| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
Correspondence: Address correspondence to A. Lynn Snow, VAMC, 2002 Holcombe (HSR&D152), Houston, TX 77030. E-mail: asnow{at}bcm.tmc.edu
 |
Abstract |
|---|
 TOP Abstract An Overview of Pain...Application of the Existing...A Testable Model of...Applying the ModelTesting the ModelReferences |
|---|
Key Words: Dementia • Pain • Assessment
At the root of these assessment and research difficulties is the decreased or absent self-report capacity of individuals with dementia due to the language, memory, and abstract thinking deficits inherent in the disease (White & Cummings, 1997). Theoretical explanations of the pain experience have evolved toward increasingly intricate multidimensional models, yet these advances have failed to directly address the unique challenge of pain assessment in persons with dementia (Harkins & Price, 1992). Without an adequate conceptual model, research and clinical advancement in pain assessment in persons with dementia has been restricted. Although enormous strides in measurement have been made (e.g., Feldt, 2000; Hurley, Volicer, Hanrahan, Houde, & Volicer, 1992), current instruments tend to focus on a limited part of the pain experience (presence or absence of pain, and pain intensity), and there is little understanding of how these restricted results are tied into a multidimensional understanding of pain.
In this article, we present a multidimensional conceptual model of pain assessment for persons with dementia, which is based on the concepts of current pain theory. We then explain how our model serves as a guide for the development of a pain assessment strategy for these persons. Finally, we provide methodological recommendations for designing evaluations of pain assessment approaches for persons with dementia.
|
An Overview of Pain Theory |
|---|
|
TOPAbstractAn Overview of Pain...Application of the Existing...A Testable Model of...Applying the ModelTesting the ModelReferences |
|---|
In a review of major pain models, Novy and colleagues (1995a) noted that all current multidimensional theories agree regarding the following: (a) simple nociception does not equal pain; (b) pain is a multidimensional construct; and (c) pain is best conceptualized as having sensory, behavioral, emotional, and cognitive components. Although these theoretical advances have affected pain assessment in children (McGrath & Brigham, 1992) and adults (Turk & Melzack, 1992), they have not directly addressed the unique challenges posed by pain assessment in persons with dementia (Harkins & Price, 1992). In the next section, we examine the applicability of these models to persons with dementia.
|
Application of the Existing Models to Persons With Dementia |
|---|
|
TOPAbstractAn Overview of Pain...Application of the Existing...A Testable Model of...Applying the ModelTesting the ModelReferences |
|---|
In sum, the literature suggests that multidimensional pain models are applicable to individuals with dementia, but with the caveat that such individuals may be unable to correctly label painful sensations. The assessment of pain in those with dementia thus requires extrapolation of the meaning of observed behaviors by external raters, despite the uncertainty this may bring. Further, because most models of pain were developed with communicative individuals, we must be vigilant in identifying components of such models that are inapplicable to nonverbal individuals. In the next section we present a pain assessment model based on those concepts held similar by the current multidimensional pain models. Further, this model is testable and accounts for the unique features of pain assessment in persons with dementia.
|
A Testable Model of Pain Assessment in Persons With Dementia |
|---|
|
TOPAbstractAn Overview of Pain...Application of the Existing...A Testable Model of...Applying the ModelTesting the ModelReferences |
|---|
|
The sensory component refers to sensations experienced in response to a nociceptive stimulus. Researchers usually measure it by asking about pain intensity, quality (burning, throbbing, etc.), temporality (continuous, intermittent, etc.), location, or duration (Melzack & Wall, 1965). The emotional component refers to emotions that accompany pain and occur as a result of the pain. This includes emotional valences attached to the pain (e.g., is the pain horrible, awful, frightening; see Melzack, 1975), and emotional reactions to the pain (e.g., depression). Some instruments assess emotions generally (e.g., "Rate your overall mood during the past week"), whereas others are more specific (e.g., "During the past week, [has] pain interfered with your mood"; see Cleeland & Ryan, 1994; Kerns, Turk, & Rudy, 1985). The behavioral component refers to pain's effect on patient function or activity (Cleeland & Ryan, 1994; Jamison, Rudy, Penzien, & Mosley, 1994). This definition specifies deliberate or functional behaviors (e.g., dressing), as opposed to automatic or nonfunctional behaviors (e.g., grimacing), because we posit that automatic pain behaviors in persons with dementia are regarded as pain intensity indicators (part of the sensation component). For example, most observational pain measures recommend summing behavior counts or intensities, and treat this total as a global pain measure; the correlation of this total with another pain intensity measure is often cited as concurrent validity evidence (Abbey et al., 2004; Lefebvre-Chapiro, 2001; Warden, Hurley, & Volicer, 2003). A challenge in defining the emotional and behavior components is determining the extent to which pain must be established as the cause. Finally, the cognitive component refers to thoughts that accompany pain and occur as a result of the pain. Most instruments focus on cognitions that make it harder or easier to tolerate pain, such as coping, self-esteem, and self-efficacy (Turk, 1992).
Incorporating Factors Affecting Pain Assessment in Persons With Dementia
We next expand this basic model to clarify the specific factors that affect the pain assessment outcome in persons with dementia (see Figure 1). In our expanded model, we specify that a nociceptive stimulus gives rise to painful sensations (influenced by stimulus-specific factors), which influence pain perception (influenced by patient-specific factors). Pain perception is described by our basic model of pain. The internal pain experience is communicated through external signs. An external rater observes these signs (influenced by the method through which signs are observed) and interprets them (influenced by rater-specific factors). The result is the external rater's pain assessment. To emphasize the process leading to the pain assessment, all arrows point that direction, although in actuality, many of the relationships are bidirectional. The proposed model is compatible with the pain communications model, which posits that an internal pain experience is encoded in expressive behavior, which is then decoded by an observer (Hadjistavropoulos & Craig, 2002); we refer the reader to this model for a thorough discussion of pain communication from an ontogenetic, adaptation perspective. We next explore each of the categories that influence the pain assessment process.
Sensation
Determined by pain etiology and by how the body reacts to nociceptive nervous signals (Melzack & Wall, 1965), the pain experience is affected by the nature of both the nociceptive stimulus and the nociceptive sensations (Melzack & Wall, 1965): location (e.g., stomach or head), intensity, pattern (intermittent or continuous), duration (acute or chronic), frequency, and quality (e.g., crushing or stabbing; see the AGS Panel on Chronic Pain in Older Adults, 1998).
Perception
Experimental psychology defines sensation as "immediate and basic experiences generated by isolated, simple stimuli," and perception as "the interpretation of those sensations, giving them meaning and organization" (Matlin & Foley, 1992). The proposed model posits the internal pain experience to be affected by not only the nociceptive stimulus but also by the perception of that stimulus, which may be affected by several patient-specific factors. Gender, emotional states, cognitive content, and behaviors are associated with markedly different pain intensity reports in response to the exact same physical stimulus (Turk & Flor, 1999; Vallerand & Polomano, 2000). Pain history is also associated with such differences, particularly for individuals with chronic pain, because of the effect of learning on one's interpretation of the stimulus (Turk & Flor, 1999) and because the nociceptive nervous system itself can be altered by prolonged exposure to noxious stimuli (Perl & Kruger, 1996). Physical status is another important factor for similar reasons (Hunter, 2001). Finally, as discussed previously, cognitive impairment can also affect one's interpretations of painful signals.
External Signs
The patient's internal experience of pain can be displayed in several ways: verbal signs (complaining of pain or other negative emotional or physical states or making sounds such as moans or groans) nonverbal signs (e.g., grimacing or exhibiting unusual motor behavior; Bachino, Snow, Kunik, Cody, & Wristers, 2001), or physiological signs (elevated heart rate or blood pressure; Williams, 1996).
Unlike the sensory, behavioral, and emotional pain components, there are no external indications of cognitive content. The cognitive component is shaded in the proposed model because it is a "black box." This does not mean that persons with dementia do not cognitively evaluate pain, but that in the presence of self-report impairment, it is not possible to predict the effects of such evaluations. Perhaps it is possible for researchers to assess the cognitive component by using informant reports, but the black box indicates uncertainty about the validity of such reports.
Observation of External Signs
The fundamental difference between pain assessment in intact and severely demented persons is an almost exclusive reliance on external signs. Even in people with milder dementia, deciding how to weight self-report versus observation is a substantial challenge because self-report impairment is gradual and varies even within the same dementia stage. Using external rater reports necessitates several important considerations, because discrepancies consistently exist between self-reports and external reports. For example, studies comparing pain self-ratings to the ratings of others report low to moderate agreement and significant patient–rater discrepancies across a variety of raters (family, nurses, aides, physicians) and settings (postoperative recovery, palliative care, geriatric inpatient and cancer outpatient settings, intensive care units, nursing homes) (Krulewitch et al., 2000; Nekolaichuk et al., 1999; Teske, Daut, & Cleeland, 1983; Weiner, Peterson, Ladd, McConnell, & Keefe, 1999; Zalon, 1993). External raters tend to rate pain lower than the patient. Further, the magnitude and bias (overreporting or underreporting relative to the patient) of these discrepancies are affected by characteristics of the assessment method used. We next examine these method factors: construct characteristics, rater type, data-collection method, and assessment instrument characteristics (Magaziner, 1997; Neumann, Araki, & Gutterman, 2000).
Construct Characteristics
Researchers can rate some aspects of pain by using more objective procedures such as frequency counts (e.g., "How often was the patient able to make the bed last week?"). Other aspects require more subjective judgments based on the rater's experience (e.g., "Rate the intensity of the patient's pain, from no pain at all to worst possible pain."). The more subjective the construct, the larger the discrepancy. For example, studies report higher agreement between patient and informant for levels of functioning and less private physical conditions, and lower agreement for more private physical conditions, depressive symptoms, psychological well-being, and pain (Neumann et al., 2000). Thus, conclusions may vary if they are based on more objective as opposed to more subjective data types.
Rater Type
Discrepancy magnitude varies by rater type. There is evidence of greater discrepancies between physicians and patients than between nurses and patients (Nekolaichuk et al., 1999), and between licensed practical nurses and patients than between aides and patients (Engle, Graney, & Chan, 2001). Differences may be due in part to varying experiences with the patient. For example, the nursing home physician must base ratings on information derived from short, infrequent visits in which the physician is focused on examining the patient to identify problems. In contrast, the nursing assistant has frequent patient contact each day, but in this contact the assistant is focused on care tasks. Raters also bring different attitudes toward and knowledge about the construct. Finally, the patient may behave differently with different raters. Perhaps rater discrepancies do not reflect variations in accuracy but rather that truth is best estimated by assembling multiple perspectives from different raters.
Data-Collection Method
External raters can collect pain assessment data through patient subjective verbal report, informant subjective verbal report, or objective behavioral observations. These methods may evaluate different aspects of the same construct. Studies comparing self-ratings to behavioral observations, and informant ratings to behavioral observations, have demonstrated that different methods yield moderately different results (Gotay, 1996; Turk & Flor, 1987; Zanetti, Geroldi, Frisoni, Bianchetti, & Trabucchi, 1999). Self-ratings and informant ratings are more subject to rater bias than performance-based ratings, yet they allow for assessment of situations that are unlikely to be observed by a standardized rater (e.g., toileting, aggression). The breadth of a complicated construct such as pain is probably more likely to be represented by combining self-reports, informant reports, and behavioral observations.
Assessment Instrument Characteristics
Finally, even within the same data-collection method, discrepancies often exist between assessment instruments. One cause could be the scope of assessment. For example, the Brief Pain Inventory includes items measuring pain interference, whereas the McGill Pain Inventory does not (Cleeland, 1989; Melzack, 1987). Alternatively, discrepancies could reflect the superior psychometric qualities of an instrument for certain populations. For example, the 21-point box scale of pain intensity has superior psychometric qualities in persons with dementia compared with a 5-point verbal rating scale (Chibnall & Tait, 2001).
Interpretation of External Signs
Many of the same factors that affect a patient's perception of his or her pain also affect an external rater's interpretation of the patient's pain. These can include the demographic characteristics and pain history factors discussed earlier. The rater's knowledge of and beliefs about pain and dementia are also powerful. If the rater believes that pain is an inevitable part of aging, or that persons with dementia can no longer feel pain, the rater may be insensitive to subtle distress signs. Similarly, the rater's relationship with the patient can affect interpretation. Amount of time spent with the patient has been shown to affect informant–patient discrepancies (Magaziner, 1997). Finally, secondary gain can affect a rater's interpretations, unconsciously or consciously. For example, the caregiver reporting a hospice patient as in intractable pain may benefit from the emotional support provided by nurse home visits.
|
Applying the Model |
|---|
|
TOPAbstractAn Overview of Pain...Application of the Existing...A Testable Model of...Applying the ModelTesting the ModelReferences |
|---|
Verbal Pain Assessment
The majority of the pain assessment literature has focused on self-report (Melzack & Katz, 1999; Weiner, 1999). Results of studies of verbal pain assessment tools in persons with dementia have been variously touted as evidence that persons with dementia can, or cannot, validly use such tools, depending on whether one's perspective is the glass half-empty or half-full. For example, one study reported that, of five simple pain intensity scales administered to a group of cognitively impaired nursing home residents, 83% of the residents could complete at least one of the instruments; however, only 17% could complete all five instruments (Ferrell et al., 1995). The ability to respond to a verbal pain assessment is generally associated with level of cognitive impairment (e.g., Chibnall & Tait, 2001; Cohen-Mansfield, 2002; Wynne, Ling, & Remsburg, 2000). Completion rates are negatively correlated with degree of cognitive impairment (Scherder & Bouma, 2000). Two limitations of this literature are that (a) only unidimensional scales, almost exclusively pain intensity scales, have been examined, and (b) accuracy of pain self-reports in people who can complete the scales has yet to be examined. In sum, it is probably best to include verbal pain assessment for individuals with dementia, but it is not sufficient to rely solely on it, particularly in moderately and severely demented persons. Further, the assessment techniques from this literature do not meet the needs described by the proposed model (or any current comprehensive pain model) because these instruments do not assess all four pain components.
Nonverbal Pain Assessment
There are currently at least seven published behavioral observational scales with adequate to good reliability and validity evidence for persons with dementia. All require the rater (typically a nurse, although one scale has been validated with aides) to observe the patient and rate the presence or absence, intensity, or frequency of behaviors. The majority of these scales focus only on the sensation component (Feldt, 2000; Snow et al., 2004b) or on the sensory and emotional components (Abbey et al., 2004; Hurley et al., 1992; Warden et al., 2003). Only two scales include sensory, affective, and behavioral components (Lefebvre-Chapiro, 2001; Villanueva, Smith, Erickson, Lee, & Singer, 2003).
In addition, even those tools that appear to be multidimensional may be limited in their scope of the emotional and behavioral components. Emotional component evaluations are limited to using affective language to describe some rated facial expressions (e.g., frightened, anxious, or sad facial expression). Behavioral component evaluations are limited to rating independence during the past few meals and the extent to which washing, dressing, and mobility are affected. Thus, in striving for a clinical application of the proposed model, we find it important to consider alternative assessment approaches in addition to these scales.
Adapting Scales Developed for Demented Persons
As we discuss in an upcoming section (Testing the Model), it is important for researchers to be able to use the same set of measures in both demented and nondemented persons. This allows them to accomplish one kind of validation work by comparing the self-reports of nondemented persons (the gold standard) against external ratings. Therefore, it is important to consider differences that might occur when behavioral observation scales are used with nondemented persons.
One important concern is that behavior will be unduly influenced by an observer. Indeed, Hurley and Volicer (2001) found it impossible to use the Discomfort Scale with nondemented persons because of the observer's impact on the patient's behavior. However, observational instruments have been shown to be feasible, reliable, and valid in nondemented patients with chronic pain (Keefe & Block, 1982; Richards, Nepomuceno, Riles, & Suer, 1982). Keefe and Richards required the patient to engage in a series of activities, thus distracting the patient from the observation. In contrast, in Hurley's method the patient is at rest and thus free to focus all attention on the observer. Thus, observational measures that engage the patient in a competing activity during the period of observation may be better suited if nondemented persons are to be assessed (e.g., Feldt, 2000; Snow et al., 2004b).
Adapting Scales Developed for Nondemented Persons
One might assume that comprehensive pain assessment in persons with dementia could be achieved by adapting existing multidimensional tools. For example, several multidimensional assessment instruments exist for nondemented adults, such as the McGill Pain Questionnaire (Melzack, 1975), the Multidimensional Pain Inventory (Kerns et al., 1985), and the Geriatric Pain Measure (Ferrell, Stein, & Beck, 2000). However, all of these instruments rely almost exclusively on the patients' pain reports. Few elicit reports from significant others, and the content is often inappropriate for persons with dementia (e.g., "Does your significant other's pain interfere with his or her ability to mow the lawn?"; see Kerns et al., 1985). Assessment instruments also exist for nonverbal neonates and infants, such as the CRIES (Krechel & Bildner, 1995). However, in addition to having not been validated in persons with dementia, these instruments are not truly multidimensional because they focus only on the sensory component. In sum, no single instrument is available that adequately assesses sensory, behavioral, emotional, and cognitive pain components in persons with dementia—existing nonverbal tools are not multidimensional, and existing multidimensional tools are not appropriate for nonverbal populations.
Given these limitations, we recommend the following strategies for comprehensive pain assessment. First, bring together multiple measures (e.g., use one of the pain scales for persons with dementia to assess parts of the emotional, behavioral and sensory components and other measures to assess other aspects of the emotional and behavioral components). Second, adapt existing pain measures for cognitively intact persons to make them appropriate for use with persons with verbal impairments (e.g., change the wording of the Multidimensional Pain Inventory to third person so an informant can rate the person with dementia); this strategy has produced valid results in the assessment of other psychosocial constructs, including depression (Teri & Wagner, 1991), deficit awareness (Snow et al., 2004a), and functional status (Kiyak, Teri, & Borson, 1994). Third, adapt existing measures for persons with dementia to assess other pain components; for example, measure the emotional pain component by using the Affect Rating Scale (a behavioral observation scale that assesses negative affect; see Lawton, Van Haitsma, & Klapper, 1996) during a known painful episode.
These strategies are recommended because, given the current available instrumentation, the only alternative to these strategies is to not attempt comprehensive pain assessment in persons with dementia. However, the clinician should be aware that such adaptations alter the psychometric properties of instruments, and thus the reliability and validity of the adapted instrument will be unknown. Clinical hypotheses based on such assessments should therefore be considered tentative, and the clinician should look for independent corroborating evidence of these hypotheses. The researcher is encouraged to undertake psychometric studies of such adapted instruments so that adequate validated instrumentation for comprehensive pain assessment in persons with dementia will soon be available.
Assessment Strategy Recommendations
The proposed model emphasizes the effects of nociceptive stimulus and patient, method, and rater factors on the external rater's pain assessment. The rater might note such factors when performing a comprehensive pain assessment, and possible effects of such factors on the assessment outcome. In particular, a thorough history of the patient's psychosocial well-being and pain will provide helpful information. Given the effects rater and method factors can have on pain assessment, another strategy suggested by the model is the importance of using multiple methods and multiple raters when feasible.
|
Testing the Model |
|---|
|
TOPAbstractAn Overview of Pain...Application of the Existing...A Testable Model of...Applying the ModelTesting the ModelReferences |
|---|
We call the first validation paradigm the Self-Report Comparison Model (see Figure 2). This approach compares the pain latent variable from the basic pain model to pain self-reports in persons without dementia. The pain latent variable would be modeled by use of pain assessment instruments that rely on external ratings rather than self-ratings to assess each of the four pain components. Agreement between the pain latent variable and pain self-report (the typical gold standard) would serve as evidence of validity of the pain latent variable. The pain latent variable could then be modeled in patients with mild to moderate dementia. The individuals without dementia would then be compared with those with mild to moderate dementia to see if there were similar patterns of relationships between the pain assessment instruments included in the pain latent variable model. If the patterns were similar in the two groups, a comparison could then be conducted between individuals with mild to moderate dementia and those with severe dementia. Thus, there would be evidence for validity of the pain latent variable in persons with severe dementia if the patterns of relationships between other-rated pain assessments were similar in these last two groups, even though it would not be possible to obtain the typical gold standard, self-report data, in this population.
|
|
Conclusions
We have developed a conceptualization of the pain assessment process for persons with dementia, based on the current theoretical literature and the assessment challenges unique to persons with dementia (particularly the challenges of using external raters). We have provided recommendations for integrating this information into a comprehensive clinical assessment strategy. We have also given research recommendations for evaluating our model or other pain models for persons with dementia. Important questions, such as "Should pain assessment strategies vary by dementia severity?," and "Which combination of pain assessment information sources generates the most accurate yet parsimonious data for nursing home residents?," can only be answered if pain research in persons with dementia moves beyond a unidimensional model of pain assessment.
|
Footnotes |
|---|
1 Houston Center for Quality of Care and Utilization Studies (a VA HSR & D Center of Excellence and Section of Health Services Research, Department of Medicine, Baylor College of Medicine), Houston, TX.
2 Veterans Affairs South Central Mental Illness Research, Education, and Clinical Center, Houston, TX.
3 Department of Medicine, Baylor College of Medicine, Houston, TX.
4 Menninger Department of Psychiatry and Behavioral Sciences, Baylor College of Medicine, Houston, TX
5 Psychiatry and Behavioral Sciences Department, University of Arkansas for Medical Sciences, Little Rock.
6 Center for Mental Health Care and Outcomes Research, (a VA HSR & D Center of Excellence) North Little Rock, AR.
7 Department of Geriatrics, Reynolds Center on Aging, University of Arkansas for Medical Sciences, Little Rock.
8 Palliative Care Department, M.D. Anderson Cancer Center, Houston, TX.
9 Department of Anesthesiology and Psychiatry, University of Texas Health Science Center at Houston, TX.
Decision Editor: Linda S. Noelker, PhD
Received for publication September 26, 2003. Accepted for publication May 14, 2004.
|
References |
|---|
|
TOPAbstractAn Overview of Pain...Application of the Existing...A Testable Model of...Applying the ModelTesting the ModelReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
