- © 2003 by American Society of Clinical Oncology
Assessment of Quality of Life in Outpatients With Advanced Cancer: The Accuracy of Clinician Estimations and the Relevance of Spiritual Well-Being—A Hoosier Oncology Group Study
- Michael J. Fisch,
- Michael L. Titzer,
- Jean L. Kristeller,
- Jianzhao Shen,
- Patrick J. Loehrer,
- Sin-Ho Jung,
- Steven D. Passik and
- Lawrence H. Einhorn
- From the Department of Palliative Care and Rehabilitation, The University of Texas M.D. Anderson Cancer Center, Houston, TX; Oncology/Hematology Associates of Southwest Indiana, Evansville; Division of Hematology/Oncology, Indiana University; Division of Biostatistics, Indiana University School of Medicine; and Walther Cancer Institute, Indianapolis; Department of Psychology, Indiana State University, Terre Haute, IN; Departments of Medicine and Behavioral Sciences, University of Kentucky, Lexington, KY; and Department of Biostatistics and Bioinformatics, Duke University, Durham, NC.
- Address reprint requests to Michael J. Fisch, MD, MPH, The University of Texas M.D. Anderson Cancer Center, Unit 008, Rm P12.2911, 1515 Holcombe Blvd., Houston, TX 77030-4009; email: mfisch{at}mdanderson.org.
Abstract
Purpose: To evaluate the association between quality-of-life (QOL) impairment as reported by patients and QOL impairment as judged by nurses or physicians, with and without consideration of spiritual well-being (SWB).
Patients and Methods: A total of 163 patients with advanced cancer were enrolled onto a therapeutic trial, and cross-sectional data were derived from clinical and demographic questionnaires obtained at baseline, including assessment of patient QOL and SWB. Clinicians rated the QOL impairment of their patients as mild, moderate, or severe. Clinician-estimated QOL impairment and patient-derived QOL categories were compared. Correlation coefficients were estimated to associate QOL scores using different instruments. The analysis of variance method was used to compare Functional Assessment of Cancer Therapy–General scores on categorical variables.
Results: There was no significant association between self-assessment scores and marital status, education level, performance status, or predicted life expectancy. However, a strong relationship between SWB and QOL was noted (P < .0001). Clinician-estimated QOL impairment matched the level of patient-derived QOL correctly in approximately 60% of cases, with only slight variation depending on the method of categorizing patient-derived QOL scores. The accuracy of clinician estimates was not associated with the level of SWB. Interestingly, a subset analysis of the inaccurate estimates revealed an association between lower SWB and clinician underestimation of QOL impairment (P = .0025).
Conclusion: Clinician estimates of QOL impairment were accurate in more than 60% of patients. SWB is strongly associated with QOL, but it is not associated with the overall accuracy of clinicians’ judgments about QOL impairment.
QUALITY OF life (QOL) is a multidimensional concept that focuses on how disease and its treatment affect the individual.1 QOL has become an important concept in cancer care,2 and clinical trials have shown that changes in QOL are associated with changes in clinical variables including survival.3–,7 Increasingly, QOL measurement has become important in evaluating the effects of cancer treatment in clinical trials,8 and numerous QOL assessment tools are now available.9–,13 There are data supporting the idea that QOL assessment in routine practice can be accomplished in a way that facilitates communication without increasing consultation time.14 However, despite such data, these assessments are infrequently used in clinical practice. Instead, clinicians often rely on their subjective analysis of a patient’s QOL. This may be related to time constraints in the busy outpatient oncology setting, the concern that patients will find these assessments too intrusive,15 or the belief that QOL is readily discernible by the clinician without formal assessment.
One commonly used QOL instrument in clinical cancer research in the United States is the Functional Assessment of Cancer Therapy–General Version (FACT-G) scale.9 This is a self-administered health-related QOL instrument designed for use by general oncology patients and consisting of 34 items assessing physical well-being, social and family well-being, emotional well-being, and functional well-being. One dimension of health that is missing from the FACT-G and most similar QOL instruments is spirituality. This realm has been called “the last taboo” by psycho-oncologist Jimmie Holland.16 Spirituality encompasses personal views and behaviors that express a need for a connection to something greater than the self and a sense of meaning and purpose in life.17 Spirituality is distinguished from religion, which encompasses structured belief systems (often with a code of ethical behavior and specific religious practices) that address spiritual issues.18 Increasing attention has been paid recently to the importance of spiritual well-being (SWB) as an aspect of QOL in chronically ill patients, including cancer patients.19–,24 Part of the impetus for the development of instruments to measure SWB has come from qualitative research regarding patient viewpoints on QOL. Such qualitative data indicate that patients’ concern about SWB far exceeds that of their physicians,25 making SWB one of the most under-addressed aspects of patient adjustment. SWB, although correlated with total QOL, has been shown to independently contribute variance to the ability to enjoy a high QOL, despite the strain of physical symptoms such as pain and fatigue.26
There are now instruments with acceptable psychometric properties for measuring SWB in cancer patients. One such instrument is the 12-item Functional Assessment of Chronic Illness Therapy–Spiritual Well-Being (FACIT-Sp) scale (Fig 1⇓).27 Like the FACT-G, the FACIT-Sp is a simple, self-administered instrument designed for use in oncology. It was designed using a widely diverse socioeconomic and cultural sample and using patients with a range of cancer diagnoses and levels of disability. The FACIT-Sp is divided into a Meaning and Peace subscale and a Faith subscale. The Hoosier Oncology Group prospectively measured QOL and SWB as part of a randomized trial comparing an antidepressant to placebo.28
Functional Assessment of Chronic Illness Therapy-Spiritual Well-Being Scale. NOTE: Rated on a 5-point scale from “Not at all” to “Very much.” Items 9, 10, 11, and 12 relate to the Faith subscale; the remaining items are related to the Meaning/Peace subscale. Copyright 1993, 1997, David Cella.
Based on our clinical experience, we hypothesized that there would be an association between patient SWB and clinicians’ global impression of an individual’s QOL. The purpose of this cross-sectional study was to examine the relationship between QOL and SWB among patients enrolled onto a therapeutic trial testing a QOL intervention. We also sought to evaluate the relationship between patient-derived QOL scores and global impressions about individuals’ QOL from their cancer physicians or nurses.
PATIENTS AND METHODS
Subjects
Between July 1998 and October 2000, 163 adult ambulatory patients with advanced, incurable malignancy were enrolled from 15 sites of the Hoosier Oncology Group (three academic centers and 12 community sites). Patients with an expected survival between 3 and 24 months were eligible for study participation. Patient characteristics are listed in Table 1⇓. All patients were enrolled onto a Hoosier Oncology Group randomized, controlled trial that compared fluoxetine with placebo in outpatients with cancer. Inclusion criteria included an Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 3 and at least minimal symptoms of depression based on a prestudy two-question screening instrument. Exclusion criteria, which were used for the antidepressant trial, are listed in Table 2⇓.
Patient Characteristics
Exclusion Criteria
At the time of study entry, patients completed the FACT-G (version 3) and the FACIT-Sp. Patients were also asked to rate their own health as excellent, very good, good, fair, or poor. The clinicians were blinded to the patients’ QOL scores, although they were aware that the assessment was being made. At baseline, clinicians estimated the patients’ impairment in QOL as mild, moderate, or severe. The clinicians also made baseline judgments about the performance status and estimated survival of each patient. The survival estimate was categorized as 3 to 6, 7 to 12, 13 to 18, or 19 to 24 months.
Statistical Methods
Correlation coefficients were used to associate QOL scores using different instruments. The analysis of variance technique was used to compare FACT-G scores based on categorical variables such as marital status, education level, patients’ description of their own health, ECOG performance status, clinician-estimated life expectancy, and clinician-estimated level of QOL impairment. χ2 tests were used to compare the clinicians’ judgment about QOL impairment (mild, moderate, or severe) with categorized FACT-G scores.
Given that there is no standardized method to categorize FACT-G scores, we examined two methods. First, we categorized FACT-G scores into groups of three based on absolute numbers (method 1). FACT-G scores can range from 0 to 108. Thus, a score of 0 to 36 would be considered severe impairment, 37 to 72 would be considered moderate impairment, and 73 to 108 would be considered mild impairment. In addition, scores were categorized using the clinicians’ judgments about levels of QOL impairment to estimate the expected percentage of patients who would fall into each of three subgroupings of actual QOL as measured by the FACT-G (method 2).
The relationship between FACIT-Sp scores and FACT-G scores was assessed using regression analysis. The McNemar’s test was used to test if physicians tend to underestimate patients’ QOL. P values less than or equal to .05 were considered significant in all analyses. All P values reported are two-sided.
RESULTS
QOL
Of the 163 patients enrolled onto the study, FACT-G scores were completed by 159 patients, and 151 patients had full data sets available for the analysis involving clinician estimates of QOL. Some of the clinician-patient pairings were missing, mostly because these questions were considered supplemental to the clinical trial and were sometimes omitted because of time constraints or other reasons. The median FACT-G score was 65.4 (SD, ± 16.1; interquartile range, 54 to 76). The FACT-G scores are listed according to clinician-estimated QOL impairment in Table 3⇓.
Summary of FACT-G Scores Grouped by Clinician-Estimated QOL Impairment
FACT-G scores did not vary significantly in relation to marital status, education level, ECOG performance status, or clinician-estimated life expectancy (Table 4⇓). However, there was a significant association between the FACT-G scores and the patients’ rating of their own health (excellent, very good, good, fair, or poor).
Clinical Predictors of QOL
SWB
The median score from the SWB instrument (FACIT-Sp) was 35 (SD, ± 9.4; range, 7 to 48). Median scores from the Meaning and Peace subscale and the Faith subscale were 22 (interquartile range, 17 to 27) and 13 (interquartile range, 10 to 16), respectively. The distribution of the SWB scores is shown in Figure 2⇓. There was a significant correlation between the FACIT-Sp score and the FACT-G score (r = 0.56, P < .0001). This relationship was more strongly accounted for by the Meaning and Peace subscale (r = 0.64, P < .0001) than by the Faith subscale (r = 0.28, P = .0004). SWB could be categorized as low, medium, or high by dividing the range (7 to 48) into three evenly spaced categories. This produces cutoffs at scores of 21 and 34, with 10% of patients scoring less than 21 (low), 38% of patients scoring between 21 and 34 (medium), and 52% of patients scoring more than 34 (high).
The distribution of spiritual well-being scores. NOTE: The box in this box plot spans the interquartile range, and lines indicate location of first quartile, median, and third quartile. Whiskers extend to last data point within 1.5 times the interquartile range in either direction.
Relating Patient-Reported QOL to Clinician Judgments About QOL Impairment
Clinician-estimated QOL categories and actual FACT-G scores were significantly correlated (r = 0.41, P < .0001), as illustrated in Figure 3⇓. Using method 1 (categorizing FACT-G scores by absolute numbers), we found that clinician-estimated QOL impairment matched the corresponding FACT-G category in 67% of patients with mild impairment and 59% of patients with moderate impairment but only in 43% of patients with severe impairment (Table 5⇓). Mismatches occurred between the patient-assessed QOL impairment and the clinician-estimated QOL impairment in 60 of 153 patients. Actual patient impairment was underestimated by clinicians for 31 patients and overestimated for 29 patients. There was a significant relationship between SWB and clinician-underestimated QOL impairment (r = 0.39, P = .0025), as shown in Figure 4⇓. The overall relationship between SWB and the FACT-G is summarized in Figure 5⇓.
Accuracy of Clinician-Predicted QOL Impairment by Method 1
Relationship between Functional Assessment of Cancer Therapy-General scores and clinician-predicted quality-of-life impairment. NOTE: The box in each box plot spans the interquartile range, and lines indicate location of first quartile, median, and third quartile. Whiskers extend to the last data point within 1.5 times the interquartile range in either direction.
Association of spiritual well-being with clinician underestimation of quality-of-life impairment. Functional Assessment of Cancer Therapy-General was categorized by method 1. The box in this box plot spans the interquartile range, and lines indicate location of first quartile, median, and third quartile. Whiskers extend to last data point within 1.5 times the interquartile range.
The relationship between spiritual well-being (SWB) and the Functional Assessment of Cancer Therapy-General (FACT-G).
Method 2 of categorizing FACT-G results used clinician-estimated QOL impairment data to estimate the expected percentage of patients who would fall into each of three subgrouping of FACT-G scores (Table 6⇓). This method produced cutoffs where the top 40% of scores were all over 71 (mild impairment), the middle 48% of scores were between 46 and 71 (moderate impairment), and the bottom 12% of scores were less than 46 (severe impairment). Using method 2, clinician-estimated QOL impairment levels correctly matched in 61% of patients with FACT-G scores indicating mild impairment and in 58% of patients with moderate impairment, but they matched in only 25% of patients with severe impairment. Once again, there was a significant relationship between SWB and clinician-estimated QOL impairment (r = 0.47, P = .0001).
Accuracy of Clinician-Predicted QOL by Method 2
DISCUSSION
Our findings demonstrate that QOL in patients with advanced cancer receiving care in an outpatient setting does not seem to be associated with life expectancy, marital status, or level of education, and perhaps, not even by performance status. There have been so many published QOL studies using different instruments and varying inception cohorts that it is difficult to discern whether these results are consistent with any general trend. In contrast, there was an expectation that the patients’ global estimate of their own health would correlate well with their QOL scores as measured by the FACT-G. This finding has face validity and is consistent with the results of another study in which patients’ perception of the severity of their illness was found to be a determinant of QOL scores.29 These findings may have relevance to the design of clinical trials for which QOL is a major end point. Performance status and its known association with survival have led many investigators to use performance status as a stratification factor in therapeutic trials. When QOL was found to be a strong predictor of survival outcomes, there was some speculation that QOL measurement might upstage performance status.30 However, given these data, stratification of patients based on their global estimate of their own health rather than performance status may be worth exploring.
There was only modest success by clinicians in accurately estimating the actual level of QOL impairment of their patients, with an overall accuracy rate of approximately 60%. The decision to assess clinician estimates of QOL impairment using a simple, one-item categorization without specific further instructions reflected our desire to capture the current clinical reality related to such judgments. The most appropriate method for categorizing the FACT-G into relevant levels has not been established and validated;31 thus, we used two methods to categorize QOL, and our findings were consistent regardless of the method used. Other studies have also demonstrated disparities between a patient’s self-reported QOL and that reported by someone other than the patient.32–,35 The current standard for assessment of QOL entails assessment from the perspective of the patient.9,13,36,37 Interestingly, the observation in our data that the most impaired patients were the least likely to be accurately classified by clinicians is identical to the findings of Passik et al38 with respect to oncologists’ ability to detect depression in ambulatory cancer patients. In that study, accurate estimations of depression severity dropped from 79% in the mildly depressed to 33% in the moderate category and to 13% in the severe category. Perhaps the explanation for this is that patients and families strive to preserve their dignity and present themselves in the best possible way when they are threatened with significant impairment. That is, sicker patients may tend to use thicker smoke screens. Detmar et al39 have documented evidence of some reluctance on the part of patients with advanced cancer to discuss certain aspects of health, with 20% to 25% of patients showing some reluctance to talk about topics in the emotional, family, and social realms. Self-reported QOL, however, was not significantly related to preferences regarding the discussion of QOL issues with oncologists in that study. Clearly, factors influencing the pattern of patient-clinician communication in the setting of advanced cancer are complex.
We know from previous data that high levels of SWB are effective in mediating enjoyment of life despite pain and fatigue26 and that spiritual concerns are one of the most under-addressed areas of patient adjustment in oncology services.40 However, this is the first study to explore how SWB relates to the global impressions of clinicians. We found that the FACIT-Sp correlated significantly with QOL scores on the FACT-G. This confirms the findings of Cohen et al 41 who demonstrated that the existential subscale of the McGill Quality of Life Questionnaire was at least as important as subscales for the physical, psychologic, and support domains. In our data, the relationship between SWB and QOL was accounted for more by the Meaning and Peace subscale than the Faith subscale, indicating that the Meaning and Peace subscale may be more important in QOL. This could be related to the fact that the Meaning and Peace subscale tends to be more sensitive to the patient’s psychologic status, whereas the Faith subscale is related more to the patient’s intrinsic religious belief system.27 This is important because someone’s sense of meaning and peace (existential well-being) could be improved by the medical community in the last several months of life, whereas someone’s sense of faith is probably less amenable to amelioration in medical settings. Our hypothesis that the accuracy of clinician-estimated QOL would be associated with SWB was not confirmed by these data. An important limitation for this analysis, however, is the lack of well-validated cutoffs for grouping SWB into different levels. When mismatches did occur, there was a significant association between SWB and underestimation of QOL impairment. Specifically, higher levels of SWB were associated with less risk of clinician underestimation of impairment. These findings are based on exploratory analysis and will require further study for confirmation. One reason why higher SWB might reduce the risk of clinician underestimation of QOL impairment is that clinicians may feel more comfortable spending additional time and communicating with patients whose SWB is high, thus improving the quality of their assessments. Alternatively, it could be that patients with adequate or high levels of SWB do not need thick smokescreens to preserve their sense of dignity.
There is some variation in precisely how investigators define QOL.8,42–,45 This is in part because of the multiplicity of factors that influence QOL, including mental and emotional state, physical or functional ability, frequency and severity of side effects, global life satisfaction, social status, sexuality, spiritual or religious well-being, and financial well-being. Our study confirms that SWB is significantly associated with QOL and that assessment of SWB is feasible and acceptable to outpatients with advanced cancer. Incorporation of some assessment of SWB into commonly measured domains of QOL is worth serious consideration. Further research related to SWB in cancer patients is needed, particularly research related to the effect of SWB on symptom expression and clinician-patient communication and to its effect on patterns of change in QOL assessed over time.
Acknowledgments
We thank the nurses and physicians of the Hoosier Oncology Group, the patients who agreed to participate in this study, and Jeanette Sawi for her project management.
Footnotes
-
Supported in part by the Mary Margaret Walther Program for Cancer Care Research, Indianapolis, IN.
-
Presented in part at the Thirty-Seventh Annual Meeting of the American Society of Clinical Oncology, San Francisco, CA, May 12–15, 2001.
- Received June 17, 2002.
- Accepted April 23, 2003.
