Health State Utilities in Latent and Active Tuberculosis
ABSTRACT
Background: Tuberculosis (TB) remains a major public health threat worldwide. Numerous cost-effectiveness analyses of TB screening and treatment strategies have been recently published, but none have utilized quality-adjusted life-years as recommended because of the lack of utilities for TB health states.
Objective: To characterize and compare utility scores from either active TB or latent TB infection (LTBI) participants.
Methods: Consenting patients attending a population-based screening and treatment clinic were administered the Short Form 36 (SF-36), the Health Utilities Index 2/3 (HUI2/3), and a general health visual analog scale (VAS) along with demographic questions. SF-36 scores were converted to Short Form 6D (SF-6D) utility scores using an accepted algorithm. Utility results were compared across scales, and construct validity was assessed.
Results: A total of 162 TB patients (78 LTBI and 84 active TB) with available SF-36 and all four utility scores (Health Utilities Index 2, Health Utilities Index 3, SF-6D and VAS) were included in the analysis. Those with active TB had significantly lower SF-36 and utility scores than those with LTBI. Although all appeared to exhibit construct validity, the HUI2/3 and the VAS appeared to have significant ceiling effects, whereas the SF-6D had significant floor effects.
Conclusions: Health state utility values for active TB and LTBI have been determined using different instruments. The three measures did not generate identical utility scores. The HUI2/3 was limited by ceiling effects, whereas the SF-6D appeared to display floor effects.
Introduction
Although effective chemotherapy is available, worldwide tuberculosis (TB) remains a major public health problem, with approximately one-third of the world's population infected [1]. In North America, because mortality due to TB is rare, the avoidance of morbidity and maintenance of patients' quality of life has become the goal of TB management [2,3]. Health-related quality of life (HRQL) has been introduced into medical practices and evaluated in many diseases including TB. There is increasing evidence to show that TB has substantial and encompassing impacts on patients' quality of life [4–9]. Furthermore, clinical measures often correlate poorly with daily well-being and function, the areas in which patients are most interested and familiar.
Two basic approaches to assessing HRQL are available: generic and disease-specific instruments [10]. Generic instruments capture the common and important aspects of health, while specific instruments are designed to focus on unique problems associated with a disease. One category of generic instruments is “preference-based” or “utility measure,” and examples of such are the EuroQol 5D (EQ-5D), the Health Utilities Index 2 (HUI2) and the Health Utilities Index 3 (HUI3), and the Short Form 6D (SF-6D). These instruments summarize HRQL into a single index number, anchored at 0 (death) and 1 (perfect health), and are based on societal health preferences. Scores can then be incorporated into cost-utility analyses (a special type of cost-effectiveness analysis [CEA]) through the calculation of quality-adjusted life-years (QALYs) that combine the utility score for a health state (quality of life) with the duration of time spent in the particular health state (quantity of life).
Recently, there have been many CEAs of the treatment of, and/or screening for, TB infection in populations such as intravenous drug users [11,12], the elderly [13], tuberculin reactors stratified by age [14,15], immigrants [16–19], and new blood-based diagnostic strategies [20–22]. So far, none have evaluated TB health outcomes in the holistic way recommended by the influential US Panel on Cost-Effectiveness in Health and Medicine [23], namely, QALYs constructed from utilities elicited from TB patients. The main reason for the lack of QALYs as outcome measures in TB economic evaluations is the dearth of TB-related utility data in the literature.
Considering the void in the literature regarding health state utility values for active TB and latent TB infection (LTBI), the primary purpose of this article was to characterize and compare utility scores between the HUI2, HUI3, and SF-6D in a sample of participants with either active TB or LTBI.
Methods
Study Setting and Subjects
Participants were recruited from the TB Clinic at the BC Center for Disease Control. To be eligible for this study, the participant had to have a diagnosis of either active TB or LTBI within 2 months of study entry. Subjects were excluded if 1) they were less than 18 years old; 2) they were not taking medication for active TB or LTBI; or 3) they were judged to be incapable of answering the questionnaires. This study received ethical approval from the University of British Columbia Behavioral Research Ethics Board, and informed signed consent was obtained from each participant.
During the first clinic visit, relevant participants' information was collected, including 1) sociodemographics, e.g., age, sex, education, and comorbid conditions; 2) TB symptoms, medication use, adverse events, and self-reported TB symptom severity and control; 3) responses to the HRQL questionnaires (Short Form 36 [SF-36], HUI2/HUI3, and a visual analog scale [VAS]). All questionnaires were self-administered, and the participants were permitted to take them home for completion. Investigators followed up by phone call to ensure that questionnaires were returned within a week.
Assessment of HRQL
SF-36 and SF-6D. The SF-36 Health Survey is one of the most widely used generic, health profile instruments [24]. Eight domain scores (physical functioning, role limitations due to physical health, bodily pain, general health perceptions, vitality, social functioning, role limitations due to emotional problems, and mental health) and two summary scores, the physical component summary (PCS) and the mental component summary (MCS), are calculated to describe health status. Nevertheless, no single, preference-based index score can be derived directly from this widely used HRQL instrument. To obtain a preference-based utility score from it, Brazier et al. restructured the SF-36 into a health state classification system with six attributes called the SF-6D (which includes physical functioning, role limitations, social functioning, pain, mental health, and vitality). Each attribute has four to six levels, and therefore the SF-6D describes 18,000 health states. Scoring is based on responses from a random sample drawn from the UK general population.
Health Utilities Index (HUI). HUI is a family of multiattribute health status classification systems that is currently composed of the HUI2 and HUI3 [25]. HUI2 and HUI3 utility scores are derived from the same questionnaire, but they contain different health dimensions with various levels in each, and they employ different samples and scoring models to obtain preference scores. HUI2 has seven health attributes (sensation, mobility, emotion, cognition, self-care, pain, and fertility), each with three to five levels [26], while HUI3 consists of eight attributes with five to six levels within each. In total, HUI2 defines 18,000 health states, and HUI3 defines 972,000 health states [27]. Preference scores for both HUI3 and HUI2 health states were based on Canadian samples using the standard gamble (SG) and a VAS approach.
In this study, both the HUI and SF-36 questionnaires used “the past 4 weeks” as the response time frame to minimize potential differences in recall.
VAS. In this study, a horizontal line, 10 cm in length, anchored at “0” (death) at the left end and “1” (perfect health) at the right end was used. Each respondent was asked to mark on the line the point that they felt represented their current health state.
Statistical Analysis
Scores for the SF-6D, HUI3, and HUI2 were calculated for each patient from the two questionnaires, SF-36 and HUI. All patients who had completed the SF-6D, HUI3, HUI2, and VAS scores were included in the analysis. Descriptive summary statistics were used to describe demographic characteristics. Chi-square tests, Student's t test, or analysis of variance (ANOVA), as appropriate, were used to examine the demographic differences between active and latent patients. In each group, patients who were included and those who were excluded were compared as well.
The SF-36 domain scores and utility scores were summarized by groups numerically and graphically. Because the utility scores were usually not normally distributed, median value and interquartile range were also presented. Differences between latent and active patients on SF-36 scores and utility values were tested using simple linear regression, adjusting for the available demographic variables. Active TB patients were then categorized by self-graded TB severity and control, and each instrument's ability to discriminate between severity subgroups was tested.
Two-way ANOVA and paired two-sample t tests were performed to examine the differences among the four utility instruments within individuals. The ceiling and floor effects for the three indirect utility instruments (HUI2, HUI3, and SF-6D) were examined from two aspects: the frequency of possible minimum and maximum scores for each instrument, and the proportion of respondents who reported the best and worst level within each single health attribute for each instrument.
Spearman's rho was calculated to assess the correlation between various measures. The correlation coefficients were explained as follows. −0.30 to 0.30, weak correlation; −0.49 to −0.30 and 0.30 to 0.49, moderate correlation; and <−0.50 and >0.50, strong correlation [28]. Agreement among the utility scores was assessed using the intraclass correlation coefficient (ICC) generated from a two-way mixed-effect ANOVA model where the patient effect was random and the instrument effect was fixed. The single-measure ICCs were used and interpreted according to the following guidelines: ICC < 0.40, poor agreement; 0.40 to 0.75, moderate to good agreement; >0.75, excellent agreement [29]. All P-values were two sided, and P < 0.05 was considered to be statistically significant.
Results
Description of Study Sample
In a consecutive fashion, over a 12-month period, we approached 147 LTBI and 133 active TB patients to participate in the study. Of these, 119 LTBI and 114 active TB patients were deemed eligible and consented to participate in the study. Of these, 27 (20 LTBI and 7 active) withdrew from the study before completing the questionnaires. Subsequent to enrollment, three patients reported as having active TB were excluded as they did not fit the entry criteria of having active TB (under medical surveillance for previously active TB, not currently active, no documentation of prior medications, no growth on smear and cultures). Thus, 104 active TB and 99 LTBI participants were enrolled in the study. Of these, 25 (12.3%) participants failed to fully complete the SF-36; 15 (7.4%) did not fully complete the HUI, and 13 (6.4%) had no VAS results. In total, 162 (79.8%) TB patients (78 LTBI and 84 active TB) with available SF-36 and all four utility scores (HUI2, HUI3, SF-6D, and VAS) were included in the analysis. In each group, there were no significant differences on the studied demographic characteristics between patients included and those excluded.
The demographic features of both the active TB and LTBI participants are described in Table 1. Active TB participants were significantly older than LTBI participants (49.0 vs. 36.3 years, P < 0.01). Subjects with active TB were more likely to have comorbid conditions (45% vs. 24%, P = 0.02) and to report alcohol use compared with those with LTBI (86% vs. 69%, P = 0.01). Overall, of the 162 TB patients in this study, 43% were males; 87% were born outside Canada, and 56% were originally Asians or Pacific Islanders.
| Latent (N = 78) | Active (N = 84) | P-value | |
|---|---|---|---|
| Age (mean in year, SD) | 36.3 (11.1) | 49.0 (19.0) | <0.01 |
| Sex (% male) | 40 | 45 | 0.48 |
| Foreign-born (%) | 85 | 89 | 0.38 |
| Race (%) | 0.50 | ||
| Asian or Pacific Islander | 51 | 60 | |
| East Indian (South Asian) | 17 | 14 | |
| Caucasian | 13 | 15 | |
| African American | 5.1 | 3.6 | |
| Aboriginal | 3.8 | 1.2 | |
| Others | 10.1 | 6.2 | |
| Marital status (%) | 0.55 | ||
| Married | 47 | 56 | |
| Single | 33 | 27 | |
| Others | 20 | 17 | |
| Smoking (% yes) | 26 | 23 | 0.60 |
| Alcohol use (% yes) | 69 | 86 | 0.01 |
| Drug use (% yes) | 10 | 4.8 | 0.27 |
| Comorbidity (% yes) | 24 | 45 | 0.02 |
SF-36 Scores
In comparison with those with LTBI, subjects with active TB had a much lower average scores on all eight domains of the SF-36 (Table 2, all P < 0.05), after adjusting for the demographic differences between the two groups. Overall, older participants tended to score lower than those who were younger, and no sex difference was observed on SF-36 domain scores in this sample.
| Latent (N = 78) | Active (N = 84) | |||||
|---|---|---|---|---|---|---|
| Mean (95% CI) | Median | IQR | Mean (95% CI) | Median | IQR | |
| Physical functioning | 54.3 (52.7–55.9) | 57.0 | 0.0 | 41.2 (38.1–44.4) | 40.2 | 25.3 |
| Role physical | 51.9 (49.9–53.8) | 56.9 | 9.2 | 38.1 (34.7–41.5) | 37.3 | 30.0 |
| Bodily pain | 55.7 (53.4–58.0) | 62.1 | 11.0 | 50.1 (47.3–53.0) | 51.1 | 20.7 |
| General health | 52.5 (50.7–54.4) | 52.9 | 9.5 | 44.5 (42.3–46.7) | 44.1 | 14.3 |
| EN (vitality) | 52.7 (50.7–54.8) | 52.1 | 12.5 | 44.0 (41.0–46.9) | 42.7 | 18.7 |
| Social functioning | 52.7 (50.6–54.7) | 56.8 | 5.5 | 39.5 (36.3–42.7) | 40.5 | 23.2 |
| Role emotional | 50.7 (48.6–52.9) | 55.9 | 2.9 | 35.0 (30.9–39.1) | 32.6 | 35.0 |
| Mental health | 50.9 (48.7–53.0) | 52.8 | 14.1 | 44.4 (41.8–47.0) | 43.0 | 16.9 |
| Physical component summary | 54.7 (53.2–56.1) | 56.8 | 6.3 | 44.8 (42.1–47.5) | 45.9 | 18.9 |
| Mental component summary | 50.3 (48.5–52.0) | 52.8 | 8.9 | 40.1 (37.1–43.1) | 41.4 | 22.0 |
- CI, confidence interval; IQR, interquartile range; MCS, mental component summary; PCS, physical component summary.
Global Utility Scores
Table 3 summarizes the four sets of global utility scores for LTBI and active TB participants, respectively. For each utility instrument assessment, significantly lower average scores were observed in active TB participants compared with those with LTBI, after accounting for the impact of the demographic differences (all P < 0.01). In active TB participants, the four preference-based instruments (the HUI2, HUI3, SF-6D, and VAS) yielded significantly different global score within individuals, with HUI2 having the highest average score overall, followed by HUI3, SF-6D, and VAS.
| Mean (95% CI) | Median | Min | Max | IQR | |
|---|---|---|---|---|---|
| Latent (N = 78) | |||||
| SF-6D | 0.82 (0.80–0.85) | 0.85 | 0.40 | 1.00 | 0.12 |
| HUI3 | 0.90 (0.86–0.94) | 0.97 | 0.08 | 1.00 | 0.10 |
| HUI2 | 0.93 (0.90–0.95) | 0.95 | 0.49 | 1.00 | 0.08 |
| VAS | 0.87 (0.84–0.90) | 0.90 | 0.39 | 1.00 | 0.20 |
| Active (N = 84) | |||||
| SF-6D | 0.68 (0.65–0.72) | 0.64 | 0.32 | 1.00 | 0.24 |
| HUI3 | 0.76 (0.70–0.82) | 0.90 | −0.26 | 1.00 | 0.31 |
| HUI2 | 0.85 (0.80–0.89) | 0.93 | 0.13 | 1.00 | 0.19 |
| VAS | 0.66 (0.61–0.71) | 0.70 | 0.08 | 1.00 | 0.32 |
- CI, confidence interval; HUI2, Health Utilities Index 2; HUI3, Health Utilities Index 3; IQR, interquartile range; SF-6D, Short Form 6D; VAS, visual analog scale.
1-4 show the distributions of the global scores among active TB participants. SF-6D scores were distributed normally, but with limited available range, and the observed lowest score was 0.32. In contrast, HUI2 and HUI3 scores covered a wider range: −0.26 to 1.00 for HUI3 and 0.13 to 1.00 for HUI2. Nevertheless, HUI2 and HUI3 scores were highly skewed toward 1 (perfect health), demonstrating a ceiling effect. For the HUI2 and HUI3, respectively, 25% and 21% of active participants reported scores of 1 (perfect health). In contrast, 2.4% of the participants scored 1 for SF-6D. VAS scores were not as skewed as HUI2 and HUI3, and spanned the entire scale range, from 0.08 to 1.00.
Distribution of Short Form 6D (SF-6D) scores in active participants.
Distribution of Health Utilities Index 3 (HUI3) scores in active participants.
Distribution of Health Utilities Index 2 (HUI2) scores in active participants.
Distribution of visual analog scale (VAS) scores in active participants.
Table 4 describes the proportion of respondents at the best and worst level within each single health dimension for HUI2, HUI3, and SF-6D. For HUI2 and HUI3, less than 4% of active TB participants reported the worst level in all health dimensions, but large proportions of participants were at the best level. There are three common health dimensions in the HUI2 and HUI3, which are emotion, cognition, and pain. In the pain dimension, about 50% of the respondents chose the best possible level, while in the cognition and emotion dimensions, the respondents chose the best level of the HUI3 domain much less than the best level of the HUI2 domain. For SF-6D dimension scores, except for vitality, all had at least 22% at the best level. For example, 42% and 33% were at the best level in the pain and role limitation dimensions, respectively. On the other hand, 63%, 29%, and 14% were at the worst level in role limitation, vitality, and social functioning, respectively. Even though the role limitation dimension has four levels in total, 96% of participants reported either the best or the worst level, leaving a gap at the intermediate levels.
| Utility instruments | Best level, no. (%) | Worst level, no. (%) |
|---|---|---|
| HUI2 | ||
| Sensation | 51 (60.7) | 3 (3.6) |
| Mobility | 60 (71.4) | 0 |
| Emotion | 59 (70.2) | 0 |
| Cognition | 61 (72.6) | 1 (1.2) |
| Self-care | 77 (91.7) | 2 (2.4) |
| Pain | 42 (50.0) | 1 (1.2) |
| HUI3 | ||
| Vision | 53 (63.1) | 2 (2.4) |
| Hearing | 82 (97.6) | 1 (1.2) |
| Speech | 77 (91.7) | 0 |
| Ambulation | 60 (71.4) | 0 |
| Dexterity | 70 (94.0) | 0 |
| Emotion | 35 (41.7) | 0 |
| Cognition | 19 (22.6) | 1 (1.2) |
| Pain | 42 (50.0) | 2 (2.4) |
| SF-6D | ||
| Physical function | 22 (26.2) | 5 (6.0) |
| Role limitation | 28 (33.3) | 53 (63.1) |
| Social function | 25 (29.8) | 12 (14.3) |
| Pain | 35 (41.7) | 6 (7.1) |
| Mental health | 19 (22.6) | 4 (4.8) |
| Vitality | 6 (7.1) | 24 (28.6) |
- HUI2, Health Utilities Index 2; HUI3, Health Utilities Index 3; SF-6D, Short Form 6D.
In Table 5, SF-36 summary scores and each of the utility scores were summarized by the level of self-reported disease severity. Each instrument demonstrated a clear monotonically decreasing trend in scores with increasing severity level. Among the four sets of utility scores, only HUI3 scores were not statistically different between adjacent severity subgroups.
| PCS | MCS | SF-6D | HUI3 | HUI2 | VAS | |
|---|---|---|---|---|---|---|
| Severity of TB symptoms | ||||||
| Very mild | 54.5 (6.6) | 52.4 (7.6) | 0.84 (0.11) | 0.93 (0.10) | 0.94 (0.07) | 0.84 (0.14) |
| Mild | 50.0 (5.2) | 44.6 (10.1) | 0.68 (0.05) | 0.84 (0.19) | 0.93 (0.08) | 0.75 (0.09) |
| Moderate | 44.6 (9.5) | 34.5 (12.6) | 0.64 (0.10) | 0.73 (0.28) | 0.83 (0.19) | 0.65 (0.13) |
| Severe | 35.6 (12.0) | 33.8 (12.4) | 0.59 (0.15) | 0.65 (0.36) | 0.76 (0.25) | 0.54 (0.22) |
| Very severe | 32.5 (11.8) | 28.9 (10.8) | 0.54 (0.08) | 0.53 (0.37) | 0.71 (0.28) | 0.35 (0.21) |
| Control of TB symptoms | ||||||
| Very well controlled | 55.8 (6.0) | 51.8 (7.8) | 0.85 (0.09) | 0.90 (0.16) | 0.94 (0.07) | 0.84 (0.15) |
| Well controlled | 49.3 (6.3) | 47.7 (9.6) | 0.73 (0.13) | 0.92 (0.08) | 0.95 (0.07) | 0.74 (0.18) |
| Adequately controlled | 40.0 (12.7) | 32.3 (11.7) | 0.60 (0.11) | 0.68 (0.30) | 0.78 (0.21) | 0.60 (0.18) |
| Not well controlled | 33.9 (8.7) | 30.2 (11.9) | 0.56 (0.11) | 0.52 (0.38) | 0.74 (0.27) | 0.42 (0.22) |
| Not controlled at all | — | — | — | — | — | — |
- HUI2, Health Utilities Index 2; HUI3, Health Utilities Index 3; MCS, mental component summary; PCS, physical component summary; SF-6D, Short Form 6D; TB, tuberculosis; VAS, visual analog scale.
Table 6 presents Spearman's coefficients for correlation between SF-36 summary scores and the four global preference-based scores. Strong correlation was observed between the four utility instrument scores (0.52–0.86, P < 0.0001). In terms of the relationship with SF-36, SF-6D scores were strongly correlated with both PCS and MCS (0.79, 0.80), while HUI3 and HUI2 were strongly correlated with PCS (0.59, 0.66), but only moderately correlated with MCS (0.37, 0.48).
| PCS | MCS | SF-6D | HUI3 | HUI2 | VAS | |
|---|---|---|---|---|---|---|
| PCS | 1.00 | |||||
| MCS | 0.45 | 1.00 | ||||
| SF-6D | 0.79 | 0.80 | 1.00 | |||
| HUI3 | 0.66 | 0.48 | 0.71 | 1.00 | ||
| HUI2 | 0.59 | 0.37 | 0.59 | 0.90 | 1.00 | |
| VAS | 0.67 | 0.59 | 0.65 | 0.58 | 0.52 | 1.00 |
- All P < 0.0001.
- HUI2, Health Utilities Index 2; HUI3, Health Utilities Index 3; MCS, mental component summary; PCS, physical component summary; SF-6D, Short Form 6D; VAS, visual analog scale.
The overall ICC among the four utility instrument scores was 0.65, indicating good agreement. ICCs (and 95% confidence intervals) between each paired utility score are presented in Table 7. There was excellent agreement between HUI2 and HUI3 (0.84). The agreement between all other paired utility scores was good (0.53–0.67).
| SF-6D | HUI3 | HUI2 | VAS | |
|---|---|---|---|---|
| SF-6D | 1.00 | |||
| HUI3 | 0.63 (0.53–0.72) | 1.00 | ||
| HUI2 | 0.67 (0.58–0.75) | 0.84 (0.79–0.88) | 1.00 | |
| VAS | 0.66 (0.56–0.74) | 0.56 (0.45–0.66) | 0.53 (0.41–0.63) | 1.00 |
- HUI2, Health Utilities Index 2; HUI3, Health Utilities Index 3; SF-6D, Short Form 6D; VAS, visual analog scale.
Discussion
This study has generated much needed evidence regarding preference-based utility scores for subjects with active TB and LTBI that can be integrated into cost-utility analyses of TB diagnostic and treatment interventions. In addition, this study provides some comparative information regarding the different methods of estimating health utilities and aids in the choice of measure to be included in primary studies. Finally, the results show that there are significant problems with all of the instruments in terms of floor and ceiling effects especially when each domain is considered.
Other authors have estimated preferences for health states in TB either through expert opinion [15,30] or by surveying the general public [31]. Nevertheless, only one other study has attempted to measure health preferences from patients with LTBI and active, or recently cured, TB [7]. These authors relied on a small, heterogeneous sample of LTBI (N = 25), active TB (N = 17) and previous active TB (N = 8) participants to derive health state utility values, thus limiting the generalizability of their findings. Furthermore, the authors utilized the SG, a VAS, and the EQ-5D to elicit health state preferences and the SF-36 as a general health measure. In general, their findings were that those with concurrent active TB rated hypothetical health states (mild, moderate, and severe) lower than those with LTBI and previously active TB. Also SG scores correlated poorly with the EQ-5D and the SF-36, and moderately with the VAS. The authors concluded that the SG was likely a more comprehensive measure of HRQL because it was not limited to a small number of domains and levels like questionnaire-based instruments are and provides a more holistic assessment of HRQL. Nevertheless, there are recent publications that put the validity and the reliability of the SG into question when used in specific patient populations [32,33]. As such, it is not clear that the SG is appropriate for the elicitation of utility values in all cases.
From the same data set, the authors published another article whose purpose was to evaluate the feasibility and reliability of the EQ-5D and the SF-36 [6]. They found that the proportion of acceptable SF-36 questionnaires was 78% versus 90% for the EQ-5D. In addition, they found a high degree of internal consistency for the SF-36 and a high test–retest reliability. Similar to our study, the authors were concerned about the use of health status questionnaires in a multicultural sample which are not adapted and tested in the respondent's native language. Nevertheless, because the authors found a low frequency of inconsistent responses and high internal consistency, they concluded that it was valid to use these questionnaires. Similarly, authors of other articles [34,35] found similar high feasibility and strong psychometric properties for the SF-36 in nonnative English language speakers.
It was difficult to compare the VAS scores from our sample to that of Dion et al.'s because of the lack of stratification in the presentation of their results into subjects with LTBI and active TB [6,7]. Nevertheless, we found the VAS to be much more highly correlated with the PCS (r = 0.65) and the MCS (r = 0.58) than Dion et al. whose correlations were 0.54 and 0.32, respectively.
Within our results, the utility global scores were significantly different with the HUI2 generating the highest and the SF-6D providing the lowest mean scores. The distribution of SF-6D scores was much more symmetrical than that of HUI3 or HUI2, but within a very limited scale range, from 0.32 to 1.00. In contrast, HUI3 scores spanned approximately twice the range of SF-6D scores, and HUI2 distributed from 0.13 to 1.00. Although having a wider scale range than SF-6D, HUI3 and HUI2 scores were highly skewed toward 1.00, with 25% and 21% participants reporting a HUI2 and HUI3 score of 1.0, respectively, indicating a possible ceiling effect in active TB participants. There was no evidence to demonstrate a ceiling or floor effect for SF-6D global utility scores, because only 2.4% of participants reported perfect health and none reported the worst health state. Nevertheless, a dimension-specific analysis SF-6D yielded possible floor effects. The most notable result was in the role limitation dimension. In the pain dimension, all three instruments demonstrated a significant ceiling effect, with 42% at the best level for SF-6D, and 50% for HUI2 and HUI3. Compared to SF-6D, all HUI2 and HUI3 single dimensions demonstrated ceiling effects with very limited floor effects.
Discrepancies between scores of the indirect utility instruments have been observed and discussed in other patient populations [36,37]. These discrepancies may arise because of the internal differences of the utility instruments, such as the health dimensions covered, the methods of obtaining preferences (such as the SG or the time trade-off techniques), and the application of the instruments. Each instrument focuses on different health dimensions with various levels within each dimension and can have different recall periods.
Nevertheless, despite these differences, some similarities have been observed as well. The correlation between these instruments was from moderate to strong, suggesting that they are likely measuring similar concepts of health. The four utility instruments were able to discriminate TB participants with different severity levels. In general, the average utility scores showed a decrease with the increasing disease severity.
In conclusion, health state utility values for active TB and LTBI have been determined using different instruments. The three measures, the HUI2, HUI3, and SF-6D, did not generate identical utility scores in active TB or LTBI participants. The biggest concern with the HUI2 and HUI3 was the ceiling effect, and SF-6D was limited by its narrow range of utility values available and the potential floor effect. As such, the HUI2 and HUI3 may be more able to evaluate HRQL in more severe participants because the SF-6D does not appear to adequately capture health states at the lower end of the scale.
This study was funded, in part, by the National Sanitarium Association, Ontario, Canada. Dr C. Marra was funded by a Canada Research Chair in Pharmaceutical Outcomes and a Michael Smith Foundation for Health Research Scholar Award in Health Services Research. Dr J. Mark FitzGerald was funded by a Michael Smith Foundation for Health Research Distinguished Scholar Award, and a Canadian Institutes of Health Research/British Columbia Lung Association Investigator Award.
Source of financial support: None.
