Implementation Science in Thailand: Design and Methods of a Geriatric Mental Health Cluster-Randomized Trial

The study will be implemented in 16 geographical districts (i.e., clusters), and all districts will be randomly assigned to either the intervention arm or the control arm. Both arms will receive the same RDAD clinical intervention protocol. The intervention arm will receive the RDAD intervention plus the GTO implementation support protocol, while the control arm will receive RDAD only.

The intervention study period will last for 6 months, providing adequate time to observe effects. Patient and caregiver data will be collected at baseline, month 3 (posttreatment), and month 6 (3-month follow-up) (Figure 1).

This study will be conducted in Thailand’s Khon Kaen province, which has 26 districts; each district comprises from three to 18 subdistricts. Each subdistrict has one long-term care (LTC) program led by a CM (typically a trained nurse) along with several CCGs (typically community health workers with additional training in elder care) who provide health and social care to community-residing elders. Ten of the 26 districts were excluded because they were already participating in another dementia-related project (N=2), their health care systems were too new to be fully functioning (N=4), and there were too few active LTC programs to support feasibility of the intervention within the LTC system (N=4). Of the remaining 16 districts (147 subdistricts, with over 267,000 older adults), eight were randomly selected for the experimental arm and the other eight for the control arm.

Patients will be included in the study if they are age 60 or above, have “probable dementia,” have one or more behavioral or psychological symptoms, are ambulatory, and have an adult (age ≥18) family caregiver who lives with or spends at least 4 hours every day with them. Measures used to determine eligibility include the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE), a locally validated short questionnaire completed by a family relative or friend familiar with the person to assess cognitive impairment in older adults (score ≥3.47 indicates probable dementia) (13); the Neuropsychiatric Inventory (NPI-Q), a well-validated instrument translated into Thai that relies on caregiver self-report to assess the presence and severity of BPSD (14); a validated ambulatory test that assesses a person’s ability to walk independently or with an assistive device (e.g., canes or walkers) but without physical assistance for support from another person (5); and a medical history constructed from family caregivers’ responses about the presence of chronic diseases, with any positive response further evaluated by the study physician to identify persons with medical contradictions to physical exercise who should be excluded from study participation. Patients will be excluded from the study if they or their family caregivers do not assent to study participation or if their CM and/or primary care providers recommend that the patient not participate in the intervention because of concerns about a medical condition (e.g., severe or unstable cardiovascular disease) and/or frailty based on medical record.

Patients will be primarily recruited from referrals from Thailand’s LTC system, where participating CMs identify candidate study participants based on their Mini-Mental State Examination (MMSE) scores, which are routinely collected and included in patients’ LTC plans. LTC patients who have been identified as having cognitive impairment on the basis of the MMSE will be screened with the IQCODE to identify persons at high risk for dementia, referred to hereafter as “probable dementia.” Participants with an IQCODE mean score above the 3.47 threshold for probable dementia and who demonstrate at least one significant BPSD symptom as measured by the NPI-Q will be eligible for participation in the trial. On the basis of the high sensitivity (90%) and specificity (95%) of the IQCODE (15), we anticipate that the positive predictive value of the IQCODE will be >90%, thus ensuring the likelihood that we will successfully recruit the required sample in the proposed study period.

Because many persons with dementia in Thailand will not have received a formal diagnosis and because project resources do not allow us to obtain a clinical diagnosis, we will rely on an adjudication process to identify persons with probable dementia. After the initial screening based on IQCODE is completed, a team consisting of the research assistant, research director, CM, and study physician will review the results of the IQCODE, along with other medical conditions (e.g., cardiac disease, hypertension), to determine whether a person meets clinical criteria for dementia according to DSM-5 and to identify any medical contraindications to the exercise intervention.

Achieving 80% power for a medium effect (0.4) at an alpha level of 0.05 among 16 districts/clusters will require 288 dyads (each comprising a person with dementia and a family caregiver) for the whole study. We plan to recruit 340 (170 per group) dyads, allowing for 15% attrition.

Two separate consents will be obtained: one for participation in screening and one for participation in the intervention study. Elderly individuals who screen positive on the assessments (i.e., IQCODE and NPI-Q) will be eligible for participation in the randomized controlled trail (RCT). Per Thai institutional review board (IRB) laws, both the family caregiver and the person with dementia are required to consent to the participation of the person with dementia in the trial. The family caregiver, or someone who has the right to act as the legally authorized representative of the person with dementia (e.g., legal guardian or family member), will provide a signature to indicate permission for his or her relative’s study participation. If the family caregiver is not able to read and write, a thumbprint from the family caregiver is required to indicate permission for the person with dementia to participate.

The person with dementia will be read a consent form that uses simple language to explain the study. We will follow recommendations of Black and colleagues (16) for assisting persons whose capacity for consent might be diminished. Thus, study staff are trained to observe verbal willingness to participate (e.g., saying yes or being willing to do what the family caregiver consents to), behavioral indications (e.g., acting agreeable, cooperating), and emotional indications (e.g., having a positive facial expression) (16). They will also observe indications of dissent to participate, as expressed verbally (e.g., saying no), behaviorally (e.g., not cooperating, being agitated), or emotionally (e.g., showing distress or unhappiness) (16).

Persons with dementia who do not provide verbal, behavioral, or emotional dissent will provide consent through either a signature or thumbprint to confirm their agreement for study participation. Persons with dementia have fluctuating behavioral and psychological problems (e.g., agitation, apathy, paranoia) that might temporarily interfere with their ability to engage in the consent process. Thus, if they are not able to provide consent by either signature or thumbprint, they will be approached up to two additional times. If they are not able to consent after three attempts, they will not be eligible for study participation. Although the Thai procedure differs from U.S. IRB study procedures, which rely on an “assent process” facilitated by an expert assessing the capacity of the person with dementia for assent, the proposed procedure for those with diminished capacity to consent is compliant with Thai IRB laws as well as the applicant institution’s IRB.

In addition, consent will be obtained from family caregivers for their own participation in the RCT intervention study. Although caregivers are not required to participate in the intervention’s physical exercise activities, they will be actively engaged in helping ensure the safety of the person with dementia throughout the physical activity intervention. The family caregiver will also be engaged in the behavioral aspects of the intervention study, learning how to better manage BPSD. Similar to prior interventions showing reductions in stress and burden and increased quality of life for involved caregivers, this study will document and analyze changes in family caregiver outcomes.

Cluster randomization will occur at the district level. To maximize the even distribution of factors that may affect outcomes (17), we employed three steps to implement the cluster randomization procedure. First, we conducted a formative evaluation of geographical variation among districts in Khon Kaen province and identified two covariates that might affect implementation outcomes: number of subdistricts (range 4 to 18) and the ratio of active LTC systems to number of LTC systems, as discussed above. Second, we stratified all participating sites/districts by these two covariates for a 2×2 distribution of districts, resulting in an even distribution across the four conditions. We next conducted stratified randomization, using a computer-generated randomization program to randomly assign the sites in sequence to ensure equal distribution between two study arms.

The intervention and control arms will receive the same culturally adapted and evidence-based RDAD clinical intervention, in accordance with its standard implementation guidelines, as well as the “usual practices” of program implementation in Thailand. Usual implementation practice consists of a top-down centralized program announcement, with a printed description of the program and its administrative requirements (e.g., schedule for submission of “care plans” and other performance indicators) and basic consultation support from district LTC management offices. The intervention arm will receive the additional GTO intervention to support implementation of the RDAD protocol (Table 1).

To assess the feasibility of conducting the proposed intervention study in Thailand, we conducted five formal meetings with over 30 key stakeholders (including health policy makers, different types of health service providers, researchers, and care receivers) and in-depth qualitative interviews with eight health professionals. We made site visits to community-based primary care settings and residential homes (Chuengsatiansup et al., 2020, unpublished manuscript). The focus of these meetings was to determine the most appropriate service delivery system in which to implement the study.

Themes and key messages that emerged from these meetings were examined, and follow-up conversations were held to clarify issues that were raised. The research team gradually achieved consensus regarding three items. First, there is increased recognition of burden of care for older adults with BPSD. Second, there are system-wide shortages of mental health specialists in the LTC system and limited availability of trained community-based health professionals in the LTC system to deliver in-home support services to persons with dementia (Box 1). Third, intervention programs should be delivered by community-based health professionals in the existing service platform of the Thai LTC system.

Through the formative research, the original RDAD protocol was modified for local cultural, clinical, and service systems in a two-step process. First, possible modifications were explored in collaboration with RDAD’s developer during her consultation with the study team in Thailand. The research team then held a series of workshops with CMs and CCGs to obtain their feedback about the acceptability of the recommended cultural adaptations, the feasibility of implementing them for Thai culture, and any additional changes that they recommended to ensure cultural and clinically appropriate communication of key concepts. The following adaptations to the original protocol were agreed upon: using local devices (e.g., drums) and activities (e.g., walking to a Buddhist temple, local folk dance) to illustrate acceptable physical activities, using storytelling to elicit CMs’ and CCGs’ clinical experiences to increase their appreciation of the relevance of the intervention programs to their clinical practice, and shortening intervals between follow-up sessions (the original 5-month RDAD intervention was changed to 12 weeks, which matches the reporting and monitoring cycle of the local LTC service system). All of these adaptations were approved by the RDAD developer, who confirmed that all critical elements of the intervention were preserved.

The proposed protocol will be implemented over a 12-week period, for a total of 13 sessions. Each dyad will be seen in the patient’s home for all 13 sessions, with each session lasting from approximately 30 minutes to 1 hour. Besides in-person discussion at each session, a brochure showing physical activity procedures and a set of forms (e.g., appointment schedule, RDAD procedures checklist) will also be given to the family caregivers. The schedule of home visits is as follows: eight sessions (2 sessions per week) for the first 4 weeks (weeks 1–4), followed by four sessions (1 session per week) for 4 weeks (weeks 5–8), and one last session on week 12. After the first nine home visit sessions, which are primarily for teaching, coaching, and reviewing, four additional follow-up home visit sessions will occur to review the status of behavioral management practice and physical exercise, answer questions, consolidate treatment gains, and encourage long-term retention in the program.

We conducted workshops with CMs and CCGs to ensure understanding of key concepts of the GTO and RDAD intervention protocols and to ensure that they fit the clinical workflow in the Thai context. Workshops were followed by qualitative interviews with trainees. We also conducted consensus meetings in which the research team, including clinicians, medical anthropologists, and public health officials, worked closely with participating CMs and CCGs to determine the fit of the interventions within the existing clinical flow of the Thai LTC system.

Given that GTO has been shown to increase implementers’ overall capacity to implement evidence-based intervention programs, our primary hypothesis is that the intervention arm, which employs the GTO implementation support intervention, will result in significantly better implementation outcomes. Based on the prior literature (see below), we selected the following five implementation outcome indicators: higher fidelity in implementing the RDAD clinical intervention, higher dosage of RDAD received by the participants, better reach of participants in the community, better acceptability as perceived by the RDAD program implementers, and higher satisfaction of the participants (i.e., family caregivers and persons with BPSD) with the RDAD clinical intervention, as compared with the control arm.

Assuming that improved implementation of the RDAD clinical intervention will lead to better intervention outcomes, our secondary hypothesis is that the intervention arm will lead to better clinical outcomes for persons with BPSD, including greater reduction in BPSD, better general medical health, and improved overall quality of life, compared with persons with BPSD in the control arm. In addition, because the family caregiver is also a key receiver of the RDAD intervention while assisting persons with BPSD, our tertiary hypothesis is that the intervention arm will lead to better family caregiver outcomes, including reduced burden, lower psychological distress, and improved quality of life, compared with caregivers in the control arm.

Three sets of measures will be used for outcome evaluation (Table 2). Implementation outcomes (primary objective) include fidelity (the extent to which the content of the intervention is delivered as intended), assessed by percentage of scheduled sessions being completed, based on a study log documented by the family caregiver after each session and confirmed by CCGs weekly (10); dosage (the actual amount of intervention received by the participants), assessed by CCGs and documented in a study log (7); reach assessment, based on records from interventionists (i.e., CM and CCG) and administrators and documented in a study log (21); acceptability, assessed by CCG and CM responses to a brief questionnaire at months 3 and 6 (22); satisfaction, assessed by the family caregiver and person with dementia (if possible) in response to questionnaire administered at months 3 and 6 (23); and readiness for change, assessed by the Activity-Based Readiness Tool, a self-report questionnaire to be completed by CCGs and CMs at baseline, month 3, and month 6 (24).

Patient clinical outcomes (secondary objective), collected at baseline, 3 months (posttreatment), and 6-month follow-up, include behavioral and psychological symptoms, assessed by the NPI-Q (14); depressive symptoms, assessed by the Geriatric Depression Scale (25); cognitive impairment, assessed by the Thai Mental State Examination (TMSE) (26); physical functioning, assessed by the Medical Outcomes Study Short Form Health Survey–36 (27) and the Sickness Impact Profile (mobility subscale) (28); activities of daily living, assessed by the Barthel Index for Activities of Daily Living (29); and quality of life, assessed by Quality of Life in Alzheimer Disease (QoL-AD) (30). All of these scales, except for the TMSE, are based on caregiver report.

Caregiver clinical outcomes (tertiary objective), collected at baseline, month 3, and month 6, include caregiver burden, assessed by the Zarit Burden Interview (31); depressive symptoms, assessed by the Patient Health Questionnaire–8 (PHQ-8) (32) (i.e., PHQ-9 without the suicidality item); caregiver distress, assessed by symptom-related distress on the NPI-Q (14); and quality of life, assessed by QoL-AD for caregivers (30).

In addition to descriptive analyses and univariate comparisons of the experimental and control intervention group conditions at baseline, we will use repeated-measures mixed models to analyze each of the implementation and clinical outcome measures. We will also use an intraclass coefficient to reflect the expected positive within-group correlation. For the analysis, the outcomes will be calculated for each individual participant, and individual outcomes data will be nested within each district. Using an intention-to-treat approach, we will conduct analysis of covariance with multivariate linear regression modeling to test effects of the intervention at month 3 and month 6. Intercepts and treatment effect will be modeled as random effects at the participant level and as fixed effects at the district level, whereas patient characteristics (e.g., levels of dementia severity and mobility) as well as length of caregiver involvement (i.e., average hours spent per day) will be control variables. Results of statistical procedures will be presented by using adjusted means (least-squares means) with 95% confidence intervals for intervention and control groups, respectively.