Pain assessment for people with dementia: a systematic review of systematic reviews of pain assessment tools

Dementia affects an estimated 44.4 million people worldwide – a figure which is set to rise – and represents an important public health issue, costing an estimated US$604 billion each year [1]. Dementia presents a particular challenge for treatment and care due to the progressive cognitive and functional decline that are hallmarks of the condition. In particular, loss of language and the ability to communicate raises the risk of unmet need. A critical example of this issue is in the assessment of pain.

Pain is common in older adults, with up to one third of community-dwelling people over 60 experiencing regular pain and 50% of people over 80 regularly taking analgesics [2]. Although pain in dementia is difficult to assess, the literature agrees that at least 50% of people with dementia also regularly experience pain [3], which is predominantly, but not exclusively related to the musculoskeletal system. Osteoarthritis is very common in these individuals [4], and pain is also frequently caused by falls, pressure ulcers, infections and underlying neuropathy due to comorbidities [5].

In addition to significant distress and discomfort, untreated pain can also be a causative factor in key symptoms and quality of life indicators for people with dementia. Behavioural and psychological symptoms of dementia such as agitation and aggression often arise as a result of underlying pain, presenting a considerable challenge for treatment and care that may lead to institutionalisation or prescriptions of antipsychotic medications that carry serious safety concerns [6]. The accurate and timely assessment of pain is therefore of critical importance when considering the overall care for people with dementia. In general, due to the subjectivity of pain, self-report is considered to be the gold standard for pain assessment. While people with mild to moderate dementia are often able to report their pain verbally or use simple visual or numerical pain intensity assessment tools, these options are not feasible for use with people with later stage dementia in whom communication ability is severely impaired [7, 8]. As a result previous work has shown that pain is frequently under-detected and poorly managed in people with dementia, in both long-term and acute care [9‐12].

In the absence of accurate self-report it has been necessary to develop observational tools to be used in both research and practice, based on the interpretation of behavioural cues as a proxy for the presence of pain. This approach has resulted in a proliferation in the number of pain assessment instruments developed to identify behavioural indicators of pain in people with dementia and other cognitive impairment. The most structured of these are predominantly based on guidance published by the American Geriatrics Society [13], which presents six domains for pain assessment in older adults. These include facial expression, negative vocalisation, body language, changes in activity patterns, changes in interpersonal interactions and mental status changes. However, the interpretation of many of these behaviours is complex when applied to dementia due to considerable overlap with other common behavioural symptoms or cognitive deficits which may confound an assessment, manifesting from boredom, hunger, anxiety, depression or disorientation [14]. This increases the complexity of identifying the presence of pain accurately in patients with dementia and raises questions about the validity of existing instruments. The psychometric and discriminative properties and the clinical utility of currently available instruments are as yet unclear. As a result there is no clear guidance for clinicians and care staff on the effective assessment of pain, nor how this should inform treatment and care decision-making. A large number of systematic reviews have been published which analyse the relative value and strength of evidence of existing pain tools. There is a need for guidance on the best evidence available and for an overall comprehensive synthesis.

This meta-review forms part of a larger programme of research focusing on developing decision support interventions to assist with the assessment and management of pain in patients with dementia in an acute care setting. A key aspect of this work is to identify existing assessment tools with validation data pertaining to the treatment of people with dementia. This meta-review presents a thorough synthesis of current systematic review literature concerning the psychometric properties and clinical utility of pain assessment tools for the assessment of pain in adults with dementia. It provides a detailed picture of the state of the field in the complex task of assessing pain.

For ease of reference, in this paper we refer to our systematic review of systematic reviews as a meta-review; we call the systematic reviews considered for inclusion in the meta-review reviews and refer to publications included in the reviews as studies; we use the term records to refer to the bibliographic data of publications of reviews (for the most part retrieved thought online database searches); the terms (pain assessment) scales, tools and instruments are used interchangeably. The process of the meta-review followed guidance from The Cochrane Collaboration [15] and the Joanna Briggs Institute [16]. In undertaking this meta-review, our review questions were the following:

The search retrieved 441 potentially eligible unique records. After screening titles and abstracts, and further removing duplicates, we obtained the full text of 183 records and assessed these for eligibility. We identified 23 reviews as being potentially eligible for inclusion, of which 13 were excluded as they did not provide data on the psychometric properties of the tools. The remaining set included 10 records reporting data from eight reviews (Schofield et al. [30] review was reported in three separate studies [30‐32], we have combined the results of this) (Figure 2). Table 4 provides details of the eight included reviews and Table 5 details of the 13 excluded reviews.

The findings are structured as follows. First, we briefly summarise the reviews considered at the time of data extraction but excluded for lack of data on psychometric properties of the tools. We then describe the reviews included in our analysis – their methods and our quality assessment. Third, we describe the findings of these reviews, i.e. the pain assessment tools - characteristics, psychometric properties, feasibility of use and clinical utility. We conclude with the reviews’ overall assessment of these tools.

The objective of our meta-review was to identify which tools are available to assess pain in adults with dementia, in which settings and patient populations they have been used, and evaluate their reliability, validity, feasibility and clinical utility. We found a relatively large number of reviews on this topic and a considerable number of pain assessment tools available for the cognitively impaired population. Of the reviews we evaluated, only one [35] met all of the quality criteria outlined in the AMSTAR checklist, and would be considered a high quality review. The review by Herr et al. [33] has an associated website where 17 tools (specifically for use in nursing homes) are listed and commented on in detail [50].

The systematic reviews commonly situated their choice of interventions to review in the context of the challenges associated with assessing pain in patients who were cognitively impaired. Whilst there was recognition that the gold standard for pain assessment was the patient’s own assessment, this was unlikely to be possible with this patient population if the person was severely cognitively impaired, thus leading to the use of observational scales. The tools included in these reviews were therefore for the greater part observational, but showed a broad variation of measures and methods of assessing pain in adults with dementia. The foundation and focus of an observational scale is how the pain is manifested or made known such as on the basis on the American Geriatric Society (AGS) Pain indicators [13]. An additional conceptual foundation resided on a differentiation of aspects of pain – such as “the sensory-discriminative and motivational-affective aspects” [43]. Very few tools appeared to have any strong theoretical underpinning to the development.

First, given the large number of existing tools identified within our meta-review, it seems inappropriate to develop further tools if on the basis of the same conceptual foundations. Instead, researchers need to either envision new assessment instruments on the basis of different conceptual foundations, or concentrate on extending the psychometric evidence base for the existing tools. There is a need to evaluate the tools on a wider scale, across a variety of patients and clinical settings, using rigorous methods and larger sample sizes. The studies need to ensure that the definitions of cognitive impairment and type of pain that is being assessed are clearly defined, to enable comparisons across populations. While rigorous research is needed, it must be noted that the accuracy of tools is difficult to assess when an objective biological marker or other gold standard criterion is lacking - such as is the case for pain intensity [35, 48, 49].

Second, research should be conducted in clinical practice to assess the feasibility and clinical utility of the tools, and thus their potential for use in everyday clinical practice. Tools may be more useful in detecting relative changes in individual patients than differences between patients [41]. Tools that showed high reliability in research may not display to be highly reliable in routine clinical practice if not administered as intended [51]. Pain assessment tools designed for research purposes, in order to aggregate and compare data across patients, do not necessarily transfer easily and effectively to clinical settings for everyday use. In general it has been suggested that measurement tools developed for the purpose of evaluation of policy making [52, 53] or routine general screening [51] might have ‘no meaning’ at the frontline.

Third, research on the clinical utility of the tools should include evaluation of their impact in terms of choice of treatment and patient outcomes. This question was considered in one of the excluded reviews [48]: the reviewers found no evaluations of the effect of protocols such as algorithms or pathways for assessing pain in inpatients (including those with dementia) and they reported “a study in cancer patients [that] found opioid-related over sedation and other adverse effects increased substantially after implementation of pain assessment on a numerical scale routinely with other vital signs” ([48], p11, citing [54]). Herr et al. [33] reported that “the use of the ADD Protocol was associated with a significant increase in the use of pharmacologic […] and non-pharmacologic comfort interventions” (p179). Alternatively, the use of the wrong pain assessment tool might reduce the likelihood that pain treatment will be initiated or contribute to acute exacerbations of pain (as could be inferred by data from [34]), or may be found to have no effect on the quality of pain management [55].

Finally, an instrument must be relevant to “the condition, setting and participants in the health interaction, in particular the patient and clinical users” [56]. This includes the notion of user-centredness, defined as “the extent to which an instrument faithfully captures both the content of the health care user’s views and the form or ways in which their views are expressed” [57]. As a number of the systematic reviews and other literature makes clear, pain is a subjective experience; thus the associated measurement gold standard is the patient’s own assessment [24]. However, where the patient is severely cognitively impaired, alternative ways to assess the patient’s level of pain must be found. Resort is made to behavioural indicators and groups of cues [58]. Ideally, these aspects would be used to assess expressed pain when the patient is at rest and when moving. Against this context, user-centredness, in terms of the actual patient’s self-assessment of pain, is unachievable. The next best option, from a user-centred perspective, becomes the assessment of a person who is most familiar with the patient in their everyday life in a hospital or other care setting – what Herk et al. [24] call a “silver standard”. It is critical, however, that the assessor has a high familiarity with the patient. It can plausibly be argued that the more this is achieved, the greater the likelihood of a closer fit of the proxy’s view with the patient’s own experience. At the same time, over-familiarity may lead to an attenuation of the observer’s focus on potential pain clues. However, as Smith [34] commented, one is rarely in ‘an ideal situation’ where the direct care providers know, and are thus highly familiar, with the care recipient’s personal habits and history. Thus, rather than seeking to review the degree of the user-centredness of tools in this area, it is important to evaluate the guidance on a tool’s use (that is, who should administer the intervention and in what range of situations) and then explore whether or not it is actually used in the indicated way in everyday clinical practice. Exploration and findings on clinical utility and feasibility of use provide suggestive evidence of potential for use in everyday practice. Such evidence is a necessary but not sufficient condition for a tool’s actual use. It is notable that little or no insight is provided into the guidance over who should administer the tools in any of the systematic reviews. To gain insight into this aspect of use and possible additional evidence on a tool’s actual use in practice, one would need to examine the original instrument designers’ papers and validation studies, rather than rely on a systematic reviewer’s observations or summaries. Notwithstanding, it can be recommended that these two aspects of tool use (guidance and actual use) are incorporated into tools reviews and, given this patient group, development of evidence on the actual use of the tools in clinical practice.

The assessment of pain in patients with dementia is challenging for clinicians, due to some patients’ inability to verbalise the nature of their pain. This review highlights the current state of the evidence base in relation to pain assessment tools, and provides insights into current gaps in our understanding. We identified a total of 28 tools that could possibly be used in clinical practice to help with this process; however we cannot at present recommend any particular tool for use in any clinical setting, due to the lack of comprehensive evidence on the reliability, validity, feasibility or clinical utility of any one particular tool. Further research should be conducted on the psychometric properties of tools and in clinical practice to assess feasibility, clinical utility and guidance on use of the tools. These are necessary albeit not sufficient pre-requisites for actual use in routine clinical practice. Throughout, the aim remains the same: to gain as good an assessment as possible of the patient’s pain with a view to effective pain management.