Health-related quality of life of cancer and noncancer patients in Medicare managed care†
This article is a US Government work and, as such, is in the public domain in the United States of America.
Abstract
BACKGROUND
Data from the Health Care Financing Administration's (HCFA) Medicare Health Outcomes Survey (MHOS) of patients enrolled in managed care services through Medicare were analyzed. The MHOS provided baseline estimates of quality of life of cancer survivors in comparison to a frequency age-matched cohort of noncancer patients.
METHOD
In 1998, the MHOS was mailed to a random sample of 279,135 beneficiaries. Completed surveys were received from 167,096 respondents (60%). Some 22,747 respondents who had been diagnosed with cancer were frequency age matched to an equal number of noncancer patients.
RESULTS
Cancer survivors had statistically significantly poorer scores than noncancer patients on all eight subscales as well as on the Physical Component and Mental Component summary measures of the Medical Outcomes Study Short Form-36 (MOS SF-36). Comparisons by type and number of cancers for which an individual was currently in treatment showed lowest quality of life for those in treatment for lung carcinoma, followed by those who were in treatment for more than one type of cancer.
CONCLUSION
The data suggest that cancer shows negative effects on health-related quality of life that are not explainable by simple effects of age because frequency age-matched cancer survivors had statistically significantly lower scores on all 10 scores of the MOS SF-36. Effect sizes are evaluated to determine the clinical significance of these differences in health-related quality of life. The MHOS offers useful data for planning and improving cancer policy and programs. Cancer 2003;97:674–81. Published 2003 by the American Cancer Society.
DOI 10.1002/cncr.11085
Cancer occurs later in life, with nearly 80% of all cancers diagnosed among individuals at 55 years and older.1 Currently, cancer is the second leading cause of death in the United States.1 However, due to improvements in prevention, early detection, and treatment, cancer death rates for all sites combined decreased an average of 0.6% per year from 1990 to 1996.2 There are an estimated 8.9 million cancer survivors in the United States.3 The term “cancer survivor” is defined in several ways. The traditional National Institutes of Health definition of a cancer survivor was someone who had survived cancer for 5 years and was disease free. The American Cancer Society (ACS) has adopted a broader definition of a cancer survivor. Its definition includes people from the time of their diagnosis with cancer and for the balance of their lives, a definition that has been used by the National Coalition for Cancer Survivorship for a number of years.4 This is the definition used in this article.
The ACS has issued three cancer challenge goals for the year 2015, which include decreasing incidence and mortality due to cancer. Recognizing that cancer and its treatment may result in physical impairments as well as psychosocial losses, these goals also include improving the health-related quality of life (HRQOL) of cancer survivors.5 Attempting to achieve any of these three goals poses formidable difficulties. However, unlike incidence and mortality, quality of life has the additional problem of lacking a commonly agreed upon standard of measurement for assessing the current status of this variable and any progress that is made in improving it.
Over the past two decades, increased attention has been given to developing HRQOL as an operationally defined patient-report based measure, and HRQOL has become a standard outcome measure in cancer clinical trials.6, 7 Recognition of the need to include HRQOL as part of population-based assessments of health status has also increased. The ACS is beginning to conduct its own population-based studies of the HRQOL, psychosocial adaptation, and changing needs of cancer survivors. However, this will take time, and data are needed now to use as a guide for program development. The available data on the HRQOL of cancer survivors are provided by relatively small opportunistic samples. Because of the difficulties of separating the effects of cancer on HRQOL from effects that are due to comorbid conditions and other life changes, additional data on a comparable sample of noncancer patients are needed for comparison.
The available data comparing the quality of life of cancer survivors with individuals who have never had cancer show little difference. A Canadian study8 compared the HRQOL of breast carcinoma survivors 8 years after diagnosis. In that study, fewer cancer survivors reported positive quality of life than similarly aged controls, but these differences were small and nonsignificant. Another Canadian study9 compared index cancer cases with neighborhood controls of a similar age and gender. The quality of life of the cancer survivors was found to be similar to the neighborhood controls. A French population-based case–control study found that testicular carcinoma survivors with a mean follow-up of 11 years did not differ significantly in HRQOL from controls.10 A Swedish study11 comparing HRQOL in long-term head and neck carcinoma survivors with general population norms showed no significant differences except on the role-physical scale of the Medical Outcomes Study Short Form-36 (MOS SF-36). These studies did not include U.S. samples and were limited in sample size. This article compares cancer survivors with noncancer patients in a reasonably large sample of respondents.
In April 1999, the ACS hosted a conference in Arlington, VA, to bring together major government agencies to identify possible collaborations and to share information on how each was attempting to measure the HRQOL of various populations. This article is the result of a successful collaboration initiated at that meeting, i.e., a collaboration between the ACS and the Health Care Financing Administration (HCFA; recently renamed the Centers for Medicare and Medicaid Services). The collaboration led to an analysis of data from the 1998 Medicare Health Outcomes Survey (MHOS) of patients receiving managed care services through Medicare. This analysis compared the HRQOL responses of cancer survivors with those of frequency age-matched respondents who were not cancer survivors, using the definition of cancer survivor as any person who is living after a diagnosis of cancer.
MATERIALS AND METHODS
Survey Instrument
The MHOS was developed by the National Committee for Quality Assurance (NCQA) under contract to HCFA to measure the outcomes of care provided by Medicare + Choice organizations to Medicare beneficiaries enrolled in these health plans. This performance measure was included in the 1998 Health Plan Employer Data and Information Set (HEDIS 3.0) for Medicare. A technical expert panel of representatives from health plans, health services researchers, and clinicians assisted in the development of the tool.12 Originally called the Health of Seniors survey and later renamed the Medicare Health Outcomes Survey, the MHOS included a 95-item core and a five-item variable module.
Measures
The core instrument included a measure of HRQOL and questions to collect demographic information. The HRQOL measure was the MOS SF-36, a 36-item well tested, valid, and reliable self-report tool that has been used in hundreds of studies world-wide to measure HRQOL.13
- 1
Physical functioning (PF). Ten questions that ask the extent to which health limits the performance of physical activities.
- 2
Role-physical (RP). Four questions that ask individuals the extent to which their physical health limits them in their work or other usual activities in terms of time and performance.
- 3
Bodily pain (BP). Two questions that ask individuals about the severity of their pain and the extent to which pain interferes with normal work, including work outside the home and housework.
- 4
General health (GH). Five questions that ask individuals to rate their current health status overall, their susceptibility to disease, and their expectations for health in the future.
- 5
Vitality (VT). Four questions that ask individuals to rate subjective well-being in terms of energy and fatigue.
- 6
Social functioning (SF). Two questions that ask individuals about limitations in normal social functioning due specifically to health.
- 7
Role-emotional (RE). Three questions that ask whether emotional problems have interfered with accomplishments at work or other usual activities in terms of time as well as performance.
- 8
Mental health (MH). Five questions that ask how frequently the respondent experiences feelings related to anxiety, depression, loss of behavioral or emotional control, and psychological well-being.
These eight scales provide the basis for calculating two summary measures, the Physical Component Summary (PCS) and the Mental Component Summary (MCS). One additional question that asks the respondents to rate their general health compared to one year ago is also included but is not used in calculating the two summary measure scores. The MOS SF-36 is scored so that higher scores represent better functioning on both the two summary measures and on all eight subscales. Norm based scaling is used so that scores are standardized using normative values for the general U.S. population. A score of 50 represents the national average for both the summary scores and for the subscales. A score that is 10 points above or below the mean score of 50 represents a difference of 1 standard deviation from the national average.
Procedures
The survey was sent to a random sample of 1,000 Medicare beneficiaries continuously enrolled for at least six months in each plan with a Medicare contract in place on or before January 1, 1997. In plans with 1,000 or fewer Medicare enrollees, all eligible members were surveyed. The sampling frame included the aged and the disabled but excluded those eligible for Medicare because of end stage renal disease, since these patients are not eligible to enroll in a managed-care plan. To ensure the validity and reliability of the data collected, plans were instructed to contract with one of six NCQA-certified vendors for administration of the survey. The survey was administered to the first cohort at baseline (May 1998) and again to the same cohort in spring 2000. A new cohort will be selected each year for baseline measurement.
Cohort 1 data collection began in May 1998 and included 279,135 beneficiaries enrolled in 268 managed care plans covering 287 market areas. Completed surveys were received from 167,096 respondents, a raw response rate of 60%. Data files were aggregated by NCQA and forwarded to HCFA's data cleaning, analysis, and dissemination contractor, the Health Services Advisory Group Health Outcomes Team. An analytic data set was created that merged plan characteristics from HCFA administrative data files with the survey response data and patient demographic and entitlement information from the Medicare Enrollment Data Base. This augmented file was delivered to HCFA in November 1998.
For purposes of this study, two separate files were created from the MHOS Cohort 1 final data set. The first file consisted of all respondents who answered positively when asked if a physician had ever told them that they had any cancer other than skin carcinoma (Q33). This yielded 22,747 unique respondents whom we defined as cancer survivors using the ACS definition of a cancer survivor as anyone still living after a cancer diagnosis.
The second data set created was a file of respondents who indicated that no physician had ever told them that they had cancer other than skin carcinoma. A frequency age-matched sample of noncancer patients was randomly selected for inclusion in the analytic data set using 13 age categories: 29 or younger, 30–39, 40–49, 50–59, 60–64, 65–69, 70–74, 75–79, 80–84, 85–89, 90–94, 95–99, and 100 years or older. We used a frequency age-matched cohort analytic approach to better assess differences between cancer survivors and those who had never been diagnosed with cancer.
Data Analysis
Both for those who were cancer survivors and those who were not, only individuals whose Medicare entitlement was based on being aged 65 or older were included in the analyses. Individuals younger than 65 (whose entitlement was based on a disability) were not included. A single SAS data file including all individuals who met these criteria (n = 43,757) was created for use in the analyses. One-way and two-way analyses of variance performed in PROC GLM were used to compare mean MOS SF-36 standardized scores. Tukey's honestly significant difference (HSD) tests were used in post hoc testing. Chi-square tests in PROC FREQ were used for demographic comparisons. Multiple linear regression analyses to test for independent predictors of MOS SF-36 subscale and summary scores were performed in PROC REG.
RESULTS
Table 1 presents data on sample characteristics. Both the cancer survivors and those who did not report having been diagnosed with any cancer other than skin carcinoma (hereafter referred to as the no-cancer group or noncancer subjects) were equally divided between the elderly (65–74 years old) and the older elderly (75 years and older). Both groups had a somewhat higher percentage of female than male respondents and were predominately white and non-Hispanic. The majority of individuals in both groups were currently married and about 30% in each group were widowed. About 30% of both groups had less than a high school education, and 34.4% of each group reported graduation from high school as their highest educational attainment. Less than one-fifth in each group reported an annual household income of less than $10,000, whereas less than one-third of each group reported an annual household income of $10,000–$19,999. Among the cancer survivors, 11.8%, 5.4%, 3.0%, and 16.7%, respectively, reported currently being in treatment for breast, colorectal, lung, or prostate carcinoma only. An additional 2.1% reported being in treatment for more than one of these four cancers.
| Characteristic | Cancer survivors | No cancer |
|---|---|---|
| No. (%) | No. (%) | |
| Age group (yrs) | ||
| 65–74 | 10,969 (50.6) | 11,137 (50.5) |
| 75 and older | 10,716 (49.4) | 10,935 (49.5) |
| Gendera | ||
| Male | 9774 (45.1) | 9126 (41.4) |
| Female | 11,911 (54.9) | 12,946 (58.6) |
| Racea | ||
| American Indian | 143 (0.7) | 164 (0.8) |
| Asian/Pacific Islander | 231 (1.1) | 414 (1.9) |
| Black/African-American | 1059 (5.0) | 1396 (6.5) |
| White | 19,387 (91.5) | 19,079 (88.6) |
| Other or multiracial | 351 (1.7) | 481 (2.2) |
| Hispanica | ||
| Yes | 697 (3.3) | 965 (4.5) |
| No | 20,283 (96.7) | 20,284 (95.5) |
| Material statusa | ||
| Married | 12,582 (59.0) | 12,347 (56.8) |
| Divorced/separated | 1904 (8.9) | 1813 (8.3) |
| Widowed | 6387 (29.9) | 7001 (32.2) |
| Single, never married | 459 (2.2) | 577 (2.7) |
| Educationa | ||
| Less than HS | 5993 (28.2) | 6981 (32.5) |
| HS/GED | 7289 (34.4) | 7413 (34.4) |
| Some college/2-yr degree | 4717 (22.2) | 4252 (19.8) |
| Four-year college degree | 1543 (7.3) | 1371 (6.4) |
| More than 4-year college degree | 1666 (7.9) | 1477 (6.9) |
| Household incomea | ||
| Less than $10,000 | 2776 (15.8) | 3200 (18.7) |
| $10,000–$19,999 | 5501 (31.4) | 5314 (31.0) |
| $20,000–$49,999 | 7395 (42.2) | 7028 (41.0) |
| $50,000–$79,999 | 1244 (7.1) | 1066 (6.2) |
| $80,000 or more | 608 (3.5) | 535 (3.1) |
| Currently in treatment for | N/A N/A | |
| Breast carcinoma only | 2560 (11.8) | |
| Colorectal carcinoma only | 1178 (5.4) | |
| Lung carcinoma only | 651 (3.0) | |
| Prostate carcinoma only | 3627 (16.7) | |
| In treatment for more than one of the above | 456 (2.1) |
- MHOS: Medicare Health Outcomes Survey; HS: high school; GED: general education.
- a Chi-square P value = 0.001 for comparing cancer survivors with those who reported not having been diagnosed with any cancer other than skin carcinoma.
Table 2 presents mean scores for the two summary measures and for the eight subscales of the MOS SF-36 by cancer status (cancer survivor or no-cancer group) and by age group (65–74 and 75 or older). For the entire sample and within each age group, cancer survivors had significantly lower mean scores on all 10 scales than did the noncancer subjects (all P < 0.001). As would be expected, individuals aged 65–74 had significantly higher mean scores than those aged 75 and older.
| MOS SF-36 scale | Mean (SD) for MOS SF-36 scale | ||
|---|---|---|---|
| Total sample (n = 43,757) | Age group (yrs) | ||
| 65–74 (n = 22,106) | ≥ 75 (n = 21,651) | ||
| PCS | |||
| Cancer | 38.5 (12.0)a | 40.6 (12.0)a | 36.2 (11.7)a |
| No cancer | 41.2 (11.8) | 43.7 (11.3) | 38.5 (11.7) |
| MCS | |||
| Cancer | 51.4 (10.7)a | 52.1 (10.3)a | 50.6 (11.1)a |
| No cancer | 52.6 (10.0) | 53.4 (9.4) | 51.7 (10.5) |
| Physical function | |||
| Cancer | 38.3 (13.1)a | 41.1 (12.5)a | 35.4 (13.1)a |
| No cancer | 40.5 (12.9) | 43.7 (11.8) | 37.2 (13.2) |
| Role-physical | |||
| Cancer | 40.6 (12.9)a | 42.8 (12.8)a | 38.2 (12.6)a |
| No cancer | 43.0 (12.8) | 45.7 (12.2) | 40.3 (12.8) |
| Bodily pain | |||
| Cancer | 43.2 (11.4)a | 44.2 (11.2)a | 42.2 (11.4)a |
| No cancer | 45.0 (11.2) | 46.3 (10.7) | 43.6 (11.4) |
| General health | |||
| Cancer | 43.0 (11.4)a | 44.0 (11.4)a | 41.9 (11.3)a |
| No cancer | 46.0 (10.8) | 47.5 (10.6) | 44.6 (10.9) |
| Vitality scale | |||
| Cancer | 45.5 (11.2)a | 47.0 (11.1)a | 43.9 (11.1)a |
| No cancer | 47.9 (10.8) | 49.7 (10.4) | 46.1 (10.9) |
| Social function | |||
| Cancer | 45.7 (12.8)a | 47.3 (12.2)a | 44.0 (13.3)a |
| No cancer | 48.0 (11.7) | 49.9 (10.5) | 46.1 (12.5) |
| Role-emotional | |||
| Cancer | 47.2 (12.1)a | 48.8 (11.4)a | 45.5 (12.6)a |
| No cancer | 48.3 (11.5) | 50.1 (10.3) | 46.4 (12.4) |
| Mental health | |||
| Cancer | 50.2 (10.6)a | 50.8 (10.3)a | 49.6 (10.8)a |
| No cancer | 51.3 (10.2) | 52.1 (9.7) | 50.5 (10.5) |
- MOS SF-36: Medical Outcomes Study Short Form-36; MHOS: Medicare Health Outcomes Survey; SD: standard deviation; PCS: Physical Component Summary; MCS: Mental Component Summary.
- a P < 0.001 for difference in mean scores between cancer and no-cancer patients.
Effect size, which refers to the strength of a relationship in a population or the degree of departure from a null hypothesis, offers a way to judge the clinical or practical importance of a result.14 Using a SD of 10, it was determined that effect sizes were small (defined as 0.2–0.4) for PC, PF, RP, GH, VT, and SF scales, with a range of 0.22–0.30, corresponding to score differences of 2.2–3.0 points. The remaining four scales—MC, BP, RE, and MH—had effect sizes less than 0.2. Although four, the differences were statistically significant for these four scales, they are not practically significant, as judged by the magnitude of the effect.
Multiple linear regressions showed that having been diagnosed with cancer was still a significant predictor of MOS SF-36 summary and subscale scores even after the effects of a number of other variables were taken into account. Variables for inclusion in the regression models were selected based on a technical report prepared for the HCFA and NCQA.15 Nonmedical variables included in the models were age (in years), gender, race (white/nonwhite), Hispanic ethnicity (yes/no), annual income (</≥ $20,000), marital status (married/not married), education (</≥ high school graduate), home ownership (yes/no), and data collection mode (paper/phone). In addition to cancer, the models included the following medical conditions: hypertension; angina or coronary artery disease; congestive heart failure; myocardial infarction; stroke; chronic obstructive pulmonary disease, emphysema, or asthma; Crohn disease, ulcerative colitis, or irritable bowel disease; arthritis of the hand or wrist; arthritis of the hip or knee; sciatica; and diabetes.
Table 3 shows regression results for variables that were significant in the multiple linear regression models for the PCS and MH. Results for these two scales are presented because they are typical of those for all 10 scales with respect to the relative sizes of the estimated betas for cancer compared with those for other variables that were also significant in the models. In addition, the betas for cancer for these two scales represent the range of betas for cancer over the 10 scales (from −0.777 to −2.573) because they are the smallest (−0.777) and one of the largest (−1.979).
| Variable | Beta estimate | SE | t | P value |
|---|---|---|---|---|
| MOS SF-36 PCS | ||||
| Age (in single years) | −0.350 | 0.010 | −36.58 | 0.0001 |
| Education (less than high school graduate) | −1.571 | 0.140 | −11.20 | 0.0001 |
| Household income (< $20,000/yr) | −2.205 | 0.133 | −16.56 | 0.0001 |
| Married | −0.869 | 0.137 | −6.35 | 0.0001 |
| Female | −0.805 | 0.129 | −6.23 | 0.0001 |
| Data collection via mail | −0.550 | 0.221 | −2.48 | 0.0131 |
| High blood pressure | −1.451 | 0.121 | −12.02 | 0.0001 |
| Angina or coronary artery disease | −2.554 | 0.188 | −13.57 | 0.0001 |
| Congestive heart failure | −4.580 | 0.252 | −18.20 | 0.0001 |
| Heart attack | −0.516 | 0.228 | −2.27 | 0.0233 |
| Stroke | −3.880 | 0.221 | −17.59 | 0.0001 |
| COPD, emphysema, or asthma | −5.431 | 0.175 | −31.08 | 0.0001 |
| Crohn disease, ulcerative colitis, or irritable bowel disease | −1.845 | 0.246 | −7.50 | 0.0001 |
| Arthritis of the hand or wrist | −2.015 | 0.137 | −14.67 | 0.0001 |
| Arthritis of the hip or knee | −5.296 | 0.136 | −39.05 | 0.0001 |
| Sciatica | −3.598 | 0.146 | −24.62 | 0.0001 |
| Diabetes | −2.504 | 0.164 | −15.24 | 0.0001 |
| Cancer | −1.979 | 0.117 | −16.87 | 0.0001 |
| MOS SF-36 MH | ||||
| Age (in single years) | −0.062 | 0.009 | −6.80 | 0.0001 |
| Education (less than high school) | −2.663 | 0.134 | −19.91 | 0.0001 |
| Household income (< $20,000/yr) | −1.816 | 0.128 | −14.23 | 0.0001 |
| Hispanic | −1.661 | 0.303 | −5.48 | 0.0001 |
| Female | −0.896 | 0.124 | −7.24 | 0.0001 |
| Data collection via mail | −0.918 | 0.213 | −4.30 | 0.0001 |
| High blood pressure | −0.822 | 0.116 | −7.10 | 0.0001 |
| Angina or coronary artery disease | −0.941 | 0.180 | −5.24 | 0.0001 |
| Congestive heart failure | −2.305 | 0.240 | −9.62 | 0.0001 |
| Stroke | −3.375 | 0.210 | −16.08 | 0.0001 |
| COPD, emphysema, or asthma | −2.666 | 0.167 | −15.95 | 0.0001 |
| Crohn disease, ulcerative colitis, or irritable bowel disease | −3.915 | 0.236 | −16.61 | 0.0001 |
| Arthritis of the hand or wrist | −1.376 | 0.132 | −10.46 | 0.0001 |
| Arthritis of the hip or knee | −0.998 | 0.130 | −7.68 | 0.0001 |
| Sciatica | −2.034 | 0.140 | −14.53 | 0.0001 |
| Diabetes | −1.049 | 0.157 | −6.67 | 0.0001 |
| Cancer | −0.777 | 0.112 | −6.91 | 0.0001 |
- MOS SF-36: Medical Outcomes Study Short Form-36; PCS: Physical Component Summary; MH: Mental Health; COPD: Chronic obstructive pulmonary disease.
To investigate the effect of current treatment for cancer on MOS SF-36 mean scores, subjects were assigned to one of seven groups: no cancer (n = 22,072); cancer but not currently in treatment for breast, colorectal, lung, or prostate carcinoma (n = 10,930); currently in treatment for prostate carcinoma only (n = 3,627); currently in treatment for breast carcinoma only (n = 2,560); currently in treatment for colorectal carcinoma only (n = 1,178); currently in treatment for lung carcinoma only (n = 651); and currently in treatment for more than one of these four cancers (n = 456). Means and post hoc testing results for the two summary measures and eight subscales are shown in Table 4. Generally, mean scores are ordered from highest to lowest in the following sequence: no cancer; cancer not currently in treatment for breast, colorectal, lung, or prostate carcinoma; in treatment for prostate carcinoma only; in treatment for breast carcinoma only; in treatment for colorectal carcinoma only; in treatment for more than one cancer; and in treatment for lung carcinoma only. For all 10 scales, those currently in treatment for lung carcinoma only or for more than one cancer had significantly lower mean scores than any of the other five groups. Those who had not been diagnosed with cancer had the highest mean score on all 10 scales and their mean for each scale was significantly higher than the means for any of the five in treatment groups except the group in treatment for prostate carcinoma only (which differed significantly from noncancer subjects on 5 of the 10 scales). Mean scores for those in treatment for breast carcinoma only or for colorectal carcinoma only were consistently lower than the means for those being treated for prostate carcinoma only and were consistently higher than means for those being treated for lung carcinoma only or for more than one of the four listed cancers.
| MOS SF-36 scale | No cancer (n = 22,072) | Cancer but not in treatment (n = 10,930)a | Prostate carcinoma in treatment (n = 3627) | Breast carcinoma in treatment (n = 2560) | Colorectal carcinoma in treatment (n = 1178) | In treatment for more than one cancer (n = 456)b | Lung carcinoma treatment (n = 651) |
|---|---|---|---|---|---|---|---|
| PCSc | 41.2d | 39.3e | 39.0e | 37.3f | 36.9f | 34.1g | 31.0h |
| MCSc | 52.6d | 52.1de | 51.9de | 50.9e | 49.1f | 46.2g | 47.0g |
| Physical functioningc | 40.5d | 39.0d | 40.0d | 36.6e | 36.5e | 33.5f | 30.4d |
| Role-physicalc | 43.0d | 41.5e | 40.5ef | 39.9f | 38.3g | 35.3h | 33.7f |
| Bodily painc | 45.0d | 43.8d | 43.8d | 41.8e | 42.3e | 40.2f | 38.9f |
| General healthc | 46.1d | 44.2e | 42.7f | 42.6f | 40.6g | 37.1h | 34.6i |
| Vitality scalec | 47.9d | 46.1e | 46.2e | 45.0ef | 43.9f | 41.5g | 38.8h |
| Social functioningc | 48.0d | 46.8de | 46.1ef | 45.0f | 42.9g | 39.5h | 37.4i |
| Role-emotionalc | 48.3d | 48.0d | 47.7d | 46.3b | 44.6f | 40.9h | 43.0g |
| Mental healthc | 51.3d | 50.8d | 51.1d | 49.6b | 48.5e | 46.0f | 46.1f |
- MOS SF-36: Medical Outcomes Study Short-Form 36; MHOS: Medicare Health Outcomes Survey; PCS: Physical Component Summary; MCS: Mental Component Summary.
- a Not currently in treatment for breast, colorectal, lung, or prostate carcinoma.
- b Currently in treatment for more than one of the following carcinoma: breast, colorectal, lung, prostate.
- c P < 0.001 for difference in mean scores across cancer/treatment groups in one-way analysis of variance.
- d, e, f, g, h, i Tukey groupings in post hoc testing; means within the same Tukey grouping are not significantly different from one another.
Although the effect sizes for the overall cancer group compared with the noncancer group were small or nonexistent, larger effect sizes were found when using the treatment groups described above for comparison to the noncancer group. For example, the effect size for GHS for those currently in treatment for colorectal carcinoma was 0.55 (moderate) and the effect size for PCS for those currently in treatment for lung carcinoma was 1.02 (large), corresponding to score differences of 5.5 and 10.2 points. The effect sizes can be summarized as follows for each treatment group compared with the noncancer group: currently not in treatment for breast, colorectal, prostate, or lung carcinoma; 0.03–0.19 (no effect for all scales); in treatment for prostate carcinoma only, 0.05–0.34 (no effect except small effect for GHS and RPS); in treatment for breast carcinoma only, 0.17–0.39 (small effects for all except MCS and MHS); in treatment for colorectal carcinoma only, 0.27–0.55 (small effects for all except GHS and SFS, which were moderate); in treatment for more than one of the four cancers, 0.48–0.90 (large effects for GHS and SFS, moderate for all others except BPS, which was small); and in treatment for lung carcinoma only, 0.52–1.15 (large effects for PCS, PFS, RFS, GHS, VTS, and SFS, moderate for all others).
DISCUSSION
The central purpose of this article is to compare the quality of life of a large national sample of cancer survivors who completed the MOS SF-36 as part of the MHOS with that of a frequency age-matched sample of noncancer MHOS respondents. No claim is made that this sample of managed care recipients in the Medicare program is representative of the Medicare population. However, this data set provides information on a large sample of people 65 and older who have access to care through Medicare, some of whom have received a diagnosis of cancer at some point in their lives, and on an age-comparable group who have not been diagnosed with cancer.
As a basis for examining the quality of life of cancer survivors, the MOS SF-36 has a number of strong practical advantages. Among its strengths is its applicability to many disease groups as well as the general population, so that it can be used to make comparisons between cancer patients and patients with other diseases as well as healthy populations. Since most studies of the quality of life of cancer survivors lack either a non-cancer control group or broad normative data on non-cancer patients against which to compare findings, this is a very important advantage in attempting to sort out the contribution of cancer and cancer treatment to differences in quality of life versus the effects of comorbitidites and natural aging.
Research documents the responsiveness of the scales and summary scores of the MOS SF-36 to various treatment interventions.16 Findings indicate that small absolute differences in scores are clinically meaningful. For instance, a study documented that untreated asthmatics reported a PCS score 3.7 points lower than asthmatics treated with drugs. Another showed that people suffering from a migraine headaches reported an increase in PCS score of 2.09 points when their headache was treated. Therefore, the difference in mean physical functioning between cancer survivors in our study (mean PCS score = 38.5) and in the noncancer group (mean PCS score = 41.2) can be likened to the increase in physical limitations experienced when suffering from untreated migraine headache, which is clinically meaningful. Clinical significance of the differences found in our study is also evidenced by the effect sizes seen, especially with regard to the impact of treatment status on HRQOL.
Our data show that cancer patients have statistically significantly poorer quality of life than noncancer respondents as measured by all eight subscales of the MOS SF-36 and the two summary measures even after considering the effects of other variables. These differences in HRQOL, however, represented small effects for five of the eight MOS SF-36 subscales. Of the summary measures, only the PCS showed a small effect. The effect size for MCS did not indicate any practical significance. These data support the view that cancer is a uniquely traumatizing event, producing broad negative impact on a variety of dimensions of a survivor's HRQOL.
Secondary analysis of data collected for broader purposes can limit the questions one can ask of these data. One limitation of this data is the absence of information on the time since cancer diagnosis. Although these data do not allow clear analysis by stage of recovery, there is evidence in these results that patients who have access to care through Medicare managed care continue to have serious biomedical and psychosocial HRQOL problems. This should receive further attention, particularly as these quality-of-life dimensions are significantly worse for the cancer survivors than for same-age controls who have not had to deal with the physical and psychosocial sequelae of cancer.
These data also allow a comparison by type and number of cancers for which a subject is currently being treated. Analyses were made across the condition of having no cancer; having a cancer diagnosis but not in treatment for breast, colorectal, lung, or prostate carcinoma; or in treatment for one or more of these four common types of cancer. As these analyses show, those who had not been diagnosed with cancer had statistically significantly higher HRQOL scores. In contrast to the small effects found when comparing cancer survivors overall with the noncancer group, larger effects were found when the comparisons were made across treatment status. Being in treatment for lung carcinoma only or in treatment for more than one type of cancer had the most negative effect on quality of life. The general pattern showed the highest MOS SF-36 scores for individuals without cancer, followed by those not in treatment for breast, colorectal, lung, or prostate carcinoma; only prostate in treatment; only breast in treatment; only colorectal in treatment; treatment for more than one type of cancer; and only lung in treatment. For all 10 MOS SF-36 scales, being in treatment either for lung carcinoma only or for more than one type of cancer resulted in statistically significantly poorer HRQOL. Moderate or large effects were found for all 10 MOS SF-36 scores when comparing those in treatment for lung carcinoma only with the noncancer group. For those in treatment for more than one of the four cancers, moderate or large effects were found for all 10 MOS SF-36 scales except for the BPS, which showed a small effect. These data show that the severity and range of effects of particular types of cancer and the varied treatments offered for various cancer sites produce different effects on HRQOL. However, the development of several malignancies is particularly difficult for patients. Having to fight cancer on several different fronts poses a cascade of complex HRQOL challenges.
Previous studies employing the MOS SF-36 have found that older age groups (particularly 75 and older) show lower scores on the eight scales of the MOS SF-36.17, 18 Consistent with these earlier findings, the 75 and older group in this study had significantly lower scores than the 65–74 group on all 10 scales. It is noteworthy that even in the 75 and above age group, those with a cancer history continued to have significantly lower scores than the noncancer group.
There are some limitations to using the MOS SF-36 to describe the HRQOL of cancer patients. Its strength in being a generic measure that can be used to assess functional outcomes across many medical conditions, both acute and chronic, as well as with healthy populations, also provides a limitation. It does not include data on domains that measure the specific effects of particular diseases such as cancer and its various treatments. Therefore, it does not reveal as much as one might like in terms of well delineated areas for intervention.
However, the MOS SF-36 as it is included in the MHOS offers important and relevant data for those concerned with planning, evaluating, and improving cancer policy and programs. Because the MHOS was repeated again after 2 years with this 1998 cohort, when data are available for analysis, changes in scores may also reveal patterns of interest to program planners. Furthermore, the continued administration of the MHOS to new cohorts each year will provide data that may be useful in tracking changes in HRQOL related to the efforts of the ACS and its partners to improve the quality of life of cancer survivors over the next 15 years.
