The Psychometric Qualities of Four Observational Pain Tools (OPTs) for the Assessment of Pain in Elderly People with Osteoarthritic Pain

Article Outline

• Abstract
• Introduction
• Methods
  • Standardized Exercise Program to Trigger OA Pain
  • Participants
  • Measurement
  • Continuous Rater Training
  • Ethical Considerations
  • Statistical Analysis
• Results
  • Participants
  • Comprehension of Self-Reported Pain Scales
  • Internal Consistency
  • Interrater Reliability
  • Capability of the OPTs to Distinguish Pain from the Nonpain Condition
  • Convergent Validity
  • Concurrent Validity
  • Correlation Between OPTs and Proxy-Reported Pain Scales
  • Confirmatory Factor Analysis
• Discussion
  • Similar Levels of Reliability Among Four OPTs
  • Similar Levels of Validity Among Four OPTs
  • Comprehension on Self-Reported Scales in Elderly People
  • Recommendations on PAINAD and Abbey PD
  • Proxy Rating and Observational Rating Measure Pain in Different Ways
  • Limitations
• Conclusion
• References
• Copyright

Abstract 

Context

Pain in cognitively impaired elderly people (CIEP) often goes unrecognized. Observational pain tools (OPTs) have been designed, but with limited evidence to support their psychometric qualities.

Objectives

This study compared four OPTs (the Pain Assessment IN Advanced Dementia [PAINAD], Abbey Pain Scale [Abbey PS], Pain Assessment Checklist for Seniors with Limited Ability to Communicate [PACSLAC], and Discomfort Scale—Dementia of Alzheimer Type [DS-DAT]), two self-report scales, and two proxy-report scales in assessing osteoarthritic (OA) pain among CIEP.

Methods

Participants (n=124) were divided into two groups: cognitively intact and impaired. They were observed by two raters simultaneously at rest and during a standardized exercise program. Besides reliabilities, the correlation between the OPTs and the self-report/proxy-report scores was evaluated. The OPT scores collected during different activity levels were compared to establish the convergent and discriminant validity. Confirmatory factor analysis was used to evaluate the construct validity.

Results

Similar and accepted patterns of reliability/validity were obtained for all OPTs, in which better levels of psychometric properties were consistently obtained during exercise. However, a single construct (OA pain) appeared only in the PAINAD and Abbey PS after deletion of the “breathing” and “physiological change” indicators, respectively. This showed that OPTs were better used to detect OA pain when pain was triggered by movement (i.e., an exercise program).

Conclusion

The PAINAD and Abbey PS appeared to be more reliable and valid for assessing OA pain while using an exercise program among elderly people, regardless of their cognitive ability.

Key Words: Pain behavior, observational pain tool, cognitive impairment, elderly, pain assessment, psychometric, nursing homes

Back to Article Outline

Introduction 

Because of the complex and subjective nature of pain, accurate pain assessment for elderly people, especially for those who are cognitively impaired, is a major obstacle to successful pain management. During the last decade, many researchers have developed observational methods of assessing pain among cognitively impaired elderly people (CIEP).¹, ² However, self-reporting and proxy-reporting methods are still the most common assessment methods for CIEP.³, ⁴, ⁵

Completion rates of CIEP on self-report scales have been varied, ranging from 7% to 100%⁶, ⁷, ⁸, ⁹, ¹⁰ and decreasing with participants’ impaired levels of cognition. Among different pain scales, the verbal rating scale (VRS) has consistently had the highest completion rates, whereas the visual analog scale (VAS) has consistently had the lowest. However, it cannot be assumed that CIEP can correctly use the scales purely based on the scale completion rates. When correlation coefficients (r) between different pain scales were calculated, strong and significant correlations (r=0.5–0.77) could be identified among pain scales when they were used by elderly people with no to moderate cognitive impairment. However, there was no correlation among pain scales when they were used by those with severe cognitive impairment.⁷ The significant correlations among pain scales suggested that the pain scales were measuring the same construct when used among those with no to mild impairment. In contrast, the lack of correlation among pain scales used by severely CIEP suggests that this group of people is unable to comprehend the concept of quantifying pain to a scale item, leading to poor consistency between pain scales. Hence, steps should be included to explore CIEP's comprehension of the self-reporting scales before using these scales for pain assessment. Although many CIEP can give pain scores, it has been shown that almost one-fifth of severely CIEP were unable to use any scale.⁸

An alternative to self-report for CIEP is commonly proxy-report by a caregiver. These include different types of pain intensity scales that were originally designed for self-report. Literature has shown that proxy ratings may be of some value when dealing with noncommunicative patients, although health care professionals tend to underestimate the presence of pain and there may be poor interrater reliability.⁹ Although there may be a reasonable level of agreement (70%) between nursing staff and patient ratings in identifying the presence of pain, estimates of pain intensity may be poor when compared with the self-report.⁹, ¹⁰

Because CIEPs’ pain experiences may not be fully reflected by self- or proxy-report, observational pain tools (OPTs) have recently been developed for this special group of people. Their structure relies on common pain-related behavioral indicators, for example, facial expressions, body movements, verbalizations, vocalizations, physiological changes, emotional changes, and changing patterns of activities of daily living (ADL). By observing for the presence or absence of the behavioral indicators, an observer identifies the likely presence of pain in a patient. Appraisals of OPTs have been based mainly on their psychometric properties, rather than comparing them directly in a clinical setting.¹¹, ¹², ¹³, ¹⁴, ¹⁵ It is unclear which OPTs should be used for pain assessment among CIEP, because all appraisals point out that the OPTs are still under development and show only moderate psychometric qualities. Therefore, further validation of the OPTs in different clinical settings is necessary.

The limitations of previous validating studies include insufficient sample size to represent the target group,¹⁶ unsatisfactory methodology to avoid expectation bias from raters,¹⁷ incomplete psychometric properties, and a lack of effort to investigate how various levels of cognitive impairment affect the psychometric properties of the OPTs. Additionally, most validating studies have not clearly defined the types of pain that the OPTs are intended to measure. It is argued that most OPTs’ poorly defined measure construct leads to inadequate strategies to control for samples’ type of pain. This may explain the relatively low OPT scores reported by previous studies.¹⁶

This study, therefore, aimed to evaluate the psychometric properties of four OPTs (the Pain Assessment IN Advanced Dementia [PAINAD] Scale, the Abbey Pain Scale [Abbey PS], the Pain Assessment Checklist for Seniors with Limited Ability to Communicate [PACSLAC], and the Discomfort Scale—Dementia of Alzheimer Type [DS-DAT]) in a real clinical setting among nursing home residents with osteoarthritic (OA) pain, using a standardized exercise program to trigger OA pain. Research questions included:

Back to Article Outline

Methods 

An observational study was used to answer the research questions. Raters A, B, and C observed participants’ behavioral responses by using the four OPTs simultaneously and independently during two conditions: at rest and during a standardized exercise program. Observation of rest lasted for approximately five minutes, whereas standardized exercise lasted approximately 20–30 minutes. Both rest and exercise were observed in the same morning by three raters. Pain rating on the OPTs was performed after the completion of each observation. Rater A completed all four OPTs for rest and exercise, whereas Rater B completed two of four OPTs for rest and the remaining two OPTs for exercise. Rater C performed a similar procedure to Rater B but with a reverse order of the OPTs. The OPTs used by Raters B and C were selected randomly. The sequence of the administration of the OPTs was selected randomly for all the raters to minimize the possible bias because of order effects. This special arrangement was implemented to minimize possible raters’ expectation bias that might have been produced by the same rater completing the same OPTs for rest and exercise. Thus, the interrater reliability was established by comparing the OPT scores collected by Rater A to those of either Rater B or C depending on the sequence in which the OPTs were completed. This study hypothesized that levels of pain during the standardized exercise program would be higher than when participants were at rest, and that they should exhibit more pain-induced behaviors during exercise. Rater A, the principle investigator of this study, was the only rater aware of the hypothesized higher pain levels during the exercise program. To avoid the possible expectation bias produced by Rater A, the hypothesis test was calculated based on OPT scores obtained by Raters B and C during rest and exercise. These two raters were unaware of the hypothesis and thus, together with the special arrangements for pain score comparison, the expectation bias by the raters was minimized.

Another research assistant, who was not involved in the administration of the OPTs, asked the participants to rate their current pain on two self-reporting scales (i.e., a five-point VRS and the Wong-Baker FACES Pain Scale) after each observational rating. This research assistant also asked the participants’ direct contact nurses to rate the pain levels of their clients on two proxy-reported pain scales (i.e., the proxy-VRS and the proxy-FACES).

Standardized Exercise Program to Trigger OA Pain 

All participants were confirmed as having OA pain by screening the medical notes. OA pain is often associated with joint activity and relieved by rest.¹⁸, ¹⁹, ²⁰ Following literature review, a set of passive range of motion (ROM) exercises was designed, which included all major joints. The participants were then requested to do a defined set of active motions that progressed from lying to sitting, to standing, and then to a walking exercise. These active motions have been shown to be successful in triggering pain-related behavioral responses.²¹, ²²

Participants 

A convenience sample of 124 residents was recruited from 14 nursing homes in Hong Kong (HK). Residents had to be 65+ years of age, of either gender, and living in the nursing home for more than three months before data collection. They must have been diagnosed as having osteoarthritis for at least six months to ensure that this was a chronic condition.

Residents were excluded if: 1) they had any conditions known to influence OA pain sensation, for example, cancer pain or acute wound pain; 2) they had severely impaired hearing and sight that may have affected their ability to use the self-report scales; and 3) they were experiencing any distressing social circumstances, for example, the recent death of a close relative. Residents who met all the inclusion criteria but who had other kinds of chronic illnesses were still recruited into this study, providing the illnesses were neither acute nor likely to affect their pain sensation.

Measurement 

A data sheet was used to collect demographic information. The medical history, such as medications and dementia diagnosis, was obtained from participants’ medical and nursing records.

Participants’ cognitive status was evaluated using the Cantonese version of the Mini-Mental State Examination (C-MMSE).²³ The C-MMSE was adapted from the original MMSE²⁴ and was found to have good psychometric qualities in the identification of cognitive impairment in Chinese patients in HK.²³ In this study, a C-MMSE score of 18 or less was used to identify cognitive impairment.

Pain was assessed with four English-version OPTs: the PAINAD, the Abbey PS, the PACSLAC, and the DS-DAT (Table 1). The selection of these four OPTs was based on the critical appraisal of the 12 OPTs, which is described in detail elsewhere.²⁵ These four OPTs received high scores in the appraisal and were chosen for direct comparison in a clinical setting. All original OPT creators gave permission to use the OPTs and were consulted about the practical use of each OPT.

Table 1. Behavioral Items of Each OPT

The PAINAD consists of five pain behaviors: breathing, negative vocalization, facial expressions, body language, and consolability. Each behavior is rated from 0 to 2 according to the surgery of behavior exhibited, thus giving a total score of 10.¹⁶ The PAINAD was tested on different groups of patients, including posthip surgery elderly patients²⁶ and nursing home residents.¹⁶, ²⁷, ²⁸ The findings showed that the PAINAD could distinguish pain from other pain-free conditions.¹⁶ Its correlation with other pain tools varied from 0.34 to 0.91, but consistently strong interrater reliability and internal consistency was obtained in various studies.¹⁶, ²⁶, ²⁷, ²⁸ Generally, preliminary data provide support for validity and reliability in long-term care and acute settings. However, research is still needed to determine whether similar results can be obtained from other settings.

The Abbey PS includes six pain behaviors, each rated on a 0–3 point scale for the severity of the behavior, with a total score ranging from 0 to 18. The total score is interpreted as severity of pain: 0–2=no pain; 3–7=mild pain; 8–13=moderate pain; and 14+=severe pain.¹⁷ However, this interpretation has been criticized as lacking the support of empirical evidence.¹² In the original design, the rater is asked to indicate which type of pain the elderly person has: chronic, acute, or acute on chronic. This item was deleted, as this study recruited only participants with OA pain. The Abbey PS was able to detect the change of pain level before and after pain-relieving interventions and had a moderate level of correlation with nurses’ proxy-pain scores.¹⁷ With the exception of internal consistency (α=0.84), other types of reliability have not yet been tested.¹⁷ The Abbey PS is an undervalidated OPT with limited available psychometric findings, although it is recommended by the Australian Pain Society²⁹ and the British Geriatrics Society.³⁰ Thus, its psychometric qualities warrant further exploration.

The PACSLAC is a checklist with 60 behavioral indicators divided into four subscales (i.e., facial expression; activity/body movements; social/personality/mood; and physiological indicators/eating and sleeping changes/vocal behavior). Each indicator is scored on a dichotomous scale. Subscale scores are added to arrive at a total score ranging from 0 to 60. In a retrospective study, the PACSLAC was able to distinguish pain events from distressing or calm events. It was moderately correlated with a proxy 10-point scale. High levels of internal consistency were found for the total scores, although the Cronbach’s alpha for the subscales was moderate to lower. Therefore, the OPT creators recommended use of the total score only.³¹ In a prospective study, the PACSLAC showed good levels of interrater and intrarater reliability. It correlated well with proxy-VAS and self-reported VRS scores.²⁸ These correlations were better than those collected in the retrospective study. There are contrasting opinions about the appropriateness of the length of the PACSLAC. The creator of this OPT suggested that a long OPT was better than a short one, as a comprehensive tool is more sensitive in detecting pain, especially when pain behaviors are diverse and individualized. Thus, a short OPT may be unable to identify pain among people exhibiting specific pain-related behavior.² However, another study reported that 28 indicators in the PACSLAC were not observed for over 90% of the participants.²⁸ This shows that some indicators may be redundant and refinement may be needed.

The DS-DAT is a nine-item observational scale for assessing general discomfort. Besides common pain-related behaviors, it also includes two items with negative meanings regarding discomfort (i.e., “content facial expression” and “relaxed body language”). A rater scores each item on a scale of 0–3, based on the frequency, intensity, and duration of each scale item. Scores are summed, resulting in a possible range from 0 to 27.³² When febrile episodes were defined as discomfort situations, this OPT was able to detect significant differences between discomfort and nondiscomfort situations. It also demonstrated acceptable levels of interrater reliability, test-retest reliability, and internal constancy.³², ³³, ³⁴ Although, conceptually, discomfort encompasses pain, discomfort also includes symptoms that may not be an expression of pain, such as fever. In many circumstances, the terms pain and discomfort are used interchangeably. However, no conclusive result as to whether the DS-DAT could retain its psychometric quality was used for chronic pain assessment. It is worth evaluating the appropriateness of using the DS-DAT to identify pain alone.

Two pain scales commonly used in local clinical settings were chosen for assessing participants’ self-reported pain levels. Studies have suggested that CIEP find self-report scales using written words easiest.⁷, ⁸ Thus, a 5-point VRS was selected. However, it is estimated that 31% of local people aged 65 years and older have no schooling and may be illiterate.³⁴ Therefore, the Wong-Baker FACES Pain Scale was also used in this study. There was a high correlation between the FACES and the VAS, in which the correlation coefficients (r) ranged from 0.88 to 0.92,³⁵, ³⁶ suggesting that the FACES was a valid self-report scale for elderly people.

It cannot be assumed that self-reported pain scores are accurate simply because they have high completion rates among the CIEP. The ability of participants to translate their subjective pain experience into an objective pain score on both the VRS and the FACES was assessed by the following questions/instructions: 1) Which descriptor/facial expression describes the worst pain? 2) Which descriptor/facial expression describes the least pain? 3) Put the descriptors/facial expressions in sequence from no pain to the worst possible pain. A correct response on one question was equivalent to one point. For instance, if participants could put all five items in the 5-point VRS in the correct sequence, they obtained a comprehension score of 5. If they could also correctly identify the scale items that described the worst and the least pain, they could obtain two more comprehension points, making a total comprehension score of 7 for the 5-point VRS. Similarly, the total comprehension score was 8 for the 6-point FACES. The level of understanding on using the pain scales between participants with and without cognitive impairment was compared. This assessment was conducted during the predata collection interview.

Two proxy-pain scales (i.e., the proxy-VRS and the proxy-FACES) were used in this study as an additional source in an attempt to capture information about participants, especially those unable to complete the self-report scales. The designs of these two pain scales were identical to the original self-report versions. In previous studies, proxy-reported pain scales have demonstrated a reasonable agreement (70%) between nursing staff and patients’ ratings in identifying the presence of pain, but estimates of pain intensity were quite poor when compared with those of self-reporting scales.⁹, ¹⁰ In this study, the participants’ direct contact nurses were asked to score on the proxy-rating scales based on their overall perceptions of the pain levels of their clients during the previous week. We reminded the nurses to recall the pain levels that the elderly residents might have during activities such as morning exercise, bathing, transferring, etc. We used this method to collect the proxy-pain scores, as this is a common practice. We intended to correlate these with the OPT scores.

Continuous Rater Training 

Three trained raters, JYWL and two research assistants, used the English-version OPTs to assess participants’ pain levels. The use of English documents is common for health professionals in HK clinical settings. Training sessions were organized to make sure that all raters were able to use all OPTs in a correct and consistent manner. The objective of the training was to learn the techniques of delivering and scoring the OPTs. Raters practiced using the OPTs with various clinical vignettes and real nursing home residents. Discussion sessions were held after each practice to clarify the selection of pain-related indicators. The rater training was continued throughout the period of data collection to ensure the correct standard of use of the OPTs.

Ethical Considerations 

This study was given ethical approval by the School of Nursing of the Hong Kong Polytechnic University. Permission to conduct the study also was obtained from the managing directors of the nursing homes. Informed consent from either participants and/or their close relatives was obtained before data collection.

Statistical Analysis 

The data were analyzed using the Statistical Package for the Social Sciences 14.0 (SPSS Inc., Chicago, IL) for most of the analyses. Descriptive statistics were used to evaluate the demographic data of the participant, the pain scale comprehension scores, and the pain scores obtained in different conditions by different pain scales. For the inferential analysis, a value of P>0.05 was considered statistically significant when testing for associations. CFA was undertaken using the LISREL 8.7 (Scientific Software International, Inc., Lincolnwood, IL).³⁷, ³⁸, ³⁹

Back to Article Outline

Results 

Participants 

Sixty-two residents with C-MMSE scores above 18 were assigned to the cognitively intact group, whereas another 62 residents with scores of 18 or below were assigned to the cognitively impaired group. The mean age for the entire group was 85.05±6.85 (range 70–102 years), of whom 119 (96.8%) were female and 4 (3.2%) were male. Demographic data are shown in Table 2.

Table 2. Descriptive Information on Demographic Data for All the Participants

Participant Characteristics	Intact Group (n=62)	Impaired Group (n=62)
General demographic data
Agea (mean±SD)	83.02±6.01	87.06±7.01
C-MMSE scorea (mean±SD)	22.71±2.86	9.97±5.42
Years of educationa (mean±SD)	3.36±4.63	0.57±1.73
Length of stay in institutions (mean±SD)	5.00±2.81	5.07±3.36
Female:male ratio	60:2	60:2
History of OA
Years of suffering from OA (mean±SD)	8.19±8.41	7.93±5.07
Number of OA joints (mean±SD)	1.52±0.70	1.56±0.72
Prescription of analgesics (n (%))	40 (64.5%)	36 (58.1%)

A two-tailed Mann-Whitney U test indicated significant differences between the cognitively intact and impaired groups according to age (z=−3.15, P<0.001), C-MMSE score (z=−9.57, P<0.001), and years of education (z=−5.30, P<0.001). An unpaired t-test (t=0.25, P=0.91) showed no significant difference between these two groups in the length of participants’ institutionalization. In general, the characteristics of the cognitively intact group were different from those of the cognitively impaired group. Except for “the length of stay in institutions,” there were significant differences in all other items in the general demographic data. However, they had similar histories of OA pain. The period of suffering from OA pain and the number of affected joints showed no significant difference between these two groups when compared with the Mann-Whitney U test (z=−1.45, P=0.15 and z=−0.38, P=0.70, respectively). The cognitively intact group appeared to have been prescribed slightly more analgesics than the cognitively impaired group but, on testing, no significant difference could be identified (χ²=0.31, P=0.58). About 80% of analgesics were prescribed as “as needed” medications for all participants.

Throughout the data collection period, only one participant from the intact group dropped out. The reason for this was that the participant felt unwell during data collection.

Comprehension of Self-Reported Pain Scales 

Fifty-five (88.7%) and 17 (27.4%) participants in the intact group and the impaired group, respectively, were able to complete the VRS assessment. The mean score of the VRS comprehension was 6.00±1.35 and 3.06±2.66 out of a maximum of seven in the intact and impaired groups, respectively. The Mann-Whitney U test showed a significant difference between the groups in their ability to quantify their subjective pain experience in objective pain descriptors (z=−4.13, P<0.0001 [two-tailed]). The completion rates for both groups improved for the FACES comprehension assessment, in which all participants in the intact group and 48 (77.4%) participants in the impaired group could complete the assessment. The mean scores for the FACES comprehension were 7.15±1.16 and 6.69±1.68 in the intact and impaired groups, respectively, out of a maximum of eight. There was no significant difference between these two groups in understanding the FACES scale (z=−1.24, P=0.215).

Internal Consistency 

For each of the OPTs and the subscales of the PACSLAC, the Cronbach’s alpha was calculated to examine the homogeneity of the tool items. The calculations were based on the entire group, the intact group and the impaired group OPT scores collected by Rater A vs. Raters B/C at different activity levels (Table 3). The lowest alphas were found for the physiological/eating/sleeping/vocal behaviors (ranging from 0.20 to 0.49) and social/personality/mood (ranging from 0.42 to 0.65) subscales of the PACSLAC. Fortunately, the Cronbach’s alpha for the total PACSLAC scale was above 0.7. This finding is similar to the initial validating study of the PACSLAC, and it is, therefore, recommended to use only the total PACSLAC score.³¹ Most of the Cronbach’s alphas calculated based on total OPT scores for the entire group were above 0.7, which indicated that the OPTs had adequate levels of internal consistency. In other words, all items within an OPT were measuring a similar characteristic.

Table 3. Internal Consistency (Cronbach’s α) of PAINAD, Abbey PS, PACSLAC, and DS-DAT

	Rest			Exercise
	Total Group (n=123)	Intact Group (n=61)	Impaired Group (n=62)	Total Group (n=123)	Intact Group (n=61)	Impaired Group (n=62)
Scores taken by Rater A
PAINAD total	0.70	0.71	0.73	0.72	0.70	0.73
Abbey PS total	0.75	0.70	0.80	0.75	0.73	0.72
PACSLAC total	0.77	0.73	0.78	0.76	0.75	0.77
Facial expressions (n=13 items)	0.65	0.67	0.60	0.71	0.69	0.70
Activity/body movement (n=20 items)	0.51	0.50	0.52	0.68	0.54	0.60
Social/personality/mood (n=12 items)	0.50	0.45	0.40	0.55	0.52	0.53
Physiological/eating/sleeping/vocal behaviors (n=15)	0.34	0.20	0.35	0.44	0.30	0.41
DS-DAT total	0.74	0.69	0.76	0.79	0.71	0.74
Scores taken by Raters B/C
PAINAD total	0.72	0.72	0.70	0.72	0.71	0.71
Abbey PS total	0.75	0.71	0.73	0.75	0.74	0.71
PACSLAC total	0.79	0.72	0.75	0.74	0.75	0.70
Facial expressions (n=13 items)	0.62	0.60	0.59	0.73	0.65	0.70
Activity/body movement (n=20 items)	0.55	0.57	0.52	0.57	0.50	0.66
Social/personality/mood (n=12 items)	0.54	0.42	0.44	0.65	0.56	0.59
Physiological/eating/sleeping/vocal behaviors (n=15)	0.29	0.23	0.33	0.49	0.33	0.45
DS-DAT total	0.83	0.79	0.76	0.80	0.79	0.76

Interrater Reliability 

In general, moderate to good levels of interrater reliability were obtained when using all four OPTs to measure pain among the participants, regardless of their cognition (Table 4). The values of the interclass correlation coefficient (ICC) collected during exercise ranged from 0.78 to 0.87 and from 0.82 to 0.90 for the intact and impaired groups, respectively. These ICCs were slightly higher than those collected at rest (i.e., ICC=0.66–0.67 for the intact group; ICC=0.80–0.87 for the impaired group).

Table 4. Interrater Reliability of PAINAD, Abbey PS, PACSLAC, and DS-DAT

	Total Group (n=123)		Intact Group (n=61)		Impaired Group (n=62)
OPTs	Rest	Exercise	Rest	Exercise	Rest	Exercise
PAINAD	0.80	0.90	0.66	0.81	0.87	0.90
Abbey PS	0.78	0.86	0.66	0.87	0.83	0.88
PACSLAC	0.75	0.81	0.68	0.78	0.80	0.82
DS-DAT	0.79	0.85	0.67	0.85	0.83	0.85

Capability of the OPTs to Distinguish Pain from the Nonpain Condition 

The mean pain scores of all pain scales collected at different activity levels among the intact and impaired groups are listed in Table 5. Pain scores obtained during the exercise program were consistently higher than those obtained when the participants were at rest. Based on the OPT scores collected by Raters B/C, who were unaware of the hypothesis of this study, the results of the Wilcoxon signed-rank test indicated that all pain scales (i.e., four OPTs and the two self-reported pain scales) produced statistically significantly higher pain scores during the exercise program than when participants were at rest for both groups. This is evidence that the OPTs were able to distinguish pain from the nonpain condition. As there was only one type of proxy-reported pain score (i.e., the general pain level of the participants rated by their direct nurses), proxy-report pain scales were not included in this analysis.

Table 5. Mean (SD) of Pain Scores from All Pain Scales and Wilcoxon Signed-Rank Test Compared Pain Scores Obtained During Rest and Exercise

	Intact Group			Impaired Group
OPTs	Rest, Mean (SD)	Exercise, Mean (SD)	z	Rest, Mean (SD)	Exercise, Mean (SD)	z
PAINAD	1.00 (1.04)	3.06 (1.87)	−5.96a	0.76 (1.37)	5.43 (2.00)	−6.70a
Abbey PS	2.05 (1.23)	4.61 (2.56)	−5.63a	2.82 (1.68)	7.69 (3.21)	−6.70a
PACSLAC	1.85 (1.74)	6.52 (4.16)	−5.86a	3.34 (2.82)	13.64 (6.02)	−6.71a
DS-DAT	0.55 (3.27)	4.82 (3.95)	−5.78a	0.40 (5.25)	9.98 (3.81)	−6.74a
VRS	0.63 (0.70)	1.88 (0.86)	−5.46a	0.86 (0.77)	2.13 (0.83)	−2.91b
	∼n=55			∼n=17
FACES	1.05 (1.06)	2.65 (1.12)	−6.31a	1.51 (1.33)	3.08 (1.19)	−4.69a
	∼n=61			∼n=48

SD=standard deviation; ∼n = sample size of the analysis.

Convergent Validity 

The Spearman correlations among the four OPTs were established according to different raters, participants’ cognitive impairment levels and different activity levels. As expected, the results of r_s showed that all OPTs were significantly correlated with each other under different conditions (Table 6). This is evidence of convergent validity, showing that all OPTs were measuring similar domains.

Table 6. Correlation Coefficient Among OPTs

	Rest			Exercise
	Abbey PS	PACSLAC	DS-DAT	Abbey PS	PACSLAC	DS-DAT
Intact group
PAINAD	0.72–0.81	0.70–0.71	0.81–0.85	0.68–0.70	0.66–0.70	0.74–0.80
Abbey PS		0.79	0.74–0.81		0.65–0.84	0.56–0.69
PACSLAC			0.61–0.71			0.66–0.72
Impaired group
PAINAD	0.75–0.82	0.62–0.84	0.73–0.91	0.83–0.89	0.82–0.84	0.78–0.82
Abbey PS		0.56–0.90	0.69–0.83		0.73–0.85	0.68–0.79
PACSLAC			0.63–0.88			0.78–0.86

All correlations are significant at the 0.01 level (one-tailed).

Concurrent Validity 

When comparing the OPT scores to the self-report scores collected when participants were at rest, the majority of the r_s reflected poor to fair levels of correlation between pain scores collected during rest (Table 7). During the exercise program, correlations were higher, where r_s ranged from 0.38 to 0.59 and from 0.35 to 0.71 when the VRS was used in the cognitively intact group and the impaired group, respectively. The r_s ranged from 0.51 to 0.64 and from 0.60 to 0.82 when the FACES was used in the cognitively intact group and impaired group, respectively. Among the two self-reported scales, the FACES generally correlated better with the OPTs. As a limited number of participants were able to use the VRS (n=55 and 17 for the intact and impaired groups, respectively), the small sample size reduced the statistical power in detecting the significance of a data set. Therefore, the VRS tended to have lower r_s with the OPTs when compared with the r_s obtained from the FACES.

Table 7. Correlation Coefficients (r_s) of OPT Scores to Self-Reported Scores

	At Rest								Exercise Program
	Intact Group				Impaired Group				Intact Group				Impaired Group
	VRS (n=55)		FACES (n=61)		VRS (n=17)		FACES (n=48)		VRS (n=55)		FACES (n=61)		VRS (n=17)		FACES (n=48)
OPTs	Rater A	Rater B	Rater A	Rater B	Rater A	Rater B	Rater A	Rater B	Rater A	Rater B	Rater A	Rater B	Rater A	Rater B	Rater A	Rater B
PAINAD	0.2	0.26a	0.2	0.31b	0.55b	0.4	0.08	0.31b	0.44	0.53b	0.64	0.61b	0.61	0.71b	0.61	0.64b
Abbey PS	0.42	0.43b	0.51	0.52b	0.38	0.43	0.25	0.31a	0.39	0.38b	0.59b	0.59b	0.76	0.65b	0.74	0.82b
PACSLAC	0.38	0.48b	0.37	0.52b	0.50a	0.23b	0.36a	0.16	0.47	0.41b	0.62b	0.62b	0.35	0.56b	0.77	0.63b
DS-DAT	0.37b	0.32a	0.27a	0.23b	0.26	0.52a	0.05	0.23	0.56	0.59b	0.51	0.64b	0.62	0.59b	0.60	0.66b

For individual OPTs, the Abbey PS correlated well with the FACES when used in the impaired group during the exercise program (r_s=0.74–0.82, P<0.01). The r_s between the other three OPTs and the FACES was slightly lower during exercise. However, all r_s were above 0.6, which showed good concurrent validity between these pain assessment methods when using them to assess pain during exercise among cognitively impaired participants. In the cognitively intact group, comparable levels of r_s were identified between all OPTs and the FACES, which were 0.61–0.64, 0.59, 0.62, and 0.51–0.64 for the PAINAD, the Abbey PS, the PACSLAC, and the DS-DAT, respectively.

Correlation Between OPTs and Proxy-Reported Pain Scales 

The proxy-reported scores were compared with the OPT scores obtained during different levels of activity. When compared with the OPT scores collected when participants were at rest, the proxy-reported scores varied between zero and fair levels of correlation. Although the proxy-reported pain scales had slightly higher values of Spearman rank correlation coefficient (r_s) than the OPTs scores collected during the exercise program, most of r_s clustered around fair levels of correlation (Table 8), which indicated a poor correlation between these two types of pain assessment method.

Table 8. Correlation Coefficients (r_s) of OPT Scores to Proxy-Reported Scores

	Exercise Program
	Intact Group				Impaired Group
	Proxy-VRS (n=61)		Proxy-FACES (n=62)		Proxy-VRS (n=61)		Proxy-FACES (n=61)
OPTs	Rater A	Rater B	Rater A	Rater B	Rater A	Rater B	Rater A	Rater B
PAINAD	0.34a	0.26b	0.32b	0.53a	0.41b	0.44a	0.41a	0.36b
Abbey PS	0.29b	0.24	0.48a	0.57a	0.39a	0.30b	0.28b	0.41a
PACSLAC	0.49a	0.4a	0.47a	0.45a	0.4b	0.44a	0.36b	0.37a
DS-DAT	0.17	0.42a	−.43a	0.14	0.47a	0.34b	0.24b	0.38a

Confirmatory Factor Analysis 

CFA was used to identify a pattern of linkage between the observed scores and underlying factors. It was hypothesized that OA pain was more detectable by the OPTs during the exercise program. This assumption was supported by the concurrent validity, in which acceptable levels of r_s were only consistently obtained among pain scores collected during exercise. Thus, the CFA was generated according to the observed scores from each OPT collected during exercise by different raters for estimating the factor structure (i.e., OA). If the OPTs were validly constructed, all items in an OPT should contribute toward the construct (i.e., OA pain). For calculating the CFA, initially all items in an OPT were included in the first CFA (Model 1) to see whether a fit model could be identified (see the footnote in Table 8 of all goodness-of-fit indices and their acceptable levels).

A fit model was identified only in the PAINAD and the Abbey PS after deletion of the “breathing” and “physiological change” items, respectively. Table 9 summarizes the goodness-of-fit indices of the PAINAD and the Abbey PS used by Raters A and B/C during exercise. For the PAINAD, Model 1 for the first CFA included all tool items and showed an unfit model in Raters B/C, with Root Mean Square Error of Approximation (RMSEA)=0.12. Additionally, a relatively low factor loading (0.33) from the item “breathing” was identified by Rater A during exercise. Hence, Model 2, with the exclusion of the “breathing” item, was used to run the second CFA. Finally, Model 2 fit all conditions with acceptable levels of all goodness-of-fit indices and item loadings (0.45–0.90). Additionally, the Cronbach’s alphas were 0.72 for both Rater A and Raters B/C during exercise (Table 3). After deletion of the item “breathing,” the Cronbach’s alpha values had slightly improved to 0.75 and 0.76 for Rater A and Raters B/C, respectively. These findings echoed the results in the CFA.

Table 9. Goodness-of-Fit Indices of CFA of the PAINAD and the Abbey PS

OPTs	Conditions	Model	χ2 (P-value)	RMSEA (90% CI)	GFI	AGFI	ECVI
PAINAD	Rater A exercise	1	3.13 (P=0.68)	0.0 (0, 0.10)	1	0.99	0.21
		2	1.20 (P=0.55)	0.0 (0.0, 0.16)	1	0.99	0.15
	Raters B/C exercise	1	13.30 (P=0.02)	∞0.12 (0.04, 0.20)	0.98	0.95	0.28
		2	3.71 (P=0.16)	0.08 (0.0, 0.22)	0.99	0.97	0.17
Abbey PS	Rater A Exercise	1	26.84 (P=0.002)	∞0.13 (0.074; 0.18)	0.98	0.95	0.42
		2	13.41 (P=0.02)	0.069 (0.043, 0.20)	0.99	0.96	0.27
	Raters B/C Exercise	1	25.23 (P=0.003)	∞0.12 (0.067, 0.18)	0.98	0.95	0.40
		2	4.17 (P=0.53)	0.0 (0.0, 0.11)	1	0.99	0.20

Model 1 includes all tool items.

Model 2 excludes tool item “breathing” and “physiological change” in the PAINAD and the Abbey PS, respectively.

χ² (Minimum fit function Chi-square) with no significant difference at P-values>0.05=a good fit of the model.

RMSEA with value<0.05=good fit; between 0.05 and 0.08=reasonable fit; >0.08=poor fit.

GFI (Goodness-of-Fit Index) and AGFI (Adjusted Goodness-of-Fit Index) with GFI>0.90 and AGFI>0.80=a good fit of the model.

ECVI (Expected Cross-Validation Index), the smaller the better, used for comparing independence model.

∞=unacceptable level of GFI (i.e., evidence of unfit model).

CI=confidence interval.

For the Abbey PS, Model 1 for the first CFA included all tool items and showed an unfit model in Rater A and Raters B/C during exercise, with RMSEA=0.13 and 0.12, respectively. Additionally, an unacceptably high factor loading for the item “physiological change” was identified in Raters B/C. In view of this, Model 2 with the exclusion of this item was used to run the second CFA. Eventually, Model 2 fit all conditions with acceptable levels of all goodness-of-fit indices and item loadings (0.65–0.92). The Cronbach’s alphas were 0.75 for Raters A and B/C during exercise. After excluding the item “physical changes,” the Cronbach’s alpha was 0.77 and 0.76 for Rater A and Raters B/C, respectively.

No consistency of model could be identified in the DS-DAT. The items “noisy breathing,” “contented facial expression,” and “relaxed body language” continued to be problematic, with low factor loadings toward various combinations of model. The CFA data of the DS-DAT are available from the first author on request.

Ten participants for every free parameter estimated is a widely accepted guideline for estimating the sample size of the CFA.³⁹ Thus, the sample size was sufficient for running the CFA for most of the OPTs in this study, except for the PACSLAC, which contains 60 items. To avoid bias because of an insufficient sample size, the PACSLAC was not examined with CFA.

Back to Article Outline

Discussion 

Our main conclusions from the findings are as follows:

Similar Levels of Reliability Among Four OPTs 

Our findings showed adequate levels of internal consistency for all OPTs (total scores) when they were used at rest or during the exercise program, although the alpha values of two subscales of the PACSLAC were relatively low. This indicated that all pain indicators within an OPT, to a certain extent, were truly measuring the same characteristic (i.e., OA pain in this study). Hence, the other psychometric properties of the OPTs were evaluated based on the total OPT scores.

Our findings also showed good interrater reliability for all OPTs. Closer examination revealed that many OPT scores were close to zero when pain was observed at rest, which may have influenced the agreement between the raters. As a result, the interrater reliability tended to be lower at rest than during exercise. This is evident to show that it is harder to estimate a person’s pain at rest by using OPTs when stimulation such as an exercise program is absent.

Similar Levels of Validity Among Four OPTs 

This study hypothesized that participants should exhibit more pain behaviors during exercise as they should be experiencing higher levels of OA pain. Wilcoxon’s signed-rank test showed that all OPTs produced significantly higher scores during the exercise program for all the participants regardless of their cognition level. This evidence supports the notion that OPTs are more sensitive in detecting OA pain during the exercise program. By this means, OPTs are able to discriminate between pain and nonpain conditions. This is in accordance with the initial validating studies for all the four OPTs.¹⁶, ¹⁷, ³¹, ³² Additionally, our analysis of convergent validity found that all OPTs correlated well with one another, which illustrated that they were measuring a similar construct.

The findings regarding concurrent validity also suggested that all OPTs should be used to measure OA pain during the exercise program. Generally, all OPTs were moderately to highly correlated with self-reported pain scales during the exercise program among all participants. However, when OPT scores were obtained when participants were at rest, they had no to fair correlation with the self-reported scales. Thus, acceptable levels of concurrent validity among different OPTs were only obtained when the OA pain was assessed during the exercise program.

When comparing the two self-reported pain scales, the OPTs tended to correlate better with the FACES. As expected, the number of residents who were able to use the self-reported scales declined with increased levels of cognitive impairment. However, the number of residents who could use the FACES was obviously higher than those who could use the VRS in both groups. The lower correlations between the OPTs and the VRS seemed to be affected by the small sample size, leading to low statistical power to detect the significance of the data. However, this also may have been caused by the poor literacy levels of the participants, which would have affected their understanding of the VRS.

Comprehension on Self-Reported Scales in Elderly People 

The strength of this study is that we examined the participants’ comprehension on both self-reported scales. The findings in the Mann-Whitney U test showed that both groups had a similar understanding of the FACES, but the cognitively intact group seemed to have a better understanding of the VRS than the impaired group. If both methods (i.e., self-reported and observational) are measuring pain, a better correlation between these two methods should be obtained in the group demonstrating better understanding of the self-reported scales, as was found in the previous study. The PAINAD and the self-reported numeric rating scale (NRS) were used to assess post-hip surgery pain during the “likely pain conditions” in 12 cognitively impaired and 13 cognitively intact patients. That study reported that the r_s was 0.915 in the intact group, noticeably higher than the r_s=0.735 in the impaired group when the correlation between the PAINAD and the self-reported NRS was examined.²⁶ In contrast, this study found that higher levels of correlation were obtained in the impaired group, which had a poor understanding of the VRS. This may have been caused by the potential problems of using the VRS among participants with lower literacy levels, regardless of their cognition levels. Comprehension of the VRS is not only affected by one’s cognition but also by one’s literacy. In Hong Kong, illiteracy is very common among elderly people, especially women.³⁴ Among all participants, 38.7% and 83.9% in the cognitively intact and impaired groups, respectively, had no formal education and were illiterate. In this study, comprehension of the self-reported scales was conducted on a one-to-one basis in a relaxed manner before data collection. Whenever a resident was illiterate, the interviewer explained each VRS item in detail. Great efforts were made to minimize the effect of illiteracy affecting participants’ ability to use the VRS. Thus, the participants without cognitive impairment showed good comprehension of the VRS because of these detailed explanations. Therefore, participants in the intact group, with higher cognition, could achieve significantly higher scores to reflect their comprehension of the VRS. Thus, the percentages of participants who were unable to use the VRS were 11.3% and 72.6% for the cognitively intact and impaired groups, respectively, differing slightly from the percentages of illiteracy. During the real data collection procedure, the explanation was not provided in as much detail as during the interview because of time constraints. It was likely that the participants’ ability to use the VRS was hindered by their literacy levels during the real data collection periods. The cognitively intact participants might have pretended that they understood the VRS and attempted to provide a score. The cognitively impaired participants might not have bothered to pretend but simply did not use the VRS, leading to only 27.4% of subjects in the impaired group scoring on the VRS. This is a possible explanation for the diverse correlations between the OPTs and the VRS obtained in this study.

Similarly, inconsistent correlations were identified from the previous studies. A high correlation (r=0.81, P<0.01) between the self-reported VRS and the PAINAD was reported by Zwakhalen et al.²⁸ By contrast, a fair correlation (r=0.304, P≤0.005) between these two scales was reported by Leong et al.²⁷ These studies did not evaluate their participants’ comprehension of the VRS or their literacy levels. Illiteracy may not be common in Western societies, so the VRS has consistently been recommended as a suitable pain scale for assessing pain among elderly people. The current results raise concern about using the VRS among elderly people with lower literacy levels. Besides cognition and literacy, many factors affect elderly people’s comprehension of self-reported scales, such as “time spent on explanation” or “elderly people’s health status and their concentration level.” Hence, the capability of using the self-reported scales should mean more than just being able to point out a score. Elderly people’s comprehension of the self-reported scales should be tested to confirm the most suitable self-reporting pain scale for an individual elderly person. To explore the interrelationships between different factors affecting elderly people’s ability to use self-reported pain scales is beyond the major objectives of this article. However, this area is worth exploring in future studies.

Recommendations on PAINAD and Abbey PD 

The CFA showed that a single factor appeared only in the PAINAD and the Abbey PS after deletion of the indicators “breathing” in the PAINAD and “physiological change” in the Abbey PS. Coincidently, both indicators are related to “physiological changes.” This finding corresponds to the original PAINAD validating study. Exploratory factor analysis showed one main factor that explained 50.1% of variance and a minor factor (i.e., “breathing” indicator alone) that explained an additional 20.6% of variance.¹⁶ Likewise, an item-total correlation below 0.2 was obtained in “breathing” when the PAINAD was used to measure a mild acute pain caused by vaccination. This indicated that this item should be discarded.²⁸ These findings raised a concern that “breathing” might not purely reflect unspecific chronic pain during unpleasant activities,¹⁶ mild acute pain,²⁸ and OA pain triggered by the exercise program in this study. Based on the indicator definitions of the PAINAD, “breathing” is scored when hyperventilation or noisy labored breathing is observed.¹⁶ Those descriptions of breathing seem to be more relevant to the physiological responses of human bodies to intense acute pain. Other physiological responses related to acute pain include increased heart rates, perspiration, and flushing.⁴⁰ These descriptions are used as examples for the “physiological change” item in the Abbey PS and are also more relevant to intense acute pain.¹⁷

Both indicators on the OPTs were observed least often in this study for OA pain triggered by the exercise program. Not surprisingly, similar results were found in other studies evaluating the PAINAD¹⁶, ²⁸, ⁴¹ when pain was not acute. It is unlikely that these items would be commonly observed in elderly people with chronic pain.

Except the CFA, most of the psychometric findings showed that the PACSLAC and the DS-DAT, to a certain extent, were as valid and reliable as the PAINAD and the Abbey PS. However, we still recommend using the PAINAD and the Abbey PS. The findings in this study provide strong evidence that these two OPTs can be used to assess nursing home residents who are suffering from OA pain. Additionally, the standardized exercise program needs to be included in the assessment procedure while using these two OPTs. We suggest that this should be a standard procedure while using the PAINAD and Abbey PS.

Proxy Rating and Observational Rating Measure Pain in Different Ways 

Overall, the OPTs correlated less well with the proxy-reported scales. The values of r_s for proxy-reported and observational ratings in this study tended to be lower than in similar analyses in previous studies. For example, good levels of concurrent validity between the Abbey PS and the proxy-VRS (r=0.59)¹⁷ and between the PAINAD and the proxy-NRS (Kendall’s tau=0.842)²⁷ were reported by previous studies. However, some raters were asked to complete both scales in those studies, and this may have caused expectation bias. In this study, proxy-reported pain scores were collected by asking the direct contact nurses of individual participants. They rated the scores based on their recent contact with the participants. This method aimed to mimic the common practice of proxy ratings used in current clinical settings. We correlated this current practice with the OPT scores and the correlation was poor, suggesting that these two types of pain assessment procedures may be measuring pain in slightly different ways. Assuming that a high intensity of OA pain is triggered by exercise, the pain scores collected, based only on nurses’ overall perceptions, underestimated the OA pain triggered by exercise. This indicates that the pain assessment based on nurses’ overall perceptions may not completely reflect the high levels of movement-triggered pain.

Limitations 

The original measure construct of the DS-DAT (i.e., discomfort) is different from the measure construct (i.e., OA pain) in this study. As a result, no consistency of a single model can be identified in any condition when using the DS-DAT for assessing OA pain. This may be caused by overlapping meanings of pain versus discomfort. “Discomfort” can be caused by negative emotions and/or physical problems, but is not limited to physical pain. The two problematic indicators, “content facial expression” and “relaxed body language,” are unique; they are included only in the DS-DAT and not in other OPTs. Together with another problematic indicator, “breathing,” it is suspected that these three items might reflect not merely physical pain but also discomfort from other causes. The findings of the CFA showed that the DS-DAT may reflect more than just a single factor (OA pain). Without the inclusion of distress conditions other than OA pain in this study, no conclusion can be made on the possible dimensions that were measured by the DS-DAT. In the future, besides physical pain, the OPTs also should be used to measure other nonpain discomforts so as to identify how the DS-DAT may be able to differentiate pain from other types of nonpain discomfort.

Besides being unable to determine whether the PACSLAC contained only a single measure construct by CFA, which limits this study, the potential problems caused by the long item list are the major disadvantage for this OPT. Sixteen indicators in the PACSLAC were not observed at all. Similarly, another validating study reported that 28 indicators of the PACSLAC were not used for over 90% of the study participants.²⁸ This raised a concern about whether it is necessary to have such a long-item list of pain indicators. Thus, refinement is suggested for the PACSLAC.

The standardized exercise program appeared to induce behaviors that were sensitive to movement-stimulated pain, such as facial expressions, body movements, verbalizations, and vocalizations. These pain behaviors are automatic actions⁴² and their occurrence is more or less similar to the reflex action, which is an automatic process. Hence, the findings in the CFA should be viewed as evidence to support the incorporation of the standardized exercise program in an observational pain assessment procedure. The PAINAD and the Abbey PS fit this procedure the most, based on the CFA findings. This should not be taken as evidence to conclude that the PACSLAC and the DS-DAT are inferior.

Because of the use of purposive samples with purely OA pain among nursing home residents, generalization of the findings in the present study is limited to similar populations. The use of the concept of exercise-triggered OA pain may exclude the capture of types of pain that may not be sensitive to movement, such as visceral pain, neuropathic pain, and headache. However, this approach may also be used in other types of pain that are likely to be triggered by movement, such as rheumatic arthritis, different types of musculoskeletal problems, etc. Additionally, most of the participants were female. Thus, it is not certain whether findings would be similar among male residents with similar conditions. Further validation studies of the OPTs are necessary among elderly people with different types of pain conditions.

Back to Article Outline

Conclusion 

The findings of this study provide evidence of the validity and reliability of four OPTs for assessment of OA pain for nursing home residents regardless of their cognition. The CFA provided additional psychometric qualities of the OPTs, especially for the PAINAD and Abbey PS. It demonstrates the success of using the standardized exercise program in the observational pain assessment methods for OA pain. “Exercise-triggered pain-related behavioral responses” seems to be clinically significant,¹², ²⁹ but few attempts have been made to investigate the possibility of a standardized movement protocol. Not until this study did a standard exercise program involve both passive ROM exercise and action motions to trigger OA pain. It is believed that the implementation of the standardized exercise program makes the observational process more homogeneous and enhances further comparison of sufferers’ pain levels. The importance of this study is to establish standards for the use of the PAINAD and Abbey PS.

Back to Article Outline

References 

1. AGS Panel on Persistent Pain in Older Persons . The management of persistent pain in older persons. J Am Geriatr Soc. 2002;50:S205–S224
   • View In Article
   • MEDLINE
2. Hadjistavropoulos T, Herr K, Turk DC, et al. An interdisciplinary expert consensus statement on assessment of pain in older persons. Clin J Pain. 2007;23:S1–S43
   • View In Article
   • MEDLINE
   • CrossRef
3. Labus JS, Keefe FJ, Jensen MP. Self-reports of pain intensity and direct observations of pain behavior: when are they correlated?. Pain. 2003;102:109–124
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (138 KB)
   • CrossRef
4. Sengstaken EA, King SA. The problems of pain and its detection among geriatric nursing home residents. J Am Geriatr Soc. 1993;41:542–544
   • View In Article
5. Weiner D, Peterson B, Ladd K, McConnell E, Keefe F. Pain in nursing home residents: an exploration of prevalence, staff perspectives, and practical aspects of measurement. Clin J Pain. 1999;15:92–101
   • View In Article
   • MEDLINE
   • CrossRef
6. Chibnall JT, Tait RC. Pain assessment in cognitively impaired and unimpaired older adults: a comparison of four scales. Pain. 2001;92:173–186
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (811 KB)
   • CrossRef
7. Closs SJ, Barr B, Briggs M, Cash K, Seers K. A comparison of five pain assessment scales for nursing home residents with varying degrees of cognitive impairment. J Pain Symptom Manage. 2004;27:196–205
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (229 KB)
   • CrossRef
8. Ferrell BA, Ferrell BR, Rivera L. Pain in cognitively impaired nursing home patients. J Pain Symptom Manage. 1995;10:591–598
   • View In Article
   • Abstract
   • Full-Text PDF (708 KB)
   • CrossRef
9. Krulewitch H, London MR, Skakel VJ, et al. Assessment of pain in cognitively impaired older adults: a comparison of pain assessments tools and their use by non-professional caregivers. J Am Geriatr Soc. 2000;48:1607–1611
   • View In Article
   • MEDLINE
10. Weiner D, Peterson B, Keefe F. Chronic pain-associated behaviours in the nursing home: resident versus caregiver perceptions. Pain. 1999;80:577–588
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (92 KB)
    • CrossRef
11. van Herk R, van Dijk M, Barr FPM, et al. Observation scales for pain assessment in older adults with cognitive impairments or communication difficulties. Nurse Res. 2007;56:34–42
    • View In Article
12. Herr K, Bjoro K, Decker S. Tools for assessment of pain in nonverbal older adults with dementia: a state-of-the-science review. J Pain Symptom Manage. 2006;31:170–192
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (222 KB)
    • CrossRef
13. Smith M. Pain assessment in nonverbal older adults with advanced dementia. Perspect Psychiatr Care. 2005;41:99–113
    • View In Article
    • MEDLINE
    • CrossRef
14. Stolee P, Hillier LM, Esbaugh J, et al. Instruments for the assessment of pain in older persons with cognitive impairment. J Am Geriatr Soc. 2005;53:319–326
    • View In Article
    • MEDLINE
    • CrossRef
15. Zwakhalen SMG, Hamers JPH, Huijer Abu-Saad H, et al. Pain in elderly people with severe dementia: a systematic review of behavioural pain assessment tools. BMC Geriatr. 2006;27:3–17
    • View In Article
16. Warden V, Hurley AC, Volicer L. Development and psychometric evaluation of the pain assessment in advanced dementia (PAINAD) scale. J Am Med Dir Assoc. 2003;4:9–15
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (116 KB)
    • CrossRef
17. Abbey J, Piller N, Bellis AD, et al. The Abbey pain scale: a 1-minute numerical indicator for people with end-stage dementia. Int J Palliat Nurs. 2004;10:6–13
    • View In Article
    • MEDLINE
18. Altman RD. Osteoarthritis differentiation from rheumatoid arthritis causes of pain, treatment. Postgrad Med. 1990;87:66–78
    • View In Article
    • MEDLINE
19. Buckwalter JA, Salzman C, Brown T. The impact of osteoarthritis: implications for research. Clin Orthop Relat Res. 2004;427S:S6–S15
    • View In Article
20. Jakobsson U, Hallberg IR. Pain and quality of life among older people with rheumatoid arthritis and/or osteoarthritis: a literature review. J Clin Nurs. 2002;11:430–443
    • View In Article
    • MEDLINE
    • CrossRef
21. Keefe FJ, Block AR. Development of an observation method for assessing pain behavior in chronic low back pain patients. Behav Ther. 1982;13:363–375
    • View In Article
    • CrossRef
22. Weiner D, Pieper C, McConnell E, Martinez S, Keefe S. Pain measurement in elders with chronic low back pain: traditional and alternative approaches. Pain. 1996;67:461–467
    • View In Article
    • Abstract
    • Full-Text PDF (756 KB)
    • CrossRef
23. Chiu HFK, Lee HC, Chung WS, Kwon PK. Reliability and validity of the Cantonese version of the mini-mental state examination—the preliminary study. J Hong Kong Coll Psychiatr. 1994;4:5–28
    • View In Article
24. Folstein MF, Folstein SE, McHugh PR. Mini-mental state: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189–198
    • View In Article
    • Full-Text PDF (635 KB)
    • CrossRef
25. Liu JYW. The assessment of osteoarthritic pain in elderly people: comparing observational pain tools, self-reported scales and proxy-reported scales. Unpublished Doctoral thesis. 2009. The University of Leeds.
    • View In Article
26. DeWaters T, Faut-Callahan M, McCann JJ, et al. Comparison of self-reported pain and the PAINAD scale in hospitalized cognitively impaired and intact older adults after hip fracture surgery. Orthop Nurs. 2008;27:21–28
    • View In Article
    • CrossRef
27. Leong IYO, Chong MS, Gibson S. The use of a self-reported pain measure, a nurse-reported pain measure and the PAINAD in nursing home residents with moderate and severe dementia: a validation study. Age Ageing. 2006;35:252–256
    • View In Article
    • MEDLINE
    • CrossRef
28. Zwakhalen SMG, Hamers JPH, Berger MPF. The psychometric quality and clinical usefulness of three pain assessment tools for elderly people with dementia. Pain. 2006;126:210–220
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (138 KB)
    • CrossRef
29. Gibson SJ, Schere S, Goucke CR. Preliminary field testing and preparations for implementation of the Australian Pain Society and Australian Pain Relief Association Pain Management Guidelines for Residential Care Nov. 2004. Available from www.apsoc.org.au/pdfs/APRA-finalreport7.docAccessed June 1, 2009
    • View In Article
30. Collett B, O’Mahoney S, Schofield P, Closs SJ, Potter J. The assessment of pain in older people. Clin Med. 2007;7:496–500
    • View In Article
31. Fuchs-Lacelle S, Hadjistavropoulos T. Development and preliminary validation of the pain assessment checklist for seniors with limited ability to communicate (PACSLAC). Pain Manag Nurs. 2004;5:37–49
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (582 KB)
    • CrossRef
32. Hurley AC, Volicer BJ, Hanrahan PA. Assessment of discomfort in advanced Alzheimer patients. Res Nurs Health. 1992;15:369–377
    • View In Article
    • MEDLINE
    • CrossRef
33. Zieber CG, Hagen B, Armstrong-Esther C. Pain and agitation in long-term care residents with dementia: use of the Pittsburgh Agitation Scale. Int J Palliat Nurs. 2005;11:71–78
    • View In Article
    • MEDLINE
34. Census and Statistics Department . Government of Hong Kong Special Administrative Region (HKSAR). Hong Kong population projection 2001–2025 Hong Kong: Hong Kong Government Printers; 2002;
    • View In Article
35. Dallam L, Smyth C, Jackson B. Pressure ulcer pain: assessment and quantification. J Wound Ostomy Continence Nurs. 1995;2:211–218
    • View In Article
36. Freeman K, Smyth C, Dallam L, Jackson B. Pain measurement scales: a comparison of the visual analogue and faces rating scales in measuring pressure ulcer pain. J Wound Ostomy Continence Nurs. 2001;28:290–296
    • View In Article
    • MEDLINE
    • CrossRef
37. Diamantopoulos A, Siguaw JA. Introducing LISREL: A guide for the uninitiated. Bristol: Sage Publication; 2000;
    • View In Article
38. Du Toit M, Du Toil S. Interactive LISREL: User’s guide. Chicago, IL: Scientific Software International, Inc.; 2001;
    • View In Article
39. Schreiber JB, Stage FK, King J, Nora A, Barlow EA. Reporting structural equation modelling and confirmatory factor analysis results: a review. J Educ Res. 2006;99:323–337
    • View In Article
40. Strong J, Sturgess A, Unruh AM, et al. Pain assessment and measurement. In:  Strong J editors. Pain: A textbook for therapists. St. Louis, MO: Churchill Livingstone; 2002;p. 123–147
    • View In Article
41. van Iersel T, Timmerman D, Mullie A. Introduction of a pain scale for palliative care patients with cognitive impairment. Int J Palliat Nurs. 2006;12:54–59
    • View In Article
    • MEDLINE
42. Hadjistavropoulos T, Craig KD. A theoretical framework for understanding self-report and observational measures of pain: a communications model. Behav Res Ther. 2002;40:551–570
    • View In Article
    • MEDLINE
    • CrossRef