Abstract
BACKGROUND: High symptom burden is common in long-term care residents with dementia and results in distress and behavioral challenges if undetected. Physicians may have limited time to regularly examine all residents, particularly those unable to self-report, and may rely on reports from caregivers who are frequently in a good position to detect symptoms quickly. We aimed to identify proxy-completed assessment measures of symptoms experienced by people with dementia, and critically appraise the psychometric properties and applicability for use in long-term care settings by caregivers. METHODS: We searched Medline, EMBASE, PsycINFO, CINAHL and ASSIA from inception to 23 June 2015, supplemented by citation and reference searches. The search strategy used a combination of terms: dementia OR long-term care AND assessment AND symptoms (e.g. pain). Studies were included if they evaluated psychometric properties of proxy-completed symptom assessment measures for people with dementia in any setting or those of mixed cognitive abilities residing in long-term care settings. Measures were included if they did not require clinical training, and used proxy-observed behaviors to support assessment in verbally compromised people with dementia. Data were extracted on study setting and sample, measurement properties and psychometric properties. Measures were independently evaluated by two investigators using quality criteria for measurement properties, and evaluated for clinical applicability in long-term settings. RESULTS: Of the 19,942 studies identified, 40 studies evaluating 32 measures assessing pain (n = 12), oral health (n = 2), multiple neuropsychiatric symptoms (n = 2), depression (n = 8), anxiety (n = 2), psychological wellbeing (n = 4), and discomfort (n = 2) were included. The majority of studies (31/40) were conducted in long-term care settings although none of the neuropsychiatric or anxiety measures were validated in this setting. The pain assessments, PAINAD and PACSLAC had the strongest psychometric evidence. The oral health, discomfort, and three psychological wellbeing measures were validated in this setting but require further psychometric evaluation. Depression measures were poor at detecting depression in this population. All measures require further investigation into agreement, responsiveness and interpretability. CONCLUSIONS: Measures for pain are best developed for this population and setting. All other measures require further validation. A multi-symptom measure to support comprehensive assessment and monitoring in this population is required.