What are we trying to do?
“A peculiar feature of contemporary medicine is the absence of any systematic enquiry into the goals of medicine” (1).
Readers of Arthritis Care & Research are predominantly health care professionals. So am I. We have been trained in the science and art of medicine: that is, of caring for others with health problems (the derivation of the word medicine is the Latin word mederi meaning to look after or to serve). But we rarely examine critically what we are trying to do in our everyday clinical work. We tend to assume that the goals of our work are clear and unambiguous and a part of the ethos and culture of our various professions (1). We are there to help individuals with disease or illness get better.
If our job is to help individuals get better, we need to know what interventions are most likely to do that, and it follows that we have to be able to assess getting better. That used to be done by asking individuals directly, and then examining them. Not any more. Now we have to measure things. There are 3 basic types of measures that have been used to assess getting better: measures of some aspect of bodily structure or function (such as a radiograph in rheumatology); assessments of pain, function, disability, or quality of life; and social measures, such as remaining in work. The International Classification of Function (ICF), which splits these outcomes into impairments, activity limitations, and restricted participation, offers a theoretical framework on which to base such measures (2). In rheumatology, we generally try to assess pain and function, and sometimes joint structure, without generally using any clear theoretical framework such as the ICF (2, 3).
There are 2 main methods of assessing pain and function: observer-dependent and observer-independent measures. Over recent decades there has been a move towards greater use of observer-independent measures, based largely on the belief that these measures provide a fairer reflection of the patient's problems and status than measures that depend on input from an external observer. This move led to the development of a new industry of outcome measure development and validation, resulting in the routine use of disease-specific or general self-assessment instruments to assess patient responses to interventions. Systematic reviews and meta-analyses of these data now appear, telling us about effect sizes, or the number needed to treat based on the amount of change that has occurred in groups of patients treated with any given intervention. But what do such numerical data mean? How do we interpret the finding, for example, that the “whatever” standard instrument (WSI) has improved from a baseline score of 43.9 to a final score of 37.6?
We cannot. A change of 4 or 5 points on the WSI has no meaning to us or to our patients. This was recognized many years ago, leading to new developments in the field. The introduction of the minimum clinically important difference (MCID) moved us along, a development well summarized in the published proceedings of Outcome Measures in Rheumatology Clinical Trials V (4). Following the MCID was the minimum clinically important improvement (MCII), which has been defined as “the smallest change in measurement that signifies an important improvement in a patient's symptom” (5). These approaches attempt to enumerate the amount of change in health status that is regarded by patients as meaningful. In essence, what is done is to anchor numerical changes in the WSI to responses from patients who are asked whether they are better or not. More recently, we have been offered the patient acceptable symptoms state (PASS), defined quite simply as the patient feeling good (or not) (6).
In this issue of Arthritis Care & Research, another important contribution to this field is published. Led by the French group who helped introduce the PASS to rheumatology, Tubach et al, a multinational team of authors, tried to determine which of the MCII or PASS is the more appropriate patient-centered measure to use by analyzing data from 2 large studies: one of patients with osteoarthritis and the other of individuals with acute rotator cuff problems in the shoulder (7). They conclude in favor of the PASS, providing the answer in the title of their article that it is more important to feel good than to feel better. So, if our mission is to help patients get better, we need to measure their PASS to find out if they feel good. And that is simple enough: Tubach et al just asked their patients, “Taking into account all activities you had during your daily life, your level of pain, and also your functional impairment, do you consider that your current state is satisfactory?” To which my answer might be, “In comparison with what?” This seems awfully similar to the old-fashioned approach of asking someone how they are (8), or our previous reliance on a simple patient global outcome question (or what is sometimes known as the Short Form 1 [SF-1]).
But is that really what we should be doing? There are some specific problems inherent in trying to generalize findings from the French studies. These problems have been well summarized in an editorial (9) accompanying the publication of the MCII and PASS data from the French osteoarthritis study (5, 6). But I have other interrelated but more philosophical concerns with this approach. They are about the nature of the data and their derivation, and about the comparator issue, or response shift.
On the nature of the evidence
We live in an era in which evidence is everything. Therefore the findings on the MCII and PASS are evidence based. But what is the nature of this evidence? The findings were based on trials of interventions: 1 cohort design and 1 randomized controlled trial (RCT). The changes in health status were recorded numerically from standard disease-specific instruments; these were then compared with simple questions about getting better or feeling satisfied with health status.
Let us think about the outcome instruments first. They assess pain and function, but, as noted above, have no clear theoretical basis and tend to muddle impairments, activity restrictions, and reduced participation (3). Furthermore, outcome instruments take no account of other aspects of mental, spiritual, or social well-being (1). The PASS question also constrains the participants to thinking about pain and function; for example, it does not ask them to consider their spiritual well-being. The problem is probably made worse by the fact that patients tend to respond to doctors in ways that they think will be most appropriate. So, if the professionals are obsessed with pain and function, then the patients will respond accordingly, and will think that they cannot complain about other things that are wrong with their lives. How would we know if the osteoarthritis or shoulder problems in these patients were responsible for social isolation or depression that was ruining their lives in the absence of much pain or a big functional problem? So the problem with the outcome measures is that they are tightly controlled within the reductionist medical model of disease and take no account of the holistic approach or of the complexity surrounding individuals.
Now let us turn to the study designs, the trials from which the data were derived. The evidence-based health care (EBH) movement has an obsession with trials in general and with the RCT in particular. It is heresy to criticize this. But here goes. Trials are carried out with unrepresentative groups of patients in artificial contexts, in which the all-important placebo effect (arguably the most important medical weapon of the caring professional) of any intervention is deliberately factored out of the picture. EBH makes an inherent assumption that we can provide discrete therapies for sick patients with specific conditions, and that we can measure the effects of these therapies with scientific outcome measures (10). These assumptions belie the fact that many patients have multiple, complex health problems that do not fit any discrete category. More importantly (in this context), these assumptions take no account of the heterogeneity of the responses. Whatever treatment we try, some patients get better, some are not affected, and some get worse, with wide variations in the degree of getting better. However, what we do with the data is to derive between-person (group) differences from within-person (individual) changes, ignoring the variations in response or individual outcome preferences. It has been said by others that it is absurd to suggest that such data can ever be generalized (11), although the proponents of EBH contest such heresy (12).
On response shift
My main argument against the use of something like the PASS (or indeed the patient global or SF-1) comes from my experience as a clinical rheumatologist. Many of the patients I have seen over the years have slowly deteriorated in front of my eyes while insisting that all is well and that they are just fine. In addition, we are all aware of the patient who comes to the clinic in what we consider to be a dreadful state of disability, and when asked how he or she is says something along the lines of “Quite good thanks, considering all things.” Considering what? That is why I like to try to use semiobjective measures in my clinics. I like to be able to measure something of some functional meaning, record it, and show it to a patient on the next visit. Then, when they say “all is well” I can say something along the lines of “But look, last time you could put your own shoes and socks on, but this time you cannot.” (Of course, I have no such option with their perception of pain.) My experience of patients' response to the challenge that their condition is actually deteriorating when they are insisting that they are fine is varied. Some patients seem to be in denial, others say that they are adapting better to the condition, a few agree and then admit that things are far from fine, saying that they did not want to admit that. Older patients, in particular, are keen to put a “brave face” on things (13).
Adaptation to changing disease or circumstances and a consequent change in the response to questions about health status is known as response shift (14). There is a significant body of literature on response shift, and it has been suggested that response shift is a significant barrier to the use of observer-independent measures of outcomes and of quality of life (8). However, a cursory search suggests that rheumatologists have paid very little attention to this concept. Given the chronic and insidious nature of many of the diseases we deal with, I think it deserves much more attention.
The need for a more individualized approach to health care
Is any of this of any relevance to the current work on outcome measurement and the introduction of things such as the MCII and PASS? Perhaps not much. The data used in the study by Tubach et al (7) were derived over relatively short periods of time; therefore, response shift is unlikely to have been much of an issue, and it could be argued that if response shift was a problem it would have had similar effects on the standard numerical outcome measures as well as the questions used to anchor those data. Similarly, the trial technologies and outcome measures used are standard and validated.
However, I believe we must be careful about generalization and uncritically adopting something such as the PASS. What I am really concerned about is the inherent conflict between EBH and holism and the need to individualize health care. In many countries this conflict is being exacerbated by politicians. In the UK, for example, we are being exhorted to comply with evidence-based health care, and at the same time to respect patient choice, provide patients with the opportunity to access complementary and alternative therapies, and make sure that the use of any interventions are based on shared decisions between patient and professional. The conflict is a deep and real one. EBH and its technologies are good at telling us what is best for groups of people. So, we learn that if we use low-dose aspirin after myocardial infarction, lives are saved. But, because of its roots in reductionism and inability to deal with heterogeneity of outcomes, EBH is not good at telling us what is right for the individual. Furthermore, EBH works best for acute conditions of younger persons, where there is less confusion from comorbidities, age-related changes in aspirations and needs, or changing social circumstances. The paradigm is less useful for the care of older persons with chronic diseases. But that, of course, is largely what those of us involved in musculoskeletal medicine spend most of our time doing.
So what are we trying to do? Are we trying to serve the whole community or just the individuals who get to see us? Are we working in a reductionist paradigm alone or should we be embracing the holistic complexity appreciation of health and health care (15)? Should we be trying to help patients with musculoskeletal disorders get better, or should we spend more time trying to prevent their problems? If it is the former (helping patients get better), how are we to decide if someone is better or not? My view is that slavish adherence to the tenets of EBH, and to measurement, must be avoided. We need to remember that if we are working with individuals, we need to connect with their hopes and fears. The new science of medicine could push the art of medicine out of the picture. We need to fight against that. We need to have more open debate about what we are trying to do, and about the philosophical basis of our attempts to measure outcomes such as health and quality of life.