Key findings
To our knowledge, this is the first systematic review to identify and appraise assessment measures of symptoms commonly experienced by people with dementia for use in long-term care settings. Our review identified 32 proxy-completed measures of common symptoms experienced by people with dementia. Of these measures, those that assess pain possess the strongest evidence of psychometric properties. Progress on all the other measures is promising, although oral health, psychological wellbeing, and discomfort measures require further psychometric evaluation, and there have been challenges in developing a measure that accurately detects depression. Neither of the two neuropsychiatric or two anxiety measures were validated in the long-term care setting. Furthermore, we found only 11 studies where measures were validated when administered by non-clinically trained caregivers even though these caregivers are frequently in the best position to detect changes quickly due to enhanced resident knowledge and contact [
12].
Despite the extent of symptoms experienced by this population, we were unable to find any multi-symptom assessment measures validated for use in routine care as an assessment measure. Instead, we found measures that assess single symptoms or symptom groups, specifically pain, oral health signs and symptoms, multiple neuropsychiatric symptoms, depression, anxiety and psychological wellbeing, and discomfort. Assessing discomfort may alert caregivers to physical or emotional discomfort that can then be further investigated to determine the underlying cause [
68]. However, content analyses of pain and discomfort measures in dementia found significant overlap resulting in poor sensitivity in assessing these constructs [
69], a finding supported by our results with pain and discomfort frequently being used interchangeably. An alternative to assessing discomfort is to provide caregivers with measures to assess all common symptoms. This would facilitate a comprehensive symptom assessment, and alert caregivers to consider all common symptoms and sources of distress.
Caregivers’ use of a battery of single assessments (e.g. pain, neuropsychiatric symptoms, oral health) could facilitate detection and monitoring of common symptoms, but is unlikely to be feasible for regular and frequent use due to the time taken to complete multiple measures. Palliative or end-of-life measures, such as the Symptom Management at the End of Life in Dementia [
70] or the Palliative care Outcome Scale [
71] could provide a brief yet comprehensive assessment of common physical, psychological, and other distressing (such as agitation) [
70] symptoms to support detection and management of symptoms in care. The former was developed to measure outcomes and evaluate end-of-life care in dementia and has been extensively evaluated [
70,
72‐
74], although predominantly after the death of the resident. It incorporates nine symptoms in people dying with dementia and therefore has the potential for use as a clinical assessment measure for people in the dying phase. The Palliative care Outcome Scale was developed for a non-dementia population but has sound psychometric properties and is used across settings to support clinical care [
75]. It has been used to assess symptoms and the quality of palliative care to nursing home patients with and without dementia [
76] and found to have the potential to identify areas of care that require addressing. Nonetheless, there was a high level of missing scores for some items (≤59.8 %) in the dementia subgroup, suggesting adaptation is required for this population [
76]. Results of a qualitative study suggest that such multi-symptom measures used in routine care may require provision of proxy-observed behaviors or signs to assess verbally-compromised residents with dementia [
77]. Use of a single multi-symptom measure may not provide a detailed assessment of each symptom. However, multi-symptom measures may support comprehensive assessment of symptoms with minimal time burden and, if required, inform requirement for further assessment or prompt referral to health professionals.
The second major finding from our study is the lack of assessment measures to assess common symptoms. The clinical challenges and importance of accurately assessing pain in this population is apparent by substantial development in pain measures, evidenced by a recent meta-review [
78]. As a consequence, we found pain measures have the strongest psychometric evidence. Nonetheless, despite the prevalence of other common symptoms in residents with dementia, such as nausea, constipation, and dyspnea, we were unable to identify any measures to assess these. With further evaluation, the Respiratory Distress Observation Scale-Family (RDOS-Family) [
79] has potential to be an important measure for detecting dyspnea in long-term care residents with dementia. The original RDOS was designed for cognitively impaired adults unable to self-report but required clinical expertise to administer [
80]. RDOS-Family is family caregiver self-administered based on observations with a 20-minute training provided. It has good inter-rater reliability (ICC = 0.71) between family and trained research assistants when used with patients hospitalized for conditions with dyspnea.
The stringent methodological requirements of the quality criteria and the challenges of conducting research in verbally compromised people with dementia resulted in no measures achieving positive ratings for all psychometric properties in the review. In particular, this review shows that detecting depression in people unable to self-report in this setting is challenging and that caregivers’ use of observational signs may be insufficient to assess depression. Self-report, or a clinician-administered observer-rated scale for those with moderate to severe dementia, has been recommended for assessment of depression in nursing home residents [
81]. The MDS 3.0 takes this approach with the embedded Patient Health Questionnaire-9 Observational Version designed for residents unable to self-report based on observations [
82]. It is completed by trained nurse assessors through interview with a caregiver who knows the resident, thus combining clinician expertise with caregiver knowledge of the resident. The Patient Health Questionnaire-9 Observational Version demonstrated strong correlation (r = 0.84, n = 48) with trained research nurse-administered CSDD [
83].
We included studies conducted in all settings and some measures therefore require further validation in long-term care settings. Where measures do not exist for symptom assessment in long-term care, this review informs selection of measures for further validation by reporting strength of psychometric properties and potential applicability in long-term care settings.
This systematic review identifies and critically appraises measures of common symptoms in the dementia population in long-term care; however, there are a number of limitations. Screening measures are used to detect diagnoses such as depression and anxiety. Studies evaluating screening measures may not have been detected or met the inclusion criteria for this study. Furthermore, the quality criteria used in this review were not developed to evaluate screening tools. However, using the same quality criteria provided consistency of appraisal across the included measures. We limited the study to English language-validated measures only and to publications in English only. This means measures not developed in English, such as the Dutch Rotterdam Elderly Pain Observation Scale [
84], or translated measures, such as the German [
85,
86] and Chinese [
87] versions of the PAINAD, and Dutch [
88,
89] and Italian [
90] versions of the DS-DAT, are excluded. We recognize that the conclusions are therefore limited to English language measures and therefore limited to English-speaking populations and cultures, with the majority of studies conducted in English-speaking countries, predominantly the United States. This means that the most established measures with the strongest international psychometric evidence that have been validated in multiple languages, countries, or cultures are not identified as such. Finally, decisions regarding whether measures met the inclusion criteria required judgement at times. To improve objectivity, those full-texts that did not clearly meet the exclusion criteria were second reviewed and a decision reached by consensus.