Design
There were five study sites: Boston, MA; Houston, TX; Providence, RI; Oklahoma City, OK; and Beaumont, TX, with all sites located in one of two selected Veterans Integrated Service Networks (VISNs). VISNs provide a unifying administrative structure for all VA services within a given geographic region. Study sites were matched by VISN to assure uniformity in this overarching administrative structure. One of the two selected VISNs (that is, VISN 16, which includes Houston, Oklahoma City and Beaumont) was chosen because it was the location of the study’s VA principal investigator. The other selected VISN (that is, VISN 1, which includes Boston and Providence) had a similar array of VA services as VISN 16.
Within each of the two selected VISNs, VA study sites were matched to be similar in: size, services offered (both inpatient and outpatient), academic affiliations, research missions and medical-residency training programs. Alzheimer’s Associations chapters within selected VISNs were similar in size, with comparable core programs and services. After selecting matched sites within each VISN, one was randomly selected to deliver PDC and the other was deemed the comparison site that would deliver usual care (UC). Specifically, in one VISN, Boston was randomly selected to deliver PDC and Providence was selected as its matched UC site. In the other VISN, Houston was randomly selected to deliver PDC and Oklahoma City was selected as its matched UC site. Beaumont was paired with Oklahoma City as part of the same UC site to assure a sufficiently sized comparison sample. Matched PDC and UC sites, rather than within-site randomization, were used to allow PDC implementation throughout partnering organizations, without concerns about diffusion of PDC to the UC group.
Veterans and caregivers at PDC sites received the care-coordination program and an initial, basic packet of educational materials on dementia; UC-site veterans and caregivers received the same basic educational materials and usual care from the VA and chapters. Aside from PDC care coordination, there were no restrictions on services or care that could be obtained by the PDC or UC groups from the VA, chapters or other organizations.
Sample
Eligibility requirements for veterans included receiving primary healthcare from the VA, residing outside a residential care facility at the time of enrollment, living within a partnering chapter’s service area, being 60+ years of age and having at least one of the following dementia diagnostic codes from the International Classification of Diseases, Ninth Revision recorded in the VA medical record: 290.41–43, 291.2, 292.82, 294.1, 294.8 and 331.0. There was no enrollment restriction based on type or severity of veterans’ impairments, or availability of a family caregiver. VA primary care physicians confirmed veterans’ diagnoses and eligibility prior to sample selection.
The sample was recruited from 18 January 2007 to 22 July 22 2009. Research interviews occurred from 21 March 2007 to 15 September 2010. Veterans’ ability to complete research interviews that provided self-reported outcomes was determined by a telephone screening, using the short Blessed Orientation-Memory-Concentration Test [
26]. The Blessed Test comes from a longer instrument and was one of the tools recommended in a dementia diagnostic protocol published by the Agency for Health Care Research and Policy [
27]. It includes six items that assess time orientation, ability to count backwards from 20, ability to say months in reverse order and ability to repeat a simple phrase that includes a name and address. The Blessed Test was successfully used as a brief instrument for measuring cognitive status over the telephone in prior studies of BRI Care Consultation [
6]. Because of its ease of use and the lack of other tested methods for deciding whether a person with dementia could complete a research interview focused on self-reported outcomes, the Blessed Test was adapted as a screening device for determining whether veteran interviews were attempted. In this adaptation, three criteria were used. First, veterans had to be able to give answers to all six questions over the telephone, even if there were errors in the answers. Second, veterans had to be able to repeat accurately at least two parts of a three-part phrase immediately after the telephone screener stated it to them. Third, veterans either had to count backwards accurately from 20 to 15 or accurately name at least three consecutive months in reverse.
Sample-size estimates were based on a one-tailed significance test, a modest effect size (.12), .80 power, an alpha of .05, 10 independent variables in a regression equation and an estimated squared multiple correlation of .55. A minimum sample size of 247 was determined based on these assumptions.
PDC model intervention
Two half-time care coordinators, with part-time administrative assistant support, delivered PDC at each intervention site. One care coordinator worked in the local VA medical center (healthcare organization) and the other worked in the partnering Alzheimer’s Association chapter (community service organization). Although from different organizations, the two care coordinators worked as a team, with one shared electronic Care Coordination Information System (CCIS) and regularly scheduled planning and case-conference meetings. Care coordinators had bachelor’s or master’s degrees in social work, nursing or other helping professions.
The care coordinator from the VA medical centers had primary responsibility for assisting veterans with medical-related concerns (for example, medications, accessing medical services, disease management) while the care coordinator from the Alzheimer’s Association chapter had primary responsibility for assisting caregivers with nonmedical concerns (for example, care-related strain, accessing family support and information services). The VA care coordinator also focused on helping families access VA services and benefits, whereas the Alzheimer’s Association care coordinator focused on helping families use community services, including those offered by the Alzheimer’s Association. This division of labor between care coordinators capitalized on the complementary strengths of each partner organization and represented a bridge between health care and community services.
Training for care coordinators consisted of a 1.5-day initial session on the PDC philosophy, service-delivery protocol and the CCIS that guides service delivery. Additionally, one- to two-hour biweekly refresher trainings were completed throughout the study period. These sessions focused on case reviews to monitor fidelity to the intervention protocol, strategies for working with a partner organization, using the CCIS and handling difficult cases. Continuing education also was provided on special topics, such as differences among illnesses that cause dementia, helping families respond to emergencies and respite for caregivers.
PDC is a coaching model driven by consumer choice, with care coordinators helping find solutions to concerns that are the priorities of veterans and caregivers. PDC followed a set, standardized protocol that required a minimum of at least one contact between care coordinators and consumers per month; more-frequent contacts occurred as needed. The protocol required care coordinators to discuss with veterans and/or caregivers a broad range of medical and nonmedical concerns, although the specific content was customized to consumers’ preferences and needs.
PDC is a low-cost service delivered by telephone, mail and e-mail, with in-person contacts rarely needed. The two half-time care coordinators from the partnering organizations (one full-time equivalent (FTE)) maintained caseloads of 75 to 125 families. All expenses to deliver PDC (that is, salaries, benefits, equipment, supplies, training, software, licensing, supervision, administrative overhead) can be recovered by charging a fee of $60 to $80 per month per family.
PDC gives equal attention to preferences and needs of veterans and caregivers. Veterans with dementia are engaged in the program whenever possible, despite their impairments. Veterans without caregivers are able to use PDC, so long as they can communicate by telephone. If veterans are too impaired to communicate by telephone, their caregivers can be the sole participant in the program.
PDC has three main components: 1) initial assessment, 2) action plan, and 3) ongoing monitoring and reassessment.
Initial assessment
The initial assessment is completed gradually during the first four weeks of enrollment. It is designed to be brief, with the action plan to address assessed concerns implemented simultaneously with or prior to completion of the entire initial assessment. The initial assessment covers a broad range of domains or potential problem areas: 23 for veterans (for example, coordinating and accessing VA services, medication management, getting and understanding the diagnosis) and 16 for caregivers (for example, finding and accessing community services, care-related strains and depression). The required initial assessment consists of a single-item trigger question for each domain; trigger questions can be formally asked or covered informally during conversations. More extensive detailed assessment questions are available for each domain as optional tools, if additional probing is necessary to clarify a problem.
Action plan
The action plan is the core of PDC. It comprises simple behavioral tasks called action steps that, if accomplished, move veterans and caregivers toward solutions to concerns they identified as important. Action steps should be easy to complete and include, for example, calling an organization to inquire about the availability of a service, reading an educational resource on a topic of concern or contacting another family member to ask whether he or she is willing to help with a caregiving task. With coaching and guidance from care coordinators, veterans and caregivers determine the content of action steps, who will complete the action steps and the projected dates of completion. New action steps are continuously added and build upon prior action steps. Multiple action steps, spread over a period of weeks or months, often are needed to find solutions to specific problems. As action steps are completed, veterans and caregivers move toward solutions and gain confidence in their self-management abilities. Copies of action plans are mailed to veterans and caregivers and summarized in the larger medical record.
On average, each veteran and his or her caregiver had more than seven action steps. The most common pertained to accessing and coordinating services and benefits available from the VA, Alzheimer’s Association or other community agencies. Specifically, 78% of veterans and caregivers had action steps related to accessing VA services or benefits, 59% to accessing Alzheimer’s Association services and 76% to accessing other community organizations. Other common action steps focused on improving care from the informal network (57%), managing symptoms (40%), improving communication with healthcare providers (33%) and home safety (29%).
Ongoing monitoring and reassessment
The hallmark of PDC is a long-term relationship to provide continuous support to veterans and caregivers. Ideally, care coordinators become knowledgeable and familiar experts who are trusted by families. On average, families had over 14 contacts with coordinators during the twelve-month study period, which focused on completing the required initial assessment and reassessments, adding new action steps and checking the disposition of pending action steps, and completing required routine checking.
Reassessments involved readministering trigger questions used in the initial assessment. They were required at least every six months. More frequent reassessments for selected domains are recommended for persistent or ongoing problems. Reassessment helps care coordinators and consumers gauge progress in finding solutions to problems.
Consistent with the design of PDC, the most contacts between care coordinators and veterans or caregivers were by telephone (80%) and regular mail and e-mail (16%), with a small number in person (4%). The number of contacts was evenly split between care coordinators from the VA and the Alzheimer’s Association, which reflected PDC’s team-based delivery model. Care coordinators initiated approximately 90% of contacts; veterans or caregivers initiated 10%. (For a more detailed description of PDC, see Judge
et al. [
28]).
Outcomes measures
Five self-reported outcomes from veterans with dementia (that is, 1) unmet needs, 2) embarrassment about memory problems, 3) isolation, 4) relationship strain and 5) depression) were examined after six and twelve months. As described below, psychometric properties of these outcomes were tested with the current sample of veterans to reaffirm their reliability and structural validity, which had been established in previous studies for all but one outcome (citations provided below). Overall, psychometric analyses showed individual items comprising these outcomes had good internal reliability, based on Cronbach’s alpha, and good structural validity, based on high factor loadings on a single factor and low cross-loadings on factors representing other outcomes. Thus, psychometric analyses reinforced the feasibility of collecting information from persons with dementia, despite mild-to-moderate levels of cognitive impairment. However, a limitation of the outcomes was the use of simple ‘yes/no’ response choices for most questions comprising these measures to facilitate administration with cognitively impaired respondents [
29].
Unmet needs
Developed for this study, this outcome was based on 24 dichotomous questions that were summed to measure veterans’ perceptions of unmet need across eight domains: 1) understanding dementia, 2) daily living tasks, 3) accessing VA and other services, 4) legal and financial issues, 5) organizing family care, 6) alternative living arrangements, 7) emotional support and 8) medications. This measure had good structural validity, with factor loadings from .83 to .62 on a single factor and good internal reliability (Cronbach’s alpha .93, .90 and .93 at baseline, six months and twelve months, respectively).
Embarrassment about memory problems
This previously published outcome [
30] was the sum of three dichotomous items that asked whether veterans felt embarrassed about memory problems, uncomfortable telling others about memory problems and uncomfortable accepting help for memory problems (Cronbach’s alpha .75, .74 and .74 at baseline, six months and twelve months, respectively). Factor analysis confirmed the independence and structural validity, with loadings from .53 to .67.
Isolation
This previously published outcome [
30] included four dichotomous items and asked veterans whether their health problems and need for assistance made them feel isolated from other people, less able to participate in group activities, less able to participate in church or religious activities, and less able to visit with family and friends (Cronbach’s alpha .77, .77 and .82 at baseline, six months and twelve months, respectively). Factor loadings for these items ranged from .39 to .84.
Relationship strain
This outcome was adapted from a published family caregiving measure [
31] and was the sum of four dichotomous items focused on veterans’ perceptions of the quality of the relationship with their caregivers. Questions asked whether, because of their health problems and need for assistance, veterans felt that their caregiver tried to manipulate them, felt that the relationship with the caregiver was strained, felt resentful toward the caregiver or felt angry toward the caregiver (Cronbach’s alpha .78, .77 and .84 at baseline, six months and twelve months, respectively). Factor loadings for these items ranged from .42 to .85.
Depression
Veteran depression was measured by the 11-item Center for Epidemiologic Studies Depression Scale, which had good reliability at all three data-collection waves (Cronbach’s alpha of .76, .79 and .78 at baseline, six months and twelve months, respectively) [
32].
Analytic strategy
Multiple regression equations tested whether veterans in the PDC and UC groups differed across the five outcomes: 1) unmet need, 2) embarrassment, 3) isolation, 4) relationship strain, and 5) depression. Two equations were estimated for each: one representing change from baseline to six months and one representing change from six to twelve months. In each equation, the effect of PDC on an outcome was represented by the regression coefficient for a dichotomous variable that distinguished between veterans in the PDC and UC groups. In addition, two product or interaction terms tested the conditional-effects hypothesis that posited the benefits of PDC would be greater for more impaired veterans (for example, those with more cognitive impairment and/or personal-care dependencies). Product terms were constructed by multiplying the dichotomous intervention variable by the two variables representing impairments. To facilitate interpretation, impairment measures were centered to have a mean of zero [
34].
Regression coefficients for the PDC variable and product terms were of primary interest to this study. When there were no significant product terms and the regression coefficient for the PDC variable was significant, it meant the difference between the PDC and UC groups pertained to all veterans, regardless of level of impairment (that is, no conditional effects). When there was a significant product term, the regression coefficient for the PDC variable represented the difference between the PDC and UC groups among veterans with mean levels of cognitive impairment or personal-care dependencies. The regression coefficient for a significant product term represented the average difference in an outcome between PDC and UC groups among veterans with higher- and lower-than-average levels of impairment. (See Cohen and Cohen [
35] or McClendon [
34] for a detailed discussion of interpreting product terms in regression analysis).
Final regression equations included the dichotomous PDC variable and any significant product terms, measures of veterans’ cognitive impairment and personal-care dependencies, the prior-wave’s version of the dependent variable and four background characteristics to control for significant baseline differences between the PDC and UC groups. By including the prior-wave’s version of the dependent variable, equations focus on ‘changes’ in outcomes from baseline to six months or from six to twelve months. Based on the coding of the intervention variable (that is, 1 = PDC group; 0 = UC group), results consistent with hypotheses were indicated by significant regression coefficients with negative values. Analyses were run using IBM SPSS Statistics, Version 21 (Armonk, NY).