Rating Scale, Standard Gamble, and
Time Trade-off for People With
Traumatic Spinal Cord Injuries
Background and Purpose. The rating scale (RS), standard gamble (SG), and time trade-off (TTO) for people with traumatic spinal cord injuries (SCIs) have not been reported. This study compared psycho-metric performances of these preference-based measures among peo-ple with SCIs in Taiwan. Subjects and Methods. In total, 187 subjects from a nationwide registry of people with traumatic SCIs were inter-viewed by telephone. Score distributions, interrater reliability, discrimi-nant ability, and convergent validity for the RS, SG, and TTO were compared. Results. The mean (median) values of the RS, SG, and TTO were 0.67 (0.70), 0.64 (0.75), and 0.53 (0.50), respectively, and their corresponding intraclass correlation coefficients for intrarater and interrater test-retest reliability were .92 and .89, .78 and .73, and .91 and .78. Compared with the SG and TTO, the RS had fewer floor and ceiling values as well as percent changes, more missing observations, a larger effect size, and better discrimination ability. The results of Spearman correlation and factor analysis showed that the SG strongly converged with the TTO (r⫽.65, and in a single common factor), but they weakly converged with the RS (r⫽.33 with the SG and r ⫽.27 with the TTO). In the linear regression models, the RS was significantly associated with neurological severity, employment, educational level, and self-care ability; the SG was associated with neurological severity and employment; and the TTO was associated only with neurological severity. Discussion and Conclusion. The RS generally performed better than the SG and TTO among people with traumatic SCIs; however, the underlying construct measured by the RS differed considerably from those measured by the SG and TTO. [Lin MR, Hwang HF, Chung KP, et al. Rating scale, standard gamble, and time trade-off for people with traumatic spinal cord injuries. Phys Ther. 2006;86:337–344.]
Key Words: Health-related quality of life, Preference, Rating scale, Spinal cord injury, Standard gamble,
Time trade-off.
Mau-Roung Lin, Hei-Fen Hwang, Kuo-Piao Chung, Chingchaw Huang, Chih-Yi Chen
Research
Report
䢇
ўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўў ўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўўў ўўўўўўўўўўўўўўўўўўP
based on their experiences with different patients and on the needs of an individual patient.1 These clinical decisions often are subjectiveand difficult to quantify. However, when identifying and incorporating specific patients’ needs as assessed by the health-related quality of life (QOL), including their health profile and preference-based measures,2 the
effi-ciency and objectivity of clinical judgments can be substantially improved.3,4
Because improvements in early postinjury care have eliminated or minimized many of the complications that typically occur in people with traumatic spinal cord injuries (SCIs), the health-related QOL outcomes among those surviving with such injuries have become an important concern.5,6Health profile measures often
are applied to assess the impact of traumatic SCIs, according to functioning and well-being in multiple domains.7–15 However, these measures have difficulty
capturing the total clinical picture of traumatic SCIs or other chronic diseases and integrating components of related medical treatments.2,3Furthermore, they do not
relate health states to death in order to allow compari-sons across patient populations.
Preference-based measures summarize a broad range of relevant outcomes among people with SCIs into an overall health-related QOL outcome, called “utility” or “preference.”16An individual preference is reflected as a
single number along a continuum that usually extends from 0, anchored as death, to 1, anchored as full health.16,17 In medical decision analyses and
cost-effectiveness analyses for optimal treatment choices, an
life years (QALYs) can be further calculated by the preference times for each year of life to incorporate the effects of both QOL and quantity of life.18,19For
exam-ple, experts draw different conclusions on whether or not thrombolysis can be recommended for people with acute ischemic stroke because of a problem of weighing the improvement in functional ability against the risk of increased mortality. By including a patient’s preference, thrombolysis with a tissue plasminogen activator gave an additional 0.13 QALYs over standard supportive care for an “average” patient with stroke; it was further identified that the thrombolysis should be restricted to patients who assign low preferences of⬍0.6⬃0.7 to major post-stroke disability to increase the gain as well as the margin of safety.20
Three commonly used preference-based measures are the rating scale (RS), the standard gamble (SG), and the time trade-off (TTO).21,22 The RS provides a simple
technique for assigning a numerical value for a certain health state. The SG reveals an individual’s preference by offering a choice between 2 alternatives: living in a health state with certainty or taking a gamble on a new intervention for which the outcome is uncertain. The SG has attracted attention because it is based on the axioms of utility theory proposed by von Neumann and Morgen-stern.23The TTO offers a choice of living defined as a
fixed length of time in full health or a variable length of time in an alternative state that is less desirable such as with an SCI.
These preference-based QOL measures have not been reported for people with SCIs. In addition to describing
MR Lin, PT, PhD, is Associate Professor, Institute of Injury Prevention and Control, Taipei Medical University, 250 Wu-Hsing St, Taipei 110, Taiwan, Republic of China (mrlin@tmu.edu.tw). Address all correspondence to Dr Lin.
HF Hwang, MS, RN, is Instructor, Department of Nursing, National Taipei College of Nursing, Taipei, Taiwan, Republic of China.
KP Chung, PhD, is Associate Professor, Institute of Health Care Organization Administration, College of Public Health, National Taiwan University, Taipei, Taiwan, Republic of China.
C Huang, PhD, is Assistant Professor, Department of Health Care Management, National Taipei College of Nursing. CY Chen, MS, is Research Assistant, Institute of Injury Prevention and Control, Taipei Medical University.
Dr Lin and Dr Chung provided concept/idea/research design. Dr Lin provided writing, fund procurement, subjects, and facilities/equipment. Dr Lin, Ms Hwang, and Mr Chen provided data collection, and Dr Lin, Dr Huang, and Mr Chen provided data analysis. Dr Lin and Ms Hwang provided project management. Dr Lin, Ms Hwang, and Dr Chung provided institutional liaisons. Mr Chen provided clerical support. Dr Chung and Dr Huang provided consultation (including review of manuscript before submission).
This research was reviewed and approved by the Institutional Review Board of Taipei Medical University.
This work was supported by the National Health Research Institute (NHRI-EX93-9204PP) and the National Science Council (NSC 90-2320-B-038-042-M56), Taiwan, Republic of China.
the preferences among people with traumatic SCIs in Taiwan, this study also compared psychometric perfor-mances of the RS, SG, and TTO, including score distri-butions, test-retest reliability, discriminant ability, and convergent validity.
Method
Subjects and Procedure
Potential subjects were identified from a nationwide registry, which consisted of 809 traumatic SCI cases collected by the Head and Spinal Cord Research Group of the Neurological Society in Taiwan during the 4-year period from July 1, 1992, to June 30, 1996.24As adopted
from Kraus and colleagues,25traumatic spinal cord injury
was defined as an acute, traumatic lesion of the spinal cord resulting in any degree of sensory or motor deficit or paralysis, including injury to the nerve roots or cauda equina. The registry was developed by reviewing medical records of 113 hospitals in Taiwan considered by the Head and Spinal Cord Research Group to have the ability to manage traumatic SCIs, and included patients who had not been transferred from other hospitals and who were coded with the rubric numbers of the
Interna-tional Classification of Disease26as 806.0 to 806.9 and 968.0
to 968.9. In addition to the neurological severity of injury (incomplete tetraplegia, complete tetraplegia, incomplete paraplegia, and complete paraplegia) using the classification of the American Spinal Cord Injury Association,27patient demographics (age and sex), time
since injury, and cause of injury (motor vehicle crash, fall, and “other”) also were recorded in the registry. We used national identification numbers and names to search mortality data from 1992 to 2001 in the Depart-ment of Health, Executive Yuan, Republic of China, and 64 people in the SCI registry were matched. Of those remaining in the registry, there were telephone numbers for 603 people. Of the 142 people who did not have telephone numbers recorded in the registry and those who could not be reached by their original telephone numbers, 10 people were identified and contacted by searching the member lists of national and local SCI associations. These telephone numbers were used to conduct telephone interviews. When a subject agreed to participate in the study during the first call, we asked him or her to provide a time session of approximately 30 to 40 minutes within 1 week for our telephone interview. In the interview, 3 preference-based QOL measures— the RS, SG, and TTO—were administered; current infor-mation on education, marital status, and employment also was collected. Before the interview, a questionnaire was mailed to each subject to provide better comprehen-sion of the questions in order to shorten the interview time.
As a result, 187 subjects were interviewed, 370 could not be reached by existing telephone numbers, 40 had died, 1 was in a vegetative state, and 15 declined to be interviewed. Compared with all people in the SCI regis-try, the 187 respondents did not significantly differ in age (45.6 versus 42.9 years), sex (76% versus 81% men), time since injury (7.8 versus 7.4 years), cause of injury (58% versus 57% motor vehicle crashes), neurological severity (40% versus 35% complete lesion), or associated injuries (43% versus 47% being positive). Verbal consent was obtained from all participants.
Instruments
With the aid of the previously mailed questionnaires, the RS method asked subjects to rate their current state of health on a 0-to-100 scale (ie, a 100-point thermometer), with 0 representing death during the study and 100 representing full health. Then the subjects told the interviewers what rating they chose. The RS value was the raw score divided by 100.
The SG method asked subjects to imagine that they would live in their current health states for an average life expectancy of people in Taiwan’s general population of 25 years (ie, 25 years for people at age 43 years) (the average age of people in the SCI registry was used to estimate the life expectancy based on Taiwan’s general population).28They then were told that a hypothetical
new treatment was available that could improve their QOL to full health. However, the treatment had a risk of immediate death. The probability of full health was varied until the subject expressed indifference between the alternatives and the preference score of his or her SCI was equal to the probability. For instance, if a subject was indifferent between 25 years with SCI for sure and a gamble with a probability of 0.7 of full health for 25 years and a complementary probability of 0.3 (1 ⫺ 0.7) of immediate death, the preference of his or her SCI was equal to 0.7.
In the TTO method, subjects were told that they would live in their current health states for an average life expectancy of 25 years. Furthermore, subjects were asked to choose between 2 alternatives: living the entire duration in the specified health state or living for a shorter interval in which health would always be perfect. The length of the interval was varied until the subject expressed indifference about the decision. The duration was used to estimate the TTO score of the health state. For instance, if a subject was indifferent to 25 years with SCI and 15 years in full health, the preference of his or her SCI was equal to 0.6 (15/25).
Data Analysis
Score distribution. Distributions of the mean and
median values, the percentages of floor and ceiling values (number of subjects who had a floor or ceiling value divided by the number of respondents), and the percentage of missing values (number of subjects who had difficulty responding divided by 187) for each preference-based measure were calculated. Further-more, the normality of the preference scores was tested using the Shapiro-Wilk method.29
Test-retest reliability. A random sample of 20 subjects
was selected to assess their original responses within 2 weeks. Seven of 10 subjects completed the retest for the assessment of intrarater test-retest reliability, and 6 of another 10 subjects did so for the assessment of interra-ter test-retest reliability. The intraclass correlation coef-ficients for both types of test-retest reliability were calcu-lated.30The intraclass correlation coefficient, defined as an
estimate of the proportion of the total measurement variability among subjects, combines information on both the linear correlation and the systematic differ-ences between repeated tests.30
Discriminant validity. The ability of these 3
preference-based measures to discriminate among groups preference-based on 9 sociodemographic and medical characteristics (age, time since injury, sex, education, marital status, employ-ment, neurological severity, cause of injury, and level of ability for self-care) was tested using the Mann-Whitney test31because these preference scores were not normally
distributed. These selected characteristics have been reported to be associated with QOL among people with SCIs.7–15For these characteristics, effect sizes were
cal-culated by the difference in each preference score between subgroups divided by the standard deviation of scores among all people with SCIs. An effect size of 0.2 to 0.5 was considered to be small, 0.5 to 0.8 to be moderate, and 0.8 or greater to be large.32
Convergent validity. Factor analysis and Spearman
cor-relation were used to test the convergence among the 3 preference-based measures. Factor analysis examined
strong factor loadings (ⱖ0.4),33 onto a single common
factor. Convergence indicates that the 3 measures have same underlying construct. Two criteria, factor eigen-values greater than 1 and scree test, were used to select the number of common factors.34 Furthermore, it was
assumed that those measures that were conceptually related would be relatively strongly correlated, whereas those measures with less in common would show weaker correlations.
A linear regression model was used to determine factors associated with each preference score in order to under-stand differences among underlying constructs of the 3 measures. We used Statistical Analysis Software, version 8.02,* for all statistical analyses.
Results
Of the 187 subjects, the age at injury and time since injury, on average (⫾SD), were 42.9⫾17.3 and 7.4⫾2.0 years, respectively. Of these subjects, 48 (26%) had incomplete tetraplegia, 28 (15%) had complete tetraple-gia, 73 (39%) had incomplete parapletetraple-gia, and 38 (20%) had complete paraplegia. Furthermore, 151 (81%) of them were male, 23 (12%) had an education level of senior high school or above, 96 (51%) had an education level of junior high school, and 68 (36%) had an education level of elementary school or no formal education, 54 (29%) were single/divorced/widowed, 101 (54%) were unemployed, and 99 (53%) could care for themselves independently. Cause of injury was asso-ciated with motor vehicle crashes for 106 subjects (57%), with falls for 62 subjects (33%), and with “other” for 19 subjects (10%).
As shown in Table 1, the median of each preference score was not equal to the mean, and the Shapiro-Wilk test for normality produced a P value of less than .05 for all preference scores (ie, these scores were not normally distributed). The percentage of the minimum possible values ranged from 0% for the RS to 10.8% for the TTO. Similarly, the percentage of the maximum possible values ranged from 0.9% for the RS to 15.3% for the
Instrument X SD Median Percentage of Floor Values Percentage of Ceiling Values Percentage of Missing Values Intrarater Reliability ICCa Interrater Reliability ICC Rating scale 0.67 0.19 0.70 0.0 0.9 6.0 .92 .89 Standard gamble 0.64 0.34 0.75 3.6 9.8 3.8 .78 .73 Time trade-off 0.53 0.34 0.50 10.8 15.3 5.1 .91 .78
TTO. The percentages of missing value ranged from 3.8% for the SG to 6.0% for the RS. The intraclass correlation coefficients for intrarater and interrater test-retest reliability were .92 and .89 for the RS, .78 and .73 for the SG, and .91 and .78 for the TTO, respectively.
The ability of the 3 preference-based QOL measures to discriminate between groups with respect to the 9 sociodemo-graphic and medical characteristics is shown in Table 2. The RS significantly discriminated between groups regard-ing 3 characteristics (education, employment, and self-care), the SG dis-criminated between groups regarding 2 characteristics (employment and self-care), and the TTO discriminated between groups regarding 1 character-istic (self-care). Furthermore, the effect size was small or greater for 5 charac-teristics on the RS (ie, sex, education, time since injury, employment, and self-care ability) and for 3 characteris-tics on both the SG (ie, employment, level of injury, and self-care ability) and the TTO (ie, sex, employment, and self-care ability).
The factor loadings of the RS, SG, and TTO onto the first common factor were 0.09, 0.59, and 0.61, respectively, indi-cating that the RS was not convergent in the same dimension as the SG and TTO. Spearman correlation coeffi-cients were .33 between the RS and the SG, .27 between the RS and the TTO, .65 between the SG and the TTO. Results of the linear regression analyses are shown in Table 3. After adjusting for other variables, the RS scores of subjects who had complete paraplegia and complete tetraplegia were 0.14 and 0.13 points lower, respectively, than those of subjects who had incomplete paraplegia. Moreover, subjects who were employed, had higher educa-tional levels, and had self-care ability had larger RS scores than those of their counterparts. Compared with the SG scores of subjects who had incomplete paraplegia, the SG scores of subjects who had complete paraplegia and incomplete tetraplegia were 0.34 and 0.12 points lower, respectively. In addi-tion, the SG scores among employed subjects were 0.11 points larger than those of unemployed subjects. Com-pared with the TTO scores of subjects with incomplete paraplegia, the TTO scores of subjects with complete paraplegia and complete tetraplegia were 0.26 and 0.19 points lower, respectively.
Table 2.
Discriminant Ability of Scores for the Rating Scale, the Standard Gamble, and the Time Trade-Off With Respect to Sociodemographic and Medical Characteristics of Subjects
Characterisic Statistic Rating Scale Standard Gamble Time Trade-off
Age at injury (y)
ⱕ43 Mean 0.69 0.64 0.53 ⬎43 Mean 0.66 0.63 0.52 P .189 .880 .918 Effect size 0.18 0.02 0.04 Sex Male Mean 0.67 0.63 0.50 Female Mean 0.71 0.68 0.61 P .182 .440 .057 Effect size ⫺0.23 ⫺0.13 ⫺0.32 Education
Junior high school or below Mean 0.65 0.64 0.51
Senior high school or above Mean 0.72 0.64 0.55
P .011 .876 .372
Effect size ⫺0.36 0.02 ⫺0.13
Marital status
Single/divorced/widowed Mean 0.65 0.59 0.50
Spouse present Mean 0.68 0.66 0.54
P .317 .209 .415 Effect size ⫺0.15 ⫺0.18 ⫺0.12 Employment status Yes Mean 0.75 0.70 0.56 No Mean 0.61 0.59 0.49 P .000 .011 .105 Effect size 0.82 0.32 0.21
Time since injury (y)
ⱕ7.6 Mean 0.69 0.65 0.53 ⬎7.6 Mean 0.65 0.63 0.52 P .093 .698 .765 Effect size 0.23 0.05 0.04 Level of injury Tetraplegia Mean 0.66 0.60 0.50 Paraplegia Mean 0.69 0.68 0.55 P .236 .096 .204 Effect size ⫺0.16 ⫺0.23 ⫺0.17 Cause of injury
Motor vehicle crash Mean 0.68 0.65 0.52
Other Mean 0.65 0.62 0.55 P .193 .543 .517 Effect size 0.19 0.09 ⫺0.10 Self-care ability Yes Mean 0.74 0.73 0.58 No Mean 0.61 0.55 0.47 P .000 .000 .009 Effect size 0.74 0.53 0.35 ўўўўўўўўўўўўўўўўўўўўўўўўўўўў
Discussion
This study provides preference-based QOL scores among people with traumatic SCI, with the RS, on average, giving the highest score, followed by the SG and the TTO. Compared with the other 2 measures, the RS produced fewer floor and ceiling values, more reliability, a larger effect size, and better discrimination ability, as well as more missing observations. Furthermore, similar to the results in other patient groups,35–37the SG scores
in our study converged strongly with the TTO scores, and they converged weakly with the RS scores. This study further revealed that the factors associated with the RS differed greatly from those associated with the SG and the TTO.
Because the RS method can easily be self-administered, whereas the SG and TTO methods considerably depend on the judgment, motivations, and cognition of sub-jects,38 it is often appealing as an alternative to the SG
and TTO methods for collecting health state prefer-ences from large populations such as patients in clinical trials.35 Despite having displayed better psychometric
properties and being easier to administer, the meaning of preferences using the RS method in which no alter-native choice is available may differ from those of the other 2 measures in which an alternative choice is available. Furthermore, contrary to the findings of stud-ies of other patient groups,39,40the preference derived
from the RS in our study was uncommonly higher than those derived from the SG and TTO. One possible explanation is that people with traumatic SCIs are more willing to trade off their life or life years than are other patient groups, because the 2 methods use death as the reference health state. That is, people with SCI might be more likely than other patient groups to risk ending up dead rather than choosing to remain in a
less-than-ence may have confounded the results (ie, people with SCI might be more willing to trade off their life or life years for the longer life expectancy of 25 years of this study than for the commonly used 10 years and shorter life expectancy).35,37,39,42,43
Differences between the SG and the TTO methods often are attributed to the subjects’ risk attitude for the hypothetically new treatment. Specifically, subjects are familiarized with having risk aversion in the SG method and risk neutrality in the TTO method.16,21,44 Recently,
an additional 3 factors—probability weighting (eg, peo-ple may underweight moderate and high probabilities or the distant future while overweighting low probabilities or the near future), loss aversion (ie, people are more sensitive to losses than to gains), and contingent weight-ing (ie, people tend to pay more attention to the probabilities or durations and relatively less attention to health outcomes)—also have been identified.45 The
current study further revealed that the SG scores increased by 0.11 point when the subjects were employed, implying a possible relationship between employment status and risk attitudes among people with SCI. Additionally, the longer life expectancy of 25 years may not have the same effect for the ST and TTO methods, and subjects seemed to be more willing to trade off their life years than to risk their life for full health.
There are several limitations to the study. First, general-izing these results to all people with SCIs needs to be done in a mindful way, because a substantial number of people in the SCI registry could not be reached for participation in the study. Despite no statistically signif-icant differences being detected in any characteristics recorded in the registry, more nonparticipants tended to
Scale, the Standard Gamble, and the Time Trade-off
Characteristic Rating Scale RD (95% CI) Standard Gamble RD (95% CI) Time Trade-off RD (95% CI) Intercept 0.68 0.67 0.59 Neurological severity
Incomplete paraplegia (reference group) 0.0 0.0 0.0
Complete paraplegia ⫺0.14 (⫺0.23, ⫺0.05) ⫺0.34 (⫺0.51, ⫺0.16) ⫺0.26 (⫺0.43, ⫺0.08)
Incomplete tetraplegia ⫺0.02 (⫺0.07, 0.03) ⫺0.12 (⫺0.22, ⫺0.03) ⫺0.07 (⫺0.16, 0.03)
Complete tetraplegia ⫺0.13 (⫺0.21, ⫺0.05) ⫺0.15 (⫺0.31, 0.01) ⫺0.19 (⫺0.35, ⫺0.04)
Employment (yes/no) 0.09 (0.05, 0.14) 0.11 (0.02, 0.20)
Education
Elementary or below (reference group) 0.0
High school 0.02 (⫺0.03, 0.07)
College or above 0.09 (0.02, 0.16)
marginally older, possibly indicating a poorer health status than those participating in the study. Further-more, of the 187 study subjects, those who had difficulty responding to the preference-based measurement were older and had complete lesions. As a result, the prefer-ences of people with SCI may have been underestimated in this study. In addition, we had no choice but to conduct the interviews by telephone because our sub-jects lived nationwide and their physical health was such that it was inconvenient for them to leave their homes for personal interviews. Nevertheless, a recent study46
showed that telephone interviews yielded similar prefer-ences on the SG and TTO compared with personal interviews. Finally, the responsiveness (ie, the ability of an instrument to detect clinically important changes over time) was not evaluated in the study and needs to be further compared among the 3 measures.
References
1Kassirer JP. The principles of clinical decision making: an
introduc-tion to decision analysis. Yale J Biol Med. 1976;49:149 –164.
2Guyatt GH, Feeny DH, Patrick DL. Measuring health-related quality
of life. Ann Intern Med. 1993;118:622– 629.
3Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost-Effectiveness in
Health and Medicine. New York, NY: Oxford University Press; 1996.
4Patrick DL, Erickson P. Health Status and Health Policy: Quality of Life
in Health Care Evaluation and Resource Allocation. New York, NY: Oxford University Press; 1993.
5Dijkers MP. Correlates of life satisfaction among persons with spinal
cord injury. Arch Phys Med Rehabil. 1999;80:867– 876.
6Scherer MJ, Cushman LA. Measuring subjective quality of life
following spinal cord injury: a validation study of assistive technology device predisposition assessment. Disabil Rehabil. 2001;23:387–393.
7Westgren N, Levi R. Quality of life and traumatic spinal cord injury.
Arch Phys Med Rehabil. 1998;79:1433–1439.
8Andresen EM, Fouts BS, Romeis JC, et al. Performance of
health-related quality-of-life instruments in a spinal cord-injured population. Arch Phys Med Rehabil. 1999;80:877– 884.
9Putzke JD, Richards JS, DeVivo MJ. Quality of life after spinal cord
injury caused by gunshot. Arch Phys Med Rehabil. 2001;82:949 –954.
10Post MWM, Van-Dijk AJV, Van-Asbeck AWA, et al. Life satisfaction
of persons with spinal cord injury compared to a population group. Scand J Rehabil Med. 1998;30:23–30.
11Krause JS. Dimensions of subjective well-being after spinal cord
injury: an empirical analysis by gender and race/ethnicity. Arch Phys Med Rehabil. 1998;79:900 –909.
12Clayton KS, Chubon RA. Factors associated with the quality of life of
long-term spinal cord-injured persons. Arch Phys Med Rehabil. 1994;75: 633– 638.
13Kennedy P, Rogers B. Reported quality of life of people with spinal
cord injuries: a longitudinal analysis of the first 6 months post-discharge. Spinal Cord. 2000;38:498 –503.
14Kannisto M, Merikanto J, Alaranta H, et al. Comparison of
health-related quality of life in three subgroups of spinal cord injury patients. Spinal Cord. 1998;36:193–199.
15Manns PJ, Chad KE. Determining the relation between quality of
life, handicap, fitness, and physical activity for persons with spinal cord injury. Arch Phys Med Rehabil. 1999;80:1566 –1571.
16Torrance GW. Utility approach to measuring health-related quality
of life. J Chronic Dis. 1987;40:593– 600.
17Froberg DG, Kane RL. Methodology for measuring health-state
preferences, II: scaling methods. J Clin Epidemiol. 1989;42:459 – 471.
18Torrance G,W Feeny D. Utilities and quality-adjusted life years. Int
J Technol Assess Health Care. 1989;5:559 –575.
19Glasziou PP, Simes RJ, Gelber RD. Quality adjusted survival analysis.
Stat Med. 1990;9:1259 –1276.
20Hallan S, Asberg A, Indredavik, Wideroe TE. A decision analysis of
thrombolytic therapy compared with standard therapy in acute isch-aemic stroke. J Intern Med. 1999;246:549 –559.
21Torrance GW. Social preferences for health states: an empirical
evaluation of three measurement techniques. Socio-Econ Planning Sci. 1976;10:129 –136.
22Drummond MF, O’Brien B, Stoddart GL, Torrance GW. Methods for
the Economic Evaluation of Health Care Programmes. 2nd ed. New York, NY: Oxford University Press; 1997.
23von Neumann J, Morgenstern O. Theory of Games and Economic
Behavior. Princeton, NJ: Princeton University Press; 1944.
24Chen HY, Chiu WT, Chen SS, et al. A nationwide epidemiological
study of spinal cord injuries in Taiwan from July 1992 to June 1996. Neurol Res. 1997;19:617– 622.
25Kraus JF, Franti CF, Riggins RS, et al. Incidence of traumatic spinal
cord lesions. J Chronic Dis. 1975;28:471– 492.
26 Manual of the International Statistical Classification of Diseases, Injuries,
and Causes of Death, 9th Revision. Geneva, Switzerland: World Health Organization; 1975.
27 International Standards for Neurological and Functional Classification of
Spinal Cord Injury–Revised 1992. Chicago, Ill: American Spinal Cord Association; 1992.
28Department of Health. Health Statistics. Taipei, Taiwan: Executive
Yuan, ROC; 2001.
29Shapiro SS, Wilk MB. An analysis of variance test for normality
(complete sample). Biometrika. 1965;52:591– 611.
30Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater
reliability. Psychol Bull. 1979;86:420 – 428.
31Mann HB, Whitney DR. On a test of whether one of two random
variables is stochastically larger than the other. Ann Math Stat. 1947; 18:50 – 60.
32Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed.
Hillsdale, NJ: Lawrence Erlbaum Associates; 1998.
33Nunnally JC, Bernstein IH. Psychometric Theory. 3rd ed. New York,
NY: McGraw-Hill; 1994.
34Stevens J. Applied Multivariate Statistics for the Social Sciences. 2nd ed.
Hillsdale, NJ: Lawrence Erlbaum Associates; 1992.
35O’Leary JF, Fairclough DL, Jankowski MK, et al. Comparison of
time-tradeoff utilities and rating scale values of cancer patients and their relatives: evidence for a possible plateau relationship. Med Decis Making. 1995;15:132–137.
36Maor Y, King M, Olmer L, et al. A comparison of three measures:
the time trade-off technique, global health-related quality of life, and the SF-36 in dialysis patients. J Clin Epidemiol. 2001;54:565–570.
37Revicki DA. Relationship between health utility and psychometric
health status measures. Med Care. 1992;30:MS274 –MS282.
1984;4:315–329.
39Bleichrodt H, Johannesson M. An experimental test of a theoretical
foundation for rating-scale valuations. Med Decis Making. 1997;17: 208 –216.
40Jansen SJT, Kievit J, Nooij MA, Stiggelbout AM. Stability of patients’
preferences for chemotherapy: the impact of experience. Med Decis Making. 2001;21:295–306.
41Kahneman D, Tversky A. The psychology of preference. Sci Am.
1982;246:160 –173.
42Dolan P, Gudex C, Kind P, Williams A. The time trade-off methods:
results from a general population study. Health Econ. 1996;5:141–154.
471– 480.
44Hornberger JC, Redelmeier DA, Petersen J. Variability among
methods to assess patients’ well-being and consequent effect on a cost-effectiveness analysis. J Clin Epidemiol. 1992;45:505–512.
45Bleichrodt H. A new explanation for the difference between time
trade-off utilities and standard utilities. Health Econ. 2002;11:447– 456.
46Van Wijck EEE, Bosch JL, Hunink MGM. Time tradeoff values and
standard-gamble utilities assessed during telephone interviews versus face-to-face interviews. Med Decis Making. 1998;18:400 – 405.
