Quality of Life in Health and Illness

              Quality of Life in Health and Illness M. Powell Lawton, Ph.D. How does a person estimate the worth of his or her life? Or death? A more answerable question with which I shall be concerned today is, "What factors influence a person's judgment of QOL and how do such judgments enter into the person's wishes for life, life-extending treatments, and perhaps for death?" Such questions are relevant to understanding the natural conditions of the good and the bad life and to the way we deliver services and particularly the way we design the contexts in which people's lives end. My focus is, of course, older people, representing the full range from good health to frailty. I shall begin by presenting my view of QOL and contrast it with the very current concept "health-related QOL." I shall then suggest that all aspects of QOL potentially influence the older person's estimate of how long they would like to live through a construct that I've called "valuation of life." Such estimates are also relative to the person's present state of health as predicted by the cognitive processes implied by adaptation level theory. I shall use the results of my recent research on QOL and health to illustrate some of these processes. The research itself should be thought of as an early search for the operationalization of some of the constructs, and a series of problem-directed, not always successful, attempts to test some of the relationships. Quality of life Quite a while ago (Lawton, 1981) I suggested a partitioning of what I then called "the good life," this term having been succeeded by the presently more modish term QOL. A simple categorization of all in a person's life that might be evaluated on the positive to negative valence Invited address delivered at the distinguished research award session of the Section on Aging of the Division of Health Psychology, American Psychological Association, Chicago, August 15, 1997. Much of the substance of this address also comprised the Novartis Award lecture of the International Association of Gerontology, Adelaide, Australia, August 19, 1997. 1 c:\...\ms\qolhe&il.ms (8/4/97) continuum resulted in 2 relatively objectively measurable classes that I called behavioral competence and environment (standards for their evaluation being social-normative and physical respectively) and 2 subjective aspects of QOL, domain-specific perceived QOL and psychological well-being. Fig. 1 illustrates some typical measures of these 4 sectors of QOL. Each sector has an internal structure, illustrated here by behavioral competence, (Fig. 2) whose components are hierarchical in terms of complexity; this structure helps locate externally observable behaviors, any of which is capable of being measured in social-normative terms. Environment is more resistant to the imposition of an internal organization, but at least one can name the familiar aspects of environment-physical, personal social, suprapersonal, and social-institutional. Perceived QOL is as Campbell, Andrews and their colleagues thought of it 20 years ago, the subjective evaluation of a person's marriage, family, housing, work, leisure activity, and many other separate and limited domains of everyday life-like environment, PQOL consists of more a list than a coherent structure. Psychological well-being, representing positive and negative affect states and the cognitive aspects of well-being, is the subjective aspect of mental health. The two objective sectors of QOL, behavioral competence and environment, may be seen as antecedents of the subjective sectors. They are no less important aspects of QOL, but causally prior to the ultimate outcomes, which are intrapsychic. I might note in passing that the ubiquitous rankings of cities or countries terms of their quality of life are based on objectively measurable facets such as employment rate, crime rate, housing cost, presence of cultural amenities and so on. The assertion that such environmental aspects constitute a facet of QOL is based on a probabilistic prediction that for people in the aggregate a significant increment in positive affect or diminution of negative affect will result from such objective environmental features. Similarly the likelihood of a favorable subjective outcome is increased by the presence of greater objective personal 2 c:\...\ms\qolhe&il.ms (8/4/97) competence in the person. PQOL includes the cognitive appraisals of one's resources for shaping the external world and for responding to stress. There is thus a causal path from PQOL to psychological well-being, rather than these 2 sectors being co-equal aspects of QOL. The elements of this conception thus have a causal priority and additional elements as contrasted with the good life (Fig. 4). Note that direct paths from behavioral competence and environment to psychological well-being as well as indirect paths through PQOL are hypothesized. Health and OOL Now what about the place of health in this system? I would say that the hottest topic in medicine today may be health-related QOL. The objective presence of illness and sometimes its treatments may cause major changes in the other domains of behavior competence, sometimes on one's environmental context, and most certainly on PQOL and psychological well-being. Longer life can mean compromised quality if the additional years involve enduring distressing or painful symptoms or if the treatment itself produces such side effects. Useful metrics such as "quality-adjusted life years" (QALYS) have been developed to estimate the value of additional years of life discounted for the conditions that compromise their quality. The recognition of such tradeoffs has been the starting point for consideration of the costs of extended life, expressed in terms of dollar costs of treatment in return for years of life discounted for their reduced quality. Inevitably such accounting leads to ethical questions regarding priorities in treatment, including the consequence that lowered life quality may be used as a reason for withholding expensive forms of treatment. I find three faults with the concept and uses of IIRQOL: The limitation of this concern for QOL to the health domain, the inclusion of only negative intrusions into QOL, and the frequent use of QOL criteria based on the judgments of only healthy younger adults. Health-relatedness. It is 3 c:\...\ms\qolhe&il.ms (8/4/97) difficult to find in the literature a clearly stated rationale as to why the study of QOL and health requires such circumscription of the larger concept of QOL. The most I have seen is the simple statement that there are other domains of QOL which are not affected by health or are not themselves aspects of health and they are excluded by definition in the HRQOL discussion. A methodological anomaly also besets much HRQOL research. The "health-related" quality is operationalized by asking respondents to respond to the QOL items using the criterion preamble "Because of your health are you. . . (unable to shop, depressed, isolated, etc.)?" One of the most frequently used measures, the Sickness Impact Profile (SIP), frames all questions in this manner, while some others do not. It seems to me that we add error to our predictions when we invite the respondent to make such attributions of lowered QOL to health. An alternative and preferred route is to measure each domain of QOL in absolute form (i.e., without the health attribution) and let the health-QOL relationship be determined by their statistical correlation. Emphasis on the negative. The second problem is intimately associated with the first: Why does HRQOL research consider primarily negative deviations from a neutral level and rarely positive deviations above average? The answer seems easy at first glance: Because poor health only degrades QOL; Good health is the norm that we assume and therefore we rarely differentiate health in the average-tosuperb range and health therefore is little involved in a consciously perceived relationship to high levels of QOL. The tradeoff principle used in HRQOL research allows no input from the positive side. In HRQOL research it is a tradeoff between more chronological years and fewer quality years. The tradeoff that I see as equally important is the balance between losses as measured by the usual negative HQOL indicators and the gains that accrue from nonhealth domains of everyday life. My 4 c:\...\ms\qolhe&il.ms (8/4/97) assertion is that such gains may offset the distresses consequent to poor health, for some people, some of the time, and in different mixes depending on the person's hierarchy of personal needs and goals. If we measure only HRQOL, or worse yet, set medical-care policy only on the basis of HRQOL, we give no opportunity for the individual's own balancing act to emerge. An example from my own circle of friends is a man with terminal cancer of the esophagus who could no longer eat through the mouth. He spent his last month enduring almost daily radiation treatment, which left him extremely weak. He had a stomach tube and experienced a moderate to moderately severe amount of pain, sleeplessness, and other symptoms. He was at home most of this month, with the steady attention of his wife, many visiting friends, and the continuous presence of one or more of his out-of-town daughters. It was very clear that his strong personal attachments were a factor in his willingness to accept the gastric feeding tube, the strain of going to the hospital for treatment, and the side effects of treatment. Picture what the conclusion might be if this person's QOL were to be assessed with the Quality of Weil-Being Scale, one of the major measures of HRQOL. This consists of a series of 30 sources of decrement in QOL scaled from the responses of very large samples of healthy-adult participants whose responses to these items were calibrated and assigned weights by analysis against their ratings of scenarios for which they rated their willingness to give up years of life in return for a smaller number of better-quality years. Using the derived weights, the QOWB Scale items may be rated by an observer or expert to produce an estimate of an individual's probable responses to the vignettes. The validity of the QOWB Scale derives from the standardization data demonstrating that it does predict the individual's own QALYS score. My friend's QOL as rated by the observer on the QOWBS would have been very low. The QOWBS presents no possibility of factoring in the other class of conditions, the possible countervailing sources of pleasure, affirmation, fulfillment, 5 c:\...\ms\qolhe&il.ms (8/4/97) attachment, and so on, which are inadequately represented not only in the QOWBS but in most other HRQOL metrics. These first two major problems with HRQOL are thus related to one another: The reason why the health-related aspect is too limiting is because this restriction eliminates nonhealth-related sources of positive quality and therefore the possibility of trading some losses for other gains in assessing one's overall QOL. A third criticism of the HRQOL stream of research is its assumption that the values attached to good and poor health states are constant across all individuals. The weights established for the QOWBS, for example, used adults in good health as the standardization group. How would older people or those in poor health value life under compromised health states? Kaplan has suggested that good and poor health are judged by normative, rather than idiosyncratic, standards and therefore a judge's own state is not a major factor in such judgment. The evidence on this point is scant and mixed. In a study of older arthritis subjects performing the same health-state judgments as those of the younger standardization subjects, Balaban's ( ) findings supported Kaplan's hypothesis: The original weights were produced by his elders. On the other hand, Sackett & Torrance ( ) found people undergoing kidney dialysis to be more tolerant of extended life under reduced QOL conditions than healthy adults. Older normals were also more willing to accept a longer period of hospitalization in return for longer life than were younger people. O'Brien ( ) found that 60% of cognitively intact nursing home residents were likely to wish CPR to be applied in case of cardiac arrest and 33% would wish tube feeding even with permanent brain damage. Unfortunately no data are at hand to allow the direct comparison with the wishes of healthier elders or younger people. Therefore this question seems still worthwhile to explore further: Do people's standards for what constitutes an acceptable QOL show signs of an adaptation process whereby lesser levels become more tolerable as their own health decreases? The adaptation-level hypothesis would predict such 6 c:\...\ms\qolhe&il.ms (8/4/97) a process of downward adjustment. On the other hand, if as Kaplan suggests, people utilize normative more than personal standards when responding to such research inquiries, the adaptation effect might be minimized. Some Research Findings Relevant to QOL & HRQOL Issues I shall describe progress that research done at PGC has made in clarifying some of these issues. I'll discuss (1) our attempts to improve our ability to measure positive QOL; (2) the derivation of a measure of Valuation of Life to represent the person's intrapsychic attachment to life; (3) the results of our attempt to measure directly years of desired life under a variety of hypothetical health-compromised conditions; and (4) tests of a model of QOL which takes account of older people's current health states. Indicators of positive OOL: Objective and subjective Domains of potential well-being were chosen from among those identified as being relatively close to the core of overall well-being by Campbell et al. (1976): Family, friends, and activities performed in one's spare time. Although these domains were measured by Campbell et al., the items did not seem suited to older people and some were not phrased in such a way as to elicit a generalized evaluation of each of those domains. Putatively objective indicators of actual behavior in these 3 domains were derived from the MAI: The numbers and interaction frequencies with family members and friends and their frequency of participation in a list of 20 activities (the Activity Scale of the MAI). These reports are termed "objective" in the sense that externally observable and accountable events, rather than qualitatively evaluated experiences, are being reported. In the absence of existing measures for the subjective aspects of QOL in these domains we developed a parallel act of questions asking the person to evaluate the frequency and quality of their lives in each 7 c:\...\ms\qolhe&il.ms (8/4/97) of these 3 domains. 7 items were used to elicit estimates of overall satisfaction in each of these 3 domains, which were called Friends Quality, Family Quality, and Time Quality. Coherent factors emerged which, although not always meeting confirmatory fit criteria, were close enough to be serviceable as multi-item indicators. Valuation of life Our concept of VOL accepts that under many negative mental and physical health conditions, people may feel a diminished attachment to life. At the same time we sought a construct that could also allow the person to account for positive features that might counteract the negative. Valuation of life was the term used that would additionally express elusive concepts such as hope, purpose, sense of future, personal goals, and sense of self. Valuation of life is defined as the subjectively experienced worth of the person's present life, weighted by the multitude of positive and negative features whose locus may be either within the person or in the environment. The distinction between positive and negaitive inputs to VOL expresses the frequent clinical observation that to some people, life appears very dear in spite of poor mental or physical health, while to others it appears of lowvalue. Concepts like purpose in life (Crumbaugh, 1972), meaning in life (Werner & Williams, 1987), optimism (Scheier & Carver, 1985) and hope (Gottschuch, 1985; Snyder, ) have been developed into scales, but no existing instrument folly represented the important facets noted earlier without also including items that expressed obviously positive or negative mental health. Individual items were incoiporated from a variety of sources, including 8 of the 10 items of Snyder's Hope Scale ( ); the majority were written for this study and subjected to pilot testing to remove ambiguous items or those with extremely skewed distributions. A 4-point scale (agree very strongly, agree, neither agree nor disagree, disagree) was applied to each of 22 items. The items are seen to 8 c:\...\ms\qolhe&il.ms (8/4/97) represent positive personal goals (items 1, 6, 12, 15, 22), meaning of life (2, 5, 7, 9, 10), hope (3, 4, 13), wish to live (8, 16), sense of a future (11, 14), and self-motivation (17, 18, 19, 20, & 21). To say that the items' content was purged of presumptions regarding mental health is a matter of degree. Obvious terms like happy, satisfied, discouraged, or depressed are absent. Yet most items can obviously be characterized as representing desirable or undesirable qualities. The extent to which this important outcome is measured independently of its hypothesized antecedents is relative, certainly not absolute. Two factors emerged, with confirmatory indices that fall just short of the .90 level, neatly sorting the items into those phrased positively and those phrased negatively~not what one would have hoped for, but clearly a result duplicating similar analyses of many other scales. There is an interesting story to be told by determining the different correlates of the positive VOL and negative VOL items. I don't have time for this intriguing side trip today, so my discussion today will be confined to the item set with less response error, positive VOL. RELATIONSHIPS AMONG HEALTH, QOL, AND VOL The early state of development of a model relating these constructs to one another is reflected in the piecemeal approach to data analysis that is possible at this point. The rationale I've presented for the VOL construct makes some good armchair sense. Its usefulness may be examined empirically by detailing and testing some hypotheses regarding the way VOL behaves in relation to better-understood constructs. If VOL behaved properly, 1. VOL is related to health but influenced at least equally by nonhealth-related aspects of QOL. 2. Both objective and subjective indicators of QOL are related to VOL independently of the indicators of mental health, positive affect and depression. 9 c:\..Ams\qolhe&il.ms (8/4/97) 3. The positive aspects of mental health (PA) are the intervening construct that mediates the sources of QOL, affect, and VOL. The data I'll discuss today are from a study of community-resident people 70+ recruited in a way that attempted to represent both the very healthy and the frail older person. Subgroups were targeted to include equal numbers of men and women, African American and white, and higher and lower-income people, as well as groups with diagnoses of arthritis, cardiovascular problems, bowel cancer, and diabetes. I'm reporting on data that are still being generated, so the differing Ns represent different data batches rather than systematically differing subgroups. 1. VOL and health. All measures of health agreed at a zero-order level in showing an association between positive VOL and good health. The health conditions measure was used in later analyses because it overlapped minimally with the measures of subjective well-being. In regression Model 1, all background characteristics and the measures of objective and subjective QOL were controlled. In Model 2, Positive Affect and CESD were added to the equation. In neither model was there an independent relationship between health conditions and VOL, although small indirect effects through QOL and MH may be discerned. 2. OOL influences VOL independently of mental health. These results are best displayed in a path diagram which for clarity omits those associated with background factors and health and shows only the significant paths. For the most part this prediction was not borne out. Although, as other literature would predict, domain-specific objective QOL is related to its subjective analogue, no objective QOL domain and only one subjective domain contributed directly to VOL, the Time Quality index. 3. Positive affect as the central construct that relates OOL to VOL. The same path diagram displays the convergence of the domain-specific indicators on PA 10 c:\...\ms\qolhe&il.ms (8/4/97) and the substantial relationship between PA and positive VOL. This is contrasted with the relatively isolated position of Depression in relation to both QOL and VOL. (One may note also the one major "crossover" relationship between Time Quality and Depression; I suggest that our Time Quality measure hit too close to the core of depression and that a more domain-specific index focused more on leisure and other discretionary behaviors, rather than time in the abstract, is needed). To summarize: The VOL positive measure met a number of criteria posited to be indicators of its usefulness in a model of QOL. The bottom line is that VOL adds different content from that subsumed under either PA or depression; VOL shows some independent contribution from aspects of QOL beyond positive affect; and VOL displays partial independence from the direct influence of physical and mental health. (Parenthetically I might add that the negative indicator of VOL showed the identical covariance patterns but unlike positive VOL it was also substantially related to all 4 background factors). THE CRITICAL OUTCOME VARIABLE: A STILL MOSTLY-MISSING LINK The only true outcome variables in a model specifying attachment to life are the behaviors that come at the end of life. First, those who value life the most should live longer if there is truth in the theory. An outcome more easily measured and likely to be predictable is the set of decisions and behaviors that have to do with prolonging life: On a univariate level, those who value their lives more might be more willing to tolerate more intrusions on their QOL. In our present data, although we are following participants until their death and documenting through family informants, the measures taken or not taken to prolong life, few deaths have occurred to date. Thus the only genuinely useable outcome variables are a set of questions that requested people's current assessment of their tolerance of a set of hypothetical QOL-limiting conditions. These questions asked about 11 c:\..Ams\qolhe&il.ms (8/4/97) the extent to which they might wish to curtail their time of remaining life under conditions of distress and disability. The QOWBS, the time-tradeoff, and a variety of more direct methods for asking people to make such estimates have appeared in the literature. Some are very complex (i.e., timetradeoff), while others have been done by older people but require a longer and more searching inquiry than we felt could be undertaken while still gathering all the other necessary information. We thus used a slight adaptation of Kaplan's QOLQ, which inquires about 3 types of QOL degradation: Functional disability, cognitive impairment, and pain. (I'll call this "Years of Desired Life" in order to reduce the number of references to different varieties of QOL). A rough order of impairments was created within each domain by using as the best QOL indicator the years desired if "You are conscious and able to recognize others and make decisions for yourself and are able to carry on with no limitations on activities" (no limit). For analytic purposes, answers to this question were used as the baseline point on 3 ordered scales, as indicated on each slide. I call your attention first to the fact that these were difficult questions for our respondents. In fact there were 3 types of nonresponse: First, those who could or did not answer (called "missing"), second, those whose response was "only God can decide" or a related response, and those whose response could be categorized as "The maximum possible" or similar response. Numerical responses in days or years were categorized, and "maximum possible" used as the longest estimate, with "missing" and "God's will" excluded from the analyses. Even the missing responses tell us something about the respondent, however. First, those who provided no answer were clearly very distressed, both physically and mentally. Across all of the hypothetical conditions of compromised HRQOL, nonrespondents reported more physical symptoms, poorer self-rated health, lower functional health (IADL), less involvement in activities and family contact, more depression, lower mastery, less positive affect, and lower positive VOL. 12 c:\...\ms\qolhe&i!.ms (8/4/97) Nonresponse within this clearly more frail group could be a matter of cognitive uncertainty about the possible effect of their physical condition on remaining lifespan or a self-protecting disinclination to think in such focused fashion about an increasingly negative future state. The "God's will" correlates were somewhat different. Slightly poorer health was also a correlate, but African American respondents were much more likely to respond in this manner. (PL: more on # God's will," missing here?). It thus would seem that cultural factors relating to the central position played by religion among African Americans accounts for this type of response. Recognizing the limitations in using measures that elicit so much missing data, two types of questions may be asked regarding the relationships of Years of Desired Life to the other elements of the model. The first is that a higher present QOL should lead people to wish to live longer under both present and presumptive compromised-health conditions. VOL as the penultimate explanatory concept should in particular show such a relationship. Second, even though healthy people may wish to live longer, those who are currently in poorer health may be increasingly more tolerant of lowered QOL as the severity of future health-comprising conditions increases. OOL. VOL. and Years of Desired Life One way to display these results is to show the zero-order correlation matrix between the indicators of QOL and the different conditions under which years of desired life were estimated. For some of these conditions, if there were a Bonferroni-like test for the significance of an excess number of nonsignificant relationships, these results would surely qualify. Don't really look at the correlations. It would seem that meaningful relationships were exhibited only with respect to the least-distressing situations: As one is now, no limitations, and mild pain without physical or cognitive disability. In the case where there is least need to assume a future health perspective--the "now" condition-better QOL clearly leads people to wish to live longer; the indicators of social 13 c:\...\ms\qolhe&il.ms (8/4/97) relationships, contrary to prediction, are the only ones that do not show such relationships. It is notable that present state of health rarely determined one's projected wishes for the future. To some extent this validates Bob Kaplan's assertion that people respond to such questions in a way that is determined by normative conceptions of disability that are relatively free of influence from personal factors. The only situation where looking across the row is relevant is for Positive VOL, the next-tolast row, where a higher score was associated with wishing to live longer under most of the healthcompromised conditions other than those of cognitive impairment. In multivariate analysis where the prior variables in the model were controlled, virtually no QOL variables other than Activity were independently associated with Years of Desired Life. Positive VOL remained as an independent predictor for the Now, No Limits, and the two ADL-dependent situations. Differential tolerance for reduced OOL among the less-healthv. Adaptation-level theory deals quite well with the effects of a new anchoring stimulus on the range of stimulus values characterized as "much" or "little." In the cognitive-decision-making realm, Kahneman & Tversky's (1979) prospect theory would suggest that the perceived gain of a small quantum of quality-years or months would be greater for those whose present quality is closer to that of the hypothetical quality than would be true for those more distant (i.e., healthier), but that the proportional greater tolerance would be magnified as the magnitude of the risk increased. Such processes have been illustrated in gerontological research focused on people whose life space has become physically constricted in terms of being housebound or institutionalized. Rubinstein has referred to this phenomenon as "miniaturization." I suggest that concurrently with miniaturization comes increasing differentiation among characteristics of the miniature environment. On a cognitive level, the scales denoting the differences among degrees of disability or among the pleasures associated with small uplifting events become stretched at the low-quality end. Previously subthreshold differentiations become part of one's life. My hypothesis is that such changes mean 14 c:\...\ms\qolhe&il.ms (8/4/97) that as health diminishes, one's world and one's personality are no longer the same as they were. Enough for theory; on to empirices. The hypothesis that people's expectations shift in accordance with the reality of their present state was tested by statistical interactions. I regret to say that none of the predicted interaction effects occurred when we used any of our health measures to separate the healthy and the less-healthy and examined their differences in Years of Desired Life as a function of increasing severity of impairment in the domains of ADL, cognition, and pain. I'll put a footnote here: My colleague Laraine Winter did find the predicted interactions in all 3 domains among elders in a different sample, using self-rated health as the health classifying variable. I don't wish to claim this as a hypothesis confirmation, only as a caution that there is a lot more to be investigated. As a summary of our findings regarding the only set of outcome variables available for study at this time, the Years of Desired Life, they provided some validation of VOL as a central cognitive mediating variable between other aspects of QOL and more specific cognitions regarding end-of-life behavior. The more specific interaction hypothesis that greater frailty would lead to progressively greater tolerance for reduced QOL as limitations increased received little support, although our other study's data produced confirmatory findings. I conclude especially that measurement is a first-level problem for which a great deal of additional effort is appropriate. Over the longer range, only longitudinal and prospective data that examine people's expressed wishes from their times of good health through poor health, and includes the actual behaviors and expressions of wishes during the last period before death, can fully test the effects of this set of hypothesized antecedents. CONCLUSION Both social policy and individual decisions about the conditions under which life prolongation is desirable and worth paying for will be benefited by continuing research on the factors 15 c:\...\ms\qolhe&il.ms (8/4/97) that erode and enhance QOL. How people perform the cognitive calculus that leads to their cognitions and behaviors concerning the end of life continues to be a mystery. Such questions have only recently begun to be asked. It is therefore unreasonable to think that the primitive methods used to estimate concepts like VOL or Years of Desired Life are able to provide closure on the important questions being asked. My concluding plea is that such limitations, particularly those involved in accepting at face value people's expressed wishes regarding a state through which they have not yet passed, be always in the forefront as we ponder how to fulfill individual wishes or form policies for society regarding the worth of life and its prolongation. 16 c:\...\ms\qolhe&il.ms (8/4/97)