Table 2.

Six commonly used nonverbal pain tools for older adults with cognitive impairment

Scale	Description	Psychometric Properties
Checklist of Nonverbal Pain Indicators (CNPI) [322]	An observational scale monitoring pain behaviors in 6 behavioral items (vocal complaints, nonverbal sound, facial grimace/winces, bracing, rubbing, and restlessness) at rest and during movement. An item is rated 0 or 1 based on the absence or presence of a pain behavior. The presence of any of the pain behavior indicates pain. There are no cutoff scores to represent pain severity.	Nursing home residents. Good internal consistency (α = 0.92 to 0.97 at rest; α = 0.74 to 0.90 during movement); Good construct validity but having a great floor effect at rest [323] Good construct validity against NOPPAIN, PACSLAC, and PAINAD (r = 0.66 to 0.71) [324] Interrater reliability for behaviors (K ranged from 0.63 to 0.82) [322] moderate inter-rater reliability at rest (K = 0.43); Fair inter-rater reliability with movement (K = 0.25) [323] Test-retest reliability ranged from 0.44 to 0.56, inter-rater reliability ranged from 0.58 to 0.71, internal consistency α ranged from 0.76 to 0.82, and factor analysis revealed that CNPI might have more than 1 factor [325]. Older patients with hip fracture in a surgical ward internal consistency (α = 0.54 at rest; α = 0.64 with movement) [322]
The Abbey Pain Scale (APS) [326]	For people with end-stage dementia. Comprises 6 questions regarding the facial expression, vocalization, change in body language, behavioral change, physiological change, and physical changes. Each question can be given a score from 0 to 3, where 0 means absence while 3 means severe. Higher total scores indicate higher pain intensity.	The Australian Pain Society has endorsed this scale for evaluating pain in older people with dementia [226]. Nursing home residents. A strong agreement (66.1 to 78.3%) between proxy-reported APS scores and presence of self-reported pain, moderate correlation between self-reported pain intensity and APS (r = 0.56; p < 0.01) residents with cognitive impairment; 25.4% above chance to correctly identify cognitively impaired patients with pain [300]. Concurrent validity between the APS and nurse’s holistic pain assessment was acceptable (Gamma = 0.59; p < 0.01), internal reliability (α = 0.74–0.81), inter-rater reliability was modest (but no actual statistics) [326] Moderate to good construct validity against PACSLAC and PAINAD at rest and exercise (r = 0.56–0.85) [327] Test-retest reliability (r = 0.62–0.68), inter-rater reliability (ICC = 0.70–0.75), internal consistency (α = 0.65–0.80). Factor analysis only revealed 1 single factor [325]. Patients in Geriatric wards It has been translated into Danish and tested on severely demented and non-communicative older patients in geriatric wards. There was a poor agreement between APS and verbal rating scale (k = 0.42), interrater reliability was good (ICC = 0.84). Fair internal consistence (Cronbach’s α = 0.52) [328].
The Doloplus 2 [329]	10-item scale evaluating three domains: (1) somatic, (2) psychomotor, and (3) psychosocial; Each item has four potential scores, where 0 means normal behavior and 3 indicates high levels of pain-related behavior. It is administered by a trained nurse.	It was originally developed in French but has been translated into English. Two systematic reviews rated Doloplus 2 as a scale with high-psychometric properties [226]. Nursing home residents. Internal consistency (α = 0.82–0.87) [325, 330]. Criterion validity between Doloplus-2 score rated by a geriatric expert nurse and pain evaluation conducted by a pain expert (R2 = 0.54); inter-rater reliability (ICC = 0.74–0.77). Small but significant correlation between the expert’s pain in movement score and the Doloplus-2 item for protective body at rest score and for the expert’s pain at rest score (R2 = 0.12; p < 0.01) and between Doloplus-2 item and pain complaints (R2 = 0.13; p < 0.01) [330]. Factor analysis only revealed one single factor [325]. Test-retest reliability (r = 0.71), inter-rater reliability (ICC = 0.73–0.81) [325].
Noncommunicative Patient’s Pain Assessment Instrument (NOPPAIN) [331]	A nursing assistant-administered observation tool for recognizing and rating of extent of pain behaviors. Contains four sections considering six pain behaviors (pain-related words, facial expression, pain noises, rubbing, bracing, and restlessness) during common care conditions (e.g., bathing). Each pain response can be rated from 0 to 5 on a surrogate Likert scale, where 0 indicates the lowest possible intensity and 5 means the highest possible intensity.	The National Nursing Home Pain Collaborative acknowledged the scale in evaluating pain behaviors but reported that the complexity of NOPPAIN might limit its clinical use [225]. It has to be validated in clinical setting. Nursing home setting. Excellent agreement (k = 0.87) for assessing video tape results [300]. Strong agreement (69.2 to 80.0%) between proxy-rated pain behaviors and self-reported presence of pain [300]. Moderate correlation between self-reported pain intensity and NOPPAIN (r = 0.68; p < 0.01) in residents with cognitive impairment. There was 25.4% above chance to correctly identify cognitively impaired patients with pain [300]. Good construct validity against CNPI, PACSLAC, and PAINAD (r = 0.71–0.78) [324, 327]. High intra-rater reliability (k = 0.70–0.86), high inter-rater reliability (k = 0.72–1.0) [324, 331, 332].
Pain Assessment Checklist for Seniors with Limited Ability to Communicate (PACSLAC) [333]	PACSLAC evaluates 60 pain behaviors classified into four subscales: (1) facial expression, (2) social behavior mood and personality, (3) physical activity and body movement, and (4) physiological changes, eating or sleeping changes, and vocal behaviors. PASCLAC-II consists of 31 items by removing items that may be mixed with signs of delirium [334].	Both PACSLAC and PACSLAC-II cover all observational pain assessment domains recommended by the American Geriatrics Society Guideline [301, 335]. Two systematic reviews also suggest PACSLAC as one of the psychometrically strongest assessment tools [226, 227]. Nursing home settings. PACSLAC internal consistency (α = 0.62–0.92) [336, 337]. PACSLAC-II internal consistency (α = 0.74–0.77) [334]. Moderate correlations between PACSLAC scores and global pain intensity ratings (r = 0.39–0.54) [337]. Good correlation between PACSLAC scores and VDS (r = 0.81) and VAS (0.72–0.86) for unblended rating of acute influenza injection [336]. Both PACSLAC and PACSLAC-II have demonstrated good differentiation between painful and non-painful states in patients (p < 0.01) [324, 334]. Good construct validity against NOPPAIN and PAINAD (r = 0.66–0.78) [324]. Good construct validity against APS (r = 0.79) [327]. PACSLAC and PACSLAC-II have strong correlation (r = 0.81–0.89) and the NOPPAIN (r = 0.73) [334]. Inter-rater reliability at rest and during movement (ICC ≥0.76) [299, 324, 327]. Inter-rater reliability (k = 0.63) for PACSLAC-II [334]. Excellent inter-rater reliability (ICC = 0.93–0.96); Intra-rater reliability (ICC = 0.86) for unblended rating of acute influenza vaccination [336].
The Pain Assessment in Advanced Dementia (PAINAD) Scale [228, 338]	A 5-min observation during activity. It evaluates five behaviors (breathing, negative vocalization, facial expression, body language, and consolability) as five indicators of discomfort rated on three levels: 0=absent, 1=present but not constant or severe, 2=severe/constant.	The National Nursing Home Pain Collaborative recommended the PAINAD for clinical use [225]. It has been validated in acute care setting and nursing homes [339]. Nursing home settings. High internal consistency (α > 0.70) [323, 327]. It can detect the presence or absence of pain but not the severity of pain [340]. Strong agreement (66.1 to 73.3%) between PAINAD and proxy-rated pain behaviors or self-reported presence/absence. There was 19.2% above chance to correctly identify cognitively impaired patients with pain [300]. High correlation between PAINAD scores and nurses’ pain reports (Kendall’s τ = 0.84) [341]. PAINAD scores decreased following administration of analgesics and changes with potentially painful activity [324, 336]. Good construct validity with CNPI, APS, NOPPAIN, and PACSLAC at rest and during exercise (r = 0.56–0.90) [324, 327] High inter-rater reliability (r = 0.80–0.97) and test-retest reliability (r = 0.90) [228, 342, 343].