Setting of the LOCK Intervention
The LOCK sleep intervention will be implemented in a total of 27 nursing homes. We will recruit 3 nursing home corporations that have (1) a minimum of 12 nursing home sites (8 will eventually participate in the study), (2) 50+ beds, (3) ≥30 long-stay residents with dementia diagnoses, (4) centralized corporate training staff that can devote 50% time to implement the intervention (corporate coaches). Once the corporations are confirmed, eligible nursing homes within the corporations will be identified for recruitment of Phase 1 and 2. We will finalize nursing home eligibility criteria with the selected corporations using the nursing home’s Minimum Data Set (MDS) and Certification and Survey Provider Enhanced Reporting (CASPER) self-reported data as the trial progresses (see Data Collection for further details). Leaders at each participating site (i.e., administrator, chief nurse, medical director, nurse managers) will commit to (a) researcher-determined scheduling of intervention participation as per the randomization schedule, (b) taking on identity and responsibility of site leadership team, and (c) work with the study staff to execute secure data transfer/shipping protocols. Working with corporations provides the advantage of shared corporate infrastructure across nursing homes to facilitate screening, implementation, and data collection.
Nursing home recruitment for Phase 1 will require cooperation from the 3 enlisted corporations to identify 3 individual nursing homes (1 per corporation) to participate in the piloting and refining of the LOCK sleep intervention. Once identified, the study intervention team (study co-investigators and consultants with clinical and implementation expertise with the LOCK sleep intervention) will lead the Phase 1 implementation of the intervention. Stakeholder engagement will occur at both corporate and the nursing home levels during this phase. At the corporate level, each corporation will assign a corporate coach to attend all nursing home visits (which may be virtual visits) with the implementation team at their corporation’s selected Phase 1 nursing home. This allows training of corporate coaches to ensure preparedness and organizational support for the nursing home when they enter study Phase 2. At the nursing home level, each nursing home will create site leadership and implementation teams to support involvement of all staff in the LOCK sleep intervention.
Nursing home recruitment for Phase 2 includes expanding our sample to an additional 24 nursing homes, 8 from each corporation, using the same inclusion criteria as Phase 1. Once selected, each nursing home will receive the piloted and refined LOCK sleep intervention.
Resident Sample
In Phase 1, we anticipate enrolling 57 residents with dementia who meet inclusion/exclusion criteria (approximately 19 per each of 3 nursing homes). In Phase 2, we anticipate enrolling 456 residents with dementia (approximately 19 per each of 24 nursing homes). An approximation of 19 residents per nursing home is calculated according to an averaged 50% prevalence of dementia within nursing homes and conservative 50% consent rate based on our study team’s experience with previous nursing home dementia trials (69, 70). We have included an estimate of approximately 20% attrition across the two study phases to maintain statistical power (71).
The potential participant pool from each selected nursing home will include all residents aged >=50 years with a dementia diagnosis. Because the LOCK sleep intervention allows staff to focus on residents with dementia who have the greatest sleep problems, without differentiating by dementia severity, we will include residents across the range of dementia severity. Nursing home staff will be trained to identify nursing home residents with high risk of obstructive sleep apnea using the STOP-Bang screening tool (72-74), as well as trained on appropriate procedures for referring any positively screened residents for medical evaluation. We will exclude residents with a high risk of obstructive sleep apnea who are not yet being treated for said condition due to evidence of inaccurate actigraph measurements in that population (75). Staff will also exclude residents who have a persistent bilateral resting tremor or paralysis in both arms (a subset of persons with Parkinson’s disease and related significant tremor-causing diagnoses), due to similar actigraph measurement inaccuracies (76).
Due to inclusion of residents with dementia diagnoses, a two-step resident consent process will be used to address the ethical balance between beneficence, assuring that residents who do not have the capacity to consent are protected by their legal authorized representative’s oversight, and promotion of autonomy for residents who do have capacity to consent. First, after leadership teams from each nursing home compiles a list of all residents with a dementia diagnosis, they will send a letter to each resident’s legally authorized representative informing them of the study and inviting the legally authorized representative to opt out if they do not wish to be contacted by research staff (77-80). Research staff will then contact all representatives who do not opt out, inviting each to consent to the resident’s participation. Once legally authorized representative consent is collected, the second consent stage for nursing home residents with mild dementia will occur. Site leadership team members will work with our study team to identify those residents with a mild MDS Brief Interview for Mental Status (BIMS) score (i.e., a score above 10) and a resulting increased likelihood of capacity to consent (81). Nursing home leadership teams will then arrange a telephone interview between study staff and these residents to ascertain the resident’s ability to consent using comprehension questions. If the resident is determined to have the ability to consent during this interview, the resident consent process formally commences at that time, with a nursing home staff person serving as in-person witness.
Importantly, if a resident can respond appropriately to all capacity screening questions but does not wish to participate, they will not be enrolled in the study, regardless of their legally authorized representative’s consent. For nursing home residents without capacity to consent that become enrolled in the study, their assent to participate will be monitored during the study at each step, and their decision to withdraw at any time (whether expressed verbally or by resistance to participation) will be honored.
Staff Sample
We will enroll 60 frontline staff in the Phase 1, piloting and refinement year, of our clinical trial. In Phase 2, we will enroll 480 staff in our clinical trial across the 24 nursing homes. We will invite all frontline NH staff roles interested in the training and intervention to enroll (e.g., Nursing Assistants, Licensed Practical Nurses, Registered Nurses, and other interdisciplinary frontline staff members). We will recruit 20 staff per nursing home to participate in study interviews about their experiences with the LOCK sleep intervention and Fitbits at mid- and at six-months post-implementation.
To increase enrollment and retention of staff, we will communicate the compelling, evidence-based benefits of the LOCK sleep intervention on improved work environments and quality of care for nursing home residents. The regular phone mentoring of nursing home corporate coaches and leadership teams will also facilitate staff retention and intervention fidelity. Consent is not required from staff members directly participating in the LOCK sleep intervention as the program itself consists of clinical practices common in nursing home QI efforts that fall within the scope of staff duties. However, it will be made clear that participation is strictly voluntary and no adverse consequences to the person’s job status or any other adverse consequences will occur if the person declines to participate in the LOCK sleep intervention at their nursing home facility. At the end of Phase 1 and Phase 2, research staff will obtain consent from the sub-sample of nursing home staff participating in interviews.
Measures, Data Collection, and Analysis Methodology
Study Design.
The LOCK sleep intervention. In Phase 1, the LOCK sleep intervention will initially use intervention materials (assistance for developing teams and implementing the intervention, checklists, role-play scenarios, and videos) developed from the pilot 6-site VA nursing home implementation [34]. Throughout this phase, we will make additional refinements based on our intervention team’s experiences from their previous work. We will also carefully adapt our materials to the non-VA nursing home setting, which differs in terms of resident population (VA is mostly male), staffing (VA has higher professional nurse staffing ratios), and leadership (VA medical director is usually on site).
Phase 2 will test the refined LOCK sleep intervention through an incomplete, stepped, wedge cluster randomized controlled trial (RCT) design (82) with 24 additional nursing homes, each acting as its own control (6 staggered steps, with 4 nursing homes per step) (See Table 1: Stepped Wedge Design with Measurement Periods). During Phase 2, each cluster of 4 nursing homes will participate in its own 22-week intervention cycle (see Figure 2: Timing and Spacing of Intervention Implementation and Sleep Outcomes Measures) after being assigned first to the control condition and then phased into the intervention at 7-week intervals. We will randomize nursing homes within corporations to steps after matching on bed size and number of long-stay residents with dementia diagnoses. Residents within each nursing home are followed longitudinally, i.e., from pre- through post-intervention measurement periods.
Outcome Measures. The primary outcome for the study is actigraph-measured resident sleep. Total sleep time will be measured using data from Micro-Mini Motionlogger Actigraphs worn by residents at night (Ambulatory Monitoring Inc., Ardsley, NY) (83). We will define nighttime as 10pm to 6am and will compute total sleep time as the total number of minutes asleep during a nighttime period. To further evaluate the impact of the LOCK sleep intervention for resident sleep quality, we will also examine the following: wake after sleep onset (total number of minutes awake during nighttime), how often the resident awoke during the nighttime, sleep efficiency (the ratio of minutes asleep to minutes awake over the period), and sleep fragmentation (an index of restlessness computed as the percentage of one-minute epochs scored as awake).
Due to imperfect correlations between actigraphic sleep data and self- or other ratings of resident sleep (84, 85), four supplementary sleep measurements will be collected from staff to triangulate the primary sleep outcome data: (a) Sleep via staff rating: a clinical, global impression of resident sleep change at the end of each 7-night actigraph sleep measurement period and at the end of each week of the 6-week sleep intervention period (86-93); (b) Staff-identified sleep-related concerns: Huddle facilitators lead the huddle team to identify and track up to 2 additional symptoms or behaviors of concern potentially related to sleep (e.g., nighttime agitation, pain), measured on the same schedule as the ‘sleep via staff rating’; (c) Inter-resident sleep variability: Medical record data to be collected at the end of the 22-week trial measurement period on the following: (1) changes in any sedating medications and/or dosages; (2) incidents of delirium; (3) any urinary tract infections; and (4) doses of any sedating medications, including as needed ones; (d) Sleep from the Minimum Data Set (MDS): Two MDS items pertaining to sleep will be included: (1) “trouble falling or staying asleep or sleeping too much.”; (2) “over the past 5 days, has pain made it hard for you to sleep at night?”
Our secondary outcomes will come from the nursing homes’ MDS 3.0. Every Medicare-certified nursing home is required to complete an MDS assessment for every resident at admission, quarterly, and at discharge, in addition to whenever the resident's status changes. We will be using the MDS to collect details about the residents’: (1) psychotropic medication use in the last seven days; (2) self-reported pain index (94, 95); (3) adherence to a regimen or isolated use of analgesic medication in the last five days, noting use of additional non-medication pain interventions; and (4) indication of decline in activities of daily living (ADLs) when considering past MDS assessments and overall functional ability (96-99). (See Table 2: Primary, Secondary, and Other Outcome Measures).
Semi-structured interviews will also provide the qualitative data needed for our team understand the fit and effectiveness of the LOCK sleep intervention during both Phases 1 and 2.
Data Collection. Data collection for the primary sleep outcome measures will follow the same timeline in Phases 1 and 2, incorporating three measurement periods: pre-intervention, intervention, and sustainment. We anticipate an initial 8-week period in which each nursing home builds its frontline staff huddle practice across all units (pre-implementation period). In the first and last week of this period, nursing home staff will conduct a week (7 days) of continuous actigraph measurement on all enrolled residents in each nursing home. These will be the pre-baseline and baseline measurements. After this, each nursing home will begin the LOCK sleep intervention on all its units (implementation period), which involves no actigraph measurement. An 8-week sustainment period will follow. In the first and last weeks of the sustainment period, nursing home staff will also conduct a week (7 days) of continuous actigraph measurement on all enrolled residents in each nursing home. These will be the post-intervention and sustainment measurements. (See Figure 2: Timing and Spacing of Intervention Implementation and Sleep Outcomes Measurement).
Nursing home staff will assist residents to wear both an actigraph and a Fitbit (if applicable during Phase 2) side by side on a wrist for the 7 days during each of the 4 measurement periods (weeks 1, 8, 15, and 22—see Figure 2) (100-102). As part of their intervention training, staff will be instructed on behavioral signs of distress. If a participant exhibits significant behavioral distress due to sleep measurement methods, staff will be instructed on modifications and alternatives. If the distress does not subside, staff will be instructed to discontinue the assessment. There will therefore be no alternative treatments due to the stepped-wedge, cluster randomized, controlled clinical trial design.
To record collected sleep and secondary data, the nursing home leadership team will use a researcher-provided, tracked, express mail service to send the research team all weekly completed staff rating forms. These mailed packages will also include any actigraphs and Fitbits for which the assessment periods are complete. For supplementary sleep data collection, research staff will additionally extract relevant information from each enrolled resident’s medical record.
For qualitative data collection, we will recruit a sample of nursing home staff engaged in huddles to participate in mid- and postimplementation semi-structured interviews (phone or in-person, as situations permit) to explore staff perceptions of the LOCK sleep intervention’s effectiveness, feasibility, facilitators, and challenges. During Phase 1, reviews of the Fitbits and suggestions to improve the interview process for Phase 2 (i.e., change in length, location, etc.) will also be collected.
Data Management. The principal investigator will be responsible for ensuring participants’ safety on a daily basis. An independent Data and Safety Monitoring Board (DSMB) will meet at least once a year by teleconference (and more often as needed), acting in an advisory capacity to the NIA Director by (1) monitoring participant safety; (2) evaluating the progress of the trial; and (3) reviewing procedures for maintaining the confidentiality of data and the quality of data collection, management, and analyses. The DSMB meetings will be guided by the NIA’s DSMB Report Template, including but not limited to: study status and stopping guidelines, recruitment status, data quality status, interim results, and safety information. This study’s DSMB members have been approved by the trials’ assigned NIA program officer as independent investigators with no scientific, financial, or other conflict of interest with the study.
Personally identifiable information from the trial will be only be stored in a secure and securely backed up University of Alabama data server and University of Texas Health Data Center in password protected files. Other collaborating sites (Boston University, Brown University, and University of Texas at Austin) will only work with collected data identified by subject number. All data, and referenced personal information therein, will be monitored by the DSMB during and after the trial. HIPAA-compliant virtual visits will assist in streamlined data collection during reduced in-person access to nursing homes due to the COVID-19 pandemic.
Analysis. In Phase 1 we will use standard microlongitudinal analysis methods to test the sensitivity of secondary outcomes and supplementary sleep data to individual differences and changes in residents’ total sleep time. For the qualitative interviews, all interviews will be audio recorded, transcribed, and processed through rapid appraisal template analysis (103, 104). The Consolidated Framework for Implementation Research (CFIR) constructs, a widely used theory for evaluating intervention implementation (71, 105), and the Relational Coordination framework will form the foundations of the analysis template (106, 107) (See Figure 3: Relational Coordination Theory). Results from this step will suggest if the value and feasibility of using the Fitbits in addition to the actigraphs supports their continued use in Phase 2.
After Phase 2 data collection is complete from all 24 nursing homes, we will examine differences in the primary outcome (total sleep time) across the pre-baseline through each nursing homes sustainment period (See Figure 2). We will use the nursing home’s separate average total sleep times across each of the four measurements. We will use a multi-level mixed models approach and either generalized linear mixed models (GLMM) or generalized estimating equations (GEE). Both GLMM and GEE approaches enable the estimation of models with random and fixed effects and the use of outcome measures having the range of distribution types likely to be observed in primary and secondary outcomes (105).
The level of analysis will be the individual nursing home resident nested within nursing home corporations, with each nursing home serving as its own control and nursing homes stratified by corporation. Data available from the pre-baseline, baseline, postintervention, and sustainment measurements (See Figure 2 and Table 1) will enable estimation of (a) within individual variation over time (pre-baseline [week 1] through sustainment [week 22]), (b) between individual variation incorporating individual characteristics (e.g., demographics, dementia severity level), and (c) facility level variation incorporating nursing home characteristics (e.g., corporate membership, facility bed-size, and staffing) (108, 109).
Since using actigraphs to measure sleep time will likely be a new practice in the nursing homes, the two pre-intervention measurement periods (pre-baseline [week 1] and baseline [week 8]) will enable us to examine the impact of measuring sleep time prior to the intervention separate from the effects of nursing home staff implementing the LOCK sleep intervention. The two post-implementation periods (post-intervention [week 15] and sustainment [week 22]) will enable us to test for longer-term intervention effects (sustainability).
Our primary analysis for trial evaluation will be a partial intention-to-treat analysis using total sleep time as the outcome. We will include all residents with both pre-baseline and baseline total sleep time assessments to properly estimate any effect of using the actigraph to measure sleep. To adjust for mid-trial attrition, we will use multiple imputation to estimate missing data, assuming no change from their last available total sleep time assessment and incorporating an estimate of random variation based on observed data. We will apply our model approach described above to the multiply-imputed datasets and results will be combined using Rubin’s Rule (110). A complete case analysis (individuals with data from all measurement periods) will be conducted to test the sensitivity of our primary analysis results.
Participating nursing homes will be treated as random effects with residents clustered within nursing homes. By clustering within nursing home, we can control for site-level effects and plausibly treat the individual residents as independent (71). We will estimate both unadjusted and adjusted models. We will include individual-level and nursing home-level characteristics in our adjusted models, as described above. A Type I error rate of 5% (α < .05) will be used to identify statistically significant associations. Raw and adjusted total sleep times will be reported for each period. We will use a similar approach for testing the impact of the LOCK sleep intervention on our secondary outcomes and our supplementary sleep data. For our secondary outcomes from the MDS, we will use data from each resident’s last available MDS assessment pre-intervention and the first available MDS post intervention. Additional MDS data may be included to extend the pre- and post-implementation assessment windows, increase our ability to distinguish individual variability from intervention-related change, and provide greater statistical power for these analyses. The form of the models tested will depend on the completeness of MDS data available for each participant.
We will triangulate the quantitative and qualitative data using thematic analysis according to responses associated with variation in the program’s implementation and sustainability (103).
Implementation Assessment. For Phase 2, we will assess implementation success using CFIR (59, 60) according to the entire sample of 24 nursing homes. Our analysis will (1) characterize nursing home variation in implementation measures based on selected constructs within 4 CFIR domains (61) and (2) use multiple data sources to document and evaluate the implementation sustainment (Table 3: CFIR Implementation Constructs and Data Sources). In addition, we will use methods outlined by Keith et al. to create a measure of variability (111) using both qualitative and quantitative data. This measurement will reflect how the intervention was used in practice, how it was sustained, and identify influences for our results. The analysis methods will apply to both the post-intervention and the sustainment analyses. We will triangulate perspectives on the CFIR constructs, to the extent possible, using the different data sources. For the qualitative data, triangulation requires first using the taxonomy of themes to assign quantitative ratings and averaging them to create a ranking. For the quantitative measures, we will empirically test for correlation and create a rating by weighting the correlated measures and creating a summary ranking.
Summary rankings will enable us to describe the variability in implementation among the trial’s nursing homes, both at the overall nursing home level and at the individual CFIR construct level. We will average the rankings across all constructs for each nursing home and then rank all nursing homes based on the average scores. This ranking process, crossed with staff impressions of the Fitbits, and rich qualitative data to provide context, will provide an overall summary of variability across nursing homes (112-114). As a final step, we will examine how this study’s implementation metrics align with quantitative results for our primary, secondary, and supplementary sleep data outcomes, helping us understand how the nursing homes implemented the LOCK sleep intervention and how the program was sustained. This will inform expansion of the intervention as well as inform other nursing home QI initiatives.
