Abstract
Differences in preference-weighted health-related quality of life (HRQOL) scores by race/ethnicity may be due to social factors. Here, Short-Form Six-Dimension (SF-6D) scores are analyzed among men in a prostate cancer prevention trial to explore such differences. Selenium and vitamin E cancer prevention trial participants who completed the SF-6D at baseline, and in at least 1 of follow-up years 1, 3, and 5 were included. This study compared mean SF-6D scores across race/ethnicity at each point using a linear mixed model controlling for demographic and clinical characteristics. At baseline, 9691 men were eligible for analysis, of whom 7556 (78%) were non-Hispanic White, 1592 (16.4%) were non-Hispanic Black, and 543 (5.6%) were Hispanic. Hispanic and White participants had higher unadjusted mean SF-6D scores than Black participants at every time point (P < 0.05), while white participants had lower mean scores than Hispanic participants at every time point after baseline (P < 0.05). After adjusting for covariates, statistically significant differences in HRQOL among the 3 groups persisted. Hispanic participants had higher preference scores than White participants by 0.073 (P < 0.001), 0.075 (P < 0.001), and 0.040 (P < 0.001) in follow-up years 1, 3, and 5, respectively. Black participants had lower scores than White participants by 0.009 (P = 0.004) and 0.008 (P = 0.02) in follow-up years 1 and 3, respectively. The results suggest there is a preference-weighted HRQOL difference by race/ethnicity that cannot be explained by social and clinical variables alone. Understanding how individuals belonging to different racial/ethnic categories view their own HRQOL is necessary for culturally competent care and cost-effectiveness analyses.
Keywords: preference-weighted quality of life, health disparities, prostate cancer
Introduction
Health-related quality of life (HRQOL), or “an individual's…self-perception of physical and mental health,”1 can be an important predictor of health outcomes and is frequently considered an outcome in its own right.1–5 HRQOL can be measured with a number of surveys including the Short Form-36 (SF-36) and its variants, which are among the most widely used instruments for this purpose.5,6
Differences in HRQOL by race and ethnicity have been previously documented, as measured by the SF-36,5 and other instruments.7–9 It is also a well-documented phenomenon that social factors can affect the way that people self-report their own health states.10,11 Findings from research on preference-weighted HRQOL vary with respect to the magnitude and direction of differences in the relative scores of Black, Hispanic, and White respondents with and without adjustments for demographic and socioeconomic factors.12,13
Pereira et al compared preference-weighted scores of Black and White men and women in the United States, ages 38–89 years, at a single time point using a host of instruments, and found that sociodemographic and socioeconomic variables explained the observed differences in scores among Black and White women.13 However, where differences among men's scores were observed, they were income dependent.13 Shaw et al also found differences by race and ethnicity among people in the United States in preference-weighted HRQOL scores using the EQ-5D that could be explained by socioeconomic variables (eg, age, income, education).12 These studies do not make longitudinal preference-weighted comparisons among White, Black, and Hispanic participants.
The appropriate and equitable delivery of health care services requires understanding how patients value their own health. In addition, to the extent that preference-weighted HRQOL is a key input for cost-effective analysis, it is important to understand how racial/ethnic disparities in measures of HRQOL may impact economic valuations of a given intervention for a disease. For instance, if certain racial groups systematically value the given intervention more highly than other groups do, then an economic valuation of the intervention that ignores such disparity may systematically overestimate the value of the intervention among other groups.
This analysis seeks to examine whether such racial/ethnic disparities may exist as captured by a commonly used HRQOL measure and consider whether such disparities can be explained by differences in socioeconomic and clinical variables.
The selenium and vitamin E cancer prevention trial (SELECT) was designed to test prostate cancer chemoprevention strategies for men >50 years old. The HRQOL data collected as part of the trial afford the opportunity to compare preference-weighted HRQOL scores by race and ethnicity. The trends in scores of ∼10,000 trial participants were examined over time and cross-sectionally.
Methods
Patient population and study design
This study was a retrospective analysis of a fully de-identified existing secondary dataset that had been collected in a previous study (clinical trial identifier: NCT00006392). Therefore, this was considered an exempt study for IRB purposes.
Data were obtained from SELECT conducted by the SWOG Cancer Research Network, a member of the National Cancer Institute's (NCI's) National Clinical Trials Network and the NCI's Community Oncology Research Program. Details of the study design have been previously published.14,15 In brief, SELECT was a randomized, placebo-controlled trial enrolling 35,533 men with no previous prostate cancer diagnosis aged 50 years or older for Black men, and 55 years or older for all others.
Patients were randomized to 1 of 4 intervention arms: selenium with matching vitamin E placebo, vitamin E with matching selenium placebo, both selenium and vitamin E, or matching placebos. The primary end point of the trial was evaluating the effect of selenium and vitamin E on the incidence of prostate cancer. The trial activated in July 2001 and blinded follow-up continued until October 2008. A subsample of SELECT participants who were randomized from a Community Clinical Oncology Program or Veterans Affairs (VA) sites were invited to complete the HRQOL questionnaire.
Quality-of-life questionnaire
The SF-36V was used as the HRQOL instrument for the SELECT trial. The questionnaire was administered in English and Spanish at the following time points: before randomization (baseline) and at years 1, 3, 5, and 7. Data were collected at annual clinic visits, and participants could mail questionnaires to the site if their clinic visit was missed.
The SF-36V is a modified version of the SF-36 that aimed to improve the precision of the results for patients in a VA setting16 while maintaining the same measurement properties as the SF-36.16–18 The SF-36 measures HRQOL across 8 domains: physical functioning, role physical, bodily pain, general health, vitality, social functioning, role emotional, and mental health.19 Each domain is scored from 0 to 100, with 0 being the worst possible state and 100 being the best possible state. Scores across domains are not comparable (eg, a role physical score cannot be compared with a role emotional score).
In addition, 2 scores in a given domain are ordinal but not cardinal: a higher number is unambiguously better than a lower number, but the difference between any 2 values within a domain is not necessarily the same (eg, 58 is better than 57, and 68 is better than 67, but 68 may be better than 67 by “more” than 58 is better than 57). The overall effect of this is that SF-36V scores cannot be compared between any 2 people. To make comparisons, the SF-6D can be used; it gives a single, preference-weighted, composite score from 0 (worst possible) to 1 (best possible), and the scores are both cardinal and ordinal.
That the SF-6D is preference weighted means the scores reflect how much a person values a given health state. To convert the SF-36V to the SF-6D, Brazier et al conducted time trade-off interviews to estimate how many years of perfect health respondents would trade for 10 years of a given state.20 Brazier et al provide an algorithm for converting SF-36V responses to SF-6D scores.
Statistical analysis
SELECT participants were included in this analysis if they completed the SF-36V at baseline, and at least 1 of the 1-, 3-, and 5-year follow-ups. Excluded from the analysis were data from the 7-year follow-up as the distribution of non-Hispanic White, non-Hispanic Black, and Hispanic respondents deviates from the distribution at the prior time points. In addition, participants who selected “non-White Other” as their race/ethnicity were excluded from the analysis because they represent a small sample size. Each SF-36V response was converted to an SF-6D score using the algorithm developed by Brazier et al20 and made available to researchers through the University of Sheffield.20–22
Counts of baseline participant characteristics were calculated at each time point. The outcome variable was the SF-6D score, calculated as already described. Unadjusted mean SF-6D scores were calculated by the race/ethnicity group and time point. This analysis models the association of HRQOL with race/ethnicity longitudinally using a linear mixed model to account for correlations between observations within each participant. Random effects were specified for the intercept and slope. Models included race/ethnicity, assessment year, and an interaction of race/ethnicity and year.
Other covariates included age at baseline (50–64 years, 65–74 years, 75 years, and older), SELECT intervention arm (selenium, vitamin E, selenium + vitamin E, placebo), baseline education level (high school or General Education Development (GED) tests, some college or vocational education, college degree or higher), body mass index (BMI), categorization at time t (<25 kg/m2, 25 to <30 kg/m2, ≥30 kg/m2), and current smoking status at time t (yes vs. no).
The authors also examined models controlling for socioeconomic status (SES) and rurality as additional covariates. Insurance status was used as a proxy for SES where those with Medicaid, dual Medicare and Medicaid, or no insurance were classified as low SES. Rurality was defined using the U.S. Department of Agriculture Economic Research Service rural–urban continuum codes, based on zip code of participant residence when available, or zip code of study site when residence was not available.23 All analyses were completed using SAS (version 9.4; SAS Corporation, Cary, NC) and R (version 3.6.1; R Foundation for Statistical Computing). P value <0.05 was reported as statistically significant.
Results
At baseline, 9691 men completed the SF-36V, of whom 7556 (78%) were White, 1592 (16.4%) were non-Hispanic Black, and 543 (5.6%) were Hispanic (Table 1). The distribution of race/ethnicity remained stable over the 5-year period of the study. A majority of respondents were between 50 and 64 years old: 57.7% were 55–64 years old and 5.2% were 50–54 years old. The distribution of participants across intervention arms was well balanced at all time points. Rates of smoking and adherence to study supplements decreased at each time point, and slightly fewer men were overweight or obese at later time points, but the distribution of rates of all other measures remained similar over time indicating dropout was approximately equal across subgroups and study treatment arms.
Table 1.
Participant Characteristics at Baseline and Years 1, 3, and 5
| N with data required for SF-6D scoresa | Baseline |
Year 1 |
Year 3 |
Year 5 |
|---|---|---|---|---|
| (N = 9691) | (N = 8939) | (N = 8076) | (N = 7309) | |
| N with any QOL data | 10,964 | 10,173 | 9575 | 8959 |
| Race/ethnicity, n (%) | ||||
| Hispanic | 543 (5.6) | 471 (5.3) | 365 (4.5) | 358 (4.9) |
| Non-Hispanic White | 7556 (78.0) | 7071 (79.1) | 6447 (79.8) | 5793 (79.3) |
| Non-Hispanic Black | 1592 (16.4) | 1397 (15.6) | 1264 (15.7) | 1158 (15.8) |
| Age at baseline, years, n (%) | ||||
| 50–64 | 6091 (62.9) | 5624 (62.9) | 5113 (63.3) | 4630 (63.4) |
| 65–74 | 3033 (31.3) | 2792 (31.2) | 2516 (31.2) | 2278 (31.2) |
| ≥75 | 567 (5.9) | 523 (5.9) | 447 (5.5) | 401 (5.5) |
| SELECT intervention arm, n (%) | ||||
| Selenium alone | 2428 (25.1) | 2245 (25.1) | 2029 (25.1) | 1843 (25.2) |
| Vitamin E + selenium | 2408 (24.8) | 2232 (25.0) | 2030 (25.1) | 1840 (25.2) |
| Placebo | 2419 (25.0) | 2215 (24.8) | 1984 (24.6) | 1788 (24.5) |
| Vitamin E alone | 2436 (25.1) | 2247 (25.1) | 2033 (25.2) | 1838 (25.1) |
| Education level, n (%) | ||||
| ≤High school graduate or GED | 2320 (23.9) | 2115 (23.7) | 1925 (23.8) | 1730 (23.7) |
| Some college/vocational school | 2900 (29.9) | 2655 (29.7) | 2402 (29.7) | 2170 (29.7) |
| ≥College graduate | 4419 (45.6) | 4119 (46.1) | 3710 (45.9) | 3374 (46.2) |
| Unknown/missing | 52 (0.5) | 50 (0.6) | 39 (0.5) | 35 (0.5) |
| Clinical profile/health status, n (%) | ||||
| Family history of prostate cancer | 1671 (17.2) | 1544 (17.3) | 1387 (17.2) | 1251 (17.1) |
| Family history of lung cancer | 1276 (13.2) | 1176 (13.2) | 1068 (13.2) | 952 (13.0) |
| Family history of colon cancer | 1097 (11.3) | 1010 (11.3) | 925 (11.5) | 839 (11.5) |
| BMI at time t, n (%) | ||||
| Normal (BMI <25 kg/m2) | 1833 (18.9) | 1604 (17.9) | 1488 (18.4) | 1388 (19.0) |
| Overweight (BMI 25 to <30 kg/m2) | 4594 (47.4) | 4172 (46.7) | 3589 (44.4) | 3197 (43.7) |
| Obese (BMI ≥30 kg/m2) | 3227 (33.3) | 2948 (33.0) | 2656 (32.9) | 2335 (31.9) |
| Unknown/missing | 37 (0.4) | 215 (2.4) | 343 (4.2) | 389 (5.3) |
| Current smoker at time t | 876 (9.0) | 784 (8.8) | 638 (7.9) | 520 (7.1) |
| Adherence to SELECT study supplements at time t | 7714 (86.3) | 6547 (81.1) | 4574 (62.6) | |
At each time point, the population includes all men with SF-6D indexed at both baseline and at that time point.
BMI, body mass index; GED, General Education Development tests; QOL, quality of life; SELECT, selenium and vitamin E cancer prevention trial; SF-6D, Short-Form Six-Dimension.
Using insurance status as a proxy for SES, 95.8% of Hispanic participants, 97% of White participants, and 96% of Black participants are considered to be of high SES (Table 2). Type of insurance differs by race/ethnicity: 61.5% of Hispanic participants, 52.6% of Black participants, and 11.2% of White participants have VA/military insurance (Table 2).
Table 2.
Insurance Status, Geographic Location, and Age at Baseline by Race/Ethnicity
| Hispanic (N = 543) | Non-Hispanic White (N = 7556) | Non-Hispanic Black (N = 1592) | |
|---|---|---|---|
| SES based on insurance,a n (%) | |||
| Low SES | 15 (2.8) | 188 (2.5) | 58 (3.6) |
| High SES | 520 (95.8) | 7328 (97) | 1529 (96) |
| Unknown | 8 (1.5) | 40 (0.5) | 5 (0.3) |
| VA/military insurance,b n (%) | |||
| VA/military | 334 (61.5) | 850 (11.2) | 837 (52.6) |
| All others | 201 (37) | 6666 (88.2) | 750 (47.1) |
| Unknown | 8 (1.5) | 40 (0.5) | 5 (0.3) |
| SES excluding VA/military,c n (%) | |||
| Low SES | 15 (7.2) | 188 (2.8) | 58 (7.7) |
| High SES | 186 (89) | 6478 (96.6) | 692 (91.6) |
| Unknown | 8 (3.8) | 40 (0.6) | 5 (0.7) |
| Geographic region,d n (%) | |||
| Rural | 21 (3.9) | 1107 (14.7) | 47 (3) |
| Urban | 479 (88.2) | 6243 (82.6) | 1504 (94.5) |
| Unknown/missing | 43 (7.9) | 206 (2.7) | 41 (2.6) |
Medicaid, Medicare + Medicaid, and no insurance (self-pay or no means) are classified as low SES; all other insurance types as high SES.
Insurance types VA, military, and VA/military (old) classified as VA/military.
Excludes participants classified as VA/military above.
Based on zip codes mapped to U.S. Department of Agriculture Economic Research Service's rural–urban continuum codes.
SES, socioeconomic status; VA, Veterans Affairs.
The unadjusted mean SF-6D score at baseline was 0.80 (95% confidence interval [CI] = 0.79–0.80) for the entire sample. Unadjusted mean scores were 0.81 (95% CI = 0.80–0.82), 0.80 (95% CI = 0.80–0.81), and 0.76 (95% CI = 0.75–0.76) for Hispanic, non-Hispanic White, and non-Hispanic Black participants, respectively (Table 3).
Table 3.
Unadjusted Mean Short-Form Six-Dimension Scores at Baseline and Years 1, 3, and 5
| Race/ethnicity | Baseline |
Year 1 |
Year 3 |
Year 5 |
|---|---|---|---|---|
| (N = 9691) | (N = 8939) | (N = 8076) | (N = 7309) | |
| All participants | 0.80 (0.79–0.80) | 0.79 (0.79–0.79) | 0.79 (0.78–0.79) | 0.78 (0.78–0.78) |
| Hispanic | 0.81 (0.80–0.82) | 0.88 (0.86–0.89) | 0.87 (0.86–0.89) | 0.83 (0.81–0.84) |
| Non-Hispanic White | 0.80 (0.80–0.81) | 0.79 (0.79–0.80) | 0.79 (0.79–0.79) | 0.79 (0.78–0.79) |
| Non-Hispanic Black | 0.76 (0.75–0.76) | 0.74 (0.73–0.75) | 0.73 (0.73–0.74) | 0.74 (0.73–0.74) |
Hispanic and non-Hispanic White participants had higher unadjusted mean SF-6D scores than non-Hispanic Black participants at baseline and at every subsequent time point (P < 0.05; Table 3). Non-Hispanic White participants had lower mean scores than Hispanic participants at every time point (Table 3; Fig. 1).
FIG. 1.
Unadjusted mean Short-Form Six-Dimension scores by race and ethnicity at baseline and years 1, 3, and 5.
Hispanic ethnicity was associated with a higher mean SF-6D score (P = 0.003) than non-Hispanic White respondents after controlling for age, education level, study arm, smoking status, and BMI (Table 4, Model 1), whereas non-Hispanic Black participants had a lower adjusted mean score (P < 0.0001). The trajectory of preference-weighted HRQOL scores trended downward over time with scores expected to be lower than baseline by 0.010, 0.014, and 0.022 in follow-up years 1, 3, and 5, respectively (P < 0.0001 for all estimates). The statistical significance of the interaction terms between race/ethnicity and year indicates that the mean SF-6D score trajectories differed over time by race/ethnicity groups despite adjusting for participant characteristics (Table 4, Model 1).
Table 4.
Linear Mixed Model Results
| Covariate | Original Model (1) |
Original Model (2): (1) + SES |
Original Model (3): (2) + Rurality |
|||
|---|---|---|---|---|---|---|
| Coefficient | P | Coefficient | P | Coefficient | P | |
| Race/ethnicity | ||||||
| Hispanic | 0.015 | <0.01 | 0.016 | <0.01 | 0.014 | <0.01 |
| Non-Hispanic White | Ref | Ref | Ref | |||
| Non-Hispanic Black | −0.033 | <0.0001 | −0.033 | <0.0001 | −0.034 | <0.0001 |
| Measurement period | ||||||
| Baseline | Ref | Ref | Ref | |||
| Year 1 | −0.010 | <0.0001 | −0.010 | <0.0001 | −0.010 | <0.0001 |
| Year 3 | −0.014 | <0.0001 | −0.014 | <0.0001 | −0.014 | <0.0001 |
| Year 5 | −0.022 | <0.0001 | −0.021 | <0.0001 | −0.022 | <0.0001 |
| Interaction effect | ||||||
| Year 1 × Hispanic | 0.074 | <0.0001 | 0.073 | <0.0001 | 0.074 | <0.0001 |
| Year 3 × Hispanic | 0.076 | <0.0001 | 0.076 | <0.0001 | 0.076 | <0.0001 |
| Year 5 × Hispanic | 0.039 | <0.0001 | 0.038 | <0.0001 | 0.040 | <0.0001 |
| Year 1 × non-Hispanic Black | −0.009 | <0.01 | −0.008 | <0.01 | −0.008 | <0.01 |
| Year 3 × non-Hispanic Black | −0.008 | 0.02 | −0.008 | 0.01 | −0.007 | 0.02 |
| Year 5 × non-Hispanic Black | 0.003 | 0.36 | 0.003 | 0.39 | 0.004 | 0.25 |
| Age at baseline, years | ||||||
| 50–54 | −0.013 | 0.02 | −0.013 | 0.02 | −0.013 | 0.02 |
| 55–64 | Ref | Ref | Ref | |||
| 65–74 | −0.011 | <0.0001 | −0.011 | <0.0001 | −0.011 | <0.0001 |
| ≥75 | −0.045 | <0.0001 | −0.045 | <0.0001 | −0.045 | <0.0001 |
| Education level | ||||||
| ≤High school graduate or GED | Ref | Ref | Ref | |||
| Some college/vocational school | 0.007 | 0.01 | 0.007 | 0.01 | 0.007 | 0.01 |
| ≥College graduate | 0.036 | <0.0001 | 0.036 | <0.0001 | 0.036 | 0.00 |
| Unknown/missing | 0.037 | 0.01 | 0.036 | <0.01 | 0.037 | 0.01 |
| SELECT intervention arm | ||||||
| Selenium alone | −0.002 | 0.53 | −0.002 | 0.58 | −0.001 | 0.60 |
| Vitamin E + selenium | 0.001 | 0.71 | 0.001 | 0.65 | 0.001 | 0.62 |
| Placebo | Ref | Ref | Ref | |||
| Vitamin E alone | −0.003 | 0.33 | −0.002 | 0.38 | −0.002 | 0.38 |
| BMI (kg/m2) at time t | ||||||
| Normal (BMI <25) | Ref | Ref | Ref | |||
| Overweight (BMI 25 to <30) | 0.002 | 0.30 | 0.002 | 0.27 | 0.002 | 0.33 |
| Obese (BMI ≥30) | −0.014 | <0.0001 | −0.014 | <0.0001 | −0.015 | 0.00 |
| Missing/unknown | −0.009 | 0.02 | −0.008 | 0.03 | −0.009 | 0.01 |
| Smoking status | ||||||
| Never smoke | Ref | Ref | Ref | |||
| Ever smoke | −0.031 | <0.0001 | −0.030 | <0.0001 | −0.030 | 0.00 |
| Socioeconomic status | ||||||
| Low | Ref | Ref | ||||
| High | 0.005 | 0.38 | 0.005 | 0.38 | ||
| Unknown | 0.010 | 0.49 | 0.011 | 0.44 | ||
| Geographic area | ||||||
| Rural | Ref | |||||
| Urban | 0.000 | 0.98 | ||||
| Missing/unknown | 0.012 | 0.07 | ||||
Several covariates had a statistically significant association with SF-6D scores. Compared with an education up to and including high school, having some college/vocational school is associated with a 0.006 (P = 0.02) increase in SF-6D score; similarly, having at least a college degree is associated with a 0.036 (P < 0.0001) increase in SF-6D score. Obesity is associated with a 0.014 (P < 0.0001) decrease in SF-6D score compared with being a normal weight, and being a smoker is associated with a 0.031 (P < 0.0001) decrease in SF-6D compared with nonsmokers. Compared with those that are 50–64 years old, those being 65–74 and ≥75 years are associated with decreases in SF-6D scores of 0.010 (P < 0.0001) and 0.045 (P < 0.0001).
The results remained similar with the addition of covariates for SES (Table 4, Model 2) and for rural geography (Table 4, Model 3).
Discussion
This analysis of longitudinal, preference-weighted HRQOL data builds on the existing body of HRQOL research, displaying differences in self-reported scores by race/ethnicity are persistent over time, including among healthy men. The results suggest that Hispanic participants had the highest preference-weighted HRQOL scores at every time point. Despite having baseline SF-6D scores similar to those of Hispanic participants, White participants had much lower scores at every subsequent time point, whereas Black participants' scores were lower at baseline and subsequently paralleled the score trajectory of White participants.
This study provides additional insights on this topic because the data are longitudinal with 5 years of follow-up; no previous study on this topic has used such longitudinal data. Moreover, contrary to findings by Pereira et al13 and Shaw et al,12 sociodemographic and socioeconomic variables do not explain the observed differences in preference-weighted HRQOL. The magnitude of observed differences among Black and White men in this analysis is statistically significant (and exceeds the minimally important difference for the SF-6D), whereas previous estimates (Shaw et al12) were not statistically significant. In addition, the direction of the difference is opposite of the earlier estimates by Shaw et al12 (preference-weighted scores were higher for Black men, here they are lower).
There are several possible explanations for the observed change in SF-6D scores from baseline to the year 1 follow-up among Hispanic participants. It is possible that participants with lower scores dropped out of the study at a higher rate than others, representing an informative missing data problem. However, it is unclear why such a trend might be unique to Hispanic participants. Furthermore, it is also possible that other factors, such as family support or health conditions, are changing differentially over time between participant groups. Unfortunately, these data are also unavailable, so this study cannot account for this possibility in the statistical analysis.
In an earlier study, Shaw et al12 reported that preference-weighted HRQOL scores were higher for Black and Hispanic respondents than for White respondents. They used the EQ-5D as the instrument, a preference-weighted measure of HRQOL analogous to the SF-6D, with controls for SES.12 Like Shaw et al, this analysis shows that Hispanic men have higher preference-weighted HRQOL than their White counterparts. However, contrary to Shaw et al, this study finds that preference-weighted scores are lower for Black men.
Notably, Shaw et al conclude that the magnitude of their point estimates for the effect of race/ethnicity on preference-weighted HRQOL indicates that “race/ethnicity is a statistically significant but practically unimportant” factor.12 Here, the point estimates for race/ethnicity cross the threshold of the statistical minimally important differences for the SF-6D that have been previously reported.24,25
Re-examining the preference weights for HRQOL using preferences that are derived from a US sample may be considered. The preference weights used for conversion of the SF-36V to the SF-6D as well as the Shaw et al EQ-5D study are based on interviews with the UK general public.12 As such, different value sets can be generated that are country specific.26 Craig et al have generated a value set based on data collected from the US population and have found that the preferences derived from their US sample are highly correlated with those derived from the United Kingdom.6 However, only 3% of the sample in the study by Craig et al has self-identified as Hispanic, whereas ∼11% of the total US population does so; this limits the extent to which the results can be generalized.6
Understanding differences among racial and ethnic groups in self-reported preference-weighted HRQOL is essential not only for providing culturally competent and equitable health care but also for conducting accurate cost-effectiveness analyses of new interventions and programs. There are many disease domains for which there is a positive association between the use of self-reported HRQOL and clinical outcomes, including mental health and palliative care.27–29 These findings are highly relevant to cancer prevention efforts given the surveys were administered as part of a cancer prevention trial.
Among these, cancer is a disease domain for which the effect of HRQOL on outcomes is particularly well studied. Higher self-reported HRQOL in patients with advanced cancer is associated with better progression-free survival,30 and the use of patient-reported outcomes in clinical care improved survival among patients with metastatic cancer.31 Adjusting for culturally-divergent preferences in HRQOL could enhance the clinical benefits of considering HRQOL in care.
Further research is needed to elucidate the reason for these observed differences to design and implement interventions that can eliminate disparities. For example, it may be fruitful to address patients' specific QOL concerns in the clinical encounter and offer solutions accordingly. These may include focus groups, exercise programs, or behavioral therapy. McMullen and Luborsky propose that the self-appraisal of health is dictated by culturally acceptable views on health and health identities.10 Furthermore, they posit that self-appraisal of health is driven by the desire to maintain identity.10
Expanding upon this study with a more diverse study population should be considered. Yin et al report that, after controlling for other socioeconomic factors, race/ethnicity mediate the relationship between physical functioning and self-reported HRQOL among older adults, with non-Hispanic Black and Hispanic respondents reporting lower HRQOL for a given level of physical functioning than their White counterparts.11 They conclude that social comparisons mediate self-reported HRQOL scores.11 This mechanism may also explain the divergent scores observed in this study.
This study is subject to several limitations. First, the study population includes only healthy men over the age of 50 years; moreover, because of the recruitment strategy, the only participants who are 50–54 years old are Black. Similar analyses should be conducted using data sets that include both men and women, as well as a wider range of age, enabling comparisons by gender and between younger and older adults. In addition, the SELECT collected data only about Hispanic ethnicity and a sample of Asian participants that was insufficiently large for analysis. Because this study is limited by the availability of the data elements included in the secondary data set, future studies that include a more comprehensive representation of races, ethnicities, and nationalities are needed.
Second, comorbidities may also explain the differences in HRQOL at baseline and over time; though history of some conditions was included in intake data forms, a more complete clinical picture at baseline and in follow-up years should be included in a future study.
Third, the results may be confounded due to sample attrition over the study period. That is, if the attrition was associated with the race/ethnicity of the study participants (eg, Blacks may be more likely than Whites to drop out of the study), then the estimates shown in Figure 1 are likely to be biased because of over-representation of the remaining races/ethnicities. If such biases were real, however, it is unclear why they would be more pronounced among Hispanic participants relative to the other groups, as suggested by Figure 1. Unfortunately, the available data provide little insight in this regard.
At any rate, Table 1 shows that the distribution of race/ethnicity over the 5-year period remained stable, which suggests that this potential bias is unlikely. Moreover, Appendix Table A1 suggests that 33% of the evaluable sample was lost over the 5-year period and that relatively small percentages of the dropouts were due to death (less than 1% of the total sample size at baseline) and incidence of prostate cancer (less than 2% of the total sample size at baseline), which would have been indicative of lower health status and thus lower preference-weighted QOL. Therefore, it is unlikely that the results are reflecting significant biases due to sample attrition.
Finally, the effect of SES on health care and health outcomes has been well established for decades.32 Prior research also indicates that having insurance is associated with better HRQOL.33,34 The present analysis is constrained by the lack of income data and insurance status beyond enrollment. Given the data available here, insurance type (Medicaid vs. other) is used as an imperfect but suitable proxy for SES.35,36
By this measure, a high percentage of participants are considered to be of high SES (Table 2), however, this does not capture changes over time. Further research using data on the social determinants of health including, but not limited to, granular income data and more detailed information about insurance status will contribute to illuminating the mechanism by which HRQOL is affected by race/ethnicity.
Differences in preference-weighted QOL by race/ethnicity could inform clinical practices and decision making. That patients may value a given health state differently on account of ascriptive characteristics could be a useful data point in clinical decision making. Future research should focus on understanding the mechanisms that drive the observed differences.
Appendix
Appendix Table A1.
Dropout Analysis
| Baseline | Year 1 | % change from baseline | Year 3 | % change from year 1 | % change from baseline | Year 5 | % change from year 3 | % change from baseline | |
|---|---|---|---|---|---|---|---|---|---|
| Number of men with any QOL data | 10,964 | 10,173 | −7.8 | 9575 | −5.9 | −12.7 | 8959 | −6.4 | −22.4 |
| Number of evaluable men at this time pointa | 9691 | 8939 | −8.4 | 8076 | −9.7 | −16.7 | 7309 | −9.5 | −32.6 |
| n | % of baseline | n | % of year 1 | % of baseline | n | % of year 3 | % of baseline | ||
|---|---|---|---|---|---|---|---|---|---|
| Dropouts from previous time point due to… | |||||||||
| Death |
|
0 |
0.0 |
75 |
0.7 |
0.7 |
91 |
0.9 |
0.8 |
| Prostate cancer |
|
3 |
0.0 |
105 |
1.0 |
1.0 |
195 |
1.9 |
1.8 |
| Transferred from QOL site to non-QOL site |
|
2 |
0.0 |
42 |
0.4 |
0.4 |
3 |
0.0 |
0.0 |
| Another reason/unknown | 747 | 6.8 | 641 | 6.3 | 5.8 | 478 | 4.7 | 4.4 | |
Evaluable refers to men with all required data to calculate Short-Form Six-Dimension scores.
QOL, quality of life.
Disclaimer
Some initial data in this article were presented at the SWOG Spring Conference (April 2021) and the ASCO Quality Care Conference (September 2021).
Data Availability Statement
The data underlying this article will be shared on reasonable request to the corresponding author.
Author Disclosure Statement
Dr. Ramsey reports a consulting/advisory role at Bayer Corporation, AstraZeneca, Merck & Company, Inc., GRAIL, Seattle Genetics, Flatiron Health, and Biovica. Dr. Ramsey reports his institution serves as an advisor for Genentech and receives research funding from Bayer Corporation and Genentech/Roche. Dr. Ramsey receives travel accommodations from Bayer Corporation. The authors have no other disclosures.
Funding Information
Funding for the SELECT was provided by the NIH/NCI/NCORP. NIH/NCI/NCORP grant awards UG1CA189974 and UM1CA182883; and the Hope Foundation for Cancer Research (Dr. Unger).
References
- 1. Zhang L, Ferguson TF, Simonsen N, Chen L, Tseng TS. Racial/ethnic disparities in health-related quality of life among participants with self-reported diabetes from NHANES 2001–2010. Diabetes Educ 2014;40(4):496–506. [DOI] [PubMed] [Google Scholar]
- 2. Reyes ME, Ye Y, Zhou Y, et al. Predictors of health-related quality of life and association with survival may identify colorectal cancer patients at high risk of poor prognosis. Qual Life Res 2017;26(2):319–330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Quinten C, Coens C, Mauer M, et al. Baseline quality of life as a prognostic indicator of survival: a meta-analysis of individual patient data from EORTC clinical trials. Lancet Oncol 2009;10(9):865–871. [DOI] [PubMed] [Google Scholar]
- 4. Braun DP, Gupta D, Grutsch JF, Staren ED. Can changes in health related quality of life scores predict survival in stages III and IV colorectal cancer? Health Qual Life Outcomes 2011;9(1):1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Bukavina L, Zaramo CEB, Tarabonata L, Modlin CS. Perceptions of health-related quality of life: a minority men's perspective. Am J Mens Health 2017;11(4):990–998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Craig BM, Pickard AS, Stolk E, Brazier JE. US valuation of the SF-6D. Med Decis Mak Int J Soc Med Decis Mak 2013;33(6):793–803. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Kandula NR, Lauderdale DS, Baker DW. Differences in self-reported health among Asians, Latinos, and non-Hispanic Whites: the role of language and nativity. Ann Epidemiol 2007;17(3):191–198. [DOI] [PubMed] [Google Scholar]
- 8. Ye J, Shim R, Garrett SL, Daniels E. Health-related quality of life in elderly and white patients with cancer: results from Medicare Managed Care Population. Ethn Dis 2012;22(Summer):302–307. [PMC free article] [PubMed] [Google Scholar]
- 9. Skarupski KA, de Leon CFM, Bienias JL, et al. Black–white differences in health-related quality of life among older adults. Qual Life Res 2007;16(2):287–296. [DOI] [PubMed] [Google Scholar]
- 10. McMullen CK, Luborsky MR. Self-rated health appraisal as cultural and identity process: African American Elders' Health and Evaluative Rationales. Gerontologist 2006;46(4):431–438. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Yin H, Lin SJ, Kong SX, et al. The association between physical functioning and self-rated general health in later life: the implications of social comparison. Appl Res Qual Life 2011;6(1):1–19. [Google Scholar]
- 12. Shaw JW, Johnson JA, Chen S, Levin JR, Coons SJ. Racial/ethnic differences in preferences for the EQ-5D health states: results from the U.S. valuation study. J Clin Epidemiol 2007;60(5):479–490. [DOI] [PubMed] [Google Scholar]
- 13. Pereira CC, A, Palta M, Mullahy J, Fryback DG. Race and preference-based health-related quality of life measures in the United States. Qual Life Res 2011;20(6):969–978. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Lippman SM, Klein EA, Goodman PJ, et al. Effect of selenium and vitamin E on risk of prostate cancer and other cancers: the Selenium and Vitamin E Cancer Prevention Trial (SELECT). JAMA 2009;301(1):39–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Lippman SM, Goodman PJ, Klein EA, et al. Designing the Selenium and Vitamin E Cancer Prevention Trial (SELECT). JNCI J Natl Cancer Inst 2005;97(2):94–102. [DOI] [PubMed] [Google Scholar]
- 16. Klein EA. SELECT Protocol. SELECT Biorepository. SWOG. https://www.swog.org/clinical-trials/biospecimen-resources/select-biorepository, Accessed November 3, 2022.
- 17. Kazis LE. The Veterans SF-36 Health Status Questionnaire: development and Application in the Veterans Health Administration. Monitor 2000;5(1):18. [Google Scholar]
- 18. Kazis LE, Lee A, Spiro A, et al. Measurement Comparisons of the Medical Outcomes Study and Veterans SF-36® Health Survey. Health Care Financ Rev 2004;25(4):43–58. [PMC free article] [PubMed] [Google Scholar]
- 19. Ware JE, Snow KK, Kosinski M, Gandek B.. SF-36 Health Survey: Manual and Interpretation Guide. Lincoln, RI: Quality Metric, Inc., 1993. [Google Scholar]
- 20. Brazier J, Roberts J, Deverill M. The estimation of a preference-based measure of health from the SF-36. J Health Econ 2002;22:271–292. [DOI] [PubMed] [Google Scholar]
- 21. Kharroubi S, McCabe C. Modelling health state preference data using a non-parametric Bayesian method. 2006. https://eprints.whiterose.ac.uk/10933/ Accessed November 8, 2021.
- 22. Brazier JE, Roberts J. The estimation of a preference-based measure of health from the SF-12. Med Care 2004;42(9):851–859. [DOI] [PubMed] [Google Scholar]
- 23. USDA ERS—Rural-Urban Continuum Codes. https://www.ers.usda.gov/data-products/rural-urban-continuum-codes/ Accessed July 5, 2021.
- 24. Walters SJ, Brazier JE. What is the relationship between the minimally important difference and health state utility values? The case of the SF-6D. Health Qual Life Outcomes 2003;1:4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Khanna D, Furst DE, Wong WK, et al. Reliability, validity, and minimally important differences of the SF-6D in systemic sclerosis. Qual Life Res Int J Qual Life Asp Treat Care Rehabil 2007;16(6):1083–1092. [DOI] [PubMed] [Google Scholar]
- 26. Tsuchiya A, Ikeda S, Ikegami N, et al. Estimating an EQ-5D population value set: the case of Japan. Health Econ 2002;11(4):341–353. [DOI] [PubMed] [Google Scholar]
- 27. Knaup C, Koesters M, Schoefer D, Becker T, Puschner B. Effect of feedback of treatment outcome in specialist mental healthcare: meta-analysis. Br J Psychiatry 2009;195(1):15–22. [DOI] [PubMed] [Google Scholar]
- 28. Greenhalgh J, Long AF, Flynn R. The use of patient reported outcome measures in routine clinical practice: lack of impact or lack of theory? Soc Sci Med 1982 2005;60(4):833–843. [DOI] [PubMed] [Google Scholar]
- 29. Gondek D, Edbrooke-Childs J, Fink E, Deighton J, Wolpert M. Feedback from outcome measures and treatment effectiveness, treatment efficiency, and collaborative practice: a systematic review. Adm Policy Ment Health 2016;43:325–343. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Hui D, Darke AK, Guthrie KA, et al. Association between health-related quality of life and progression-free survival in patients with advanced cancer: A secondary analysis of SWOG clinical trials. JCO Oncol Pract 2022;18(4):e442–e451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Basch E, Deal AM, Dueck AC, et al. Overall survival results of a trial assessing patient-reported outcomes for symptom monitoring during routine cancer treatment. JAMA 2017;318(2):197–198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Adler NE, Boyce WT, Chesney MA, Folkman S, Syme SL. Socioeconomic inequalities in health. No easy solution. JAMA 1993;269(24):3140–3145. [PubMed] [Google Scholar]
- 33. Institute of Medicine (US) Committee on the Consequences of Uninsurance. Effects of Health Insurance on Health. In: Care Without Coverage: Too Little, Too Late. National Academies Press (US); 2002. https://www.ncbi.nlm.nih.gov/books/NBK220636/ Accessed September 21, 2021. [PubMed]
- 34. Penson DF, Stoddard ML, Pasta DJ, Lubeck DP, Flanders SC, Litwin MS. The association between socioeconomic status, health insurance coverage, and quality of life in men with prostate cancer. J Clin Epidemiol 2001;54(4):350–358. [DOI] [PubMed] [Google Scholar]
- 35. Ayanian JZ, Kohler BA, Abe T, Epstein AM. The relation between health insurance coverage and clinical outcomes among women with breast cancer. N Engl J Med 1993;329(5):326–331. [DOI] [PubMed] [Google Scholar]
- 36. Foraker RE, Rose KM, Whitsel EA, Suchindran CM, Wood JL, Rosamond WD. Neighborhood socioeconomic status, Medicaid coverage and medical management of myocardial infarction: atherosclerosis risk in communities (ARIC) community surveillance. BMC Public Health 2010;10:632. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data underlying this article will be shared on reasonable request to the corresponding author.