Chapter 7. Outcome Measurements in Pain Medicine

Central to the evaluation of health care is the measurement of health. Until the first part of the twentieth century, health was defined as the absence of disease and was measured in terms of morbidity and mortality. This simple approach to health status was rejected in 1948 with the expansion of the concept of health by the World Health Organization (WHO), which defined health as, “A state of complete physical, mental and social well being and not merely the absence of disease or infirmity.”1 This definition reflected the multidimensionality of health and considered not only biologic markers, but also the ability to perform physically, psychologically, and socially in the everyday environment.

This change in the definition of health gave rise to the current outcomes movement. Other factors, such as the reversal of the proportion of care rendered for acute illnesses versus chronic diseases, technological advancements in health care, rising health care costs, the emerging concept of quality of care, and increased recognition of the importance of patients’ views about their care and health, have further fostered the growth of this movement. Outcomes research seeks to define which treatments are effective, at which point in the natural history of medical conditions these interventions should be provided, for which patients, and at what cost, thus providing objective evidence to guide treatment.

Conscientious health care providers use both individual clinical expertise and the best available objective, external evidence for treatment. The best available external evidence for treatment is defined as clinically relevant research, often from the basic sciences of medicine, but especially from patient-based clinical research into the accuracy and precision of diagnostic tests (including the clinical examination), the power of prognostic markers, and the efficacy and safety of therapeutic, rehabilitative, and preventive regimens.2 The integration of clinical expertise and best available external evidence for treatment is the practice of evidence-based medicine. With the plethora of current and relevant literature, it is impossible for most clinical care providers to keep abreast of the latest developments in their field. Because of this problem, structured approaches to literature synthesis, such as that organized by the Cochrane Collaboration, have arisen to summarize, with the least possible bias, the best available research on a specific topic (see later discussion). The goal is to make relevant information widely available, and evidence-based practice unencumbered for all health care providers.

This chapter focuses on common terminology used in outcomes assessment, and on outcomes measurement tools used in research on and treatment of patients with chronic noncancer pain, as they are part of the foundation of what will become evidence-based practice and perhaps eventually guidelines for treatment. Although the comprehensiveness and validity of outcome measures for the treatment of all types of pain lags behind those of equally high-impact conditions that affect the public’s health, this lag is even more pronounced for cancer pain than for noncancer pain.3 Much more research has addressed functional assessment, and how pain management influences function, in patients with acute or chronic noncancer pain than in those whose pain results from malignancy.4

We, as health care providers, treat patients to make them “better.” How is “better” defined, and by whom? “Better” from the point of view of the practitioner, the patient, or society? Does “better” equate to less pain, increased physical functioning, decreased disability (as judged by the physical therapist), improved quality of life (as judged by the patient), or decreased cost of worker’s compensation charges and fewer health care visits (as judged by payers)? Does the same intervention that benefits one patient benefit a group of patients with similar conditions? How do we know whether it does? These are questions that the outcomes assessment movement is trying to address.

Outcomes research studies the results of medical care.5 It involves “the rigorous determination of what works in medical care and what does not” and states that “outcomes research, by informing the content of policy positions, payment rules, and practice guidelines, presumably solves both the problems of quality and cost that beset health care and does so by scientific rather than political means” (p. 1268).6 Outcomes research is the foundation of evaluation of the quality and costs of health care delivery. Adoption of an evidence-based approach to health care, exemplified by the Cochrane Collaboration,2,7,8 has been accompanied by a shift towards emphasis on patient-centered health outcomes.9 This broadened perspective has heightened the need for tools to monitor and adjust treatment and to approach clinical decision making from a viewpoint that is evidence based and patient centered.10 The pressing need to know which treatments reduce chronic pain, which improve functional status (including return to work and social activities), whether or not they change pain intensity, and, in particular, which treatments are worth paying for, has fueled the development of a number of instruments. These instruments are intended to capture in a simple, speedy, and robust fashion the health status of patients.11

This section discusses common terminology used in outcomes assessment and provides several examples of assessment tools used in measuring outcomes during the treatment of patients with chronic noncancer pain. Health assessments focus on three broad categories of measures: traditional biologic, general (or generic), and disease specific.12 Traditional biologic measures may be primary, such as morbidity and mortality, or surrogate, such as a decrease in blood pressure in patients given an antihypertensive drug. Measures used for patient-centered outcomes generally estimate persons’ health-related quality of life (HRQOL) and their ability to function and to do the things they want to do. These measures may be generic, evaluating overall health status, or disease-specific, focusing on the effect of a given condition on a person’s life.

HRQOL assessment is the measurement or evaluation of the health of an individual or a patient. HRQOL may include biologic markers, but it emphasizes indicators of physical functioning, mental health, social functioning, and other health-related concepts, such as pain, fatigue, and perceived well-being.13 Concepts included in some commonly applied HRQOL instruments are presented in Table 7-1.

Quality of life includes HRQOL, but is a broader term that includes nonmedical aspects of life that reflect the aggregate impact of food, shelter, safety, living standards, and social and physical environmental factors.13

Patient based outcome measures are indicators of patients’ evaluations of both changes in patient health status, including HRQOL and mortality, and the quality of health care. The importance of patients’ views has been increasingly recognized in health care.14 One might even argue that the increased interest in palliative care and pain control in recent years is the direct result of a power shift in which patients and their families—the consumers of health care—have much greater autonomy and power than under the previous disease-centered model of care.15 Clinicians’ taking the patient’s view into account is associated with greater patient satisfaction with care,16 better compliance with treatment programs,17 and an increased likelihood of maintaining a continuous relationship during health care.18

The distinction between disease-based clinical investigation and patient-centered outcomes research is analogous to that between measures of efficacy and measures of effectiveness. In an ideal setting, such as a randomized, controlled clinical trial, the efficacy of a treatment may be derived as the dose-response relationship for a given physiologic effect assessed under well-controlled conditions. In controlled trials, the end points of interest are usually biologic measures, such as changes in blood glucose levels or blood pressure. However, equally important to practitioners and patients is the effectiveness of a treatment, which refers to the outcomes of this treatment when applied in typical practice settings, measured over the course of disease, and including measures that matter most to patients (patient-centered outcomes).19 Outcomes research is more likely to be generalizable to typical medical practice than are controlled clinical trials. Terminology commonly employed in outcomes research is presented in Table 7-2.

Because the purpose of this overview is to present a few widely applied outcomes measurement tools and the context in which they are used, we next describe the criteria used to select one from among available instruments rather than how to create a new questionnaire.

Selection of a specific outcomes tool will depend on the population of interest and the ability of the measurement tool to detect changes within the domain of interest. The selection of an instrument consists of two phases. The first has to do with the condition(s) for which this instrument will be used; the second has to do with the psychometric properties of the instrument.

Choosing a domain-specific, condition-specific, or generic instrument depends on the aim of the study. If one specific domain is of interest, such as pain intensity or depression, a domain-specific instrument can be used (e.g., the McGill Pain Questionnaire or the Beck Depression Inventory). In general, a condition- (or disease-) specific instrument will have a narrow focus but will provide considerable detail in the area of interest. If the interest is in general HRQOL, comparison with different conditions, or with healthy people, a generic instrument can be used. Generic and condition-specific HRQOL instruments can be used together to supplement the information collected.20 Using a condition-specific survey or module together with a generic scale may provide more insight into aspects of health that are not well measured by either type of instrument.21–23 Comparison of the impact of pain on health status with the impact of other chronic illnesses on general health status, for example, allows researchers to conduct trials of various treatments so as to make clinical decisions in medical practice and inform health care policy.12

Important psychometric properties to consider include the following24–26:

• Test-retest reliability: the extent to which the measure generates consistent results. How closely do the results of repeated applications agree with each other?
• Internal reliability: (quantitated by Cronbach α*) the sensitivity of the number of items that make up the measure and the degree of intercorrelation between the items. A Cronbach α of 0.9 or higher is generally preferred for measurement in a single person, whereas Cronbach α of 0.7 or higher is preferred for group measurement.27
• Validity: the extent to which the instrument actually measures what it claims (i.e., the correspondence between what the instrument reports and reality).

• Responsiveness: the ability of an instrument to detect changes, particularly clinically important changes, over time in individuals or in groups of subjects.
• Applicability: the appropriateness of the instrument’s use in the specific study population.
• Practicality: the likelihood that an instrument can be applied readily, without excessive burden to patient or investigator, and produce data that can be easily analyzed and applied.

*Cronbach alpha = a coefficient of reliability (or consistency) used to measure how well a set of items (or variables) measures a single unidimensional latent construct. It ranges from 0 to 1. (For details see: http://www.ats.ucla.edu/stat/Spss/faq/alpha.html.)

Of the many generic instruments available to assess HRQOL, four validated, widely used questionnaires stand out. Brief descriptions of these instruments are given in Table 7-3.

Pain, in general, and chronic and persistent pain, specifically, is a unique challenge to outcomes research because of the importance of subjective information. Unlike the majority of other medical conditions, chronic pain may not involve a distinct organ system, pathophysiologic process, or specific discipline. Although pain is characterized as a symptom, it is, in fact, a subjective experience, a perception.41 This perception not only depends on nociceptive transmission and modulation within the CNS, but is integrated with psychological, social, and other environmental factors.42 Physical functioning, work, family, and social relationships are usually impaired by chronic pain. Comorbid conditions, such as depression or anxiety, often accompany chronic pain.43 For these reasons, it is argued that the assessment of patients with chronic pain should be accomplished within a multidimensional framework.44 Assessment of chronic pain should provide clinicians with relevant information to formulate a treatment plan, and also allow for measurement of the outcome of treatment interventions. The generic HRQOL instruments discussed earlier are mostly epidemiologic tools and as such are able to measure change in large samples of patients. By design they are not intended, nor are they sufficiently sensitive, to measure changes in a single subject. Furthermore, these instruments do not provide information on items frequently assessed in pain management, such as solicitous responses,45,46 coping ability,47,48 fear avoidance,49–51 and the extent of disablement from pain.

Many instruments are used to assess the impact of pain on patients’ lives. Ideally the instrument should provide relevant information to all clinicians within an inter- or multidisciplinary team, have a low respondent burden, and be sensitive enough to detect changes at both group and individual levels. Widely used methods to assess pain and its influence range from domain to condition specific. Some of the most frequently used tools are presented next.

Domain-Specific Measurements

Pain Intensity (or Pain Relief)

The three most commonly used methods to assess pain intensity are the verbal rating scales, visual analog scales, and numerical rating scales (Table 7‐4). Von Korff52 cautions that multiple factors influence patients’ pain reports, including time of day. Aggregated pain measures have, therefore, been shown to be more reliable and more sensitive to treatment effects than single items.53 Aggregated pain measures are scores that are created from multiple measures. For instance, the average of three concurrent responses to a 100‐mm visual analog scale of pain intensity ratings of current, average, and best pain can be taken.54 A composite measure shown in cancer pain patients to have high internal consistency (Cronbach α > 0.8) consists of an average of ratings on a 0‐to‐10 scale of current, least, and average pain.55 Jensen and Colleagues56 report that individual 0‐to‐10 pain intensity ratings have sufficient psychometric strength to be used in chronic pain research, especially in studies with large sample sizes, but composites of 0‐to-10 ratings may be more useful when maximal reliability is necessary (i.e., in studies with small sample sizes or in the monitoring of an individual patient).

Verbal Rating Scales (VRSs)

These scales are positively and significantly related to other measures of pain intensity.26 Jensen and colleagues57 reported on the potential clinical utility of classifying pain as mild, moderate, or severe based on the impact of pain on quality of life. There is a nonlinear relationship between pain intensity and pain interference. Pain intensity begins to have a serious impact on functioning when it reaches a specific threshold; about 5 on a 0-to-10 scale in patients with cancer pain.55 To explore in greater detail the relationship between pain severity and interference in patients with cancer pain, Serlin and colleagues55administered the Brief Pain Inventory to a total of 1897 patients from numerous sites in the United States, France, China, and the Phillippines. In this classic study, they gathered self-reported data on pain severity as well as interference by pain with enjoyment of life, activity, walking, mood, sleep, work and relations with others. These four diverse populations had “fairly consistent patterns relating pain severity to pain interference.” Statistical analyses showed that pain severity on a 0–10 verbal numerical rating scale could be stratified according to the degree of interference it produced as mild (1–4), moderate (5–6), and severe (7–10).

Visual Analog Scales (VASs)

These are simple tools to assess intensity and other dimensions of pain, such as anxiety, efficacy of treatment, and emotional responses26,58,59 (Fig. 7-1). Patients mark the scale at a point that represents the severity of their pain at a specified time point, or within a well-defined interval (e.g., the past 24 hours). Variations of these techniques request that patients circle a number from 0 to 10, or place a mark through one of these numbers. VASs are more sensitive and precise than descriptive scales. They are also easy to use and interpret; however, they are limited to expressing only one dimension of the complex experience of pain. It may be difficult for patients to imagine the worst pain imaginable, or they might report their pain as being outside the 0-to-10 limits, saying that it is a 20, for example.

The validity of VASs is supported by their positive relations to other measures of pain intensity.24,60 They are sensitive to treatment effect and are distinct from measures of other subjective components of pain.52

Numerical Rating Scales (NRSs)

These scales were demonstrated to provide sufficient levels of discrimination for patients with chronic pain to describe their pain intensity.61 Like VRSs and VASs, NRSs demonstrate positive and significant correlations with other measures of pain intensity.26,60

Pain Affect

The McGill Pain Questionnaire (MPQ)

The questionnaire provides estimates of the sensory, affective, and evaluative dimensions of pain.62 It is one of the most frequently used instruments for pain measurement and is considered useful for evaluating pain treatments and as a diagnostic aid.26,63–66 In addition to collecting information about diagnosis, drug therapy, pain and medical history, and other symptoms and modifying features, the MPQ also contains a list of words that describe pain, divided into groups pertaining to the sensory, affective, and evaluative dimensions of the pain experience.

The MPQ is available in several languages, as well as extended (Dartmouth Pain Questionnaire, McGill Comprehensive Pain Questionnaire) and shortened versions. Components of the MPQ have been incorporated into other instruments.26 Although the MPQ is one of the leading pain assessment tools, and is considered the gold standard of pain assessment tools, it has some limitations.67 For the purposes of this discussion, clinicians should keep in mind that it may be difficult to discriminate between types of pain syndromes in persons who are very anxious or who have other psychological morbidity.

Pain Distress Scales

The Acute Pain Management Guideline Panel68 recommends, among other tools, the use of the scales shown in Figure 7-2.

Brief Pain Inventory (BPI)

The BPI was originally developed for use in persons with cancer, although it is also used to assess pain in people with other diseases.26 The purpose of the BPI is to assess the severity of pain and the impact of pain on daily functions. Assessment areas include severity of pain, impact of pain on daily function, location of pain, pain medications, and the amount of pain relief in the past 24 hours or the past week.69 The internal consistency ranges from Cronbach α 0.77 to 0.91. The form takes about 10 minutes to complete. It is valid for use in Chinese (Mandarin),70 Filipino, French, Hindi,71 Italian,72 Japanese, and Vietnamese, among other languages. It is also available in a shortened form, the BPI-SF73,74 that takes 5 minutes to complete. When applying the BPI to persons with chronic noncancer pain, the clinician should keep in mind that interpretation may be difficult if the questions asked do not reflect the patient’s experience. Furthermore, questions about functioning are subject to both floor and ceiling effects. The BPI is copyrighted, but permission to use it is routinely granted at no cost after providing a short description of intended use. (Samples of the form, both the short and the long version, can be found at http://prg.mdanderson.org/bpicopy.htm.)

Multidimensional Pain Inventory (MPI)

The MPI (formerly the West Haven-Yale Multidimensional Pain Inventory, or WHYMPI) was developed by Kerns and colleagues.45 It is a 64-item, self-report questionnaire comprising three parts and 12 subscales. The first two parts are related to patients’ appraisals of pain and its impact on different domains of their lives, and patients’ perceptions of the responses of significant others to their distress and suffering. The third part assesses how frequently patients perform 18 common daily activities. Internal consistency ranges from Cronbach α 0.70 for outdoor activities to 0.90 for interference.

Using cluster-analytic and multivariate classification methods, three homogeneous subgroups of patients with chronic pain have been identified and replicated across a wide range of medical diagnoses (back pain, temporomandibular disorders, headache). The three groups’ distinct profiles were labeled “dysfunctional,” “interpersonally distressed,” and “adaptive coper.”

Burton and colleagues75 suggested that a psychometric battery consisting of the MPI and the BPI76 was useful for both identifying problem areas that might impede treatment of patients with chronic pain and assessing treatment outcome. Moreover, the MPI has been shown to be predictive of chronicity of pain following acute symptom onset.77,78 The MPI has demonstrated reliability and validity in patients with chronic low back pain.79

A recent study compared the redundancy, reliability, validity, and sensitivity to change among the Short-Form Health Survey (SF-36), Oswestry Disability Index (ODI, see later in this chapter), and MPI on a cross-sectional sample of N = 424, with a follow-up sample of N = 87 of patients with chronic pain, seen in an interdisciplinary pain clinic.80 The Cronbach α ranged from 0.69 to 0.92 for the MPI, from 0.79 to 0.91 for the SF-36, and was 0.86 for the ODI. Three concepts overlapped between the SF-36 and the MPI: pain, interference/social functioning, and mental health. Both the SF-36 and the MPI contributed unique scales (e.g., the MPI “significant other” scales; R2 range 0.03–0.16). Significant changes following treatment were observed for the MPI pain severity, interference and outdoor work activities; the SF-36 physical and social functioning, bodily pain; and the ODI. The MPI is used widely, and has been translated into Spanish, Portuguese, French, Swedish, Dutch, and Italian.44,81,82

Treatment Outcomes of Pain System (TOPS)

Because, when applied to individual patients, the SF-36 lacked measurement reliability for assessment of treatment outcomes, lacked sensitivity to upper extremity or facial pathology, and failed to separate limitations of work versus everyday activity, two of us (Wittink and Carr) and colleagues conducted a two-part study to develop an outcomes instrument suitable for measurement of change in individual patients with chronic pain. A novel group of scales derived from responses to 61 questions (which include the SF-36) proved sufficiently reliable for routine follow-up of individual patients during treatment for chronic pain. This new instrument, the TOPS, allows assessment of individual patient outcomes and aggregate or individual clinician performance during interdisciplinary treatment of chronic pain.22,23

In addition to the SF-36, this instrument contains demographic data and 14 scales of which 7 fit into the Nagi framework.83 The remaining 7 scales are considered to be mediating factors between the domains of pain, functional limitation, and disability. The 7 main scales include pain symptom, perceived and objective family disability, work limitations, objective work disability, and upper and lower body limitations. The Cronbach α of these scales ranges from 0.70 for objective work disability to 0.92 for lower body functioning and 0.93 for perceived family/social disability. The mediating scales include fear-avoidance, passive coping, life control, and solicitous responses. In addition, two scales measure patient satisfaction with care and outcomes. The final scale is the total pain experience scale, which is a composite of pain intensity, pain interference, physical functioning, and disability.

For evaluating a single patient, the TOPS pain symptom, perceived family/social disability, and total pain experience scales are considerably more sensitive to individual change than the SF-36 bodily pain (BP) scale. Their increased sensitivity can be attributed to their greater measurement reliability. For example, the Cronbach α equals 0.93 for total pain experience compared with 0.84 for the SF-36 BP scale. The TOPS scales also provide a clearer and more clinically relevant set of concepts for the pain clinician who may have objectives apart from a simple reduction in pain intensity, such as the reduction of suffering or the reduction of disability in spite of pain. This high internal consistency allows for the measurement of change during treatment of an individual patient.22

Pain clinic normative values were established based on a sample of 1230 administrations of the tool in interdisciplinary pain clinics in Boston and Salt Lake City. The instrument was translated and validated in French Canadian and 8 European languages.84 The TOPS takes 10 to 15 minutes to complete.

Many patients with chronic pain learn to function normally despite their pain, continue to work productively, and rarely seek medical care. Factors such as coping ability,85–87 fear avoidance,49,88 self efficacy89,90 and catastrophizing90–92 have been associated with adjustment differences in patients with chronic pain. The beliefs or cognitions that patients have regarding their pain problem are hypothesized to have a direct impact on mood. For instance, negative thoughts about pain are strongly related to depressive symptomatology.93 Depression has been associated with high health care utilization and costs94 and is prevalent in patients with chronic pain. Patients with chronic pain may develop a variety of psychological problems, including depression, anxiety, sleep disorders, and disruptions in family life. Because of the importance of psychological factors in the experience of chronic pain, adding an assessment of mental health and cognitive factors to generic and condition-specific HRQOL may help patients and health care providers work together more effectively toward their common goals of pain relief and improved functioning. Some commonly used mental health assessment tools are listed in Table 7-5.

Moderate to strong associations have been identified between coping responses, pain severity, psychological well-being, and physical functioning.95,96 Fear avoidance beliefs correlated significantly with self-reported disability in activities of daily living and work loss49 and were shown to be a significant predictor of chronic pain40,97 in patients with musculoskeletal disorders. Catastrophizing was shown to predict depression,86,90,91 perception of pain,87,98 lower self-efficacy for pain, lower spousal ratings of self-efficacy for control of fatigue or mood symptoms,99 and disability.100 Patients’ beliefs and self-appraisals thus play a large part in shaping the outcome of treatment. Many instruments exist that measure patient beliefs; here, we limit ourselves to discussing a few widely used tools.

Coping Strategies Questionnaire (CSQ)

Questionnaires that address cognitive factors include the widely used CSQ and the Pain Beliefs and Perceptions Inventory (PBPI). The CSQ was designed to help identify methods of coping used by persons with chronic low back pain.101 It contains six types of cognitive strategies, two types of behavioral mechanisms, and two effectiveness ratings. The CSQ was found to be internally reliable when used to assess pain coping strategies. The authors also found that praying, hoping, and coping self-statements were used frequently, while others such as reinterpretation of pain sensations, were not. Overall, cognitive coping and suppression, helplessness, and diverting attention or praying explained much of the variance in coping strategies. The CSQ has been studied widely to better describe its factor structure, its utility in persons with low back pain or cancer pain, and its utility for prediction of patient and spouse ratings of patients’ self-efficacy.87,99,102–106

Pain Beliefs and Perceptions Inventory (PBPI)

The PBPI assesses three aspects of pain beliefs: self-blame, perception of pain as mysterious in origin, and beliefs about pain duration.107 These authors found that the belief that pain will last is associated with greater pain intensity and decreased compliance with psychological and physical therapies. The PBPI contains only 16 items; thus, respondent burden is very low. Like the CSQ, the PBPI has also been used widely and has been translated for use in the United Kingdom.108–110

Fear Avoidance Beliefs Questionnaire (FABQ)

The FABQ is a 16-item instrument developed by Waddell and colleagues.49 Within the 16 items are two fear-avoidance scales. The first scale (7 items) concerns fear-avoidance beliefs about work and the second scale (4 items) concerns fear-avoidance beliefs about physical activity. The internal consistencies (Cronbach α values) for these two scales were 0.88 and 0.77, respectively. Test-retest reliability had a κ equal to 0.74.

Pain Catastrophizing Scale (PCS)

The PCS is a 13-item instrument developed in 1995 at the Dalhousie University Pain Research Centre to facilitate research on the mechanisms by which catastrophizing has an impact on the pain experience.111 The items on the PCS were drawn from previous experimental and clinical research on catastrophic thinking in relation to pain experience.101,112,113

The PCS yields a total score and three subscale scores assessing rumination (“I can’t stop thinking about how much it hurts”), magnification (“I worry that something serious might happen”), and helplessness (“There is nothing I can do to reduce the intensity of the pain”).114 A total PCS score of 38 represents a clinically relevant level of catastrophizing.114

The Cronbach α for the total PCS is 0.87; for rumination, 0.87; for magnification, 0.66; and for helplessness, 0.78. Test-retest reliability across a 6-week period was r = 0.75.111 The PCS takes about 5 minutes to complete. Support for good internal consistency and validity of the PCS was provided by others.115,116

Disease-specific instruments reflect particular limitations or restrictions associated with specific disease states. These instruments are designed to be sensitive in determining the effects of treatment on or the spontaneous longitudinal course of a single disease or condition. Disease-specific measures have been developed for almost each imaginable condition. For an overview of pain specific tools, see Table 7-6.

A few further examples of disease-specific instruments are instruments that measure the impact of migraine,117 shoulder pain,118–120 knee pain121 (KOOS, a knee injury and osteoarthritis outcome tool, and Lysholm scales),122 neck pain,123 and back pain. Two of the most commonly used disease-specific tools for back pain are the Oswestry Disability Index (ODI)124 and the Roland-Morris Disability Questionnaire (RDQ).125 The ODI and RDQ scores are highly correlated, with similar test-retest reliability and internal consistency. Floor and ceiling effects determine the choice of instruments. A greater proportion of patients score in the top half of the distribution of RDQ sores than in the top half of the ODI scores. The ODI is, therefore, recommended in patients who are likely to have persistent severe disability and the RDQ in patients who are likely to have relatively little disability.126

Oswestry Disability Index (ODI)

The ODI124 is one of the most frequently used tools in back pain research. It consists of ten sections that include pain intensity, personal care, lifting, walking, sitting, standing, sleeping, sex, social life, and traveling. Each section is scored on a 6-point scale (0–5), with 0 representing no limitation and 5 representing maximal limitation. The subscales combined add up to a maximum score of 50. The score is then doubled and interpreted as a percentage of patient perceived disability (the higher the score, the greater the disability).

The ODI has excellent test-retest reliability (r = 0.99,124 ICC = 0.83,127) and clinical face validity. The internal reliability, Cronbach α, was found to be 0.71 for version 1.0.128 Two studies that determined Cronbach α of version 2.0 found it to be 0.7614 and 0.87.129

Roland-Morris Disability Questionnaire (RDQ)

The RDQ125 was derived from the Sickness Impact Profile (SIP). The generic SIP was modified to become disease-specific by adding “because of my back pain” to each item. Twenty-four items were selected from the SIP by the original authors because they related specifically to physical functions that were likely to be affected by low back pain. These items include walking, bending over, sitting, lying down, dressing, sleeping, self-care, and activities of daily living. The Cronbach α for the scale has been estimated to be between 0.84 and 0.93.126 The RDQ correlates well with the SF-36 physical subscales, the SIP,130 and pain ratings.131 It is available in 12 languages, and translations are available from the author (mroland@man.ac.uk).

Work Limitations Questionnaire (WLQ)

The WLQ was developed by Lerner and Amick with support from GlaxoWellcome, Inc. The WLQ is a 25-item, self-administered questionnaire that evaluates the degree to which health problems interfere with ability to perform job roles. It was designed to assess groups of individuals who are currently employed. The WLQ indicates the degree to which health problems interfere with specific aspects of job performance (on-the-job disability), and the impact of these limitations on workers’ productivity.

The WLQ items ask respondents to rate their level of difficulty (or, on one scale, their level of ability) to perform 25 specific job demands. These demands have four defining features: (1) a wide range of jobs in the United States include these demands; (2) a wide variety of physical and emotional health problems can make it difficult to perform these demands effectively; (3) the demands are considered important to their jobs by workers who hold these jobs; and (4) losses in individual work productivity are frequently related to the degree to which these job-related demands are not met.

Responses to the 25 items are combined into four work limitation scales: the Time Management scale (Question 1), the Physical Demands scale (Question 2), the Mental/Interpersonal Demands scale (Questions 3 and 4) and the Output Demands scale (Question 5). The Cronbach α ranged from 0.88 for Output Demands to 0.91 for Mental/Interpersonal Demands. The WQL was shown to have high reliability and validity.132 The instrument takes 5 to 10 minutes to complete. (For more information, contact wlq@lifespan.org.)

The gold standard of current evidence-based literature synthesis is the Cochrane Collaboration. The Cochrane Collaboration was developed in response to a call by Archie Cochrane, a British epidemiologist, for systematic, up-to-date reviews of all randomized controlled trials on health care. Several centers have been established throughout the world, and collaborative review groups prepare and maintain systematic reviews. At the beginning of 1997, the existing and planned review groups (more than 40) cover most of the important areas of health care. Relevant to this chapter, Cochrane collaborative review groups have been formed to assess spine problems, musculoskeletal pathology, and pain (palliative and supportive care).8 The Cochrane group aims are “preparing, maintaining and disseminating systematic reviews of the effects of health care” to provide reliable, unbiased, up-to-date information to health care providers worldwide to permit informed decisions about the specific effects of health care interventions.8 As stated earlier, in making decisions about the care of individual patients, the results of these reviews must be integrated with the clinician’s expertise, which has been acquired through experience and practice. The results of the reviews must also be integrated with patients’ understanding and preferences, which derive from their knowledge of their condition (particularly if it is a chronic or recurrent health problem), the treatments offered, and the responsiveness or otherwise of the former to the latter (http://www.cochrane.org/cochrane/cc-broch.htm#GDAHC). The integration of the results of the reviews, clinician expertise, and patient feedback and participation is considered evidence-based practice.

Methods for scoring the quality of research reviewed have been established.133 If studies are combinable, a meta-analysis can be performed. A meta-analysis is a synthesis, usually understood to be quantitative, of the results of several studies.2 The precision of such a meta-analysis is greater than any one of its component studies because of the aggregation of patient numbers. Cumulative meta-analysis recalculates aggregate treatment effects and confidence intervals as each new relevant study is published. This technique of ongoing recalculation permits early decisions as to whether a treatment is efficacious or not, thereby averting the need for subsequent unneeded, costly, time-consuming (and some would say, unethical) clinical trials. Cumulative meta-analysis has documented that it may take years for statistically significant conclusions from randomized controlled trials to diffuse into textbooks and narrative review articles.134

One meta-analysis published outside of the Cochrane library identified published studies of treatment in multidisciplinary pain clinics (MPCs) between 1960 and 1990.135 These authors identified 65 studies that met inclusion criteria. They concluded that multidisciplinary pain treatment resulted in large effect sizes that were maintained for more than 6 months, that MPCs were efficacious, and that the effects were not limited to patients’ perceptions but also extended to objective behavior, such as return to work or decreased use of health care resources. One analysis evaluated whether return to work could be predicted after MPC treatment.136 These authors concluded that although prediction of return to work is an increasingly important topic, few of the studies they evaluated met appropriate design and statistical criteria. They were unable to clearly identify which variables were useful predictors of return to work.

One of the challenges unique to evaluating treatment at MPCs is that the criteria used to define success are often not uniform.137 Despite this problem, Turk and colleagues137 found that reported pain reduction ranged from 14% to 60%, and that reductions appeared to be well maintained at follow-up, although some studies reported no improvement during treatment. Treatment was reported to result in decreased opioid use in nearly three fourths of persons, whereas untreated people generally reported no change. Treatment was also noted to improve activity levels and return to work, lower use of the health care system (including hospitalization and surgery), and increase the percentage of disability claims that were settled. This last point suggests that many persons with chronic noncancer pain can return to work, even after an extended period of being disabled.

Finally, one group of researchers performed a meta-analysis of meta-analyses.138 These researchers concluded that MPC treatment is consistently effective for most outcomes, including return to work, although many different outcome variables were assessed in individual meta-analyses and ultimately, it was “unclear what combination of treatments is necessary for an effective [treatment] package.” Meta-analytic results depend, however, on the studies included within the analysis. Thus, owing to a variety of methodologic problems that affect pain treatment studies, the results of this literature synthesis should be interpreted cautiously. Indeed, some experts have argued that a single meta-analysis of heterogeneous trials of a single intervention applied to diverse groups of patients with a complex clinical condition may be frankly misleading.139

The College of Physicians and Surgeons of Ontario, Canada, was the first to publish guidelines on the treatment of chronic, noncancer pain based on the best available evidence in November 2000. Clinical practice guidelines have been defined as “systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances.” In the establishment of guidelines, levels of evidence generally are stratified as follows:

• Level I Strong evidence from at least one systematic review of multiple well-designed randomized controlled trials.
• Level II Strong evidence from at least one properly designed randomized controlled trial of appropriate size.
• Level III Evidence from well-designed trials without randomization, single group pre-post, cohort, time series, or matched case-controlled studies.
• Level IV Evidence from well-designed nonexperimental studies from more than one center or research group.
• Level V Opinions from respected authorities, based on clinical evidence, descriptive studies, or reports of expert committees.

Often, guideline committees will add a second dimension that documents not only the nature of the evidence that forms the basis of the recommendations, but also the strength and consistency of the evidence. Guidelines thus incorporate Cochrane or other systematic reviews when available, but go one step further by making specific recommendations with the definite intent to influence what clinicians do. (Evidence reports and clinical practice guidelines can be found on http://www.ahrq.gov.)

In this chapter, we have presented an overview of the terminology used in HRQOL outcomes assessment and examples of instruments that can be applied in clinical practice. Systematic measurement and documentation of HRQOL is a useful, clinically relevant approach to incorporate patient preferences into front-line medical decision making. Doing so is expected to improve overall patient satisfaction with care.10 No tool will be used if it is too burdensome (too long or too difficult to understand). Instruments must be easily understood, administered, and interpreted by both clinicians and patients. No instrument is ideal for all intended uses; questionnaires are available in a variety of forms and many of these can be readily incorporated into clinical care and research. Outcomes assessment is a dynamic area, particularly as it stands at the interface between routine practice and new standards for pain assessment and treatment applied by the Joint Commission for the Assessment of Healthcare Organizations. The dissemination of ever-simpler and more powerful means to capture data electronically in everyday health care opens new opportunities for understanding which treatments are effective, and for whom, and may provide irrefutable evidence that we who treat pain add patient-centered value to the health care enterprise.

Web Sites to Visit for Measurement Tools

http://www.york.ac.uk/inst/crd/welcome.htm

Web Sites to Visit for Evidence-Based Research

• http://www.cochrane.org: main site for the Cochrane Collaboration
• http://www.herts.ac.uk/lis/subjects/health/ebm.htm: Web sites and databases
• http://www.jr2.ox.ac.uk/bandolier/: bandolier EBM
• http://www.medicine.ox.ac.uk/bandolier/painres/PRintro.html: pain research at Bandolier EBM
• http://www.iwh.on.ca/: Institute for Work and Health, Canada
• http://www.ahcpr.gov/: Agency for Healthcare Research and Quality

Web Sites to Visit for Guidelines on Managing Pain

• http://www.ampainsoc.org/pub/bulletin/nov00/clin1.htm
• http://www.guideline.gov: research on health care in general