Association between self-reported and observational pain measured by PainChek app in residents with dementia

Abstract

Purpose The PainChek application (app) with artificial intelligence is promising in assessing pain in people with dementia (Atee, et al., 2018). It has six domains with 42 items, including facial expressions using automated facial recognition technology, vocalizations, movement, behaviors, activity, and body language observed and documented by assessors (presence = 1, absence = 0, total score 0 - 42), with a score > 6 indicating the presence of pain. While not all individuals with dementia can reliably self-report pain (Hadjistavropoulos, et al., 2014), clinical guidelines advocate individual self-reporting pain (the gold standard of pain assessment) (Herr, et al., 2019). Notably, around 80% of nursing home residents retain the ability to self-report pain (van der Steen, et al., 2021), including some with severe dementia (Pautex, et al., 2006). However, the relationship between self-reported pain and observational pain measured by PainChek remains unclear in people with dementia. Hence, this study aims to explore the potential of PainChek to classify self-reported pain and the association between these two assessments for residents living with dementia in nursing homes who have chronic pain.

Method Data was analyzed from a cluster randomized trial conducted between July 2021 and October 2022. Observational pain was initially assessed using PainChek, followed by self-reported pain via the embedded numeric rating scale (NRS) whenever residents were capable of self-reporting. Linear mixed model analyses were conducted to explore the association between the paired PainChek and NRS score, adjusting for age, gender, timing (rest or movement), cognitive impairment (measured by Mini-Mental State Examination, MMSE), and culturally and linguistically diverse backgrounds.

Results and Discussion A total of 2149 paired pain assessments from 84 residents (mean age 84.9, 70.2% female, mean MMSE 11.6, Mean PainChek score 3.82) were included for analysis. Residents self-reporting pain and without pain had mean PainChek scores of 4.77 (±2.7) and 3.4 (±2.2), respectively (p < 0.001). A weak correlation was observed between the PainChek and NRS score (r =0.32; p < 0.001). Kwon et al. (2021) reported a similar correlation (r = 0.36) between self-reported NRS and the observational Pain Assessment in Advanced Dementia. Controlling for confounding factors and repeated measures, a significant correlation was still observed (β = 1.23, 95% Confidence Interval: 0.99 – 1.48, p < 0.001). Notably, for residents self-reporting pain, the PainChek with a cut-off of six suggested a potential risk of underestimating pain when solely relying on PainChek observations, a concern previously highlighted as the observational pain assessment scale could underestimate pain severity compared with self-assessment scales (Pautex, et al., 2006). Thus, a combination of self-report and observational pain assessments should be concurrently employed to assess pain in individuals with dementia. Although PainChek might be used as a complementary tool supporting caregivers for pain assessment, more research is warranted to establish a reliable cut-off to minimize the risk of underestimation of pain in people with dementia.