, e.g., only by approximations or figures, 2) target health status was complex and determined by multiple attribute, 3) the reference point was not death or perfect health, e.g., best health or worst health, 4) there was no clear definition of zero or one for the RS.In case of duplicate measurements on the same subjects in the same studies, we used the report published earlier.

Retrieval Results
From MEDLINE search, 29044 articles were retrieved.The first tier screening found 376 relevant articles for which full texts were obtained.After a critical review of these articles and relevant articles, 164 articles were judged to meet the inclusion criteria.These articles are classified in Appendix I according to the disease category 11-174)

Contributed Journals
The number of studies on utility has been increasing almost exponentially in the last decade (Figure 1).The most frequently reporting journal was Medical Decision Making (Figure 2), far ahead in this field, having published 29 utility studies so far.Quality of Life Research, which is designed specifically for the study of HRQOL studies, followed.

Focus of Interest
Medical Decision Making, Quality of Life Research and Medical Care mostly published utility studies conducted for the purpose of technical assessment of utility measurement method or elicitation process itself (25/44, 57%).On the other hand, utility data published in clinical journals were mostly related to clinical trial (29/120, 24%), health outcome measurement (57/120, 48%), cost-effectiveness analysis (18/120, 15%), and decision analysis (4/120, 3%).

Number and Methods of Utility Measurements
In total, 993 utility measurements were reported in the retrieved articles.As to the methods of eliciting utilities, TTO was used for 401 (40% of the total) utility measurements, followed by RS for 304 (31%) and SG for 288 (29%) utility measurements.

Sample Characteristics
The subjects from whom utilities were elicited included, in decreasing order, patients with the disease at issue, healthy volunteers, and patients with the disease unrelated to that at issue.Far majority (55%) of utility data were elicited from patients with the disease at issue (Figure 3).These patients were not recruited for utility measurement in the first place but rather for clinical trials.Healthy volunteers were usually medical staffs or students.Utility values obtained from patients with the disease at issue were significantly higher than those obtained from  The time trade-off was most frequently used method, followed by rating scale and standard gamble.The most frequently studied subjects were patients with the disease at issue, followed by healthy volunteers and patients with the disease unrelated to that at issue.range bar were those reported in a single study.In many cases, utilities for a given health status differed if they were measured by different group of researchers.The values often depended on the subject's characteristics, the method of measurement, and the expression of health status among others.These figures and tables show only a part of utilities listed in Appendix 1 in order to represent the range of reported utilities as many as possible (31% of the total).For example, utilities of coronary heart disease, ranging from 0.533 to 1, were frequently measured by TTO.Classified by their severity, utilities for mild disease ranged from 0.88 to 1.0, those for moderate disease 0.832 to 0.997, and those for severe disease 0.533 to 0.929.As to the relationship between measurement methods and utility values, there was a strong tendency for RS to yield the lowest and SG to yield the highest.For example, Read et al. reported that utility value corresponding to moderate angina pectoris obtained by SG was 0.93, which was significantly higher than 0.832 obtained by TTO.Furthermore, RS yielded 0.718, which was significantly lower than TTO 32).measurements.For this reason, the number of studies on utilities has been increasing almost exponentially in the last decade (Figure 1).In addition, the number of articles entitled costeffectiveness analysis increased from seven in 1989 to 48 in 1999 in accordance with the social needs to consider costeffectiveness in health policy as well as in clinical practice.
This tendency indicates the importance for health service researchers and clinicians to be aware of HRQOL .
Our review on previously published utility studies showed that Medical Decision Making is the leading journal in this field (Figure 2).It reflects the leadership role of this journal in the development of methodologies related to medical decision Table 2. Utility values elicited from patients with the disease unrelated to that at issue.RUQ: right upper quadrant; RS: rating scale; SG: standard gamble; TTO: time trade-off .making in the past 19 years .Since a single cardinal measure of HRQOL , i.e., utility, is required for decision analysis and cost-effectiveness analysis , many of which were published in clinical journals 5) .On the other hand, Medical Decision Making , Quality of Life Research and Medical Care published far less number of studies of cost-effectiveness analysis and accordingly less data on utility in spite of much data on technical assessment of utility measurement method.
We found that TTO was the most frequently used method , followed by RS and SG.This reflects the fact that RS is easy to use but subject to unreliability and that SG is difficult to use but subject to good validity 10).In this regard, TTO , going in the middle, is fairly easy to use and subject to fair validity .
Our review shows that most of the study subjects from whom utilities were elicited were patients with the disease at issue and that these patients were not recruited for utility measurement in the first place but rather for clinical trials .It reflects that patients' preference has been widely taken into account in the assessment of outcomes of clinical trials.On the other hand, the majority of healthy volunteers were medical staffs or students.Reasons may include the convenience of access by researchers, quick apprehension of rather difficult concept of utility, and clear imagination of hypothetical clinical situations.
A chronic health status seems to be easier to imagine and assess its utility than acute health status for persons without the disease at issue.A chronic illness lasts for long time and the study subjects can exchange utility values for a given time or gamble their entire life.It is inevitable that real patients in acute distress have difficulty cooperating to answer utility inquiries.
Utility values compiled in the current study for a given health status were wide in range.This is due to the nature of the combined data for a given disease, in terms of disease stage, severity, symptoms and treatment.It is quite natural that individual patient differs from other patients with the same disease in terms of the characteristics of the disease and their own health preference.Therefore, it may be ideal to elicit utility values from particular patient when physicians conduct decision analysis and cost-effectiveness analysis in particular situation.However, comprehensively compiled data such as this could serve as a benchmark or substitute when individual utility values were not available.
We excluded multiattribute utility methods such as EuroQol, Quality of Well-Being or Health Utilities Index, from this review 2, 3) because they themselves generally do not afford one-to-one corresponding utility values of a given disease and are difficult to use for a quantitative research at the present time.However, utility values of several distinct diseases were recently measured using such multiattribute utility methods and validated 175,176).Further studies are necessary in this regard.
The growing necessity of making clinical decisions based on quantitative data should further prompt utility studies in the years to come.We hope that our current review will serve as reference of utility studies ever done in the past 30 years.To be continued .

Figure 1 .Figure 2 .
Figure 1.The chronological trend of studies reporting utility values in 1973-1999.The number of articles on utilities has been increasing, especially in the last decade.

Figure 3 .
Figure 3.The methods of utility elicitation according to subject character.
Figures 5 through 7 and Tables 1 through 3 show the weighted mean, minimum, and maximum utility values for a variety of health status according to type of samples and methods of measurement.Utilities shown only as a point without a

Figure 4 .
Figure 4.The frequency of utility elicitation according to disease category.Diseases related neurology, cardiology, nephrology and gastroenterology & hepatology were most frequently measured.

Figure 5 .
Figure 5. Utility values elicited from patients with the disease at issue.Rhombus shows weighted mean utilities.The horizontal bar shows the range of utility values.AIDS: acquired immunodeficiency syndrome; HIV: human immunodeficiency virus infection; AP: angina pectoris; ASO: arteriosclerosis obliterans; PTA: percutaneous transluminal angioplasty; CRF: chronic renal failure; HD: hemodialysis; CAPD: continuous ambulatory peritoneal dialysis; HRT: hormone replacement therapy; RS: rating scale; SG: standard gamble; TTO: time trade-off.

Figure 6 .
Figure 6.Utility values elicited from patients with the disease unrelated to that at issue .Rhombus shows weighted mean utilities.The horizontal bar shows the range of utility values.GS: gallstone; AS: asymptomatic; EA: enlarged abdomen; RS: rating scale; SG: standard gamble; TTO: time trade-off .

Table 1 .
Utility values elicited from patients with the disease at issue.

Table 3 .
Utility values elivited from healthy volunteers.