Using observational facial descriptors to infer pain in persons with and without dementia

BACKGROUND: For patients with advanced dementia, pain diagnosis and assessment requires observations of pain-indicative behavior by others. One type of behavior that has been shown to be a promising candidate is the facial response to pain. To further test how pain-indicative facial responses are, we investigated the predictive power of observational facial descriptors to (i) predict the self-report of pain and (ii) to differentiate between non-painful and painful conditions. In addition, the expertise of the observers (nurses vs. healthy controls) and the cognitive status of the observed (dementia vs. cognitively healthy) were considered. METHODS: Overall 62 participants (32 nurses and 30 control subjects) watched 40 video-clips, showing facial expressions of older individuals with and without dementia during non-painful and painful pressure stimulation. After each clip, participants were asked to rate the videos using commonly used facial descriptors of pain and also to provide global pain estimate ratings of how much pain the observed individual might have experienced. RESULTS: Out of the 12 facial descriptors used, only 7 were able to differentiate between non-painful and painful conditions. Moreover, participants were better in predicting the pain self-report of the observed individuals when using facial descriptors than when using global pain estimates. Especially, the anatomically-orienting descriptors (e.g. opened mouth, narrowing eyes) showed greatest predictive power. Results were not affected by pain-expertise of the observers (nurses vs. control subjects) or diagnostic status of the observed (patients with dementia vs. cognitively unimpaired subjects). CONCLUSIONS: The fine-grained and specific observation of facial responses to acute pain appeared to provide valid indication of pain that is not compromised when patients with dementia are observed. The regular professional training does not put nurses at advantage to detect pain via facial responses.