Main Site Specific Outcome Paper

Paper Proposal

Effects of an automated telephone support system on Alzheimer’s family

Caregivers: Findings from the REACH for Telephone-Linked Care© intervention

study.

Main outcome paper, Boston

Mahoney, Diane Ph.D.

Tarlow, Barbara Ph.D.

Jones, Richard Sc.D.

Submitted 11/12/01

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 1 Abstract

Objective: To evaluate the effects of a computer mediated automated interactive voice response (IVR) intervention designed to assist family caregivers managing persons with

Alzheimer’s related disruptive behaviors.

Design: An 18 month randomized controlled study of 100 caregivers; 51 in the usual care control group and 49 in the technology intervention group. Measures of the caregivers appraisal of the bothersome nature of caregiving, anxiety, depression and caregiving mastery were recorded at baseline, 6, 12, and 18 months. The analyses were conducted using a repeated measures approach for longitudinal data.

Results: There was no significant main effect of the intervention in reducing caregiver bother, anxiety or depression. There was a significant intervention effect for participants with lower mastery at baseline on all three outcomes: bother (P=0.04), anxiety (P=0.01) and depression

(P=0.007). Additionally, wives exhibited significant intervention effect in the reduction of the bothersome nature of caregiving (P=0.02).

Conclusion: Wives who exhibited low mastery and high anxiety benefited the most from the

REACH for TLC intervention.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 2 Effects of an automated telephone support system on Alzheimer’s family

Caregivers: Findings from the REACH for Telephone-Linked Care© intervention

study.

I. Introduction

Over the last decade the integration of technology into our daily lives has pivotally affected the way we communicate with each other. In our society, technology applications provide critical linkages to information, services, and people who are important to us. A “digital divide” however, has arisen that separates those with and without access to technology; specifically the Internet, now considered the “information superhighway”.1 Early research into the attributes of those lacking access to technology revealed that they were more likely to be of lower socio-economic status, members of racial/ethnic minorities, female, and older, characteristics already associated with underserved populations of historical concern to public policymakers.1

In 1994, the NIA issued a request for proposals aimed at helping family caregivers of people with Alzheimer’s disease (AD), and solicited innovative technological interventions in addition to traditional counseling and educational interventions.2 In response, we developed an application that integrated for the first time an automated Interactive Voice Response (IVR) system, rich with Alzheimer’s information content, with a voice mail telecommunications system for messaging with peers and AD professionals. We purposely designed the computer- mediated system to operate with the participants’ ordinary telephone and telephone line to facilitate access, minimize intervention costs, and to avoid altering the home environment. In this paper we report the results of using an IVR system to help reduce stress and increase caregivers’ mastery.

II. Background

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 3 Caring for a relative with AD and related disorders involves substantial personal, social and health costs.3-5 Recent evidence documents the greater demands faced by dementia caregivers when compared to caregivers of persons with other chronic diseases.6-8 Persons caring for family members with AD are reported as spending more time providing care than other caregivers, experience poorer health,9-10 increased levels of burden,11,12 anxiety and depression.13,14 Numerous studies confirm the association between dementia caregiving and both anxiety and depression.4,15-17 Disruptive behaviors, such as wandering and verbal assault common in middle stage AD, are frequently reported by caregivers as the primary cause of their burdensome feelings.18-20 Disruptive behaviors vary greatly in their frequency, intensity and consequences for all family members, contributing to the substantial variation in the level of burden reported by caregivers. Researchers have reported finding no positive linear relationship between care recipient impairment and appraisals of burden and substantial individual differences in how well people adapt to caregiving demands are commonly reported.4,11,21

The caregiving burden people experience is mediated by other factors, such as caregiving relationship and perceived mastery. Among the relationships, wives have constituted the majority of AD caregivers7 and experienced more physical and emotional stress than husbands.4,22,23 Wives have provided more help and assistance with a wider range of tasks, frequently providing the necessary personal care activities that have been found to increase feelings of stress more than other caregiving tasks.24,25 Husbands have remained a sizable minority of spousal caregivers, 36%.26 Husbands were reported as being more stressed when their wives were no longer able to be their confidants and primary sources of social and emotional support.12,27 It is reported that 27-30% of all caregivers are male.28,29 Male caregivers have been more likely than females to use instrumental problem solving strategies as part of their successful coping efforts.30,31 Adult children as a class of caregivers also responded differently to the demands of caregiving. Adult children appeared to be less adversely affected

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 4 by caring for a parent than were spousal caregivers.32 Lawton, Moss (1991) reporting on the differential effects from the stresses of caregiving, found that caregiving satisfaction correlated with positive affect and burden correlated with depression for spousal caregivers, but was not for adult children.33 Thus, within caregiving families, one’s relationship to the care recipient remains an important consideration.

In addition to relationship, mastery has emerged as an important characteristic that affects the caregiving experience. Pearlin and Schooler (1978) first included this construct in their research as a means for understanding the psychological resources inherent in the personality characteristics of people that enabled them to cope with the stressful consequences of social strain.16 Mastery, the sense that one regards one’s life chances as being under one’s own control, has been found to help people cope with life strains such as caregiving.34,35 The sense of mastery is enhanced when caregivers perceive a positive aspect to their caregiving.33,35

The association between problem solving strategies and positive affect has benefited caregiving husbands more than wives, who more often have used emotion focused coping strategies that were associated with low mastery and negative affect.22,36

The care recipients’ status and the caregiving context change over the course of the disease within a family and varies greatly among families as well.19,37 It is unlikely that any single approach to supporting families of persons suffering from AD would be useful to all families or even to the same family over a long period of time. Understanding the variability in caregiver outcomes and developing individualized and flexible interventions remain important research objectives.

Interventions using IVR technology programs have been designed previously as a means for clinicians to monitor community dwelling persons affected with chronic diseases.

Automated systems have been successfully implemented to manage hypertension,38 weight loss,39 diabetes,40 mental health,41 and to monitor the activities of daily living for homebound elders.42 At the time of this proposal there had been only one other computer-based program

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 5 designed for AD caregivers. ComputerLink was developed by Brennan, Moore and Smyth

(1991) to provide information, communication, and decision support for AD caregivers via an

Intranet based computer system.43 Caregivers were able to access an electronic bulletin board for public and private messaging, information on AD, and a decision support program to help caregivers analyze their own issues using keypads linked to terminals installed in their homes.

Caregivers varied in their use of system components, number of contacts, and duration of system use.44 Subsequently, Brennan and associates analyzed the effects of a similar program on caregivers of persons with AIDS and noted that AD caregivers used the system much less frequently and for shorter periods, emphasizing the need to design short, easy to use interventions for this category of caregiver.45 Schneider, Schwartz and Fast (1995) explored how IVR systems should interact with their users to have the greatest impact and they found that users preferred interactions with computers that were conversational in nature, personalized and responsive to their particular expectations.46 Bourgeois, Schulz and Burgio

(1996) reviewed six behavioral intervention studies and concluded that multi-component interventions, that offered AD caregivers multi-faceted support and services tailored to their particular needs, were more likely to be effective.47

We built upon the lessons learned from the previously cited technology research and developed for the first time a system that integrated voice mail telecommunications with IVR technology. The Reach for TLC project was a pioneer technology application. Within the IVR field, this project was the first to target a caregiver rather than the person with the condition of interest. The Reach for TLC system provided multiple modules that allowed the user to choose what best suited his or her needs. The voice mail component offered multiple internal options and easily linked to the IVR subsystem modules. The interactions were personalized, kept brief and used a conversational style to deliver information and advice.

Programmatic development of the system was theory driven. From the technology field, innovation-diffusion theory suggested that there were variations in acceptance of any new form

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 6 of technology.48 Reported organizational and personal barriers to technology adoption have included resistance to change, fear of substituting computers for people, and technical complexity.49 These insights provided the rationale for using a known and commonly useful technology, the telephone, as the intervention hub. The ordinary home telephone became the interface to a computer-mediated program of support for AD family caregivers. The program design and features flowed from the innovation and diffusion model and these have been detailed previously.50

From the caregiving literature, stress process theory guided the intervention development. Pearlin’s model of AD caregiver’s stress provided the conceptual framework for this study.51 In this model the relationships among caregiving factors contribute to the development of caregiver stress. Our specification of the model’s domains was as follows: the background and context of stress (caregiving relationship), plus the stressor (care recipient’s behaviors), as mediated by caregiver mastery, resulted in manifestations of caregiver stress

(appraised bothersome nature of the care recipient’s disruptive behaviors, anxiety, and depression). Based upon both the technology and caregiving theories, we anticipated that both the participants’ level of mastery and caregiving relationship would influence the response to the intervention. From this premise we tested the following hypotheses:

1) Caregivers in the intervention group will have less caregiving stress (due to care recipient’s disruptive behaviors, anxiety, and depression) than control group members.

2) Caregivers in the intervention group with low mastery will experience more intervention effects than those with high mastery.

3) Caregiving wives will show greater reduction in caregiving stress than husbands or children.

III. Methods

A. Study Design and Sample

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 7 The Reach for Telephone-Linked Care (Reach for TLC©) project was an 18-month randomized controlled study designed to test the feasibility and effect of a computer mediated telecommunications system with a sample of AD caregivers. Caregivers were eligible to participate if there were: over the age of 21, provided four or more hours per day of assistance or supervision for a minimum of six months to a family member with AD who had at least two impairments of instrumental activities of daily living (e.g. driving, shopping, managing money) or one activity of daily living (e.g. toileting, bathing, eating) and exhibited at least one AD related disturbing behavior. Alzheimer’s disease was defined either as having a medical diagnosis of probable AD, or cognitive impairment with a Mini-Mental Status Exam (add cite for MMSE) of 23 or less.52 Caregivers were required to have touch-tone telephone service, and a touch-tone phone was provided if needed. Persons were excluded if they had plans to institutionalize the family member within six months, were participating in another clinical intervention study for caregivers, were terminally ill, or refused consent.

The participants were recruited by referrals from cooperating health and social agencies, including large urban medical centers and small clinic practices, as well as individual responses to media notices about the study.53 Participants were recruited from the community over a 22- month period between 1997-1999 and included eastern Massachusetts, southern New

Hampshire and northern Rhode Island. These efforts resulted in 143 contacts of which 118 met the eligibility criteria and 100 participants were enrolled, an 85% recruitment rate (yield). The majority of persons declining participation (18) said they were too busy to participate in the project and several reported their own physical health problems as reasons for not participating.

Of the 100 enrollees, 49 were randomly assigned to the intervention group and 51 to the control group.

Institutional Review Board (IRB) approvals were obtained from two academic medical centers and two research institutes prior to the start of the field work, and yearly IRB reviews and approvals were maintained over the course of the study. A research assistant, who was a

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 8 certified project interviewer, visited participants in their homes to obtain a signed informed consent and to administer the baseline interview. Participant randomization occurred after completion of the baseline data. Two separate computer generated random assignment lists, one for males, one for females were generated for each recruitment site, ensuring that each intervention and control group was balanced by gender and site. Those assigned to the intervention group received instruction on use of the intervention and were given a User’s

Manual. The research assistant demonstrated use of the system and had the participants practice until they could demonstrate correct access to the system and each of the modules.

This process took, on average, twenty minutes. The research assist also rated (low to high) the participants’ proficiency level when they used the system, and low proficiency users were offered additional training. Participants were encouraged to use all modules, with emphasis on the Weekly Caregiver Conversation, as this module was designed to help caregivers manage the specific disruptive behaviors they reported as being their primary source of caregiver stress.

Control group participants were given a reference booklet, containing similar content to Module

1 of the intervention, that provided strategies to mange AD related disruptive behaviors. All participants were subsequently interviewed by phone at 6, 12 and 18 months time points.

B. Description of Reach for TLC System

The intervention consisted of an integrated telephone network system and an Interactive

Voice Response (IVR) computer network system. An IVR system is essentially a talking computer that speaks in response to signals received from touch-tone telephones. The caller dialed into our telephone network and selected one of the five modules to use. The caregiver heard a digitized human voice (we used a professional radio announcer as the narrator to ensure a pleasing voice recording) that spoke a computer-mediated script. The IVR system contained the necessary hardware to process the narrator’s messages for encoding in the databases used to store and update files. The telephone network operated the Personal

Mailbox and Bulletin Board modules, similar to standard voice messaging systems.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 9 The technology application developed for this project used a multi-system telecommunications based approach to help caregivers manage Alzheimer’s related disturbing behaviors commonly associated with caregiver stress such as wandering, hitting, resistance, and outbursts. Because caregivers have diverse needs and preferences, the intervention was designed to offer multiple components with flexibility in order to effectively appeal to a variety of users. Participants choose the type of component, frequency, duration and timing of the usage.

Caregivers used a confidential password to access the intervention system and to protect their anonymity. Caregivers dialed in from any standard touch-tone telephone and heard the narrator greet them by name, review the menu options and provide the service they requested.

Caregivers responded to the questions and comments they heard by pressing designated numbers on their touch-tone key pad, such as 1 for yes and 2 for no. Depending on the user’s response, the narrator provided tailored information specific to the disruptive behavior reported.

The intervention was available 22 hours a day; except for 2 hours during the night for network file backup. A Technical Helpline allowed caregivers to report system problems. A detailed description of the intervention architecture and system operations, as well as the content development process have been described previously.50

The following system modules were available to participants:

1. Weekly Caregiver’s Conversation: the automated IVR conversation that monitored the caregiver’s stress levels. The system narrator queried the caregiver about any disturbing behaviors exhibited by the care recipient, and if desired, provided the caregiver with information and strategies to reduce the reported disruptive behaviors. For example, if they agreed (1) that

“losing things” was currently a disturbing behavior they would be offered tips on ways to specifically manage that kind of behavior in their family member. Subsequent calls assessed the responses to the problem solving strategies and if the behavior continued, offered additional strategies and a suggestion to use the voice mail support system. If caregivers reported that the problem behaviors continued and their stress levels significantly increased during a three

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 10 week period, the computer generated an alert to our system manager and if confirmed, she faxed to the designated provider, a report that graphed the increased stress by week, as well as the types and frequency of bothersome AD behaviors during this period. The module also contained a monthly caregiver self-health assessment component, as well as reminders about the advice and resource features.

2. Personal Mailbox: the voice mail module allowed caregivers to send and receive confidential communications anonymously among themselves or to communicate with a clinical nurse specialist who directly answered or triaged questions to the Ask-the-Expert panel.

3. Bulletin Board: an in-home support group that provided personal and caregiver group voicemail, similar to a computer chat group. System users could anonymously post messages and receive back responses from other users.

4. Activity/Respite Conversation: a care recipient distraction module designed to reduce disruptive behaviors and to provide caregiver with respite time. When initiated by the caregiver, this module offered the care recipient a personalized pleasant conversation designed to engage the listener in a safe, comforting and non-demanding activity. The conversation lasted 18 minutes and would repeat once if not disconnected. While offering a generic conversation, the person’s favorite hobbies, foods, holidays, friends and songs were included as “donuts”, that is the computer would automatically insert personally tailored data triggered by the user’s password entry. For example, one person would hear “I understand your favorite food is

______(chocolate cake), while another would hear (meatloaf) depending upon the caregivers suggestions. It’s nice to have a favorite food. I like ______too. It taste’s so good.”

Similarly the conversation was spoken with the appropriate male or female pronouns and was available in English and Spanish.

C. Measures

1. Background variables: Standard demographic measures and salient background information measures were obtained, including age, gender and years of education for both the

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 11 caregiver and care recipient. Caregivers were asked about their ethnicity and employment status. Caregiver relationship was self-reported by caregivers as wife, husband or child.

Additional information was also obtained about the care recipient: Cognition was measured by

Folstein’s Mini Mental Status Exam score (MMSE)52 which varies from 0 to 30, with a cutoff point of 23 or less indicating cognitive impairment; Activities of Daily Living (ADL’s)54 scores vary from 0-8, higher scores indicating greater functional impairment; and Independent Activities of

Daily Living (IADL’s) scores vary from 0-6 with higher scores indicating greater functional impairment.55 The results are reported in Table 1.

2. Outcome variables

Three outcome variables related to caregiver stress were examined; bothersome nature of the care recipient’s disruptive behaviors, anxiety, and depression.

Bother was measured by The Revised Memory and Behavior Problems Checklist (RMBPC).56

This measure assessed the type and frequency of care recipient disruptive behaviors and the caregiver’s appraisal of distress generated by those behaviors. Caregivers were asked whether their care recipients manifested any one of 24 problem behaviors, yes or no, during the past week. The 9 item disruptive behaviors subscale has been used in this analysis since the intervention was designed to help reduce and manage care recipient disruptive behaviors. The

RMBPC’s Bother scale contained questions asking how much potential disruptive behaviors might bother the caregiver. If caregivers reported the presence of a disruptive behavior, they were asked how bothered or upset they were by each behavior, using a five point scale ranging from 0, “not at all” to 4 “extremely” bothersome. Average upset scores were calculated, assigning “no upset” to behaviors not manifested. Scores ranged from 0 to 36, higher scores indicating more distress. Cronbach’s alpha coefficient for the disruptive behaviors subscale of the RMBPC was 0.66.

Anxiety was measured by the State Anxiety Inventory (STAI).57 This ten-item measure was taken directly from Speilberger’s larger State-Trait Personality Inventory (STPI:1979).

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 12 Respondents were asked how much they agreed with statements such as, “I feel nervous” for the time interval, “during the past week.” Response options ranged in intensity from (1) “not at all” to (4) “very much.” Four items were scaled negatively; a “very much “ response would indicate a low level of anxiety, such as “I am relaxed.” Cronbach’s alpha coefficient for the anxiety measure was 0.89.

Depression was assessed with the Center for Epidemiological Studies-Depression scale (CES-

D)58 a 20 item measure that asks about the frequency with which respondents experienced depressive symptoms within the past week. Response options ranged from 0 to 3: (0) “rarely or none of the time (less than 1 day)”, (1) “some or a little of the time (1-2 days)”, (2) “occasionally or a moderate amount of time (3-4 days)”, and (3) “most or all of the time (5-7 days)”.

Cronbach’s alpha coefficient for the CES-D was 0.89.

Mastery16 assessed one aspect of the caregiver’s psychological resources for coping with stress, measuring the extent to which the participant regards his/her life chances as being under his/her own control. Participants were asked, How strongly do you agree/ disagree that: “I have little control over the things that happen to me.” Responses were indicated on a five point ordinal scale of (1) “Strongly Disagree” to (5) “Strongly Agree”, with lower scores indicating greater mastery (Cronbach’s alpha = 0.71). We used this measure of mastery to identify participants that were comfortable with their caregiving role, and those who were more uneasy, by classifying individuals with mastery scores 0.5 standard deviations above the mean as high, and the remainder as low-mid mastery caregivers. We used this low-mid vs. high mastery variable to investigate possible interaction between baseline mastery and the effectiveness of the intervention in terms of caregiver ratings of bothersome behaviors, depressive symptoms, and anxious complaints.

D. Analysis

Bivariate analyses were used to provide descriptive data regarding caregivers and care recipients randomized to the intervention group and control group, and to evaluate the

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 13 adequacy of randomization. We used a repeated measures approach to longitudinal data to evaluate the extent to which participation in the intervention condition modified the trajectory of caregiver reported behavioral problems (RMBPC-disruptive behavior subscale), depressive symptoms (CESD), and anxiety (STAI). We used a generalized estimating equations (GEE) approach with the four (baseline, six month, 12 month, and 18 month follow-up) assessments for RMBPC disruptive sub-scale as repeated observations, with time, intervention status and their interaction as explanatory variables. Analyses were repeated with CESD interpersonal and state anxiety as dependent variables. An exchangeable correlation structure was modeled to accommodate unbalanced design and to model the non-independence of repeated observations within subjects. We explored other autocorrelation structures (e.g., autoregressive) and obtained qualitatively identical results. Results with the exchangeable autocorrelation matrix have been reported because this model permits including all randomized participants, while others require listwise deletion of observations with missing follow-up. We also explored linear, quadratic and piecewise normative trajectories for the three outcomes, and found that a linear function provided the best fit to the observed data (based on -2×log- likelihood). Therefore, all results presented describe a linear effect of time. Main effect models were generated regressing the repeated observation of RMBPC, CES-D, or STAI on three independent variables: intervention status, time, and the interaction of time and intervention status. The associated parameters provided difference in baseline outcome for members of the intervention group, change over time for participants, and the difference in change over time for participants randomized to the intervention group relative to controls (i.e., the treatment effect).

Following these main effect models, the models were re-estimated according to our hypotheses by stratifying on caregiver relationship (wives, husbands, and children) and level of mastery at baseline. Analyses were conducted with the STATA statistical software package. Results of these models are summarized with figures displaying the model-implied trajectory for the outcome variables.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 14 IV. Results

As Table 1 indicates, there were no significant differences in characteristics between the control and intervention groups for either the caregivers or the care recipients enrolled in the study. Just under 80% of the caregivers in both the intervention and control group were women, with a mean age of approximately 62 years. The majority of caregivers in both the intervention and control groups were white, and the largest percent of intervention and control group participants were retired. Intervention and control group caregivers were very similar in education and in relationship to care recipient: most being the spouse of the care recipient.

The care recipients in the intervention and control groups included nearly equal numbers of women, and were similar in age. The care recipients were very similar in MMSE scores, ADL scores and IADL scores. Differences were noted in educational achievement, although none were significant. More of the control group participants had either less than or more than a high school education.

------

Insert table 1 about here

------

Intervention Effect

Hypothesis 1 postulated that the intervention group would experience reduced caregiver stress and this was not supported. Overall, there was no significant main effect of the intervention in reducing bother scores (RMBPC), depression (CES-D) scores, or state anxiety

(STAI) (figure 1). Trajectories did not differ overall for control group and intervention group participants, as displayed in figure 1.

Hypothesis 2 tested whether treatment effects would vary among participants with low to mid mastery versus those with high levels of mastery. Subgroups based on Pearlin's mastery

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 15 measure included those with low to mid levels of mastery (up to 0.5 standard deviations above the overall sample mean) and those with high level of mastery (a mastery level more than 0.5 standard deviations above the mean) at baseline. There was a significant intervention effect for all three outcomes for participants with low-mid mastery at baseline. These effects are summarized in figure 2. Participants with low-mid mastery at baseline experienced a greater decline in RMBPC bother scores, CESD depressive symptoms, and STAI anxious complaints over the eighteen-month study period relative to controls. There was no significant difference

(p>.05) between rates of change in the study outcomes for participants with high mastery at baseline (results not shown).

Hypothesis 3 suggested that wives would differ in response from other caregivers and this was partially supported. There was a significant intervention effect for caregivers who were wives, for RMBPC scores (p=0.023), results not shown. There was no significant difference in treatment effectiveness for any of the other outcomes within any of the other relationship groups.

V. Discussion

The present study examined whether an automated telecommunications system designed for Alzheimer’s caregivers exerted positive effects on caregiver outcomes. The system aimed to reduce caregivers’ stress related to managing their care recipients bothersome behaviors associated with Alzheimer’s disease. We did find significant intervention effects on caregivers with lower mastery at baseline and for those who were wives. As previously cited research indicated, wives provide the majority of family caregiving, and interventions that support their role and reduce negative caregiver outcomes are important. The IVR intervention did have the effect of reducing the bothersome nature of dementia related behavioral disturbances for wives. Findings from this research lend support to further considerations of using the telephone as a means to reduce this particular form of caregiver stress. Moreover, another important category of participants, those with low caregiving mastery, demonstrated the

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 16 most reduction in caregiver stress related outcomes of bothersome behaviors, depression, and anxiety. Prior IVR technology research has found beneficial results for people who are uncomfortable with in-person face-to-face counseling situations.41 Findings from this study contribute insights into the characteristics of the subgroups of users who most benefit from IVR interventions. Previous research has suggested the need for technology developers to study and share such characteristics to help advance the technology field.45,59

Another positive but unexpected finding occurred among a subset of non-users who reported that they were satisfied with the automated system. A content analysis of their comments revealed that these participants perceived no immediate need for the system, but they valued knowing they could have accessed it if the need arose. The theme that emerged was a sense of security and comfort from knowing that a source of information and respite was available at any time, irrespective of usage. This finding suggests that availability can exert a positive influence on caregiver satisfaction and highlights the necessity to both assess and differentiate intervention availability from actual usage. By contrast, unsatisfied non-users unequivocally expressed their strong preference for personal interactions with humans.

Moreover, they had expressed this reluctance to use IVR technology at the baseline interview and this was highly predictive.

There are several possible explanations for the lack of an overall main effect from the intervention. One of our study eligibility criteria required that the caregivers report at least one bothersome behavior at minimum, and we expected many more. Participants, however, not only reported on average, low baseline total bother scores (RMBPC, 8.7 M and 0.5 s.e.), but also low depression scores (CES-D, 13.1 M and 1.1 s.e.). With such low values, we experienced floor effects, wherein there was little room to further lower these values in a meaningful way. Others have found the majority of caregiver intervention studies have had either non-significant or very modest results despite the participants’ high satisfaction rates or positive evaluations.60,61

Whitlatch, Zarit, & von Eye (1991) specifically noted that someone with an initial low score of

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 17 depression can only worsen and suggested that a regression to the mean results in an underestimate of the impact of the intervention.62 Also, the intervention was designed to follow a consumer choice approach that offered users a selection of four modules to choose from, and we expected differential usage based upon the users’ preferences. Consequently, the intensity of the intervention delivery varied even among users. Biegel, Sales and Schulz (1991) have suggested a greater likelihood of effectiveness from interventions that offer both high intensity and long duration.63 Our caregivers chose the opposite pattern, one of low intensity and short duration.

Another factor was that the majority of participants enrolled through referrals from academic medical centers where they had access to Alzheimer’s specialty teams for information and assistance. It is plausible that a sample without similar easy access to Alzheimer’s experts would have used the system more and in a different manner. This has some modest support based on our observation that greater usage occurred among a small subset of users from rural

New Hampshire and many caregivers from academic medical centers reported no difficulty in obtaining information and advice from their AD specialists, physicians and nurses.

A constraint we faced in participating in a multi-site collaborative study was the need to adhere to a recruitment schedule and a longitudinal design that became uniquely problematic in our region. We found that many caregivers would not commit to an eighteen month caregiver intervention study and loose the opportunity to participate in one or more of the numerous

Alzheimer’s clinical drug treatment trials that were being conducted at the same time in our locale.53 Additionally, the TLC system failed to record data on some of the subjects, further reducing the analytic sample.64 Thus, while our sample size was adequate for the technical feasibility analysis, it was less than desired for the caregiver outcome analyses, and this also contributes to the possibility of a Type II error, or underestimate of an effect.

Although not directly tied to the intervention, the fact that there was very little change in reports of bothersome-ness of behaviors, or depressive and anxious symptoms over the 18

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 18 month period in AD caregiver's careers was a notable finding. This observation speaks to the need to more fully understand the developmental trajectory of ratings of bothersome behaviors and destabilization of mood among AD caregivers. While it is widely reported in the literature that AD caregivers suffer greater depression than age matched controls,8,10 the findings reported here suggest that bother and mood destabilization do not occur rapidly or continuously over the course of AD caregiving careers. Appropriate targeting of interventions requires identifying a suitable population and that may include important dimensions such as nature of relationship, length of caregiving, and sense of mastery over the caregiving situation. Future research is needed that explores the natural course of important dimensions of caregiver's biopsycho-social responses over the duration of caregiving careers, results of which are critical for timing and maximizing intervention benefits.

Moreover, findings from our prior analysis of TLC features and usage complements the present outcome analyses and together these studies provide important insights to further understanding of the strengths and limitations of this technology based intervention.64 For example, previously we did not find the integrated voice mail bulletin board, personal mail box, and professional triage features to be necessary components and we believe streamlining them in future applications would decrease the system’s development cost, complexity and likelihood of technical problems.64,65 The respite call, however, was a feature highly valued by those who perceived the need for and allowed the care recipient to use it. We did find that users needed to perceive a personal reason or clear purpose to motivate them to adopt using the system.

Training until proficiency was demonstrated became a critical component of this adoption and is strongly recommended. Users preferred short conversation style interactions with advice and counseling on demand for up to a four month time period, after which there was negligible usage.64

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 19 B. Summary

In summary, this study uniquely contributes to both the fields of Alzheimer’s CG intervention and technology research. Our findings add important insights into the understanding of who most benefits from automated IVR telecommunication technology for AD caregivers. The intervention was innovative in several aspects; it was completely automated, operated with immediate availability seven days a week, and was not limited by traditional daytime working hours or the forty hour a week work schedule. REACH for TLC offered consistent high quality information without the variability associated with different interventionists and human factors such as fatigue or forgetfulness. The technology permitted the delivery of the content through a consumer choice approach without the need for or control by professionals. The system was not limited by staff availability or turnover. From the technology side, the REACH for TLC system was the first application that demonstrated the feasibility of merging voice mail subsystems systems with IVR technology, and, beyond that, went on to determine the components that were most useful. It was the first IVR application to study Alzheimer’s caregiver outcomes and to identify both the characteristics of the users, as well as their patterns of usage over a twelve-month period.

Future caregiver telecommunications research should seek to find the optimal mix of technology, content information, and human interface. Would an IVR system with more intensity via scheduled sessions or special sessions mediated by a human interventionist exert stronger outcome effects and appeal to more of the non-users? What similar or different outcomes would occur if the IVR content were adapted to a web-based interaction? For a clinical application, we suggest integrating the system into a network of clinical providers, linking to staff members for advice instead of an unfamiliar external expert panel. This would diminish the providers’ resistance to technology, due to fears of computer substitution for their caregiver services and loss of clients, that we and others have encountered.64,66 For maximal effectiveness, we suggest that clinicians should target caregivers, who foresee the need for

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 20 either the present or future use of the technology, and portray the user characteristics associated with positive outcomes.

VI. Conclusion

The outcome analyses revealed that those most likely to benefit from the REACH for

TLC IVR system were wives who exhibited low mastery and high anxiety. The system was able to decrease the bothersome nature of the disturbing behaviors associated Alzheimer’s caregiving in this subset of critically important caregivers and they appear particularly suited to this IVR application.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 21 Acknowledgments

We acknowledge the efforts of Sharon Tennstedt, Ph. D., Denis Cohen and staff at the New

England Research Institutes, Watertown, MA for conducting the participant interviews and managing the local REACH collaborative data system, Robert Friedman, M.D. and staff at the

Medical Information Systems Unit, Boston Medical Center, for oversight of the technology operations and IVR data, Ladislav Volicer, M.D. for his content and research expertise, Clare

Wohlgemuth RN, MS, the intervention triage nurse and the panel of experts.

Funding Statement

The Resources for Enhancing Alzheimer’s Caregiver Health Project is supported by the

National Institute on Aging and the National Institute of Nursing Research (Grants: Burgio,

NR13269; Burns, AG13313; Eisdorfer, AG13297; Gallagher-Thompson, AG13289; Gitlin,

AG13265; Mahoney, AG13255; Schulz, AG13305)

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 22 Reference List

1. Irving L. Falling through the Net: Defining the Digital Divide. Washington, DC: Department of Commerce, National Telecommunications and Information Administration; 1999a: 1-29. 2. National Institute on Aging. Request for Funding Proposals.AG-94-003 R.1994. 3. Hooker K, Manoogian-O'Dell M, Monahan DJ, Frazier LD, Shifren K. Does type of disease matter? Gender differences among Alzheimer's and Parkinson's disease spouse caregivers. Gerontologist. 2000;40:568-73. 4. Schulz RE. Handbook on Dementia Caregiving; Evidence-Based Interventions for Family Caregivers. New York, NY: Springer Publishing Company; 2000. 5. Schulz R, O'Brien AT, Bookwala J, Fleissner K. Psychiatric and physical morbidity effects of dementia caregiving: prevalence, correlates, and causes. Gerontologist. 1995; 35: 771-791. 6. Hooker K, Monahan DJ, Bowman SR, Frazier LD, Shifren K. Personality counts for a lot: predictors of mental and physical health of spouse caregivers in two disease groups. J Gerontol B Psychol Sci Soc Sci. 1998;53:73-85. 7. Ory MG, Hoffman RR, 3rd, Yee JL, Tennstedt S, Schulz R. Prevalence and impact of caregiving: A detailed comparison between dementia and nondementia caregivers. The Gerontologist. 1999;39:177-85. 8. National Alliance for Caregiving and American Association of Retired Persons. Family caregiving in the US: Findings from a national survey. Final Report. Bethesda, MD: National Alliance for Caregiving; 1997. 9. Baumgarten M, Battista RN, Infante-Rivard C, Hanley JA, Becker R, Gauthier S. The psychological and physical health of family members caring for an elderly person with dementia. J Clin Epidemiol. 1992;45:61-70. 10. Schulz R, Beach SR. Caregiving as a Risk Factor for Mortality: The Caregiver Health Effects Study. Journal of the American Medical Association. 1999;282:2215-2219. 11. Vitaliano P, Russo J, Young H, Becker J, Maiuro R. The screen for caregiver burden. The Gerontologist. 1991;31:76-81. 12. Zarit S, Todd P, Zarit J. Subjective burden of husbands and wives of caregivers: Alongitudinal study. The Gerontologist. 1986;26:260-266. 13. Cohen D, Eisdorfer C. Depression in family members caring for a relative with Alzheimer's disease. Journal of American Geriatric Society. 1988;36:885-889. 14. Gallant MP, Connell CM. Predictors of decreased self-care among spouse caregivers of older adults with dementing illnesses. J Aging Health. 1997;9:373-95. 15. Lawton MP, Moss M, Kleban MH, Glicksman A, Rovine M. A two-factor model of caregiving appraisal and psychological well-being. Journal of Gerontology. 1991;46:P181-189. 16. Pearlin LI, Schooler C. The structure of coping. J Health Soc Behav. 1978;19:2-21. 17. Kiecolt-Glaser J, Glaser R. Chronic Stress and Mortality Among Older Adults, Editorial. Journal of the American Medical Association. 1999;282:2259-2260. 18. Gaugler JE, Davey A, Pearlin LI, Zarit SH. Modeling caregiver adaptation over time: the longitudinal impact of behavior problems. Psychol Aging. 2000;15:437-50. 19. National Institute on Aging. Progress Report on Alzheimer's Disease, taking the next steps. . Silver Spring, Maryland 20907-8250: National Institute on Aging, National Institutes of Health; 2001:1-62.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 23 20. Teri L. Behavior and caregiver burden: behavioral problems in patients with Alzheimer disease and its association with caregiver distress. Alzheimer Dis Assoc Disord. 1997;11:S35-8. 21. Haley WE, Levine EG, Brown SL, Bartolucci AA. Stress, appraisal, coping, and social support as predictors of adaptational outcome among dementia caregivers. Psychol Aging. 1987;2:323-30. 22. Rose-Rego SK, Strauss ME, Smyth KA. Differences in the perceived well-being of wives and husbands caring for persons with Alzheimer's disease. Gerontologist. 1998;38:224- 30. 23. Horowitz A. Sons and daughters as caregivers to older parents: differences in role performance and consequences. Gerontologist. 1985;25:612-7. 24. Miller B, Cafasso L. Gender differences in caregiving: Fact or artifact? Gerontologist. 1992; 32: 498-507. 25. Dwyer JW, Coward RT. A multivariate comparison of the involvement of adult sons versus daughters in the care of impaired parents. J Gerontol. 1991;46:S259-69. 26. Stone R, Cafferata GL, Sangl J. Caregivers of the frail elderly: a national profile. Gerontologist. 1987;27:616-26. 27. Chappell NL. Aging and social care. In: Binstock RH, George LK, eds. Handbook of aging and the social sciences. 3 ed. New York, NY: Academic Press; 1990:438-454. 28. Cantor MH. Strain among caregivers: a study of experience in the United States. Gerontologist. 1983;23:597-604. 29. Chang CF, White-Means SI. The men who care: An analysis of male primary caregivers who care for frail elderly at home. The Journal of Applied Gerontology. 1991;10:343-358. 30. Seltzer MM, Greenberg JS, Krauss MW. A comparison of coping strategies of aging mothers of adults with mental illness or mental retardation. Psychol Aging. 1995;10:64- 75. 31. Rose JM, DelMaestro SG. Separation-individuation conflict as a model for understanding distressed caregivers: psychodynamic and cognitive case studies. Gerontologist. 1990;30:693-697. 32. George LK, Gwyther LP. Caregiver well-being: a multidimensional examination of family caregivers of demented adults. Gerontologist. 1986;26:253-259. 33. Lawton MP, Kleban MH, Moss M, Rovine M, Glicksman A. Measuring caregiving appraisal. Journal of Gerontology. 1989;44:P61-71. 34. Peters-Davis ND, Moss MS, Pruchno RA. Children-in-law in caregiving families. Gerontologist. 1999;39:66-75. 35. Narayan S, Lewis M, Tornatore J, Hepburn K, Corcoran-Perry S. Subjective Responses to Caregivng for a Spouse with Dementia. Journal of Gerontological Nursing. 2001;27:19-28. 36. Kramer B. Differential predictors of strain and gain among husbands caring for wives with dementia. The Gerontologist. 1997b;37:239-249. 37. Gitlin L, Corcoran M, Winter L, Boyce A, Hauck W. A randomized, controlled trial of a home environmental intervention: Effect on efficacy and upset in caregivers and on daily function of persons with dementia. The Gerontologist. 2001;41:4-14. 38. Friedman R, Stollerman J, Mahoney D, Rozenblyum L. Virtual visit: Using telecommunications technology to take care of patients. Journal of the American Medical Informatics Association. 1997;4:413-425. 39. Friedman RH. Automated telephone conversations to assess health behavior and deliver behavioral interventions. J Med Syst. 1998;22:95-102. 40. Piette JD, Mah CA. The feasibility of automated voice messaging as an adjunct to diabetes outpatient care. Diabetes Care. 1997;20:15-21.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 24 41 Baer L, Greist JH. An interactive computer-administered self-assessment and self-help program for behavior therapy. J Clin Psychiatry. 1997;58:23-28. 42. Mahoney D, Tennstedt S, Friedman R, Heeren T. An automated telephone system for monitoring the functional status of community-residing elders. Gerontologist. 1999;39:229-34. 43. Brennan PF, Moore SM, Smyth KA. ComputerLink: Electronic support for the home caregiver. ANS Adv Nurs Sci. 1991;13:14-27. 44. Brennan P, Moore S, Smyth K. The effects of a special computer network on caregivers of persons with Alzheimer's disease. Nursing Research. 1995a;44:166-172. 45. McClendon M, Bass D, Brennan PF, McCarthy C. A computer network for Alzheimer's caregivers and use of support group services. Journal of Mental Health and Aging. 1998;4:403-420. 46. Schneider S, Schwartz M, Jeffrey F. Computerized, telephone-based health promotion: II. Stress management program. Computers in Human Behavior. 1995;11:205-214. 47. Bourgeois M, Schulz R, Burgio L. Interventions for caregivers of patients with Alzheimer's disease: A review and analysis of content, process and outcomes. International Journal of Aging and Human Development. 1996;43:35-92. 48. Randles F. On the diffusion of computer terminals in an established engineering environment. Management Science. 1983;April:465-476. 49. Trice AW, Treacy ME. Utilization as a dependent variable in MIS research. DATA BASE. 1988:33-41. 50. Mahoney D, F., Tarlow B, Sandaire J. A computer mediated intervention for Alzheimer's caregivers. Computers in Nursing. 1998b;16:208-216. 51. Pearlin L, Mullan J, Semple S, Skaff M. Caregiving and the stress process:An overview of concepts and their measures. The Gerontologist. 1990;30:583-589. 52. Folstein MF, Folstein SE, McHugh PR. "Mini-mental state". A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research. 1975;12:189-98. 53. Tarlow BJ, Mahoney D, M. The cost of recruiting Alzheimer's disease caregivers for research. Journal of Aging and Health. 2000;12:519-539. 54. Katz SF, A B/Moskowitz, R W/ Jacobsen, B A/ Jaffe, M W/. Studies of illness in the aged: The index of ADL: A standardized measure of biological and pyschosocial function. Journal of the American Medical Association. 1963;185:914-919. 55. Lawton MP, Brody EM. Assessment of older people: self-maintaining and instrumental activities of daily living. Gerontologist. 1969;9:179-86. 56. Teri l, Truax P, Longsdon R, Uomoto J, Zarit S, Vitaliano P. The Revised Memory and Behavior Problems Checklist. Psychology and Aging. 1992;7:622-631. 57. Spielberger CD, Gorsuch RL, Lushene R, Vagg PR, Jacobs GA. Manual for the State- Trait Anxiety Inventory (Form Y). . Palo Alto, CA: Consulting Pyschologists Press; 1985. 58. Radloff L. The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement. 1977;106:203-214. 59. Miller S, Vermeersch PE, Bohan K, Renbarger K, Kruep A, Sacre S. Audio presence intervention for decreasing agitation in people with dementia. Geriatr Nurs. 2001;22:66- 70. 60. Zarit SH, Teri L. Interventions and services for family caregivers. In: Schaie KW, Lawton PM, eds. Annual review of gerontology and geriatrics. New York, NY: Springer Publishing Company; 1992:287-310. 61. Toseland RW, Rossiter CM. Group interventions to support family caregivers: a review and analysis. Gerontologist. 1989;29:438-48. 62. Whitlatch CJ, Zarit SH, von Eye A. Efficacy of interventions with caregivers: a reanalysis. Gerontologist. 1991;31:9-14.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 25 63. Biegel D, Sales D, Schulz R. Family caregiving in chronic illness. . Theoretical perspectives on caregiving. Newbury Park, CA: Sage; 1991:29-61. 64. Mahoney D, Tarlow B, Jones R, Tennstedt S, Kasten L. Factors affecting the use of a telephone-based intervention for caregivers of people with Alzheimer's disease. Journal of Telemedicine and Telecare. 2001;7:139-148. 65. Mahoney DF. Developing Technology Applications for Intervention Research: A Case Study. Computers in Nursing. 2000;18:260-264. 66. Slack WV. Cybermedicine - How computing empowers doctors and patients for better health care. San Francisco: Jossey-Bass Publishers, 1997.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 26 Table 1. Care giver and recipient characteristics, intervention and control groups,

Boston Resources for Enhancing Alzheimer's Caregiver Health (REACH) Study (n=100).

Telephone

Linked Care Control

(intervention) Group

N or % or N %

Mean sd Mean sd

Caregiver characteristics

Age in years (mean (sd)) 61.4 13.6 63.7 11.8 F(df)=0.8(1) p=0.378

female 40 82 38 75

male 9 18 13 26 2(df)=0.7(1) p=0.390

Ethnicity

White 38 78 41 80

Black or African American 7 14 9 18

Hispanic 2 4

Other ethnicity groups 2 4 2 (df)=4.4(3) p=0.226

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 27 Employment

full-time 8 16 9 18

part-time 5 10 12 24

homemaker 9 18 7 14

retired 23 47 22 43

unemployed 4 8 1 2 2 (df)=5.0(4) p=0.290

Relationship to Care Recipient

spouse 27 55 27 53

child 17 35 21 41

sibling 2 4 3 6

other relation 3 6 2 (df)=2.4(3) p=0.497

Years of education

< High school 3 6 5 10

High school diploma 15 31 16 31

>High school 31 63 30 59 2 (df)=0.5(2) p=0.775

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 28 table 1, continued

Care recipient characteristics

Age in years (mean (sd)) 76.7 9.0 79.2 7.6 F(df)=2.3(1) p=0.136

female 23 47 25 49

male 26 53 26 51 2 (df)=0.0(1) p=0.835

Years of education

< High school 11 23 15 31

High school diploma 21 45 13 27

>High school 15 32 21 43 2 (df)=3.5(2) p=0.178

MMSE score (mean (sd)) 11.6 7.6 11.2 9.0 F(df)=0.1(1) p=0.817

ADL score (mean (sd)) 3.5 2.0 3.3 2.2 F(df)=0.1(1) p=0.706

IADL score (mean (sd)) 7.6 1.1 7.3 1.2 F(df)=0.9(1) p=0.342

df, degrees of freedom; sd, standard deviation

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 29 control (n=51, circle), intervention (n=49, triangle) control (n=51, circle), intervention (n=49, triangle)

24.1 24.6

21.2 21.8 ) ) s t s

i 18.3 19.1 t i n n u

D 15.4 16.3 D S S 5 5 2 .

12.4 2 13.5 . 0 ( 0

C 9.5 10.8 D P S B E M

6.6 C 8.0 R

3.1 4.7

0.8 2.5 baseline 6 mo 12 mo 18 mo baseline 6 mo 12 mo 18 mo observation observation treatment effect, total sample: p=0.140 treatment effect, total sample: P=0.323 control (n=51, circle), intervention (n=49, triangle)

) 23.8 s t i n u

22.1 D S

5 20.4 2 . 0 (

I 18.7 A T

S 17.0

13.5 baseline 6 mo 12 mo 18 mo observation treatment effect, total sample: p=0.377

Figure 1. Main intervention effect for ratings of bothersome behavior (RMBPC bother, top left panel), depressive symptoms (CESD, top right panel) and anxious symptoms (STAI, bottom left panel). The range of the y-axis includes the mean ±1 standard deviation (sd), marked in 0.25 sd units based on the distribution of the total sample at baseline. Lines plotted are model- implied trajectories for the outcomes for treatment and control groups derived from a generalized estimating equations model with exchangeable correlation structure. P-values refer to the significance of the interaction between the intervention group and time. Significant values (p<0.05) imply the rate of change is different in the control and intervention groups.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 30 RMBPC control (n=32, circle), intervention (n=32, triangle) CESD control (n=32, circle), intervention (n=32, triangle)

24.1 24.6

21.2 21.8 ) ) s t s

i 18.3 19.1 t i n n u

D 15.4 16.3 D S S 5 5 2 .

12.4 2 13.5 . 0 ( 0

C 9.5 10.8 D P S B E M

6.6 C 8.0 R

3.1 4.7

0.8 2.5 baseline 6 mo 12 mo 18 mo baseline 6 mo 12 mo 18 mo observation observation treatment effect, low & mid mastery: p=0.040 treatment effect, low & mid mastery: p=0.007 STAI control (n=32, circle), intervention (n=32, triangle)

) 23.8 s t i n u

22.1 D S

5 20.4 2 . 0 (

I 18.7 A T

S 17.0

13.5 baseline 6 mo 12 mo 18 mo observation treatment effect, low & mid mastery: p=0.012

Figure 2. Intervention effects, limited to participants with low to mid mastery (Pearlin, 0.5 standard deviations (sd) above the total sample mean or less) at baseline for bothersome behavior (RMBPC bother, top left panel), depressive symptoms (CESD, top right panel) and anxious symptoms (STAI, bottom left panel). The range of the y-axis includes the mean ±1 standard deviation (sd), marked in 0.25 sd units based on the distribution of the total sample at baseline. Lines plotted are model-implied trajectories for the outcomes for treatment and control groups derived from a generalized estimating equations model with exchangeable correlation structure. P-values refer to the significance of the interaction between the intervention group and time. Significant values (p<0.05) imply the rate of change is different in the control and intervention groups.

D:\Docs\2018-04-04\05b19384ed6fd02f16bfcc0b1d6eb7f7.doc04/27/184:07 上午 31