Assessing and Treating Chronic Pain in Community-Dwelling Older Adults with Dementia: A Single-Subject Approach

AS SESS ING AND TREATING CHRONIC PAIN IN COMMUNITY-DWELLNG
OLDER ADULTS WITH DEMENTIA: A SINGLE-SUBJECT APPROACH














By

AMANDA FLOETKE ELLIOTT


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2006

































Copyright 2006

by

Amanda Floetke Elliott
















ACKNOWLEDGMENTS

I would like to express my profound appreciation to my dissertation chair and

mentor, Dr. Ann Horgas, for her constant support and dedication to guiding me through

the wonderful world of research. Her direction and faith in me throughout these past five

years have made it possible to complete the journey through this program. I would also

like to acknowledge Dr. Jennifer Elder, Dr. Michael Marsiske, and Dr. Meredeth Rowe

and thank them for their time and support and for sharing with me their knowledge and

expertise throughout the fruition of this research proj ect. Many thanks go to the families

that participated in this study and who welcomed me into their homes each day. I have

learned so much more than I ever expected too. I feel truly blessed to have had the

privilege to know each of them. I would like to thank my family and friends for

everything that they have done to see me through this program. I thank my parents and

brother for constantly supporting me and encouraging me to succeed. I thank my friends

for all of their many different kinds of support; I needed them all. Finally, I especially

thank my husband for putting up with me these last five years and seeing me through this

program, even though it meant a lot of time apart from each other. I truly could not and

would not have made it to here without his love and support.












TABLE OF CONTENTS


page

ACKNOWLEDGMENT S .........._.... .............._ iii.._... .....


LI ST OF T ABLE S .........._.... .............._ viii...__ ....


LIST OF FIGURES .............. .................... ix


AB STRAC T ................ .............. xi


CHAPTER


1 INTRODUCTION ................. ...............1.......... ......


Background and Significance of the Problem ................. .............. ......... .....2
Rationale and Need for this Study ................. ......... ...............6.....
Theoretical Framework.................. .................7
Statement of Problem and Study Specific Aims ................. .......... ................1 1

2 REVIEW OF THE LITERATURE ................. ...............14........... ...


Pain Prevalence ................. ...............14.................
Pain Assessment .............. ...............16....
Self-Report .............. ...............16...
Observational Strategies ................. ............. ...............19.......
Advantages of observational methods............... ...............20
Limitations of observational methods ................. ................ ......... .21
Pain Behaviors ................. ...............22........... ....
Identifying Pain Behaviors ................. ...............23........... ....
Testing Pre-identified Behaviors ............... .... ........... ........... ............2
Underassessment and Treatment of Pain in Persons with Dementia. ................... ......26

Caregiver Assessments ................ .... ....... ............2
Informal Caregivers versus Trained Ob servers ................. ....___ ..............28
Informal Caregivers versus Participant Self-Report ................. ... .. ............ .....29
Strategies to Strengthen the Validity and Reliability of Caregiver Report ................33
Excess Disabilities related to Dementia and Chronic Pain............... ..................3
Pain Management and Treatment Guidelines............... ...............3
Sum m ary ................. ...............41.......... ......

3 M ETHODS .............. ...............42....


Study Design............... ...............42.
Sample .............. ..... .. ...............45
Recruitment Procedures ................... ....... ...... ... ........ .. .. .. .......4
Inclusion and Exclusion Criteria for Participants with Dementia ................... ....46
Inclusion and Exclusion Criteria for Caregivers .............. ....................4












Participant Pair 1, Caregiver 1 (CG 1) and Participant 1 (P 1) .........................47
Participant Pair 2, Caregiver 2 (CG 2) and Participant 2 (P 2) ...........................47
Participant Pair 3, Caregiver 3 (CG 3) and Participant 3 (P 3) ...........................47
Participant Pairs Withdrawn from the Study ....._._._ ........___ ........._.....48
M measures ................ ......... ...... ......... .............4

Comprehensive Intake Assessment ................. ............ ........... ...... .........4
Caregiver Ratings of Pain Behaviors and Excess Disabilities ................... .........49
Trained Observer Ratings of Pain Behaviors ......._.__ .... ........ ..............52
Treatment Protocol .............. ...............55....
Statistical Power ............ ..... .._ ...............56...
Proc edure s............... ....._ ...............57....
Phase A l, Session 1.............. ...............60....
Phase A l............... ...............61...
Phase Al-Final Session............... ...............62
Phase B1 (Eight Sessions). ............. ..... ..._ ................62.
Phase A2- Return to Baseline (Range of 4-8 Sessions) ............. ...............63
Phase B2- Reimplementation of Intervention (Range of 4-8 Sessions) ..............63
Data Coding ............... ... ... ....__ .......__ ............6
Data Analysis Procedures and Variables of Interest. ....._____ ..... ... ..............65
RM BPC .............. ...............65....
PADE .........__ ......... ._ ...............66...
Activity Protocol ................ ...............66...
Additional Caregiver Questionnaire .............. ........__ .....__ ...........6
Overview of the Analysis of Single-Subj ect Design Data ................. ................ ...66
Statistical Analysis of Each Aim ................ ...............71........... ...

4 RE SULT S .............. ...............74....


Descriptives .............. ...............74....
Initial Intake M measures .............. ...............74....
Final Outtake Measures ................. ....... .. ...... .... .. ..... ............ .........7
Effect of Acetaminophen on Self-Reported Pain Intensity and Observable Pain
Behaviors ............... ........ ..............7
Total Number of Pain Behaviors ....__ ......_____ .......___ ...........7
Participant 1............... ...............78...
Participant 2............... ...............80...
Participant 3................ ...............82..
Total Duration of Pain Behaviors ....__ ......_____ .......___ ...........8
Participant 1............... ...............84...
Participant 2............... ...............85...
Participant 3............... .......... ......... .......8
Caregiver and PI Pain Ratings on the PADE .............. ...............89....
Participant 1............... ...............89...
Participant 2............... ...............90...
Participant 3............... ........ ......... ........9
Effect of Acetaminophen on Excess Disabilities. ................ ......................__.93
Total RMBPC Score ................. ...............94........... ....












Participant 1............... ...............94...
Participant 2............... ...............96...
Participant 3............... ...............97...
PADE Part III Subscale Score ....._........___...... .....__ ...........9
Reliability of Caregivers' Ratings of Pain Behaviors .................. ................. ... 100
Pain Behaviors Frequently Displayed by Persons with Dementia .........._...............102
Frequent Pain Behaviors determined by the Coded Activity Protocols ............103
Frequent Pain Behaviors determined by Caregivers' Ratings on Part I of the
PA D E ............. .. ... .. .. ... ......... .. ...........10
Frequent Pain Behaviors reported by Caregivers on the Free Response
Question ............. .............. 104... .....

5 DI SCUS SSION .........__........_. .............. 106...


Effect of Acetaminophen on Self-Reported Pain Intensity and Observable Pain
Behaviors .............. ...............107....
Self-Report of Pain ............ ..... .._ ...............107...
Ob servation of Pain B ehaviors ...._.._.._ ........__. ...._.._ ..........10
Proxy Pain Assessments ........._.._.... ........._._. ......._._.......... 11
Effect of Acetaminophen on Excess Disabilities ......___ .........__ ................113
Reliability of Caregivers' Ratings of Pain Behaviors .................. ................. ... 1 17
Pain Behaviors Frequently Displayed by Persons with Dementia .........................119
Validation of the Study Findings ....__ ........._.... ......... ............2
Treatment Integrity ................. ...............120.............
Validity of the Findings .....__................. ...............121 ....
Nursing Implications .............. ...............123....
Clinical Si gnif icance............... .............12
Lim stations ................. ...............128.............
Recruitm ent .............. ...............128....
Sample Size ........._.__......_.._ ...............129....
Tool Choi ce ........._.__......_.._ ............... 13 0...
Feasibility .............. ............... 13 1...
Future Directions ........._.._.._ ....__. ..............._ 132..


APPENDIX

A PERFORMING GRAPHICAL ANALYSES FOR SINGLE-SUBJECT DESIGN
DATA ............ ...... ...............137..


B INFORMED CONSENT TO PARTICIAPTE IN RESEARCH-CHRONIC PAIN
PARTICIPANT .............. ...............143....

C INFORMED CONSENT TO PARTICIPATE IN RESEARCH-CAREGIVERS ....154


D DAILY CAREGIVER QUESTIONNAIRES ......____ ..... ... ._ .............. ..163

LIST OF REFERENCES ........._... ...... ..... ...............169...












BIOGRAPHICAL SKETCH ................. ...............180......... ......


















LIST OF TABLES

Table pg

1-1 Theoretical and Study Constructs .............. ...............12....

3-1 Study Constructs and Measures .............. ...............53....

3-2 Study Design .............. ...............60....

3-3 Behavioral Definitions .............. ...............67....

4-1 Descriptive Characteristics for Participants .............. ...............76....

4-2 Descriptive Characteristics for Caregivers ................. ....................___......76

4-3 Mean Duration (in Seconds) of Individual Pain Behaviors ........._..... ........_......86

4-4 Caregiver and Trained Observer PADE Part I Correlations .............. .................101

A-1 Calculating the M ean .............. ...............137....

A-2 Calculating the Level Change between Phases .............. ....__. ..............138

A-3 Determining the Latency of Change between Phases ................. .....................140

A-4 Calculating 15% Above and 15% Below the Mean ................. ............ .........141


















LIST OF FIGURES


Figure pg

1-1 Communications model of pain ................. ............. ......... ........ .......9

4-1 Total number of pain behaviors for Pl ................. ...............79........... .

4-2 Mean frequencies for individual behaviors for P 1. ............. .....................8

4-3 Total number of pain behaviors for P 2 ................. ................ ......... ...._81

4-4 Mean frequencies for individual behaviors for P 2. ............. .....................8

4-5 Total number of pain behaviors for P 3............. ......___.....__ ..............83

4-6 Mean frequencies for individual behaviors for P 3. ......................__ ..............84

4-7 Total duration of pain behaviors for P 1. ......................_. ......._..............85

4-8 Total duration of pain behaviors for P 2. .........__ ........... .......___..........87

4-9 Total duration of pain behaviors for P 3. ..._____ ....... .... ..__ ............88

4-10 Caregiver and TO PADE Part I pain ratings for P 1. ............. .....................9

4-11 Caregiver and TO PADE Part I pain ratings for P 2. ............. .....................9

4-12 Caregiver and TO PADE Part I pain ratings for P 3 .............. ....................9

4-13 Frequency of excess disabilities across sessions for P 1. ............. ....................95

4-14 Frequency of excess disabilities across sessions for P 2. .............. ....................97

4-15 Frequency of excess disabilities across sessions for P 3............_ .........._ .....98

4-16 Mean frequencies of excess disabilities related to physical functioning................ 100

4-17 Scatterplot for the CG' s and the TO's total scores on PADE Part I for Pl............101

4-18 Scatterplot for the CG' s and the TO's total scores on PADE Part I for P2............102

4-19 Scatterplot for the CG' s and the TO's total scores on PADE Part I for P3............102










4-20 Total frequency of pain behaviors coded from the activity protocols. ................... 103

4-21 Total frequency of PADE Part I pain behaviors across participants..........._........104

A-1 Presenting the mean for each phase in the graph. ............. ......................3

A-2 Example of a trendline produced in Excel ................ .......... ................ ..139

A-3 Example of calculating the stability of data points within a study phase............... 141
















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

AS SESS ING AND TREATING CHRONIC PAIN IN COMMUNITY-DWELLNG
OLDER ADULTS WITH DEMENTIA: A SINGLE-SUBJECT APPROACH

By

Amanda Floetke Elliott

August 2006

Chair: Ann Horgas
Major Department: Nursing

Previous research on pain assessment in persons with dementia has mostly focused

on nursing home or hospitalized populations, finding that this subgroup of older adults

experiences comparable pain to cognitively intact elders. These studies advocate an

analgesic trial to treat pain in this population as untreated chronic pain can lead to excess

disabilities, such as depression and functional decline. Currently there are pain treatment

guidelines specifically for older adults with dementia. Acetaminophen is the first line

drug recommended in pain treatment guidelines for older adults. This feasibility study

investigated the effects of acetaminophen in reducing pain behaviors and excess

disabilities in community-dwelling persons with dementia. Three participants with

severe dementia and their primary caregiver completed 24 daily sessions that allowed the

researcher to observe for changes in pain behavior over time following an ABAB study

design. During baseline phases (A), pain and excess disabilities were assessed through

the completion of daily measurement tools by trained observers and informal caregivers









and videotaped behavioral observations of the participant performing an activity protocol

(walking, standing, sitting, and lying) designed to elicit pain. Treatment phases (B)

consisted of the administration of the treatment drug, Tylenol Arthritis, 1.3 grams every 8

hours while awake in addition to Phase A procedures. Results show that acetaminophen

effectively reduced chronic pain behaviors in this sample of older adults with dementia.

The frequency of excess disabilities also showed a general decline throughout the study

period. Finally, primary caregivers proved to be moderately reliable proxy informants of

participants' pain.















CHAPTER 1
INTTRODUCTION

Older adults in today's society are living longer than they did in previous

generations. As a result, there is an increased need for understanding the specific

healthcare issues that face this segment of the population. One of the most important, yet

often overlooked, healthcare issues that needs to be addressed is that of proper

assessment and management of chronic pain. Chronic pain, sometimes referred to as

persistent pain, is defined as pain lasting longer than 3 consecutive months in duration.

Pain is a complex phenomenon that affects each individual differently. In order to

understand the exact nature of an individual's pain experience, it is necessary to elicit his

or her own interpretation of this experience. Pain has been described in a variety of

ways, such as "an individual's perception of a sensation which is noxious and

uncomfortable, and one from which escape or relief is sought" (Weissman & Matson,

1999, p. 31) or as "an unpleasant sensory and emotional experience associated with

actual or potential tissue damage, or described of in terms of such damage" (Price, 1988,

p. 6). Perhaps the most common definition of pain, suggested by McCaffery and Beebe

(1989), has been that pain is "whatever the patient says it is and occurs whenever the

patient says it does" (p. 7).

Clearly, each definition recognizes the role of the individual in interpreting the

experience of pain. Most individuals who experience pain can recognize its onset and

can outwardly express this pain, usually through verbal reports. After interpreting their

pain, they can seek pain relief through a variety of pharmacologic and/or










nonpharmacologic forms. Effective pain management can exert a positive influence in

many aspects of an individual's life as pain has been shown to be related to depression

and weight loss, and to interfere with activities of daily living, sleep routines, social

behavior, functional status, and quality of life (Ferrell, 1995;Villanueva, Smith, Erickson,

Lee, & Singer, 2003; Warden, Hurley, Volicer, 2003). However, there are groups of

people, such as those with dementia, in whom verbal reports of pain are either

unobtainable or unreliable. How does this normal encoding and decoding of the pain

experience apply to persons with dementia who, due to the nature of the disease, have

barriers in the ability to express themselves? This question serves as the foundation for

this study.

Background and Significance of the Problem

It has only been within the last 15 or so years that researchers have begun to study

the assessment of pain in persons with dementia in order to find a way to detect

meaningful behaviors indicative of pain (Hurley, Volicer, Hanrahan, Houde, & Volicer,

1992; Villanueva et al., 2003; Warden et al., 2003;). These studies have shown

promising results. Recognizing that the presence of dementia complicates pain

assessment in these persons due to both the cognitive and verbal deficits that accompany

this disease, these researchers have concentrated on identifying observable behaviors

such as facial expressions, body movements and postures, as well as vocalizations that

are considered to be indicative of pain.

The recent progress that has been made in this area has evolved from research

studying pain in general. It should be noted that the study of pain has a relatively short

history itself, gaining popularity only within the last several decades. First, researchers

focused on finding reliable ways to assess pain in the normal adult population. The most









commonly used method of pain assessment in this population is through self-report.

Simple rating scales allow a person to quantify the intensity of his or her internal painful

experience in a measurable way. From these research endeavors, tools such as the

Numeric Rating Scale (NRS), the Verbal Descriptor Scale (VDS), and the Visual Analog

Scale (VAS) have been developed to measure pain. Each of these tools consists of a

scale anchored at 0 (meaning no pain) with increasing numbers that represent increasing

amounts of pain, and in the case of the VDS, verbal descriptors that represent

incrementally higher levels of pain. Tools like these also allow for measuring the

effectiveness of pain treatment by comparing pre-treatment pain intensity ratings to post-

treatment pain intensity ratings.

As is natural in the course of research, once reliable measures of pain were found in

the normal adult population, researchers could begin to expand the study of pain

assessment into other populations and to specific causes of pain. Studies have shown that

both the VDS and NRS are reliable and easy to use with an elderly population (Gagliese

& Melzack, 2003; Herr & Mobily, 1993). Once pain research was expanded to include

older adults, researchers found a common theme in most study results in that the 65+ age

group had the highest self-reported pain prevalence rates (Gibson & Helme, 1995). Yet,

pain in this population is often dismissed as a normal consequence of aging. Recent

studies, however, have indicated that this is not the case (Edwards, Fillingim, & Ness,

2003; Gagliese & Melzack; Harkins, 1996; Helme & Katz, 2003). There is evidence that

some underlying pathology is involved, with osteoarthritis (OA) of articular j points being

one of the most common causes of chronic pain in older adults. In fact, the presence of

OA in adults over age 65 is twice that of younger adults (Gibson & Helme). Clearly,









research efforts in this area have shown that older adults do experience pain and report

higher pain prevalence rates than do younger adults.

Building on these findings, researchers have had the basis to propose that pain may

affect all older adults experiencing a disease process that typically has pain as a

symptom. Even though research studies have found the highest pain prevalence rates in

those 65 and older, most investigators believe that older adults underreport the intensity

of their pain. This belief has been validated recently by Labus, Keefe, and Jensen (2003)

in a review of 30 studies that compared the correlation between self-reported pain

intensity and direct observation of pain behaviors, which resulted in only a moderately

positive relationship (r= 0.26) on average, across the studies. Several theories have been

proposed as to why older adults underreport their painful symptoms. In cognitively intact

older adults this could be due to their diminishment of the importance of mild pain,

attributing it to the normal aging process (Gagliese & Melzack, 2003; Gibson & Helme,

1995) or because of their fear about what an intensification of painful symptoms might

mean (i.e. worsening of a disease process, need for diagnostic testing, or impending

death) (Herr, 2002). These theories led the Joint Commission on Accreditation of Health

care Organizations (JCAHO) (2001) to recognize pain assessment as the 5th vital sign,

requiring healthcare professionals to systematically ask about its presence rather than to

rely on the hospitalized patient to volunteer information. There is one fatal flaw in using

only self-report to measure pain (whether volunteered by the individual or elicited by

another party): this method is not applicable to all populations. For instance, infants can

no better report their pain level than can older adults with the compromised cognitive and

verbal abilities of dementia.









For these reasons, researchers have recently focused efforts on ways to reliably

assess pain in persons with dementia. The most valid and reliable way found has been to

look for observable behavioral indicators of pain. Keefe (1982) modeled this approach

for other investigators when he found that he could reliably assess pain by observing pain

behaviors in people with chronic low back pain, looking specifically for guarded

movements, bracing, rubbing, grimacing, and sighing. Although the participants in his

study were not cognitively impaired, his work in this area allowed researchers to take

these behaviors indicative of pain and to observe for them in persons with dementia. In

fact, Hadjistavropoulos and Craig (2002) reported findings from studies showing similar

behavioral displays of pain observed for both cognitively intact and impaired participants.

As mentioned previously, several researchers (Hurley et al., 1992; Villanueva et al.,

2003; Warden et al., 2003) have found valid and reliable behaviors that indicate the

presence of pain in persons with dementia.

While the majority of these research studies have looked at behaviors in persons

with dementia residing in nursing homes, it is important to study these behaviors in a

community-dwelling sample as well. This is especially true because of the estimated 4.5

million persons with Alzheimer' s disease, more than 70% live at home where nearly 75%

of their care is provided by family and friends (Alzheimer' s Association, 2006). Now

that pain behaviors have been identified in persons with dementia, research efforts need

to shift gears so as to focus on ways to treat this pain. While guidelines exist for pain

treatment in older adults in general, there are no specific guidelines for treating pain in

persons with dementia, largely because there has been little work done in this area.









Rationale and Need for this Study

Due to the questionable use of self-report for pain assessment in persons with

dementia, researchers have developed and employed behavioral observation strategies in

this population with success and have determined that people with dementia do

experience pain. The next logical step is to try to do something about the pain. This is an

area of research, and clinical practice as well, that warrants further development. In

general, pharmaceutical studies have excluded persons with dementia from drug trials,

thus preventing this group of vulnerable elders from benefiting from research outcomes

(Ancill, 1995). It is necessary to identify effective treatments for pain in persons with

dementia because of the possible link between untreated or undertreated chronic pain and

excess disabilities, such as depression, declines in functional performance, and sleep and

behavioral disturbances (Gibson & Helme, 2000; Villanueva et al., 2003). In relation to

persons with dementia, the term 'excess disabilities' has been used to describe "reversible

symptoms that are undesirable and temporary extensions of a specific primary disability"

(Kolcaba, 2003, p. 3). Because most studies assessing pain in older adults with dementia

have been focused on institutionalized persons with moderate to severe dementia, this

study will also fill a gap by addressing the pain management needs of persons with

dementia who remain residing in the community, typically with informal family

caregivers.

Several researchers have called for the need of an analgesic trial to treat pain in

persons with dementia. Kovach, Weissman, Griffle, Matson, and Muchka (1999)

developed the Assessment of Discomfort in Dementia (ADD) protocol which was able to

more accurately assess signs of discomfort in persons with dementia, and most

importantly, to increase the use of scheduled analgesics to reduce discomfort in these










persons. Their findings also support the belief that the pain experienced by people with

dementia is amenable to treatment. The authors note that one limitation of these results is

that only the use of scheduled analgesics, not prn or "as needed" analgesics, increased.

However, prn analgesics, such as acetaminophen, have incredible potential for treating

chronic pain, but are often not given because nurses are trained to give prn analgesics in

response to verbal reports of pain.

Acetaminophen (Tylenol) is one of the most commonly used analgesics for the

treatment of chronic pain associated with musculoskeletal conditions. Farrell (2002),

reports that pharmaceutically, acetaminophen is the most widely used analgesic and

antipyretic agent in the United States and the world. As long as dosing instructions are

followed, there are few, if any, risks associated with taking acetaminophen. In fact, two

of the maj or benefits of using acetaminophen as an analgesic are its availability over the

counter and its relatively low cost. As described below, all of the current treatment

guidelines recommend acetaminophen as the first line drug of choice for older adults.

The purpose of this study, then, is to evaluate the effectiveness of an acetaminophen trial

in reducing the expression of observable pain behaviors and excess disabilities related to

chronic pain in community-dwelling persons with dementia, as assessed by trained

observers and informal caregivers.

Theoretical Framework

The specific role of theory in the design and conduct of research studies is that of a

foundation from which to organize thoughts and hypotheses. Theories then serve as

guides for researchers to use to describe, interpret, and prescribe interventions (Meleis,

1997). Research without theory produces isolated information, but research with theory

produces science (Alligood & Tomey, 2002). The choice of a guiding theory is a critical









component of the research study design. The Communications Model of Pain

(Hadjistavropoulos & Craig, 2002), which addresses the subtleties of pain expression and

communication, will serve as the guiding framework for the proposed feasibility study.

Within an overall theoretical framework for understanding pain assessment,

Hadjistavropoulos and Craig (2002) derived a model of human communication that

incorporates both self-report and observational measures of pain. The Communications

Model is an appropriate theoretical framework for the proposed research study, as the

communication of pain is critical to pain assessment and subsequent treatment. The

authors derived the Communications Model from Prkachin and Craig's (1995) model to

assist in understanding complex social interactions between persons with pain and their

caregivers. This latter model was itself a derivation of Rosenthal's (1982)

conceptualization of non-verbal communication. Prkachin and Craig proposed an

AB C model focused on facial displays of pain, where (A) is the experience of an

internal state that may be (B) encoded into expressive behavior allowing the observer to

(C) draw inferences about the nature of the sender' s experience. Hadjistavropoulos and

Craig broadened this model to include verbal and non-verbal communication. In their

AB C model, (A) represents nociceptive or neuropathic pain processed in the brain

and experienced as pain, (B) represents the encoding of pain into either verbal or

nonverbal behavior, and (C) represents the potential for this behavior to serve as a

message that an observer can decode (see Figure 1-1).

The main empirical indicators derived from this model include self-report of pain,

the social context of behavior, and observational measures of nonverbal behaviors. Self-

report measures are verbal reports that reflect an individuals' subj ective description of the











Verbal Programs

Automatic
Intrapersonal Prcsig message
Influences Higher Clarity
Mental

Obsenrer
Bias


E te ce E Prc Deodin





Higher
Mental

Automatic
Contextual Processing Message
Influences Clarity


Non-Verbal Programs



Figurel-1. Communications model of pain (Hadjistavropoulos & Craig, 2002)
experience of pain. The social context of behavior reflects whether this social transaction

is instigated by the individual in pain or if it is in response to another' s query.


Hadjistavropoulos and Craig (2002) point out that the effectiveness of this social

transaction can depend as much upon respondent characteristics as upon the nature of the


message itself.


For persons with cognitive impairment, self-report measures of pain may not be the

most effective way to communicate their pain experience. Rather, observational


measures of nonverbal pain (such as facial behavior, limb and torso movements, and


changes in habitual patterns of daily living), which focus on signs of pain that are often

involuntary and nondeliberate, are more appropriate. One of the key concepts behind this

model is that it recognizes that observational measures of automatic expressive behaviors,


which focus on involuntary, publicly observable signs of pain, are preferred for









measuring pain when the capacity for self-report of subj ective experiences is either not

available or when the veracity of self-report is doubted (Hadjistavropoulos & Craig,

2002, p. 553). The authors suggest that this nonverbal behavior usually can be seen as

automatic in the sense that the "behavior represents stereotyped, reflexive patterns of

response to actual or impending tissue damage that are outside immediate conscious

awareness or voluntary control" (p. 553).

For the purpose of this feasibility pilot study, designed to measure the effectiveness

of an analgesic drug trial in reducing pain and its associated excess disabilities, the

presence of chronic osteoarthritic pain in conjunction with normal daily activity will

serve as the pain stimulus. Potential influences (including intrapersonal, cultural, and

contextual factors) on the expression of pain are no less relevant in persons with dementia

than in cognitively intact older adults, and previous research has found that nonverbal

pain expressions do include manifestations of affective qualities of the pain experience

(Hadjistavropoulos & Craig, 2002) (A, in Figure 1-1). The encoded verbal and nonverbal

expressions of the pain experience (B, in Figure 1-1) will then be decoded by the primary

caregiver and the trained observer in order to evaluate the effectiveness of the analgesic

drug trial (C, in Figure 1-1). Caregivers and the trained observer will also decode the

expression of excess disabilities of untreated chronic pain such as depression, declines in

functional performance, sleep disturbances, agitation, and behavioral disturbances in

order to observe the effect the analgesic trial had on these outcomes.

Given the complexities of the pain experience and the issues of validity concerning

self-report in persons with dementia, self-report measures alone cannot be expected to

capture the full spectrum of the pain experience. This is why Hadjistavropoulos and










Craig (2002), incorporate behavioral observations to assess pain in persons with cognitive

impairment. Observer reports in conjunction with participant self-report offer the most

comprehensive model of pain assessment for persons with dementia. By identifying pain,

this model will aid with pain control because the effect of an intervention can be

evaluated through subsequent pain assessments following its implementation. Assistance

with pain control is needed the most by people who are vulnerable and dependent upon

others for generalized care, including persons with dementia. However, caregivers must

rely on the availability of reliable and valid information concerning the nature of the

individual's distress in order to provide aid, which through this model, can be obtained

through behavioral observations. Therefore, the Communication Model will guide the

proposed study because it addresses both self-report and observational measures of pain,

recognizes the limitations of using self-report measures alone in persons with dementia,

and supports a multidimensional approach incorporating observational measures.

Specific theoretical constructs, study constructs, and operational definitions are presented

in Table 1-1.

Statement of Problem and Study Specific Aims

As mentioned previously, because pain is so prevalent among older adults, and

because older adults with cognitive impairment have compromised ability to reliably self-

report their pain, the use of observational strategies for pain assessment in this population

is warranted. The use of an acetaminophen trial in this population for pain control is well

supported in the literature and thus will serve as the tool for assessing the specificity of

observational pain measures. The Communications Model serves as a suitable theoretical

framework for designing studies aimed at assessing pain through self-report and










Table 1-1. Theoretical and Study Constructs
Theoretical Study Constructs Operational Definitions
Constructs


Verbal Program Pain
(Self-reported -Self-reported pain average scores on the NRS
pain)
PADE: Pain Assessment for the Dementing Elderly
ADLs: Activities of Daily Living, including dressing, feeding oneself, and transfers
RMPBC: Revised Memory and Problem Behavior Checklist
NRS: Numeric Rating Scale

observational measures, and for examining the effect of an analgesic trial on reducing

pain behaviors and excess disabilities related to untreated chronic pain. Therefore, in

order to examine the effectiveness of an acetaminophen drug trial in persons with

dementia, this study has 4 specific aims.

Aim 1: To investigate the effects of the scheduled administration of acetaminophen (1.3

grams up to three times per day) on self-reported pain intensity and the number of

observable pain behaviors exhibited by persons with dementia.


Non-Verbal
Programs
(Observed Pain)


Pain
-Caregiver reports of pain






-Trained Observer reports
of pain





-Behavioral observation
of pain

Pain related excess
disabilities
-Functional performance

-Behavioral disruptions


-Sum score of PADE Part 1:
Physical (observable facial
expression, breathing pattern, and
posture); and Part 2: Global
assessment of overall pain

-Sum score of PADE Part 1:
Physical (observable facial
expression, breathing pattern, and
posture); and Part 2: Global
assessment of overall pain

-Total number of pain behaviors
observed during a video-taped
activity protocol


-Sum score of PADE Part 3:
Functional (ADLs)
-Total and subgroup scores on
RMPBC for memory-related
problems, depression, and
disruptive behaviors









Hypothesis: Regular administration of acetaminophen will decrease the total

number of observable pain behaviors (as rated by the caregiver and trained observers)

exhibited by the participant, but will have no effect on self-reported pain intensities.

Aim 2: To investigate the effects of the scheduled administration of acetaminophen on

the frequency and severity of excess disabilities of pain in persons with dementia.

Hypothesis: Among persons with dementia, the frequency and severity of excess

disabilities of pain, such as depression, agitation, sleep disturbance, behavioral

disturbances, and impaired functional performance will decrease from baseline after

implementing regular administration of acetaminophen.

Aim 3: To determine the reliability of informal caregiver ratings' of the frequency of

pain behaviors and excess disabilities exhibited by persons with dementia.

Hypothesis: After an initial training session, informal caregivers will display

moderate to good reliability with a trained observer in ratings of the frequency of pain

behaviors and excess disabilities.

Aim 4: To determine the most frequent behavioral indicators of pain displayed by

persons with dementia.

Hypothesis: Based on findings from similar studies, the most frequently displayed

behaviors indicative of pain are expected to be guarding, rubbing, shifting and bracing

(Keefe & Block, 1982; Horgas & Elliott, 2005).















CHAPTER 2
REVIEW OF THE LITERATURE

The following discussion will demonstrate how this proposed study will build on

the information gathered by previous researchers in an effort to explore the effects of a

pain treatment regimen on observable pain behaviors in persons with dementia. In order

to understand the necessity of this treatment regimen, it is important to Birst understand

the path laid by previous researchers that supports the need for this study. First, the

degree to which pain afflicts both cognitively intact and impaired older adults will be

reviewed. Second, the validity of emerging strategies for pain assessment in cognitively

impaired older adults will be discussed and analyzed in relation to the more traditional

pain assessment strategies for cognitively intact older adults. Included in this discussion

will be an overview and definitions of the term "pain behaviors" as well as a section

describing the use of informal caregivers as proxy raters. Third, excess disabilities

related to untreated and undertreated chronic pain will be explored in order to highlight

the need for effective pain management in this population. Fourth, current pain

management guidelines for older adults will be discussed, followed by a review of studies

highlighting inadequate pain management in persons with cognitive impairment. The

proposed study will fill a gap in the existing literature by beginning to investigate pain

treatment in cognitively impaired elders.

Pain Prevalence

Adults ages 65 and over continue to constitute the fastest growing segment of

society. Because the number of conditions an individual may develop with pain as a









symptom or outcome increases with age, it is appropriate that pain assessment in this

population receive attention from both medical and research communities. The

prevalence of pain in this population is most often attributed to the presence of chronic

musculoskeletal conditions that are common in this age group. A 1998 study conducted

by Horgas and Tsai found musculoskeletal conditions, such as arthritis or osteoporosis, to

be the most frequently diagnosed painful conditions among nursing home residents. Herr

(2002) states that "osteoarthritis alone may be a source of chronic pain in as much as 80%

of the population older than age 65" (p. 21).

Weissman and Matson (1999) estimate that 80% of community dwelling elderly

have at least one chronic disease, that 70% experience some type of pain, and that only

18% take pain medication for chronic pain. Huffman and Kunik (2000) found that 86%

of rural community dwelling older adults had pain during the previous year and 59%

reported multiple pain complaints. In a study specifically comprised of community-

dwelling persons with dementia, only 32% self-reported having current pain (Shega,

Hougham, Stocking, Cox-Hayley, & Sachs, 2004), whereas Mitchell, Morris, Park, and

Fries (2004) found that 53.4% of their community-dwelling sample with dementia

receiving terminal care were experiencing daily pain. Several researchers have found

that anywhere from 45% to 83% of elderly residing in nursing homes experience some

type of pain (Blomqvist & Hallberg, 1999; Horgas & Dunn, 2001; Weissman & Matson).

Ferrell, Ferrell, and Osterweil (1990) found that as few as 29% of a sample of nursing

home residents reported having no problems with pain, while 24% reported having

constant pamn.









Pain Assessment

With such a high prevalence of pain in older adults, it is necessary to have

successful methods of assessing painful conditions in order to provide individuals with

appropriate treatment. This is important because untreated chronic pain may lead to

excess disabilities such as impaired physical and social functioning, lowered quality of

life, and/or depression (Parmelee, 1993). Eliciting an individual's self-report is the most

common method of pain assessment. However, Ferrell, Ferrell, and Rivera (1995) found

that approximately 65% of nursing home residents have a specific barrier to the ability to

self-report pain due to the presence of cognitive deficits or mental illness. Thus, it is

important to consider multiple approaches to assessing pain.

Self-Report

Several tools exist for measuring self-reported pain, such as the visual analog scale

(VAS), the verbal descriptor scale (VDS) and the numeric rating scale (NRS), and

research has consistently "supported the use of simple, self-report pain rating scales for

chronic pain patients" (Chinball & Tait, 2001, p. 173). Both the VDS and NRS have

been shown to be reliable and easy to use with an elderly population (Gagliese &

Melzack, 2003; Herr & Mobily, 1993). Typically these scales are presented either

vertically or horizontally with values ranging from 0-6, 0-10, or even 0-100, where 0

represents no pain and the highest number represents worst possible pain. When specific

questions about pain are presented, the individual is asked to choose a number on the

scale that most accurately reflects this pain.

However, there are several factors that may complicate pain assessment in older

populations and consequently question the validity of self-reports. In cognitively intact

older adults, these factors include a tendency for older adults to under-report pain, either









because they diminish the importance of mild pain and attribute pain to the normal aging

process (Gagliese & Melzack, 2003; Gibson & Helme, 1995) or because they fear what

an intensification of painful symptoms might mean (i.e. worsening of a disease process,

need for diagnostic testing, or impending death) (Herr, 2002). Older adults may also

underreport pain in an effort to maintain a positive self-concept. It is also important to

consider that the manner in which a person responds to pain is highly individualized and

the impact pain will have depends on both obj ective indicators of its severity and

stressfulness, as well as how it is interpreted for personal meaning (Markus & Herzog,

1991).

In cognitively impaired individuals, the factors that complicate the use of self-

report are vastly different. Snow, Rapp, and Kunik (2005) report that in order to

accurately self-report the presence of pain, one must possess the ability to "understand

the question in a pain rating, recall pain events in the given time frame, (and) accurately

interpret the experience of noxious stimuli as painful events" (p. 22). Communication

disorders including reduced receptive and expressive language exacerbate cognitively

impaired individuals' inability to report pain (Chinball, Tait, Harman & Leubbert, 2005).

Supporting this fact is the finding that chronic pain is less likely to be identified among

cognitively impaired nursing home residents than more alert individuals (Sengstaken &

King, 1993). It is not clear whether the decreased self-report of pain in persons with

dementia is due to actual less pain or a lessened ability to report such pain (Huffman &

Kunik, 2000). In either case, measuring pain with self-reports in this population may be

unreliable due to the deficits inherent to dementia. Such deficits include compromised

cognitive and verbal skills (i.e. memory loss, loss of judgment, confusion, and attention









and language deficits) making it difficult for persons with cognitive impairments to recall

and/or express pain. These deficits not only affect pain assessment in this population,

they also may constrain the ability to assess the effectiveness of pain interventions in that

persons with dementia may not be able to reliably report a decrease in pain either, thus

potentially leading to further undertreatment of pain in this population. (Chinball et al.,

2005).

However, it should be mentioned that some researchers have supported the use of

self-report in this population. Ferrell et al. (1995) found that at least 83% of nursing

home residents were able to successfully use at least one self-report pain scale, even

though the study participants' average Mini Mental State Exam score was 12, indicating

moderate to severe cognitive impairment. Parmelee (1994) reported that "when questions

are phrased simply and straightforwardly, even moderately demented individuals can

give valid, reliable information about their pain experience" (p. 289). In a recent study,

Pautex and colleagues (2005) found that only 12% of their hospitalized participants with

dementia did not understand any one of the four self-report scales they were

administered. However, these authors caution that "the ability to complete an assessment

does not imply reliability" (p. 527). Because of the conflicting opinions as to the validity

of self-report in persons with dementia, it stands to reason that, while it should not be

automatically dismissed in this population, neither should self-report be the sole method

of pain assessment. Alternative strategies for assessing pain in the cognitively impaired

have been the recent focus of research for several investigators. These strategies have

revolved around assessing objective behavioral manifestations of pain, and since a

maj ority of persons, even those with dementia, were able to use at least one self-reported










pain scale, these scales can serve as useful measures against which to compare

observational strategies.

Observational Strategies

If we were to accept McCaffery and Beebe' s (1989) definition that pain is

"whatever the patient says it is and occurs whenever the patient says it does," we would

have to conclude that if a person does not verbally express that they are in pain then they

must not have pain. Marzinski (1991) argues that this definition cannot be used with the

nonverbal elderly (including those with dementia) and that behavioral assessment is the

only acceptable way of assessing pain in this population. Also, since the use of self-report

indices of pain requires verbal skill and comprehension, it is necessary to focus on these

alternative measures of pain in persons with compromised verbal abilities. Therefore, to

assess pain in persons with dementia, researchers have begun to focus on automatic, less

verbal, pain displays. Weissman and Matson (1999) support this focus by finding that in

moderately to severely cognitively impaired persons, discomfort is often non-verbally

exhibited. The rationale for using observable behaviors as representative of pain is

appropriate because observational measures of pain capture behavior that is more

automatic and less subj ect to voluntary control, in contrast to self-reports of pain which

require higher mental processes and may be more susceptible to purposeful distortion

(Hadjistavropoulos & Craig, 2002). This is appropriate for persons with cognitive

impairment, where unlike with verbal abilities, there is no barrier to automatic processes.

Keefe and Block (1982) were among the first researchers to use direct observation

of pain in their work involving patients with chronic low back pain. Their observation

methodology consists of instructing the patient to "engage in a standard set of daily

activities designed to elicit pain behavior" (Keefe & Smith, 2002, p. 118). The authors









observed for the following five overt behaviors that were thought to be consistent with

chronic pain: guarded movement, bracing, rubbing the affected area, grimacing, and

sighing. Through a series of experiments, these behaviors were shown to be highly

reliable and to have good construct, concurrent, and discriminant validity. Thus, this

work established several important components of observational methods: 1) pain

behaviors can be reliably recorded by trained observers, 2) the frequency of pain

behaviors correlates with the patient' s own self-report of pain, and 3) naive observers'

ratings of pain also correlate with the frequency of pain behaviors.

Through the success of this work, a person' s observable pain behaviors have

become essential to understanding the experience and impact of pain (Keefe & Smith,

2002). Because these methods were developed in verbal, cognitively intact populations

who can also self-report pain, it is clear that the observation of pain behaviors is crucial

to pain assessment in persons with dementia who have compromised cognitive and verbal

skills, such as memory loss, loss of judgment, confusion, and attention and language

deficits, making it difficult to recall and/or express pain.

Advantages of observational methods

There are several advantages to using observational methods to assess pain in

persons with dementia. These observations are "more objective and accurate measures

than can be obtained through anecdotal observation" (Keefe & Smith, 2002, p. 123).

Another advantage is that it allows the researcher or clinician to directly observe pain

behaviors as well as the effects of chronic pain interventions by comparing the frequency

of pain behaviors before and after the implementation of the intervention. If an

intervention is an effective pain reliever, the frequency of observable pain behaviors will

decline after its implementation. Additionally, another advantage of observational









methods is that the researcher can make alternative treatment decisions if intervention is

ineffective.

A maj or design advantage of obj ective methods is that they allow the researcher to

precisely measure the effects of social, psychological, or environmental factors that may

influence chronic pain. Keefe and Hill (1985) found pain behaviors to be more likely to

occur when patients were moving, such as during walking or transferring from one

position to another. Consequently, their research plan included having participants

perform an activity-based protocol in order to ensure maximum likelihood of observing

pain behaviors. Furthermore, Keefe and Smith (2002) maintain that for persons with

musculoskeletal problems (e.g. arthritis), performing daily tasks such as sitting, standing,

walking, and reclining for 1-2 minute periods is capable of eliciting pain behavior.

Finally, and perhaps most importantly to the proposed research study, observational

methods of pain assessment allow a researcher to gather reliable and valid pain (and pain

intervention) information in populations whose verbal report of pain is compromised and

unreliable, such as is the case for persons with dementia.

Limitations of observational methods

Despite the apparent utility of observational methods to assess pain in persons with

dementia, there are limitations. One limitation is the potential for reactivity (i.e. the

tendency for the participant to alter behavior in the presence of an observer) (Elder, 1999;

Keefe & Smith, 2002). Due to cognitive decline, persons with dementia are less likely to

consciously alter their behavior than are cognitively intact persons (Hadjistavropoulos &

Craig, 2002). The potential for observer bias is another limitation of observational

methods. Keefe and Smith suggest that this limitation can be minimized by carefully










training observers, conducting frequent reliability checks, and holding periodic retraining

sessions.

Pain Behaviors

Measuring pain behaviors in nonverbal persons through direct observation can

yield otherwise unobtainable information about an individual's experience of pain. In

order to use the term pain behaviors as an outcome measure, it must first be operationally

defined in terms of its use in the study. To operationally define pain behaviors and select

appropriate measurement tools to fit the study's guiding theory and research questions,

other researchers' definitions and measures of pain behaviors must be reviewed. Pain

behaviors have been defined as "verbal or nonverbal actions understood by observers to

indicate that a person may be experiencing pain and suffering" (Loeser, 2001, p. 19).

Specifically, these actions include, but are not limited to, audible complaints, facial

expressions, abnormal postures or gait, use of prosthetic devices, avoidance of activities,

overt expressions, and verbal or nonverbal complaints of pain, distress, and suffering.

This definition captures the most frequently recognized behaviors thought to signify pain

and serves as a basis from which other researchers can select the most appropriate

behaviors to study pain in their population of interest.

All definitions of pain behavior recognize them to be observable, nonverbal

behaviors that signify pain to others (Cohen-Mansfield & Creedon, 2002; Keefe & Block,

1982). However, behavioral observation studies differ in their selection of the actual

behaviors they choose to study in their population of interest. Fordyce (1976), who

popularized the study of chronic pain behaviors, proposed that a thorough behavioral

analysis could be just as important as a medical evaluation for evaluating chronic pain

and proposed identifying the following as pain behaviors: the report of pain, low levels of









activity, taking pain medications, body posturing, and facial expressions (as cited in

Keefe, 1982). Subsequent research studies have followed one of two paths; either they

are designed to conduct a comprehensive behavioral analysis or they are designed with a

specific set of pre-identified behaviors and measures to identify pain.

Identifying Pain Behaviors

Cohen-Mansfield and Creedon (2002) further developed the definition of pain

behavior by interviewing nursing home staff regarding specific behaviors they thought

were associated with resident pain. The resulting core group of behaviors was

categorized into the following four categories: specific physical repetitive movements

(i.e. facial expressions and body postures), vocal repetitive behaviors (i.e. moaning,

crying, or screaming), physical signs of pain (i.e. skin discoloration, swollen joints, or a

change in vital signs), and changes in behavior from the norm for that person (i.e.

changes in mood, movement, or eating patterns). The authors also found that nursing

home staff rated vocalizations, rubbing the affected area, and reluctance to move to be

among the most prevalent resident pain behaviors. This work led them to develop the

following working definition of pain as it applies to evaluating pain behaviors in persons

with dementia: "pain is suffering associated with bodily injury or disease, characterized

by physical and/or emotional discomfort, which gives rise to a set of distinctive behaviors

perceived by caregivers as indicative of that discomfort" (p. 65). Alternatively, Weiner,

Peterson, and Keefe (1999) found other salient pain behaviors in nursing home residents

(on which residents and caregivers agreed) to be the use of mechanical help, shifting

weight when seated, taking or asking for pain medication, moving or walking in a

protective fashion, moving extremely slowly, limping, lying down, bracing when seated,









clutching the painful area, stiffness, and asking someone to do something to help the

pamn.

Testing Pre-identified Behaviors

Rather than conducting a comprehensive behavioral analysis, other researchers

have tested the ability of specific behaviors to signify pain. Keefe and Block (1982)

observed for specific behaviors in persons with chronic low back pain by having them

perform an activity-based protocol (consisting of activities of daily living) as a pain

stimulus. The behaviors they hypothesized would signify pain included: guarded

movements, bracing, rubbing, grimacing, and sighing. Not only did the authors find that

they could reliably observe these behaviors, but the frequency of these behaviors

correlated with patients' self-report, decreased with treatment, were observable by naive

observers, and discriminated persons with chronic low back pain from normal and

depressed controls.

Hadjistavropoulos and colleagues (1998), in an effort to develop an observational

strategy to assess pain in persons with dementia, investigated facial reactions to pain

using a combination of the extensively tested and well-validated Facial Action Coding

System (FACS), nurses' ratings, and student volunteers' ratings. They discovered that

facial reactions, especially brow lowering and chin raise, were useful indicators of pain in

this population whether rated by FACS trained observers, nurses, or students.

Observing pain behaviors can also provide daily documentation of the effects of a

pain management program (Keefe & Smith, 2002). Because pain behaviors are readily

observable and recordable by trained observers, an observational method of pain

assessment is a reasonable way to assess pain in a nonverbal or unreliable population,

such as those with cognitive impairments. Researchers have begun to study behavioral









assessments by taking these behaviors identified as indicative of pain in cognitively intact

older adults and then applying these to cognitively impaired older adults with promising

findings (Hurley et al., 1992; Villanueva et al., 2003; Warden et al., 2003). Often in

research studies using observational methods to assess behavior, participants are

videotaped so that these behaviors may be coded at a later time (Keefe & Block, 1982;

Keefe, Crisson, Maltbie, Bradley, & Gil, 1986).

In a study conducted by Hurley and colleagues (1992), the authors argue that

persons who have lost the cognitive capacity and verbal ability required to communicate

must rely on nursing staff and/or other caregivers to recognize, assess, and treat their

pain. However, how can these caregivers perform such a task without the knowledge of

what they should be looking for to indicate pain in their care recipients? This is why

these researchers have focused on behaviors commonly found in people with dementia of

the Alzheimer's type who have pain. These behaviors include facial expressions, body

movements, certain postures and gestures, and vocalizations. The authors showed that

these observed behaviors served as useful external indicators of what the person with

dementia experienced internally. Although researchers have begun to make strides in

identifying useful ways of assessing pain in persons with dementia, there still exists a

critical underassessment of pain in this population leading to large quantities of untreated

and undertreated pain. To help abate the chasm, the American Geriatrics Society (2002)

put forth a comprehensive framework regarding behavioral indicators of pain. The

framework recognizes the following 6 main types of pain behaviors/indicators : facial

expressions, verbalizations/vocalizations, body movements, changes in interpersonal

reactions, changes in activity patterns or routines, and mental status changes.









Underassessment and Treatment of Pain in Persons with Dementia

It has been recognized that assessing pain in persons with dementia presents

specific challenges, and as a result underassessed and undertreated pain is an unfortunate

reality for many people with dementia. An estimated 4.5 million people in the United

States have dementia of the Alzheimer' s type; this number is expected to grow to

approximately 16 million by the year 2050 (Carter, Rose, Palesch, & Mintzer, 2004).

Furthermore, 50% of older adults over age 85 have Alzheimer' s disease with 28% of this

population having severe disease (Herr, Bjoro, & Decker, 2006). With prevalence rates

of pain in older adults ranging from 45-83% and with no reason to assume that people

with dementia experience less pain, it is critical to assess the amount of treatment for pain

this population receives. Some factors that may contribute to this inadequate pain

assessment and treatment in the healthcare environment are due to "the presence of

comorbidities that compete for attention of healthcare staff, a tendency for observers to

discount high levels of reported pain, physician fears regarding drug use for pain, and

nurse expectations about pain coping...each of these factors is magnified with respect to

cognitively impaired older patients" (Chibnall & Tait, 2001, p. 173-4).

Several researchers have demonstrated that cognitively impaired individuals

receive lower amounts of medication for pain treatment. Horgas and Tsai (1998) found

that "nursing home residents with cognitive impairments are less likely to be prescribed

and administered pain drugs and receive lower dosages of analgesics when medicated

than are their more cognitively intact peers" (p. 240). Also, they found that while 87.6%

of participants were prescribed at least one pain drug, only 47% were given an analgesic

during the course of their study. Of these drugs, mild analgesics were most commonly

prescribed (80%) with acetaminophen being the most frequently prescribed analgesic










(59.6%). Chibnall and Tait (2001) found that 61% of cognitively intact elders were

prescribed a nonsteroidal anti-inflammatory agent (NSAID) as compared to 48.5% of

cognitively impaired elders. Morrison and Siu (2000) found that when "as needed"

analgesics are ordered in the elderly, "only 24% to 27% of the prescribed doses are

actually administered; this percentage declines with advancing age" (p. 241).

Additionally, Pautex and colleagues (2005) found that approximately one-quarter of

hospitalized persons with dementia who self-reported pain were not receiving any

analgesia. It is imperative that these treatment disparities be overcome. These disparities

in pain treatment continue to exist despite the advances made in assessing pain in persons

with dementia. Thus, educating those that care for persons with dementia about proper

pain assessment and treatment is a crucial component to obtaining appropriate pain

management outcomes in this population.

Caregiver Assessments

According to the Alzheimer's Association and the National Alliance for Caregiving

(2004), there are 8.9 million caregivers in the United States caring for someone 50 years

or older with dementia. Considering that pain control is "one of the least well managed

aspects of professional medical care (and that) untrained family caregivers are routinely

given the responsibility of managing pain at home," it is imperative that these caregivers,

who have the most day-to-day contact with persons with dementia, possess the skills to

assess, treat, and reliably report the pain that the individual experiences (Levine,

Reinhard, Feinberg, Albert, & Hart, 2003, p. 20). While many researchers have

developed assessment tools and strategies that have proven to be successful and reliable

in detecting pain in this population, not all of these tools and strategies are easily

implemented or taught to caregivers. In fact in most observational studies, data are coded










(either in real-time or from videotaped sessions) by trained research assistants. The

researchers who developed the Discomfort Scale for Dementia of the Alzheimer Type

(DS-DAT) acknowledge that this scale is most useful for research purposes because it

requires more time to train for accurate use than is realistic for caregivers (Warden et al.,

2003). In order to translate research into practice as well as to test the construct and

social validity of observational methods, these methods must be tested by naive observers

(meaning any untrained observer such as a family caregiver, nurse, or nursing assistant)

who have the most daily contact with these pain patients. Since caring for persons with

dementia is often a time-consuming and stressful job, many caregivers may not have a lot

of time to think about assessing the individual's pain. Quick, easily taught, and reliable

tools and strategies for caregivers to use to identify pain are needed.

In an effort to create a tool that caregivers could reliably use to assess pain in

persons with dementia, Villanueva and colleagues (2003) developed the Pain Assessment

for the Dementing Elderly (PADE). This tool was tested by a group of caregivers of

persons with dementia in long-term care settings. The tool assesses "facial expressions,

activities of daily living, and the caregiver' s overall judgment of the resident' s pain"

(Villanueva et al., p. 2). The researchers found valid and reliable pain assessments with

this group of caregivers after a 1-hour training session. Also with practice, caregivers

were able to complete the PADE in 5-10 minutes.

Informal Caregivers versus Trained Observers

Several research studies have tested their observational methods with naive

observers and found promising results. Hadjistavropoulos et al. (1998) compared trained

observer' s ratings with the Facial Action Coding System (FACS) to pain intensity ratings

by 10 female nurses and 10 female university students after watching videotaped









reactions of older adults undergoing venipuncture. All three of these measures were

found to be significantly correlated with one another, thus lending support to the validity

of informal caregivers' naivee observers') ratings. Likewise, Keefe and Block (1982)

asked untrained observers to review videotaped segments and rate subj ect' s pain intensity

with three different measures. The total number of pain behaviors (as coded by a trained

observer) correlated well with all three ratings from naive observers. Furthermore,

Werner, Cohen-Mansfield, Watson, and Pasis (1998) found that family caregivers' and

staff members' ratings of senior day care participants' pain severity were highly

correlated (r= 0.86).

Informal Caregivers versus Participant Self-Report

In order to rely on caregiver ratings, it is important to trust that their ratings

accurately represent what the care recipient would say themselves. Overall, the reliability

of naive observers as proxy raters has not been found to be very strong and even

healthcare professionals have been found to be ineffective at both pain assessment and

treatment (Herr et al., 2006). Horgas and Dunn (2001) found no significant association

between nursing assistants' ratings of resident pain and the resident' s own ratings. Pain

was also underdetected by nursing assistants for 37.8% of residents. Most of the research

studying caregiver ratings has been done within the context of chronic illness, especially

cancer. For cancer symptoms, Nekolaichuk, Maguire, Suarez-Almazor, Rogers, and

Bruera (1999) suggest that the best assessment and treatment outcomes for cognitively

impaired patients may be obtained by selecting caregivers who best understand and

represent the patient's experience. These authors also state that both healthcare providers

and other significant caregivers tend to overestimate symptoms of depression and anxiety

and they underestimate performance status and quality of life. Nekolaichuk et al.,









through a study comparing cancer patients, their family caregivers, and nurses' ratings of

patient' s symptoms, found that all three respondents were reasonably consistent with one

another when ratings were averaged across all symptoms at two measurement times.

However, for any one rater on one occasion, reliability estimates were less than 0.70 for

most symptoms.

Through a review of studies examining ratings of pain by cancer patients and their

caregivers, Lin (2001) found that most studies used correlation statistics which may

overestimate the actual relationship because this statistic does not take into account error

variance. The studies reviewed tended to show that patient and caregiver ratings were

significantly correlated and accurate, with some studies showing a tendency for family

caregivers to overestimate, rather than underestimate the patient' s pain. In her own study

comparing cancer patients' and family caregivers' ratings of pain intensity and

interference, Lin found that the Pearson's correlation, intraclass correlation coefficients,

and the kappa statistics ofinterrater agreement were statistically significant, lending more

support that the caregiver ratings were accurate representations of the patient' s own

ratings. However, the author cautions that the kappa values for worst pain (.25), least

pain (. 18), current pain (.20), and average pain (.25) do not represent good agreement

beyond chance and that family caregivers tended to overestimate patient's pain severity.

In a study to identify behaviors signaling pain in nursing home residents, Weiner et

al. (1999) also examined the congruence between self-reported resident and caregivers'

(nurse and family) ratings of pain behaviors and pain intensity. The authors found that

while both the nurse and family caregiver felt similarly confident in their ability to rate

the residents' pain, both had poor agreement with the residents' own ratings regarding









identification of pain behaviors and rating pain intensity. Interestingly, over 75% of

nurses and only 44% of family caregivers felt that the residents showed their pain

outwardly and 71% of nurses felt that cognitive impairment made assessments more

difficult. Cremeans-Smith and colleagues (2003) found that 55% of spousal caregivers

rated their wife's recent osteoarthritis pain severity in agreement with her own rating.

Again, for those dyads who disagreed, the caregiver was more likely to overestimate,

rather than underestimate, their spouse's pain severity. In the only study to date studying

the congruence between caregiver and patient report of pain in community-dwelling

persons with dementia, 59% of the 150 patient-caregiver pairs, agreed on the presence or

absence of patient pain. Again, caregivers were more likely to estimate their care

recipients' pain severity (Shega et. al, 2004).

There has been no standard set to gauge how accurate caregiver ratings must be in

order to be considered as valid proxies. The maj ority of research utilizing proxy reports

from caregivers has shown only moderate agreement with the care recipient, with

caregivers tending to overestimate levels of pain and excess disabilities related to chronic

pain (such as depression, anxiety, and functional disability) (Cremeans-Smith et al.,

2003; DeBettignies, Mahurin, & Pirozzolo, 1993; Lin, 2001; Nekolaichuk et al., 1999;

Seltzer & Buswell, 1994; Weiner et al., 1999; Zanetti, Geroldi, Frisoni, Bianchetti, &

Trabucchi, 1999). These more negative assessments (i.e. higher ratings of pain intensity)

by caregivers have been attributed to the distress and burden they experience when caring

for a close relative with dementia, as well as to a greater reported degree of concern over

patient disabilities (DeBettignies et al.; Seltzer & Buswell; Zanetti et al.). In situations

where relying on caregiver proxies is necessary, they should be seen as reasonably










accurate (Lobchuk, Kristjanson, Degnet, Blood, & Sloan, 1997). While the research in

this area demonstrates that caregivers' accuracy in assessing patient symptoms is not in

total agreement with care recipients, caregiver report is the best option for assessing pain

and its excess disabilities in persons with dementia. And, as has been shown, although

caregivers' ratings are not perfect, they tend to err on the side of overestimating rather

than underestimating the patients' own ratings. However, it should be noted that most of

the studies that asked caregivers to give proxy reports for the care recipient did not base

this proxy report on direct observations of the care recipient, but rather on global

perceptions.

Among the many factors that may contribute to caregivers' inability to report

accurate proxy ratings, the most important to consider is the tremendous stress and

burden that caregivers of persons with dementia are under. Many of these caregivers are

themselves elderly, 63 being the average age of caregivers caring for someone age 65+,

and one-third of these caregivers are in fair to poor health (Administration on Aging,

2004). Caregivers are at risk for "burnout" as the average person with dementia requires

69-100 hours of care each week (Kahn et al., 1998). Often behavioral disturbances in

persons with dementia result in high levels of caregiver burden and are frequently a

precipitant to nursing home placement (Kahn et al.; Lantz, 2003). Czaja, Schulz, Lee,

and Belle (2003) found that functional limitations and behavioral problems of the care

recipient were the primary stressors for the caregiver. This is especially true for

caregiving wives, who are more likely to experience caregiving distress and report more

frequent problem behaviors in their spouse than caregiving husbands (Ingersoll-Dayton &

Raschick, 2004). Some researchers have proposed that these behavioral disturbances and









functional limitations may be related to untreated chronic pain (Buffum, Miaskowski,

Sands, & Brod, 2001; Douzjian, Wilson, Shultz, Tapnio, & Blanton, 1998; Ferrell, 1995;

Gibson & Helme, 2000; Lantz; Kahn et al.; Villanueva et al., 2003). Therefore, it is

necessary to educate caregivers in both the importance of, and strategies to, assess and

treat pain in their care recipients.

Strategies to Strengthen the Validity and Reliability of Caregiver Report

The first step in overcoming methodologic problems related to proxy report is by

using observational methods (either in combination with the caregiver report or training

the caregiver to observe care recipient behavior). There are also strategies to help

increase the reliability of caregiver ratings. An initial strategy is to have the caregiver be

the person who best understands and represents what the care recipient experiences; this

would most likely be the primary caregiver who lives with and is most familiar with the

care recipient (Bohac, McNeilly, & Folks, 2003; Nekolaichuk et al., 1999). The most

useful way to increase untrained observers' ratings is through education and training.

Training does not have to be as intense as is required for formally trained observers. For

example, when creating the Pain Assessment for the Dementing Elderly tool, Villanueva

and colleagues (2003) obtained valid and reliable caregiver pain assessments after a 1-

hour training session. Researchers should also educate caregivers to time their

observations with potentially painful events in order to maximize the potential for

observing pain behaviors (Weiner et al., 1999). Bohac and colleagues suggest increasing

caregiver familiarity with the research instruments being used (with practice sessions) to

ensure greater reliability and validity of caregiver reports. Finally, as mentioned

previously, caregiver reports have been shown to be influenced by the amount of burden

and distress the caregiver feels. Perhaps a strategy aimed at decreasing caregiver burden









and distress would in turn, positively influence the reliability and validity of their reports

of care recipient behavior.

Excess Disabilities related to Dementia and Chronic Pain

Investigators whose research focuses on persons with dementia are well aware of

the excess disabilities (behavioral disturbances) that accompany disease progression.

Research in this area can hardly be conducted without acknowledging the complex

behavioral problems that accompany this disease (Teri et al., 1992). Lantz (2003) reports

that 65% of community dwelling older adults with dementia have at least one disruptive

behavior and 40% have three or more. Rowe, Straneva, Colling, and Grabo (2000) found

that 86% of persons with dementia evaluated at a community-based clinic had at least one

disruptive behavior, with 86% being verbally agitated, 55% being physically agitated,

33% being verbally aggressive, and 21% being physically aggressive. Overall, the

behavior problems commonly recognized to be associated with dementia include:

depression, agitation, aggression, anxiety, delusions, impulsivity, repetitive actions,

restlessness, disinhibition, resistiveness, physiological risks, impaired cognitive function,

difficulties with performance of activities of daily living (ADLs) and instrumental

activities of daily living (IADLs), pacing, vocalizations, altered sleep routines and time

spent in bed, wandering at night, altered social behavior, and increased health care

utilization and costs (Albert, et al. 2001; Ancoli-Israel et al., 2003; Herr et al., 2006;

Hinton, Haan, Geller, & Mungas, 2003; Kahn et al., 1998; Lantz, 2003; Opie, Doyle, &

O'Connor, 2002; Teri et al). Furthermore, Logsdon, Gibbons, McCurry, and Teri (2002)

report that caregivers for persons with severe dementia (MMSE <10) reported higher

rates of disruptive behaviors and worse physical functioning in their care recipients.










The presence of these behavioral problems and limitations in functional

performance is a source of stress and burden to the caregiver. Also, behavioral problems

in this population can negatively influence quality of life as well as lead to increased

social isolation and risk of falls (Horgas & Margrett, 2001). While previous researchers

have targeted reducing these outcomes with interventions such as light therapy (Ancoli-

Israel et al., 2003), psychosocial and behavioral interventions (Czaja et al., 2003), or

administration of psychotropic medications (Opie et al., 2002), other researchers have

found more promising results by assessing and treating underlying chronic pain

(Douzjian et al., 1998; Opie et al.). The rationale for this approach stems from the

observation that many of the behavioral problems and functional limitations that afflict

persons with dementia are also seen in cognitively intact persons who experience chronic

pain. For instance, Vlaeyen, Van Eek, Groenman, & Schuerman (1987) identified nine

components of chronic pain behavior: anxiety, attention seeking, verbal pain complaints,

medication use, general verbal complaints, distorted posture and mobility, fatigue,

insomnia, and depressive mood. Ferrell (1995) classifies depression, decreased

socialization, sleep disturbances, impaired ambulation, and increased health care use and

costs as complications from untreated chronic pain. Thus, there is considerable overlap

between excess disabilities related to dementia and those related to chronic pain.

Ferrell (1995) emphasizes that in order to maximize mobility and independence,

health care providers and caregivers need to assess ADLs because evaluating functional

status is an important measure of the success of pain management. Studying pain from

osteoarthritis (OA) in rural adults age 45 and older, Jordan, Linder, Renner, and Fryer

(1995) found that hip OA, hip pain, knee OA, and knee pain were all associated with self-










reported and observed limitations in functional performance. Even the American

Geriatrics Society's panel on persistent pain in older adults recommended in 2002 that

recent changes in function, verbalizations suggestive of pain, and nonverbal pain

behaviors should be assessed in all chronic pain patients. For those with dementia,

caregiver reports should be solicited.

Several researchers have begun to study the link between pain and specific problem

behaviors in persons with dementia. Buffum et al. (2001) found significant positive

correlations (r= .50, P= .003) between measures of discomfort (Discomfort Scale) and

agitation (Cohen-Mansfield Agitation Inventory) in nursing home residents with

dementia, suggesting that agitated behaviors could indicate painful sensations. The

authors suggest that although the correlation was strong, the discomfort behaviors may be

related to other things, such as hunger or being wet from urinary incontinence, and the

best way to evaluate the role of pain in prompting these behaviors would be to administer

an analgesic and then observe for changes in behaviors. Three groups of researchers

have done this. In an effort to reduce challenging behaviors in nursing home residents

with dementia, Opie et al. (2002) initiated 4 distinct groups of interventions: psychosocial

strategies, nursing approaches, psychotropic medications, and management of pain.

Interventions were tailored to each of the 99 participants, with the maj ority receiving

more than one intervention. Changes in pain management were recommended for 18

participants, and while the authors do not report the effects of pain management alone in

reducing challenging behaviors, overall their intervention strategies showed a statistically

significant improvement in the target behaviors of restlessness, physical aggression,

verbal disruption, and inappropriate behavior. The limitation of this study is that the









effect of any one intervention on reducing challenging behaviors is unknown; however,

their results argue that individually-tailored interventions are effective in reducing

challenging behaviors in persons with dementia.

Another group of researchers was able to show a strong relationship between

treating pain and decreasing problem behaviors. Douzjian et al. (1998) observed that

demented residents who were admitted to a skilled nursing facility (SNF) after being

discharged from the hospital were often receiving antipsychotic medications without

having any documented psychological history. Since the use of antipsychotic

medications is strictly regulated in SNFs, they hypothesized the residents' behavior may

be linked to pain, so the authors devised what they term an "informal study" by trying to

initiate a pain treatment regimen to replace the use of antipsychotic medications. Eight

residents were placed on a pain treatment regimen of acetaminophen 650 mg three times

a day. Five (63%) of these residents showed a decrease in the number of problem

behaviors and the staff was able to discontinue four orders for antipsychotic drugs and

two for antidepressants. Most impressively, the authors were able to decrease the

facility's use of antipsychotic drugs from 20% to 0% over the study period. This is

clinically significant because antipsychotic drugs often work inconsistently and can lead

to adverse events such as a higher risk for falls (Opie et al., 2002).

In perhaps the most promising study to date, Chinball and colleagues (2005) found

that 4 weeks of acetaminophen (3,000 mg/d) was more effective than placebo in

increasing participants' engagement with their environment. Specifically, while taking

the acetaminophen, participants spent significantly more time in "social interactions,

engaged with media, talking to themselves, engaged in work-like activity, and










experiencing unattended distress...and spend less time in their rooms, less time removed

from the nursing home unit, and less time performing personal care activities" (p. 1921).

However, there was no effect of acetaminophen on agitation, emotional well-being, or as

needed psychotropic medication use. Similarly, Allen and colleagues (2003) also found

that nursing home residents who had received analgesic medication during the course of

their 4-week study were more active than those who did not receive any analgesia.

Pain Management and Treatment Guidelines

Now that researchers have begun to demonstrate successful ways to assess pain in

persons with dementia and a correlation between chronic pain and behavioral problems,

the next logical step is to further examine pain treatment strategies in this population.

Treatment guidelines for the general population with chronic mild to moderate pain state

that nonpharmacologic strategies are the appropriate first line defense. These strategies

include patient education, self-management programs, and exercise, among others. For

obvious reasons, these strategies are not appropriate for persons with impaired memory

function. The American Geriatrics Society's 2002 Management of Persistent Pain in

Older Adults state that analgesic drugs are safe and effective for use by older people.

They recommend following the adage "start low and go slow" in order to assess each

individual's response to a given medication. The guidelines report that for most patients

"with persistent mild to moderate musculoskeletal pain respond favorably to around-the-

clock doses of acetaminophen" (p. S213). This is advantageous because this treatment is

capable of providing satisfactory pain relief and has a much lower risk of side effects

than NSAID drugs. The maximum dosage for acetaminophen in these guidelines is 4000

mg/24 hours, with 4-6 hour dosing (not to exceed a daily amount of 4 grams),

recognizing that the half-life for acetaminophen ranges from 1 to 4 hours. Kovach et al.










(1999) recommend that for persons with dementia who cannot report pain symptoms,

"the use of a limited trial of analgesics in response to a broad array of physical and

behavioral symptoms" may be a good method of determining whether or not the person is

experiencing pain and may provide the most appropriate intervention for the patient (p.

417). Ferrell (1995) reports that acetaminophen is the most commonly prescribed

analgesic for elderly nursing home patients, and supports its use as the preferred

analgesic choice for patients without substantial inflammation because of its lower side-

effect profile. In fact, in a study conducted by Chinball and colleagues (2005), where

3,000 mg of acetaminophen was administered in 3 daily 1,000 mg doses for 4

consecutive weeks, there were no adverse events related to the acetaminophen and all

post-study liver function tests were normal. Likewise, in a sample of 519, only one

person had an allergy to acetaminophen and one other had a drug interaction with

acetaminophen (Buffum et al., 2004).

The American College of Rheumatology (2000), the European League Against

Rheumatism (2003), and the British Medical Journal (2000) all agree that for persons

with OA experiencing mild to moderate joint pain, the appropriate first line

pharmacologic agent is acetaminophen (Tylenol). This is in agreement with the

guidelines derived specifically for older adults by the American Geriatrics Society (2002)

described above. It should be noted, however, that none of these guidelines gives

specific recommendations for persons with dementia. However, the general scientific

opinion is that dementia does not affect the experience of pain, just the memory of it.

Therefore, it follows that these guidelines should be applicable and safe in older adults

with dementia.









In summary, further investigation into pain management in older adults with

dementia warrants attention due to the promising results found by previous researchers.

Because the prevalence of musculoskeletal pain in older adults in general is large and

there is no reason to believe that persons with dementia do not experience equivalent

levels of pain as cognitively intact older adults, it is imperative that this pain be assessed

and managed effectively. Several methods of assessing pain behaviors in cognitively

impaired older adults have been shown to be reliable and valid. However, many need to

be evaluated in terms of showing sensitivity to treatment effects. The creators of the Pain

Assessment for the Dementing Elderly (PADE) state that a limitation of their validation

of this tool was that it lacked an intervention phase; specifically, they state that it would

be of interest to see if PADE scores were to decrease in response to analgesic

administration (Villanueva et al., 2003).

The other area of considerable importance is the relationship between chronic pain

and behavioral problems as well as limitations in functional performance in persons with

dementia. It is these excess disabilities that place a significant amount of stress and

burden on the caregiver. If treating the underlying chronic pain condition can have a

positive effect on behavioral problems and functional limitations, this would be of

clinical significance to the caregiver as well as lead to a better overall quality of life for

the person with dementia. Clinicians have begun to recommend that behavioral problems

in persons with dementia need to be considered in the multiple contexts from which they

can develop, many of which are other identifiable problems, such as pain, and that

treating these behavioral problems should always begin with correcting any underlying

medical conditions and alleviating any discomfort (Kahn et al., 1998; Lantz, 2003).










Therefore, the proposed pilot study aims to evaluate the effects of an analgesic trial in

reducing observable pain behaviors as well as reducing behavioral problems and

functional limitations common in persons with dementia. In order to accurately assess

and evaluate the effects of any intervention, it is imperative to know the patient' s baseline

level of behavior, compare any changes in behavior over time, initiate interventions, and

then evaluate the effect of these interventions with the same assessment tools (Buffum et

al., 2001).

Summary

A review of the literature exploring the assessment and treatment of chronic pain in

persons with dementia provided support for the need for an analgesic trial to reduce pain

behaviors in this population. Furthermore, this review highlighted several issues which

were addressed in the present study, including the use of caregiver assessments of care

recipient pain and the relationship between chronic untreated pain and excess disabilities.

Based on the recommendations of the leading treatment guidelines and the paucity of

studies investigating drug treatment for pain in older adults with chronic pain and

dementia, acetaminophen was the appropriate treatment choice for this feasibility study.

The literature review also supported investigating the effect of around-the-clock dosing

of acetaminophen on the frequency and severity of excess disabilities in persons with

dementia. Finally, the use of observational measures to assess change in pain behaviors

after the implementation of a pain treatment intervention was strongly supported in the

literature.















CHAPTER 3
IVETHOD S

Study Design

This feasibility pilot study was conducted using an ABAB single-subj ect design.

Through this design, the research participant served as his or her own control through the

use of repeated observations which controls for inter-subj ect variability (Crosbie, 1995).

This particular design was used in order to investigate the feasibility and effect of

implementing a scheduled analgesic trial on participant pain behaviors and excess

disabilities related to chronic pain, and also attempted to capture the significance of an

individuals' change in behavior. The ABAB design was able to examine the effects of

acetaminophen by alternating a baseline condition (phase A- no acetaminophen

administered) with the intervention condition (phase B). These phases were then

repeated to complete the four phase design. In this type of design, the effects of the

intervention are evident if measures of participant pain behaviors and excess disabilities

decrease during the first intervention phase, "revert to or approach original baseline

levels of performance when treatment is withdrawn, and improve when treatment is

reinstated in the second intervention phase" (Kazdin, 1982, p. 110). This longitudinal,

feasibility design to study the effects of acetaminophen in persons with dementia is

needed as there is a paucity of literature on the effectiveness of acetaminophen to

alleviate pain in community-dwelling persons with dementia. Thus, the study was

intended to provide important preliminary data about this important, but understudied

phenomenon.









One benefit of SSD is that it allows for a relatively small N, generally at least three

subj ects are required, which is convenient when the target population is difficult to recruit

for research studies (as was the case for community-dwelling persons with dementia and

their caregivers) (Kazdin, 1982). SSD analysis has traditionally consisted of visual

inspection of the graphic display of data, with occasional use of statistical tests to

evaluate the reliability of findings, especially in cases with variable or trend-affected

baselines (Greenwood & Matyas, 1990; Kazdin). This allows the researcher to view an

individual's data points during the baseline observation phase and from the intervention

phase on an ongoing basis. Then, judgment for when to switch phases, as well as about

an intervention' s effects, is made based on the overall pattern of the data (Hoerster,

Hickey, & Bourgeois, 2001; Morgan & Morgan, 2001).

A drawback of SSD is that the number of measurement occasions needed is

dependent upon how quickly baseline stability is established in order to begin the

treatment phase and then how long it takes to observe a treatment effect. Thus, the length

of a phase depends on the nature of the phenomenon being investigated. If it is relatively

stable in nature and responds quickly to an intervention, the study phases can be

relatively short. Likewise, if it is more labile in nature and slower to respond to an

intervention, the phases will need to be longer to observe baseline stability and the

intervention's effect. Since pain is generally somewhat variable in nature, and the

effectiveness of acetaminophen on reducing chronic pain in persons with dementia is

unknown, this study departed from the traditional methods for changing phases. Rather

than observing for stability in pain behaviors in the initial baseline phase before

introducing the intervention, the treatment was give after 8 days of baseline data









collection. This allowed for the collection of information on the intra-individual

variability of pain behaviors in persons with dementia. Also, due to the unknown effect

of acetaminophen or how long it may take to show any effect, the initial intervention

phase was created to last the same number of sessions (8) as the baseline phase in order

to give the design equivalence.

SSD has not been widely employed in nursing research, in part due to some of the

criticisms about determining the effect of an intervention through visual inspection.

Despite the reliable criteria for visual inspection (such as comparing the magnitude and

rate of change to the variability of performance during a phase, the duration of the phase,

and the consistency of the effect across phases or baselines), critiques of SSD persist

(Kazdin, 1982). One criticism has been that visual inspection of data caters to

interventions with strong effects (Greenwood & Matyas, 1990). If an intervention's

effects are strong, there will be convincing graphical evidence that the intervention was

responsible for the change in behavior, thus virtually eliminating any chance of a Type II

error. However, when an intervention's effects are more subtle or more variable, there is

a high risk for Type I error as the graphical analyses of this intervention' s effect may find

that the intervention failed to effect any change in behavior (Crosbie, 1995). This is,

when statistical procedures may be employed in order to test the strength of the

intervention's effects when they appear weak or non-existent based on graphical

analyses.

For this proposed feasibility study, a SSD provided the needed combination of

flexibility in sample size and measurement occasion requirements while paying rigorous

attention to the nature of individual change. Because the effect of acetaminophen in









reducing pain and its related behaviors in community-dwelling persons with dementia is

unknown, a SSD, with ongoing analysis of graphical data, allowed the researcher to

observe effects immediately. While this design does not allow for comparisons across

individuals and does not rely on complex statistics for analysis, it should be noted that the

proposed study is a feasibility study, and this design allowed the researcher to carefully

follow the trend of the data and adjust the study phases according to participant

performance. This design allowed the researcher to examine daily change in behavior

and to carefully format "the nature of change in the single case before seeking similarities

and differences across cases" (Nesselroade, 1984, p. 275).

Sample

Five community dwelling older adults (over age 65) with dementia and their

primary caregiver were recruited and enrolled to participate in this study in order to

obtain a target sample of three pairs to complete the study. Two pairs (40%) were

withdrawn by the Principal Investigator. Informed consent was obtained from the legally

authorized representative of all participants with dementia and assent was obtained from

all participants with dementia. The primary caregiver for each participant also signed an

informed consent form to participate in daily sessions with the Principal Investigator over

the course of a 4-6 week period, as well as to complete several daily measurement tools

and administering multiple doses of acetaminophen on a daily basis according to study

protocol. Each caregiver was also the health care surrogate and legally authorized

representative of the participant with dementia. Each caregiver and their care recipient

also resided together in the same dwelling.









Recruitment Procedures

Individuals with dementia and their primary caregivers were recruited from the

West Central and North Central Florida community. Initial recruitment included working

with local chapters of the Alzheimer' s Association through which the PI gained access to

present the study to members of Alzheimer' s Caregiver support groups. Flyers were

posted and announcements were placed in local church bulletins. The PI also gained

permission to attend the University of Florida' s neurology clinics and the Memory

Disorders Clinic to identify possible participants. Also, local family practice and

geriatric physicians were asked to advertise the study to potentially eligible participants.

Inclusion and Exclusion Criteria for Participants with Dementia

Inclusion criteria consist of: 1) previously established diagnosis of probable

Alzheimer' s Disease (AD) or other related dementia by a physician, 2) score of 23 or less

on the Mini Mental State Exam (MMSE), 3) age 65 or over, 4) diagnosis of a painful

musculoskeletal condition, such as osteoarthritis (OA), 5) able to swallow an oral pill (as

assessed by the principal investigator), and 6) not currently taking prescribed analgesics

on a regular basis. As advocated by previous researchers, persons receiving routine

aspirin for cardiovascular prophylaxis (81 mg/day) were allowed participate in the study

and continue this regimen (Chinball et al., 2005). Participants were excluded if they are

unable to walk or have a hypersensitivity to acetaminophen that would prohibit them

from taking this drug. Also, persons with major health problems, especially any kidney

or liver disease, were excluded.

Inclusion and Exclusion Criteria for Caregivers

Caregivers recruited for this study were required to be the primary caregiver to the

participant with dementia as well as be able to read and write the English language









fluently. To be considered the primary caregiver, this person must have provided direct

care to the individual with dementia and have had daily contact with him/her so that they

were able to observe the participant over the course of the day.

Participant Pair 1, Caregiver 1 (CG 1) and Participant 1 (P 1)

The first participant pair consisted of a 58 year-old daughter (CG 1) who was the

primary caregiver for her 79 year-old mother (P 1). Both were Caucasian and educated at

the high school level. The mother was widowed and the daughter was married. Both

resided in the same house with the daughter's husband and her adult son. At the time of

the study, the daughter had been the primary caregiver for her mother for approximately

one year and was receiving respite care one afternoon each week.

Participant Pair 2, Caregiver 2 (CG 2) and Participant 2 (P 2)

The second participant pair consisted of a 75 year-old wife caring for her 83 year-

old husband. Both were Caucasian with a technical or trade school education. They

resided together in an independent dwelling with no other residents. At the time of the

study, the wife had been functioning as a caregiver for approximately 2 years and her

husband attended an adult day care group for approximately 5 hours a day on weekdays.

Participant Pair 3, Caregiver 3 (CG 3) and Participant 3 (P 3)

The third participant pair consisted of a 67 year-old daughter caring for her 93 year-

old mother. Both were Caucasian. The mother had an 8th grade or less education and the

daughter had attended some college. The mother was widowed and the daughter was

divorced. They resided together in the daughter' s home with no other residents. At the

time of the study, the daughter was employed full-time as an administrative program

assistant and had an independent aid supervise her mother during the day. She had been

serving as her mother' s caregiver for approximately 2 years.









Participant Pairs Withdrawn from the Study

As mentioned previously, two of the five pairs that were enrolled in this study

were withdrawn before completion by the PI. One of these pairs, an African-American

daughter caring for her mother, completed the first two study phases. However, during

the initial treatment phase (described below) the caregiver was unable to administer the

treatment drug according to protocol (i.e. the treatment was administered only one time

each day instead of the minimum of two doses to satisfy the study protocol which aimed

to provide continual pain relief). The second pair withdrawn consisted of a Caucasian

wife caring for her Caucasian husband. This pair completed only the first 6 sessions of

the study (the first phase, described below, was 8 total sessions). This pair was

withdrawn after it became clear that the participant with dementia was no longer

providing assent to participate in the study and was aggravated at being asked to perform

the activity protocol by his caregiver.

Measures

Comprehensive Intake Assessment

Once the caregiver expressed willingness to participate in this study, the principal

investigator arranged an initial meeting in the residence where the individual with

dementia receives care to explain the study, obtain informed consent, and to screen

participants for eligibility based on the inclusion and exclusion criteria. Caregivers

provided their own consent for study participation and also provided proxy consent for

the person with dementia, as each caregiver was also the health care surrogate and legally

authorized representative for their care recipient with dementia. (Copies of the informed

consent forms are provided in Appendices B and C). Assent from each individual with

dementia was also obtained at each session. Each participant with dementia was given









the Mini Mental State Examination (MMSE) by a trained research assistant (Folstein,

Folstein, & McHugh, 1975). A score of 23 or less was used as a cutoff to verify

cognitive impairment in participants with a diagnosis of dementia. Caregivers' cognitive

status was judged to be intact by the PI through conversation and training procedures

during this initial session; caregivers were not asked to complete formal cognition tests.

Demographic information, including age, sex, gender, marital status, education

level, and race was collected. Also a medication log sheet was used to compile care

recipients' prescribed and over the counter medications. Both caregivers and individuals

with dementia were also screened for depression with the 15 item form of the Geriatric

Depression Scale (GDS) (Sheikh & Yesavage, 1986), or the Center for Epidemiological

Studies Depression Scale (CES-D) if they were less than 65 years of age. The Dementia

Rating Scale-2 (DRS-2) (Mattis, Jurica, & Leitten, 2001) was administered to individuals

with dementia in order to gain more accurate baseline measures of cognition as well as to

thoroughly describe sample characteristics.

Caregiver Ratings of Pain Behaviors and Excess Disabilities

The Pain Assessment for the Dementing Elderly (PADE) was used to assess pain

behaviors and pain-related disability. This tool was originally designed to assess pain in

older adults with dementia residing in long-term care facilities (Villanueva et al., 2003).

This tool was divided into three parts (i.e. subscales) and contained 24 total questions.

Responses were provided using either a Likert scale (scored 1-4, or 0 for a skipped or

N/A answer) or multiple choice (scored 1-4) to elicit information about the effects of pain

on physical, global assessment, and functional areas. Part I of this measure assessed

physical manifestations of pain such as observable facial expressions, breathing patterns,

and posture. Responses on Part I were scored so that higher scores were representative of










higher distress. Part II consisted of one question which allowed the caregiver to rate, on a

Likert Scale (none to severe), their global assessment of the individual's level of pain at

the time of observation. Part III assessed functional performance and allowed caregivers

to rate the individual's performance of ADLs. Part III was scored so that higher scores

represented less independence and higher difficulties with ADLs. (See Appendix D for a

copy of the PADE tool). Because the directions for the PADE tool asked caregivers to

base their responses to Parts I and II of this measure on a 5- to 10-minute observation of

the subj ect, caregivers were asked to complete these sections after observing the

participant with dementia perform a 10-minute activity protocol (described below).

This measure was chosen because it was designed for caregivers to use to rate

global assessments of pain, pain behaviors (e.g. facial expressions, breathing, posture),

and ability to perform activities of daily living. Another reason for including this

measurement tool is that it has shown significant correlations with measures of agitation

in the elders with dementia and has proven to be quick and easy for caregivers to use.

PADE intraclass correlation coefficients (ICC) were generally high (interrater ICC= 0.54-

0.95; stability ICC= 0.70-0.98; internal consistency (a) =0.24-0.88). The PADE was also

significantly correlated with the Cohen-Mansfield Agitation Inventory (r= .30-.42). The

PADE was also able to differentiate individuals who were judged to suffer from clinically

problematic pain from those who were not.

The Revised Memory and Behavior Problems Checklist (RMBPC) was designed to

allow caregivers to measure problems with memory, depression, and disruptive behaviors

exhibited by the person with dementia (Teri et al., 1992), and it is the most commonly

used tool to measure behavior problems in community-dwelling individuals with









dementia (Allen, Kwak, Lokken, & Haley, 2003). This measure consists of 24 items and

asks the caregiver to report the frequency of the problem in question and their reaction to

this problem (i.e., how much it bothered them) over the preceding week. For the

purposes of the proposed study, this was modified slightly to have caregivers report the

frequency of and their reaction to problem behaviors for each day of the study. (See

Appendix D for a copuy of the RMBPC). The RMBPC has shown good overall

reliability for patient behavior (a= 0.84) and caregiver reaction (a= 0.90). The measure

also has been shown to be valid as the depression subsection was positively correlated

with Hamilton Depression Rating Scale (r=.44) and the memory-related problems

subsection was positively correlated with the MMSE (r=.48).

Caregivers also completed a third, brief, free-response questionnaire composed of

the following two open-ended questions: 1) How typical was this day compared to most

days? (i.e. where there any doctor' s appointments, visitors, etc.) and 2) What behaviors

did you notice that were most representative of pain in the participant today? (Please

include anything that signaled pain to you, even if it was not part of the other two

questionnaires). (See Appendix D for a copy of this questionnaire).

These measures were selected because they were designed for use by caregivers to

rate specific behaviors related to the study's specific aims. Together, these measures also

addressed each of the 6 main types of pain behaviors described in American Geriatrics

Society's (2002) pain behavior framework. Daily administration of the PADE allowed for

frequent assessment of participants' pain behaviors in order to examine daily variations

of these behaviors (during the observation phases) and the effectiveness of

acetaminophen in reducing these behaviors (during the treatment phases). Also, because









measures of ADLs were recorded in this tool, the effectiveness of acetaminophen in

increasing participants' functional ability was also assessed. Daily assessments with the

RMBPC allowed the researcher to evaluate daily variation in participants' mood and

problem behaviors (in observation phases) as well as the efficacy that reducing pain has

in secondarily improving mood and decreasing problem behaviors in persons with

dementia (in treatment phases). See Table 3-1 for a summary of study constructs and

measures. Also, by assessing the typicality of the day, any potential influences on

participant behavior or caregiver burden, such as the increased stress of taking the

participant to a doctor's appointment, were noted. Also, it was thought that asking

caregivers what pain behaviors they are most noticing would allow the researcher to

establish which behaviors are most useful in signaling pain to community-dwelling

dementia care providers.

Trained Observer Ratings of Pain Behaviors

The trained observer (i.e., the Principle Investigator or a trained research assistant

who was familiar with administering the study tools and who was physically present at

the data collection session) also completed the PADE (Parts I and II) daily as well as the

frequency portion of the RMPBC for the first and last day of the study. (In this study, the

trained observers consisted of the PI and two trained research assistants; one a PhD

candidate in nursing and the other a senior high school student in the International

Baccalaureate program). Since both the trained observer and the caregiver based their

measures on the PADE part I and II on the same 10-minute observation of the participant

performing the activity protocol (described below), this provided a way to measure inter-

rater reliability between the two groups for assessing pain behavior.










Table 3-1. Study Constructs and Measures
Constructs Measure Subject Burden
Pain
-Caregiver and -Sum score of PADE Part 1: -PADE total time to complete
trained observer Physical (observable facial for Parts I-III is 5-10 minutes
reports of pain expression, breathing pattern,
and posture); and Part 2:
Global assessment of overall
pain
-Behavioral -Total number of pain -20 minutes total for activity
observation of pain behaviors observed during an protocol, 10 minutes to set up
activity protocol and take down equipment and
10 minutes for videotaping
-Self-reported pain -average score on the NRS 1 minute
Pain outcomes/
correlates
-Functional -Sum score of PADE Part 3: -PADE total time to complete
performance Functional (ADLs) for Parts I-III is 5-10 minutes

-Behavioral -Total and subgroup scores on -RMPBC total time to
disruptions RMPBC for memory-related complete is 10-15 minutes
problems, depression, and
disruptive behaviors
PADE: Pain Assessment for the 1)ementing Elderly
ADLs: Activities of Daily Living, including dressing, feeding oneself, and transfers
RMPBC: Revised Memory and Problem Behavior Checklist
NRS: Numeric Rating Scale

As mentioned previously, participants also performed a videotaped activity-based

protocol derived from Keefe and Block' s (1982) work with chronic low back pain

patients. Initial findings using the behavioral observation system showed a highly

significant correlation between pain rating with total frequency of observed pain

behaviors (r = .71, p < .01). This activity based protocol was previously revised to

facilitate assessment of pain behaviors in persons with dementia by Ann L. Horgas, PhD,

RN; RO1 #NR05069-01. Using this revised protocol, each participant was asked to

perform 4 normal activities of daily living (sitting, standing, lying, and walking in place)

as well as transferring to and from each of these activity states. Each activity was

performed for a period of one minute, with activities being repeated to reach a behavioral









observation time total of 10 minutes. The order of these activities was randomized for

each session and the activity protocol was videotaped. The trained observer coded each

videotaped activity protocol for the specific pain behaviors described below. Since each

participant must have a diagnosis of OA or other painful condition, activities of daily

living are deemed to be sufficient stimulants of pain because they represent activities that

people with chronic pain often have difficulty with, but that are not too difficult for

participants to perform (Keefe et al., 1986). These activities are also ecologically valid as

they represent normal and generalizable everyday activities for all older adults who are

not so frail as to be confined to a bed or wheelchair. Furthermore, previous studies have

shown that increases in observable pain behaviors during activities rather than at rest

(Herr et al., 2006).

Participants were also asked at each session to self-report their pain level using the

numeric rating scale (NRS), both before starting the activity-based protocol and after the

protocol was completed. The NRS is the self-report tool of choice for this study because

it has been considered conceptually easier to understand by the elderly. This may be

because of the added number selections on this scale as compared to other pain rating

scales (Herr & Mobily, 1993). The NRS was presented as a horizontal line with 0= no

pain as the left anchor and 10= worst pain as the right anchor with equally spaced dashes

for each number 1-9. Participants selected the number from 0 to 10 that most accurately

identifies the pain in question. This measure has been shown to be both valid and reliable

in older adults (Gagliese & Melzack, 2003). See Table 3-1 for a summary of study

constructs and measures.









Treatment Protocol

For the purpose of this study, Tylenol Arthritis Pain Extended Relief was the

analgesic of choice for use as the treatment drug. This form of acetaminophen was

chosen for its more convenient 3 times per day dosing over the 4 times per day dosing of

Tylenol Extra Strength and Regular Strength Tylenol. Tylenol Arthritis has the same

safety and side effect profile as the other forms of Tylenol. The difference is that each

650 mg pill of acetaminophen has an immediate release outer layer and a delayed

released inner core so each dose (2 pills) provides up to 8 hours of pain relief.

Furthermore, "most of the acetaminophen dose is released from the tablet matrix within 5

hours, but peak acetaminophen serum concentrations may be delayed 8 hours or longer

following ingestion" (BC Drug and Poison Information Center, 2005, p.1).

Acetaminophen is thought to cause analgesia by inhibiting prostaglandin synthesis in the

central nervous system (i.e. elevation of the pain threshold) and the therapeutic serum

levels for analgesia are 5-20 mcg/mL (Health Digest, 2006; Physician's Desk Reference,

2006). The dose response curve specific for Tylenol Arthritis medication indicate that

the peak plasma concentration level (near 9.5 mcg/mL) is reached between 1 and 2 hours

after administration, and declines to nontherapeutic levels (less than 5 mcg/mL) 6 hours

after administration (McNeil PPC, 2002).

The dosing schedule for the treatment drug was designed for each individual

participant to follow a continuous dosing pattern during the participant' s waking hours.

This means that the caregiver was asked to give two 650 mg pills to the participant every

eight hours while the participant was awake. The typical sleep and wake schedule of

each participant was taken into account when devising a typical dosing schedule. For

example, a possible dosing schedule could then be to take the first dose with the morning









meal (9-10am) and a second dose with the evening meal (5-6pm). Alternatively, if the

participant is currently taken another medication every eight hours, the caregiver was

asked to give the treatment drug along with the other medications) as long as there were

no contraindications to doing so.

The Tylenol Arthritis medication was provided to the participants free of charge by

the PI. A large enough bottle to contain the amount of pills needed for the entire study

was supplied. Participants were allowed to retain any remaining pills at the end of the

study. The dosing schedule was worked out in advance with each caregiver to fit the

daily routine for the individual with dementia. Caregivers were also provided with a

medication log highlighting this dosing schedule to serve as a reminder to administer the

treatment drug. Based on the participants' daily routines, a 2 time/day dosing schedule

(morning after waking and approximately one hour before bed) was established for

participants 1 and 3. Since participant 2 was awake more hours of the day, a 3 time/day

dosing schedule was established (morning, early afternoon, and late evening). However,

it is important to note that although there were dosing differences, each of the participants

received the maximum number of doses during waking hours in order to provide

continual pain relief without disrupting sleep patterns.

Statistical Power

Because this study was a feasibility pilot study to test the effects of an

acetaminophen trial, the power of this study was not computed. There have been few

research studies specifically looking at the statistical power of acetaminophen in reducing

pain in persons with dementia. Although Buffum, Sands, Miaskowski, Brod, and

Washburn (2004), found that 650 mg of acetaminophen given four times per day was no

more effective than prn acetaminophen in reducing discomfort in persons with dementia,









this study targeted nursing home residents with severe dementia. It is quite possible that

the results may be different in community-dwelling persons with more moderate

dementia. Likewise, in an informal study, Douzjian and colleagues (1998) found 650 mg

doses of acetaminophen three times a day to be an effective pain reliever. This lends

support that using an even stronger dosage (two 650 mg pills up to three times per day)

would also be effective in reducing pain. Furthermore, Chinball and colleagues (2005)

found moderate-to-strong effect sizes (eta2 = 0. 14-0.29) for acetaminophen in improving

outcomes assessed by the Dementia Care Mapping tool in nursing home residents with

moderate-to-severe dementia.

While acetaminophen is the first line recommended analgesic, it is not the most

powerfully acting analgesic available. However, it is reasonable to assume that given its

endorsement by leading geriatric treatment guidelines (American College of

Rheumatology, 2000; American Geriatrics Society, 2002; British Medical Journal, 2000;

European League Against Rheumatism, 2003), that it will produce at least a small effect.

Also, the validity and reliabilities of the PADE, RMBPC, and the original activity

protocol designed by Keefe and Block (1982) as well as the revised activity protocol

(Horgas, 2001), support that these tools will useful in detecting changes in participant

behavior.

Procedures

As shown in Table 3-2, this ABAB single-subj ect study design consisted of an

initial baseline daily observational phase (Al) lasting a minimum of eight sessions within

a two-week period to observe behavior under conditions before treatment was

implemented. As described above, this study departed from the traditional methods for














Table 3-2. Study Design

Variables ISession 1 Baseline Intervention Baseline Intervention
(Baseline) Session Session Session Session
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 121 22 23 24
Inae MeasuresX

Otae Measures I X

Caregiver IX X
Intuton I(Tool I(Medication
Administration) Administration)
Caregiver Rating
-PADE IX IXX XX X XX X XX XX X XX IXX XX X XX X
-RMBPC IX IXX XX X XX X XX XX X XX IXX XX X XX X
Acetaminophen IIIX XX XX X XX X XX X
Activity Protocol IX XX XX X XX IX XX XX X XX X X XX X XX X










changing phases. While the target goal was to have daily sessions, due to the

unpredictable nature of collecting repeated measurements in persons with dementia, it

was thought that it might not be convenient or possible for the caregivers to meet with PI

on subsequent days throughout the study period. For this reason, guidelines for a

minimum number of sessions within a maximum time-period were established. Typically

the initial baseline phase is continued until the participant's behavior appears to be stable

"or until it is evident that the response does not improve over time" (Kazdin, 1982,

p11.Since pain is labile in nature, it was proposed that at least eight baseline sessions

would be needed to observe each individual's typical pain behavior patterns. The

baseline phases of this design were necessary to describe the current level of behavior as

well as to predict what future behavior would have looked like without an intervention.

Once the typical pain behavior pattern was observed, the intervention phase (B1)

was initiated. In an effort to create equivalence between study phases, this phase was

also eight sessions in length (within a two-week period). Similarly to the baseline phases,

the intervention phases describe current behavior and predict future behavior if the

intervention were to be continued. However, the intervention phases also test the

predictions from the baseline phases. That is, the investigator can test whether behavior

in the intervention phase actually departs from the predicted baseline level. The return to

baseline phase (A2) began with the withdrawal of the intervention in an attempt to show

that the participant's behavior would return to baseline performance without the

intervention. This phase was planned to last a minimum of four sessions (if behavior

returned to baseline levels rapidly) to a maximum of eight sessions. The unique purpose

of this second A phase was to test the prediction from the B1 phase. If behavior returned










to A1, levels then it would be clear that the predicted and obtained performance levels

differ, suggesting that the intervention withdrawal altered behavior. Following the return

to baseline phase, the intervention was reimplemented (phase B2). This phase was also

planned to last a minimum of four sessions to a maximum of eight sessions. If in this

phase, behavior again improves, the overall pattern of results would strongly suggest that

the intervention was responsible for the change in behavior.

During the study, the Principal Investigator or a trained research assistant

conducted all activity protocols using a digital video camera to capture each session. An

effort was be made to try to schedule all videotaped sessions at the same time each day in

order to establish a routine that will cause less burden for the caregivers and to try to

control for variability in pain experienced at different times of the day. Also, it should be

noted that because the half-life of acetaminophen is 1 to 3 hours, the standard washout of

3.5 half-lives would occur within 10 hours (Buffum et al., 2004). Thus, there is no need

for a washout period between phases B1 and A2. During any days were the caregivers

were unable to meet with the research team, they were told they could complete the daily

measurement tools for that day if they chose too. However, CG 1 completed all 24

sessions within 25 days, CG 2 completed the study within 24 days; and CG 3 completed

the study within 26 days, so there were not many instances of gaps between visits by the

research team.

Phase Al, Session 1

Day 1 of the study consisted of participant and caregiver screening for enrollment.

A trained research assistant administered the IVMSE separately to the participants with

dementia to confirm cognitive impairment. The participants were also given a

comprehensive intake assessment in order to thoroughly describe and evaluate sample









character sti cs. Consent was obtained from the caregivers, who were each, also the

legally authorized representative of their care recipient and assent was obtained from the

participants with dementia themselves. Caregivers were asked to provide a current list of

all of the medications that the participant received, including any over the counter

medications. A medication log for each phase of the study listing all of the participant' s

medications was created and given to the caregivers to use as a log to record daily

administration of all medications. In an effort to increase the reliability of caregiver

ratings, caregivers also received instruction and training in administering the PADE and

the RMBPC. Sample videotapes were shown to the caregivers in an effort to increase

their knowledge of pain behavior as well as their ability to identify excess disabilities.

They were asked to administer these tools themselves in front of the principal

investigator to ensure understanding of tool administration.

Caregivers received a packet of measurement tools for each phase of data

collection. To facilitate data collection, the measurement tools were printed in 14-point

font on white paper (to facilitate reading). In addition, the packet was organized and

clearly labeled to delineate the correct measure for each session. Also, on this day, the

initial videotaped activity protocol was completed. Finally, caregivers and the trained

observer completed the PADE and RMBPC regarding participant behaviors, and the

caregivers answered the questions regarding the typicality of the day as well as if they

noticed any pain behaviors that they thought most indicated pain.

Phase Al

Throughout this baseline observation phase, the PI or trained RA completed the

activity protocol and the PADE, and the caregivers completed both the PADE and the

RMBPC during the visit. Since an effort was made to time study sessions at









approximately the same time each day, caregivers were instructed to think back over the

participant' s behavior since the previous day's session (or for the previous 24 hours if

sessions did not occur on subsequent days) and to accurately reflect this behavior on the

measurement tools. Additionally, they were asked to evaluate if this was a typical or

atypical day for the participant (i.e. were there visitors, were they ill, or did they visit the

doctor?) and to describe what behaviors they most noticed as indicating pain. Caregivers

also recorded administration of all medications on the log sheet provided.

Phase Al-Final Session

For the final session in the observation phase (8th SCSsion), the PI instructed

caregivers on how to appropriately administer the treatment drug, acetaminophen, and

worked with the caregiver to make a convenient dosing schedule/plan. After setting up

the schedule (e.g. 900 hours and 1700 hours or morning meal and evening meal) for each

individual, the caregiver administered the first dose of acetaminophen at the next

scheduled time, after the activity protocol for that day was completed. Caregivers then

received a packet for the second study phase including the daily questionnaires and

medication log sheets with an additional space for recording when the treatment drug was

given. Again, the PI and the caregivers completed the daily behavioral assessment tools

and the caregivers provided information regarding the typicality of the day as well as the

pain behaviors they noticed.

Phase B1 (Eight Sessions)

During the second study phase, (the first intervention phase), caregivers continued

to administer both the PADE and the RMBPC during the daily visits and recorded if the

day was a typical day for the participant as well as the pain behaviors they noticed in the

participant. In addition, the caregivers gave the treatment drug, as per the previously









established schedule, and recorded the time of administration on the log sheets provided.

Participants continued to perform the videotaped activity protocol and the PI or trained

RA also completed the PADE.

Phase A2- Return to Baseline (Range of 4-8 Sessions)

Caregivers continued to log all medication administration during this week but

were instructed to discontinue giving the treatment drug. Caregivers were, however,

instructed that they could administer rescue doses of pain medication as needed by the

participant, but to make sure to log these doses on the medication log sheet. Data

collection proceeded as before in Phase Al.

Phase B2- Reimplementation of Intervention (Range of 4-8 Sessions)

This phase proceeded exactly as the original intervention phase. Caregivers were

instructed to resume giving the intervention drug at the previously scheduled dosing

times, logging all medications given. Participants continued to perform the activity

protocol videotaped by the PI or trained RA, and the both the PI or trained RA and the

caregivers completed the measurement tools as previously described. Caregivers also

provided information regarding the typicality of the day as well as the most frequent

behaviors that they saw in the participants that represented pain. On the last day of this

phase, qualitative information regarding the caregivers' perceptions and satisfaction of

the pain treatment was gathered. The PI or the trained RA also administered the Geriatric

Depression Scale and the Dementia Rating Scale to the participant with dementia and the

depression scale to the caregiver as part of an outtake assessment to evaluate changes in

depression and cognitive performance over the course of the study.









Data Coding

Each videotaped activity protocol was coded using Noldus The Observer software

the day it was recorded. This enabled the PI (who was the primary coder) to evaluate

participant performance on an immediate basis in order to accurately assess behavior and

the need for changes in study phases. Since all videotaping was done using digital video

cameras, the media files were transferred to the computer using Dazzle Video Pro

software and were then opened in The Observer' s viewing window. Each media file was

saved on an external hard drive as a backup. A coding program was created in the

Observer that allowed independent coders to code the media files for specific pain

behaviors. The independent variables entered in this program were participant ID

number, study phase, and session number of the phase. Two categories of behaviors

were used in this program, state behaviors and event behaviors. State behaviors are

behaviors that occur for an unspecified amount of time, this requires that a behavior

remain 'active' for the entire duration that this behavior is observed. Event behaviors are

behaviors that are observed, but the duration of the behavior is not coded. The Observer

requires that all state behaviors in a behavioral class be mutually exclusive; that is, no

two behaviors in the same behavioral class can be active at the same time.

A second independent trained coder (who was an honor' s student in the

undergraduate nursing program) coded 25% of all videotapes to ensure inter-rater

reliability and primary coder accuracy. This second coder was blind to the phase of the

study in which videotaped sessions had occurred. Coder reliability was measured with

the Kappa statistic (K) which not only takes into account the frequency of agreements

between coders, but also accounts for the number of agreements one would expect as a

function of chance (Blomqvist & Hallberg, 1999). A Kappa of 0.80 was established as









the criterion for coder training. Coders trained on practice tapes until their intra-rater

reliabilities were greater than a kappa of 0.80 and were considered trained when their

inter-rater reliabilities were greater than a kappa of 0.80 on practice tapes when compared

with the master coder. For this study, the two raters reached an average inter-rater

agreement of kappa =0.84, or 84% on the practice tapes (K range= 0.72-0.96), which

meets the established criterion for training (Elder, 1999; Keefe, Bradley, & Crisson,

1990; Keefe & Hill, 1985). Reliability estimates for the actual study tapes were also

good at kappa =0.80 (K range=0.62-0.98). This was considered to indicative of good

reliability as there were 9 behaviors, 5 activities, and an actual total of 17 codes that

coders used for the same observation, and other researchers have proposed that reliability

coefficients between .60-.80 indicate good to very good reliability (Blomqvist &

Hallberg; Gibson & Helme, 2000). Training strategies to increase inter-rater reliability

included informal discussions about coding behaviors, independent study of the training

manual (containing definitions and instructions for coding the variables of interest), and

practice coding sessions with the 'master coder' and all research assistants

The two independent coders coded participants' activity protocols by recording the

specified pain behaviors of interest that have been adapted and modified by Horgas

(2001) from definitions of behavioral coding categories originally employed by Keefe et

al. (1986). See Table 3-3 for the specific behaviors of interest for this study.

Data Analysis Procedures and Variables of Interest

RMBPC

The RMBPC yields both a global summary score and subscores for participants'

behavioral disturbance and caregivers' reactivity. Also, frequency and reaction scores for









the three specific subscales (Memory-Related Problems, Depression, and Disruptive

Behaviors) were computed.

PADE

The PADE is divided into three parts with Part I representing the physical domain,

Part II representing a global assessment domain, and Part III representing the functional

domain. Since questions are answered on either a Likert scale (range 1-4, or 0 for

skipped answer or N/A) or with multiple choice (1, 2, 3, or 4), sum scores for each of the

three parts were computed along with a total measure score.

Activity Protocol

Coding of the activity protocol yielded information about the duration and frequency

of state behaviors (i.e. guarding, bracing, & rigidity) and the frequency of event

behaviors (i.e. sighing/nonverbals, rubbing, stopping, grimacing, shifting, and

vocalizations). Therefore, the mean time spent in a state behavior and the mean

frequencies of all behaviors were calculated by Noldus the Observer software and used in

the analyses.

Additional Caregiver Questionnaire

Caregiver responses as to the pain behaviors that they most frequently noticed to be

indicative of pain in the participant were tabulated with frequency counts of specific

behaviors listed by the caregiver.

Overview of the Analysis of Single-Subject Design Data

The overall obj ective of data analysis in any research study is to examine whether

a truthful change has been demonstrated and whether or not this change is attributable to

the intervention. In single-subject research, this is accomplished by examining the effects

of an intervention at different points over time (Kazdin, 1982). In ABAB design, the












Table 3-3. Behavioral Definitions
ACTIVITY MOVEMENT BRACING COMPLETING PAIN BEHAVIORS
(STATE) (STATE) (STATE) ACTIVITY (EVENTS)
(STATE


Sit = begin coding
when Ss' buttocks
rest on chair


Rigidity = Stiffness
of the lower body
(from back down to
feet) exhibited by
Ss maintaining the
affected lower body
part in a stiff and
abnormal position
during standing,
sitting, or lying down








Guarding =
abnormally slow,
stiff, interrupted, or
rigid movement -
during4 transfer~rtring ~rt~t~rtrt~
or walking


Bracing = holding
onto someone or
wwth,,rling for support
(e.g., walker, bed,
cane) participant-
initiated behavior.
If tester initiated
behavior, code as
bracing if there is
evidence that the
participant needed
support (e.g., was
unstable or falling.).
Can occur during sit,
stand, walk and
transfer. but not
during lying down.


Rubbing = touching, rubbing,
or holding the affected lower
body part (e.g., lower back/waist
to feet) (palms down)
Code 1 completed circular or
forward/backward movement as
one rub.












Shifting = shifting of weight.
This includes channinn position
or shifting weight (side to side
shifts) of the trunk, hips, or
lower extremities while doing an
activity. [Excludes front to back
sway.] Can occur during any
activiy; shifting during walking
would reflect purposeful shifts
in the lower back or trunk.


Stopping =
interrupting/stoppi ng
the activity state
during any activity
state except transfer ~rt~rt~t~rt~rt~
stopping the activity
they are supposed to
be doing
(Ex: while walking -
stopping to stand still
or sit down;
while standing,
stopping to sit down;
quitting the activity
state)


Stand = begin coding
when Ss is in an
upright position with
one or both feet on
the floor












Table 3-3. Continued
ACTIVITY MOVEMENT BRACING COMPLETING PAIN BEHAVIORS
(STATE) (STATE) (STATE) ACTIVITY (EVENTS)
(STATE


Walk = begin coding
when Ss is in an
upright position, and
begins moving feet in
marching/walking
pattern
Lie = begin coding
when Ss' head and
back touch the bed
(horizontal position)






Transfer = begins
when the 1-minute
activity state ends
(When the timer rings
to end the 1-minute
activity, code this as
beginning the transfer
activity). End transfer
when the next 1-
minute activity
begins


Grimacing = Obvious change in
facial expression, indicated by
frown (tightened lips, corners of
mouth pulled back or down) and
eye change (narrowed eyes and
furrowed brow).
Sighing/Non-verbal
Vocalization = an obvious
exaggerated exhalation of
breath, usually accompanied by
shoulders rising and falling or
change in lip/mouth position
OR any spontaneous non-verbal
sound such as moaning,
groaming, gasping, grunting.
Verbal Complaint =
spontaneous expression of pain
or discomfort in affected body
area (e.g., trunk, back, and lower
extremities). [Verbal responses
to questions or statements made
in the context of questioning do
not apply.]









effects of an intervention are replicated so a judgment as to the intervention's

effectiveness is made on the overall pattern of the data. If participant behavior changes

each time the intervention is applied and again when it is withdrawn, the effect of the

intervention is apparent. These judgments are made based on the visual representation of

the data. Visual inspection refers to "reaching a judgment about the reliability or

consistency of intervention effects by visually examining the graphed data" (Kazdin, p.

232). As part of the design requirements, participant data are graphed continuously over

the course of the study. These graphed data points represent the study behavior(s) of

interest. There are several characteristics of the data that are examined in visual analysis

of data. These characteristics are either related to the magnitude of change across phases

(mean and level) or the rate of change (trend and latency). In addition, overlap of data

points between phases and well as stability within a phase are also of interest (Kazdin).

A change in mean refers to a shift in the average rate of performance between

phases. For this study, higher means of pain behaviors and excess disabilities were

expected during the baseline phases and lower means were expected during the

intervention phases. A change in level refers to the shift or discontinuity of behavior

between the end of one phase and the beginning of the next phase. Since the trend or the

slope of the data shows whether behavior systematically increases or decreases over time,

a change in trend may be revealed when the intervention is applied or withdrawn. In the

present study, decreases in level were expected when shifting from a baseline phase to an

intervention phase, and increases in level were expected when shifting from an

intervention phase to a baseline phase. Likewise, the trend (slope) was expected to be in

the positive (increasing) direction during baseline phases and in the negative (decreasing)









direction during intervention phases. The latency of change refers to the period between

the onset or termination of one phase and subsequent changes in behavior and is assessed

by examining level and trend changes together. Therefore, the latency of change reveals

how long after an intervention is applied it takes to change behavior. The closer this

change takes place to the change in experimental condition, the more clear the effect of

the intervention. In summary, "visual inspection is conducted by judging the extent to

which changes in these characteristics are evident across phases and whether the changes

are consistent with the requirements of the (study) design" (Kazdin, 1982, p. 237).

Kazdin also points out that there are also specific background characteristics to examine

in relation to the visual analysis of data. These characteristics speak to the reliability of

an effect and are measured through examining the: variability of performance within a

particular phase (stability), the duration of the phase, and the consistency of the effect

across phases or baselines (overlap). (See the Appendix A for a detailed explanation for

performing these graphical analyses).

One potential drawback to visual analysis is that only interventions that produce a

large effect will be considered to have produced a change. In the case of an intervention

that produces more modest effects, statistical evaluation may be warranted. While

statistical evaluation is not the norm in single-subj ect research, there are several

circumstances in which it can reveal significant intervention effects that may have been

missed in visual analysis. Specifically, the statistical evaluation of single-subject data to

supplement visual analysis is appropriate when there are unstable baselines, new areas of

research are being investigated (intervention effects may be weak), there is increased

intrasubj ect variability (where there is little control over the environment and potential









influences), or when even small changes in behavior may be important and meaningful

(Kazdin, 1982). Appropriate statistical tests include conventional t and F tests (when

there is no serial dependency between data points) and time-series analysis.

Statistical Analysis of Each Aim

Aim 1: To investigate the effects of the scheduled administration of acetaminophen (1.3

grams three times per day) on self-reported pain intensity and the number of observable

pain behaviors exhibited by persons with dementia.

Hypothesis: Regular administration of acetaminophen will decrease the total

number of observable pain behaviors (as rated by the caregiver and trained observers)

exhibited by the participant, but will have no effect on self-reported pain intensities.

The variables of interest for these analyses included caregivers' and the PI' s or

trained RA' s daily responses on the PADE part I and the total frequency and duration of

individual pain behaviors as coded with the Observer from the videotaped activity

protocol. These totals were graphed against the day of measurement. As described

above, the standard in analyzing SSD data consists of visually inspecting dependent

measures (i.e. total number of pain behaviors) during independent variable conditions

(acetaminophen trial phase) relative to baseline measures (Morgan & Morgan, 2001).

The data analysis and graphical interpretation occurred in an ongoing manner such that

data were analyzed and graphed as soon as they were collected. At the conclusion of the

study, the graph for each participant was interpreted to look for an overall pattern in pain

behaviors during baseline phases before treatment and any change in the frequency or

duration of behaviors after initiation of pain treatment. Specifically, changes in mean,

level, trend, latency, overlap, and stability were examined. If warranted, statistical

evaluation using t and F tests to "compare whether differences in means are statistically









reliable between, or among, the different phases" (Kazdin, 1982, p. 245) or time-series

analysis to "examine whether there is a statistically significant change in level and trend

from one phase to the next" (Kazdin, p. 248) were planned.

Aim 2: To investigate the effects of the scheduled administration of acetaminophen on

the frequency and severity of excess disabilities of pain in persons with dementia.

Hypothesis: Among persons with dementia, the frequency and severity of excess

disabilities of pain, such as depression, memory problems, behavioral disruptions, and

impaired functional performance will decrease from baseline after implementing regular

administration of acetaminophen.

The variables of interest are the frequency ratings for each of the three subgroups

(memory-related problems, depression, and disruption) on the RMPBC and for the

measure as a whole which were calculated for each participant daily. Also of interest for

this aim are caregiver's responses on the PADE part III. Again, this data was graphed

and analyzed as described for Aim 1.

Aim 3: To determine the reliability of informal caregiver ratings of the frequency of pain

behaviors exhibited by persons with dementia.

Hypothesis: After an initial training session, informal caregivers will display

moderate to good reliability with a trained observer in ratings of the frequency of pain

behaviors.

The variables of interest were the caregivers' ratings on the PADE part I and the

trained observer's ratings on the PADE part I. To evaluate the reliability of informal

caregivers' ratings of pain behaviors, the caregiver' s ratings were compared to those of









the trained observer' s for each study session and the inter-rater reliability was calculated

using Pearson's correlation coefficient.

Aim 4: To investigate the most frequently displayed behavior(s) indicative of pain that

is/are displayed most frequently by persons with dementia.

Hypothesis: Based on findings from similar studies, the most frequently displayed

behaviors indicative of pain are expected to be guarding, rubbing, shifting, and bracing

(Horgas & Elliott, 2005; Keefe & Block, 1982).

The variables of interest are caregiver ratings on the PADE part I, the total number

of individual pain behaviors as coded with the Observer from the videotaped activity

protocol, and caregiver responses as to which behaviors they most noticed as being

indicative of pain in the participant. The most frequently rated and coded behaviors were

then extrapolated and this list of behaviors serves as the behaviors employed most by

persons with dementia to relay pain.















CHAPTER 4
RESULTS

The results of the present study are presented below. First, all descriptive statistics

are presented; specifically, scores on the initial cognition and depression measures for

each participant are compared to their scores on these measures from the final day of the

study. Initial and final caregiver depression scores are also presented. Next, the results

for each of the 4 specific aims are presented. Of note, for aims 1 and 2, graphical

representations of the variables of interest are presented individually for each participant.

Each graph was then analyzed by looking at changes in the mean between phases, the

latency of change between phases (which consisted of examining level and trend

changes), the overlap of data points between phases, and finally, the stability of data

within phases.

Descriptives

Initial Intake Measures

The intake Mini Mental State Exam (MMSE) score for Participant (P) 1 was 10,

indicating moderate (near severe) cognitive impairment while the initial Dementia Rating

Scale-2 (DRS-2) total score (31) indicated severe cognitive impairment. The intake

depression measure indicated no depression (Geriatric Depression Scale score = 0).

Initial depression scores for caregiver (CG) 1 indicated mild depression, (Center for

Epidemiological Studies Depression Scale score of 19).

The intake MMSE score for P 2 was 20, indicating moderate cognitive impairment,

while the initial DRS-2 total score (111) indicated severe impairment. The intake










depression measure indicated no depression (GDS = 1). Initial depression scores for CG

2 indicated no depression (GDS = 2) as well.

The intake MMSE score for P 3 was 5 indicating severe cognitive impairment and

the initial DRS-2 total score (65) indicated severe impairment as well. The intake

depression measure indicated no depression (GDS = 1) and initial depression scores for

CG 3 indicated no depression (GDS = 2). Demographic data along with these scores are

presented in Tables 4-1 & 4-2, comparing those pairs who completed the study to the two

pairs that were withdrawn from the study.

Final Outtake Measures

During the last study session, final outtake measures were taken from both the

participant and the caregiver. This was done specifically to see if there had been any

change in cognition (for the participants) or depression (for both participants and

caregivers) during the course of the study. Scores for these outtake measures are

presented in Tables 4-1 & 4-2. As can be seen outtake DRS-2 total scores were exactly

the same for Participants 1 & 2 as their intake scores. Interestingly, while there was some

variation in subscale scores between intake and outtake measures, the total score

remained the same. For P 3, the DRS-2 total score improved from 65 to 71, however,

this increase in total score did not change her severely impaired classification on this

measure.

Depression scores also did not show much change from intake assessment to

outtake assessment. Participant 1 remained at 0 both occasions, and while Participants 2

and 3 showed declines on the GDS (1 to 0 for P 2, and 3 to 1 for P 3), although neither

participant was initially classified as depressed. The same was true for Caregivers 2 and

3 (CG 2 decreased from GDS 2 to 1, CG 3 remained at GDS 2) who were not initially













Table 4-1. Descriptive Characteristics for Participants
ID Age Sex Racea Educationb Marital
Status


MMSE Initial
DRS-2
Score


Final
DRS-2
Score


31
111
71


Initial
Depression
Score (GDS)


Final
Depression
Score (GDS)


Participants
Included
Pl
P2
P3
Participants
Withdrawn


0 3


aO= Caucasian, 1= Black
bl __th grade or less, 2= 9-11th grades, 3= High School, 4= Technical or Trade School, 5= Some College
'0= Never been married, 1= Married, 2= Widowed, 3= Separated, 4= Divorced


Table 4-2. Descriptive Characteristics for Caregivers
ID Age Sex Race" Educationb Marital
Status'


Relation to
Participant


Daughter
Wife
Daughter


Initial Depression
Score



19 (CES-D)
2 (GDS)
2 (GDS)


Final Depression
Score



22 (CES-D)
1 (GDS)
2 (GDS)


Caregivers
Included
CG 1
CG 2
CG 3


F 0 3


Caregivers
Withdrawn
4 37 F 1 5 0 Daughter
5 77 F 0 3 1 Wife
aO= Caucasian, 1= Black
bl __th grade or less, 2= 9-11th grades, 3= High School, 4= Technical or Trade School, 5= Some College
'0= Never been married, 1= Married, 2= Widowed, 3= Separated, 4= Divorced


12 (CES-D)
5 (GDS)









classified as depressed. However, Caregiver 1, the only person in the study to initially

have a score in the depressed category (on the CES-D) increased from an initial score of

19 to an outtake score of 22

Effect of Acetaminophen on Self-Reported Pain Intensity and Observable Pain
Behaviors

The primary specific aim of this study was to assess the effect of scheduled doses

of acetaminophen on self-reported pain intensity and the number of observable pain

behaviors. Participants were asked to self-report their pain intensity using the numeric

rating scale (NRS) both immediately preceding and following the activity protocol. Over

the course of the study (24 sessions for 48 total occasions for self-report), P 1 self-

reported no pain on all occasions. P 2 used the NRS on four occasions (with this being

on both occasions at 2 sessions during the Einal treatment phase) to self-report mild pain

intensity (scores of 1s and 2s on the NRS). P 3 used the NRS on six occasions (twice for

both occasions at a session and twice preceding the activity protocol only) to self-report

pain intensity. On Hyve occasions, during both the baseline and treatment phases, this

pain was mild (1s and 2s on the NRS) and on one occasion during the initial baseline

phase, this pain was moderate (6 on the NRS).

The effect of acetaminophen on observable pain behaviors was evaluated by

comparing pain behaviors exhibited by participants' during baseline performance of an

activity protocol with their performance of the same protocol while receiving scheduled

doses of acetaminophen. Several methods were used to observe for changes in

performance. First, the total number and total duration of the previously defined target

pain behaviors (see Table 3-3) were coded for using Noldus The Observer software.

Changes in performance were also assessed with Part I of the PADE measure which was










completed daily by both the CG and the trained observer (TO) (i.e. either the Principle

Investigator or trained research assistant who was physically present at the data collection

session). In total, there were 4 measures employed to assess the effect of acetaminophen

on observed pain behaviors exhibited by persons with dementia. The results for each of

these measures are represented graphically for each participant. As part of the analyses

of graphed data, changes in mean, the latency of change (i.e. comparing level changes

between the end of one phase and the beginning of the next phase and the change in trend

or slope between phases), overlap of data between phases, and the stability of data within

a phase are presented. Complete descriptions of these components of graphical analysis

are presented in Chapter 3. (See Appendix A for examples of how the mean, level, trend,

latency, overlap, and stability were calculated).

Total Number of Pain Behaviors

Participant 1

Figure 4-1 displays P l's frequency of pain behavior for each session during

baseline and intervention conditions. Mean: During the initial baseline condition (Al),

the mean frequency of pain behavior exhibited during the activity protocol was 32. 1

behaviors per session. This mean frequency decreased to 18.6 after the introduction (Bl)

of the intervention (acetaminophen). Upon the return to baseline condition (A2), the

mean frequency of behavior increased to 27.5 and then decreased to 17.5 when the

intervention was reintroduced (B2).

Changes in both level and trend are analyzed together to examine the latency of

change between phases. Latency: There was an 11 point decrease in level between

phases Al and Bl, however, the trend in phase Al was initially in the unintended

direction (slope= -1.44) and flattened (slope= -0.04) in phase Bl. Between phases B1











Total Number of Pain Behaviors for Participant 1

45-Al B1 A2 B2








S20-







1 5 9 13 17 21
Sessions

Figure 4-1. Total number of pain behaviors for P 1 across sessions from the coded
activity protocols.

and A2, there was a small level change in the unintended direction (1 point decrease) but

the there was a large change in trend in the intended direction (slope= -0.04 to 5.8).

Finally, between phases A2 and B2, there was a 15 point decrease in level and a large

change in trend in the intended direction (slope= 5.8 to -2.8). Overlap: There was

nooverlap in data points between phases Al and B l, while 2 points overlapped between

phases B1 and A2, and 1 point overlapped between phases A2 and B2. Stability: There

was 100% stability in the data points in phases Al, A2, and B2, while phase B1 had

62.5% stability in data points.

The changes in mean frequencies for each individual pain behavior by phase for

Participant 1 are presented in Figure 4-2. Only behaviors which occurred greater than 25

times during the course of the study are included in these analyses. As can be seen, for










Mean Frequencyof Individual Pain Behaviors Across Phases for Participant 1







8 m Phase Al
m Phase B1
E ~OPhase A2
: O Phase B2






Mean Guarding Mean Rlgidity Mean Bracing Mean Shifting Mean
Slghing/Nonverbal
Behaviors

Figure 4-2. Mean frequencies for individual behaviors across phases for P 1.

each behavior except shifting, the mean frequencies are higher in the two baseline phases

and are lower during the treatment phases.

Participant 2

Figure 4-3 displays P 2's frequency of pain behavior for each session during baseline

and intervention conditions. Mean: The mean frequency of pain behavior in theinitial

baseline phase (Al) was 33 behaviors per session, which decreased to 22.5 during the

first intervention phase (B l). The mean frequency of behavior increased during the

return to baseline phase (A2) to 31.3 and decreased to 20. 1 when the intervention was re-

introduced (B2). Latency: There was a 9 point decrease in level from the last session in

phase Al to the first session in phase B l, and a change in trend from a slope of 1.0 in the

intended direction in phase Al to nearly no trend (slope = 0.02) in phase Bl. Although a

large change in level between phases B1 and A2 (14 point increase) was observed, the











Total Number of Pain Behaviors for Participant 2

45-Al B1 A2 B2









S20-








1 5 9 13 17 21
Sessions


Figure 4-3. Total number of pain behaviors for P 2 across sessions from the coded
activity protocols.

change in trend was in the unintended direction (slope= 0.02 to -1.5). Between phases

A2 and B2 there was a 10 point decrease, accompanied by a change in trend in the

unintended direction (slope= 0.9). Overlap: There was a 1 point overlap between phases

Al and B1 and also between phases B1 and A2. There was no overlap of data points

between phases A2 and B2. Stability: There was 100% stability in data for phases A2

and B2, 62.5% for Al, and 75% for phase Bl.

The changes in mean frequencies for individual pain behaviors by phase for P 2 are

presented in Figure 4-4. Behaviors which occurred greater than 25 times during the

course of the study are included in these analyses. As for P 1, for each behavior except

shifting, the mean frequencies of each behavior are higher in the two baseline phases and

are lower during both the treatment phases.











Mean Frequencyof Individual Pain Behaviors Across Phases for Participant 2








6- 5Phase B1
0 Phase A2







Mean Guarding Mean Bracing Mean Rubbing Mean Shifting Mean Gnmacing Mean
Sighing/Nonverbal
Behaviors



Figure 4-4. Mean frequencies for individual behaviors across phases for P 2.

Participant 3

Figure 4-5 displays P 3's frequency of pain behavior for each session during

baseline and intervention conditions. Mean: The mean frequencies of pain behavior for

P 3 were higher across all phases than for either Participant 1 or 2. The change in mean

frequencies was as follows: 57.8 behaviors per session for phase Al, 30 for phase Bl,

53.3 for phase A2, and 29.8 for phase B2. Latency: There was a 35 point decrease in

level between the end of phase Al and the beginning of phase Bl, however, the change in

trend went from a slope= 3.5 in the intended direction to a still accelerating slope of 1.2

in phase Bl. Between phases B1 and A2, there was a 9 point increase in level, however,

there was relatively little change in trend (slope= 1.2 to 0.9). There was a 28 point

decrease in level between phases A2 and B2, however, the trend continued to accelerate

in the unintended direction (slope= 2.1). Overlap: There was also no overlap of data

points between any of the phases. Stability: The data were 100% stable in phases A2

and B2, 87.5% stable in phase Al, but had 25% stability in phase B l.











Total Number of Pain Behaviors for Participant 3

90-Al B1 A2 B2









S40-








1 5 9 13 17 21
Sessions


Figure 4-5. Total number of pain behaviors for P 3 across sessions from the coded
activity protocols.

Changes in mean frequencies for each individual pain behavior by phase for

Participant 3 are presented in Figure 4-6. Behaviors occurring greater than 25 times

during the course of the study are included in these analyses. There is a clearly observed

pattern of increased mean frequency during baseline phases and decreased mean

frequency for intervention phases for guarding, rigidity, stopping, shifting, and

sighing/nonverbal behaviors. The pattern is less clear for bracing and verbal complaints,

but there is an initial decrease in these behaviors after the initial baseline phase.

Total Duration of Pain Behaviors

The following section presents study results regarding the effect of

acetaminophen on the duration of pain behaviors. As mentioned previously, only certain











Mean Frequency or Indiv idual Pain Behav iors Across Phases for Participant 3








HPhase A1
10 gPhase B1
c 8-OPhase A2
2 ~O Phase B






Mean Guarding Mean Rlgidity Mean Bracing Mean Stopping Mean Shlitng Mean Mean Verbal
Sighing/Nonverbal Complaint
Behaviors



Figure 4-6. Mean frequencies for individual behaviors across phases for P 3.

behaviors were coded both for their frequency and their duration. These behaviors

included: guarding, bracing, rigidity, and stopping.

Participant 1

Figure 4-7 displays P l's total duration of the pain behaviors (e.g. guarding,

rigidity, bracing, and stopping) in seconds for each session across all 4 phases. Mean:

The mean durations for phases were as follows: 325.6 seconds per session for Al, 120.7

seconds for Bl, 278 seconds for A2, and 155.4 for B2. Latency: In evaluating the

latency of change between phases Al and B1 there was a change in level of a 285.5

second decrease, and although the trend in phase Al was in the unintended direction


(slope= -4) there was a pronounced flattening of the trend in phase B 1 (slope= -0.2).

Between phases B1 and A2 there was a decrease of 56.3 seconds in level and a marked


change in trend from a slope=-0.2 to a slope = 93.0. Finally, between phases A2 and B2

there was a 204.6 second decrease in level as well as a trend lessening to a slope= 12.4.











Total Duration of Pain Behaviors for Participant 1

45 -Al B1 A2 B2

400-

350-

a, 300-

S250-

S 200-



100-




1 5 9 13 17 21
Sessions



Figure 4-7. Total duration of pain behaviors for P 1 across sessions from the coded
activity protocols.

Overlap: There was no overlap between phases Al and Bl, and only 1 point of overlap

between phases B1 and A2, and A2 and B2. Stability: The data were relatively unstable,

finding 62.5% stability in phase Al, 50% stability in phases A2 and B2, and only 25%

stability in phase B l.

The change in mean duration for each behavior is shown in Table 4-3. The


expected pattern of longer durations during the baseline phases and shorter durations

during the intervention phases was seen for each behavior; although stopping was an

infrequent behavior, occurring less than 5 seconds in each of the latter three phases.

Participant 2

Figure 4-8 displays P 2's total duration of the pain behaviors in seconds for each

session across all four phases. Mean: The mean duration of pain behaviors was as









Table 4-3. Mean Duration (in Seconds) of Individual Pain Behaviors across Phases and
Participants
Participant B ehavi or Phase Al Phase B 1 Phase A2 Phase B2
1 Guarding 151.2 59.3 121.5 59.7
Rigidity 129.0 48.3 110.8 81.1
Bracing 34.4 11.9 41.9 12.3
Stopping* 10.9 1.2 3.8 2.4
2 Guarding 95.0 55.6 101.2 71.3
Rigidity* 98.0 32.1 47.2 30.6
Bracing 11.5 6.7 10.8 7.1
Stopping* 0.8 3.3 5.3 3.3
3 Guarding 200.4 43.9 137.3 74.8
Rigidity 225.2 112.9 128.2 80.4
Bracing 304.8 257.0 263.7 222.8
Stopping 43.4 12.4 16.6 13.4
* Infrequent behaviors (<25 total occurrences)


follows: 205.3 seconds per session during phase Al, 97.7 seconds during phase Bl, 164.6

seconds during phase A2, and 112.2 seconds during phase B2. Latency: Between phases

Al and B1 there was a marked decrease in level (159.5 seconds), however, the trend in

the initial baseline phase was strong in the unintended direction (slope= -8.3) which

flattened in phase B1 (slope= -0.8). Between phases B1 and A2, there was an increase in

level of 138.7 seconds and an increase in trend (slope= 18.9) in the intended direction.

Between phases A2 and B2, there was a decrease in level (188.2 seconds), however, the

change in trend continued to increase to slope= 25.7. Overlap: There was a lot of

overlap between data in this participant, 4 points between phases Al and Bl, 3 points

between phases B1 and A2, and 2 points between phases A2 and B2. Stability: There

was also little stability of the data in all phases (37.5% in Al, 12.5% in B l, 0% in A2,

and 50% in B2).












Total Duration of Pain Behaviors for Participant 2

450 Al B1 A2 B2

400-

350-

vl300-

250-

S 200-

0 150-

100-




1 5 9 13 17 21
Sessions



Figure 4-8. Total duration of pain behaviors for P 2 across sessions from the coded
activity protocols.

The change in mean duration for each behavior is shown in Table 4-3. The


expected pattern of longer durations during the baseline phases and shorter durations

during the intervention phases was seen for each behavior, although rigidity and stopping


were infrequent behaviors, with stopping occurring 5 seconds or less in each phase.


Participant 3

Figure 4-9 displays P 3's total duration of the pain behaviors in seconds for each

session across all four phases. Mean: The mean total duration of pain behaviors was


higher for P 3 than for the other two participants. The change in mean for phases was as

follows: 773.8 seconds per session for phase Al, 426.2 seconds for phase Bl, 545.9


seconds for phase A2, and 391.5 seconds for phase B2. Latency: Between phases Al


and Bl, there was a 209.8 second decrease level and a trend change from a slope of 16.0











Total Duration of Pain Behaviors for Participant 3

1200 Al B1 A2 B2


1000-


a 800-


.C 600-


a

200-



1 5 9 13 17 21
Sessions



Figure 4-9. Total duration of pain behaviors for P 3 across sessions from the coded
activity protocols.

in the intended direction to a slope of -16.0 in the intended direction for the intervention

phase. This same effect followed for the latency of change between phases B1 and A2,

although the change in level was a decrease of 12.3 seconds. The change in trend was

strong with the slope increasing to 111.65 for phase A2. Between phases A2 and B2

there was a 60.6 second decrease for level change, and a change in trend to slope of -

78.1. Overlap: There was one point of overlap between phases Al and B1 and also

between A2 and B2. There were 2 points of overlap between phases B1 and A2.

Stability: Overall, the stability of phases varied, with 75% stability in phases B1 and A2,

50% in phase Al, but no stability (0%) in phase B2.

The change in mean duration for each behavior is shown in Table 4-3. Again, the


expected pattern of longer durations during the baseline phases and shorter durations

during the intervention phases was seen for all behaviors.