Improving resident outcomes in NHs remains a national priority. While effective practices are known from efficacy trials, there is a lack of knowledge about how NH staff can implement these practices 234513. Centers for Medicare and Medicaid Services contracts with QI organizations to implement QI programs, including QI collaboratives, educational programs, and toolkits to reduce geriatric syndromes such as falls, pressure ulcers, incontinence, pain, delirium, and depression 394041. Unfortunately, such efforts have not resulted in the expected improvements 67942. Complexity science suggests that a major barrier to the effectiveness of QI programs is their content focus; they do not impact the processes needed to actually implement practice change. The CONNECT for Quality study is significant because it will test a novel intervention that attempts to create the foundation needed for nursing home staff to implement content learned in QI programs such as FALLS. Thus, CONNECT has the potential to have a broad and far-reaching impact on QI efforts nationally and influence care for multiple geriatric syndromes.

Of further significance, this study uses existing staff to improve resident care, without requiring additional resources. CONNECT, which targets local interactions among staff, strengthens the interdependencies among staff and also addresses other common barriers to interdisciplinary problem solving, such as omitting Licensed Practical Nurses and Nurse Aids (NAs) from decision making 242543, poor communication between provider groups 23, and overreliance on hierarchical management 2430444546. CONNECT, if successful, thus has the potential to be generalizable to real-world nursing home settings by enhancing existing staff capacity to learn and improve.

Further, this study is significant because it puts the tools of change into the hands of direct- care staff. CONNECT will establish networks for new information about fall risk factor reduction to spread throughout the nursing home. These networks are critical because NAs provide 80-90% of the hands-on care to residents 47, and they often are the first to observe early signs of fall risk 48. Yet, NAs frequently lack the interactions with the multi-disciplinary team needed to intervene effectively 2549. CONNECT will create opportunities for more rapid information exchange and problem solving among multiple disciplines and will increase the likelihood that the NA will carry out appropriate fall prevention care.

In our prior work 161718192324252728305051, we found that both managers and staff can use MPs to influence self-organization and produce better quality of care. In CONNECT, staff will learn to consider three system parameters 22 derived from the theory of complex adaptive systems 22 to guide their use of nursing MPs. Relationship-oriented MPs in CONNECT, which are collectively called local interaction strategies, influence the three system parameters--connection between staff members, information exchange, and cognitive diversity. When staff use the MPs, they create and recreate meaning of events, change beliefs, foster creativity, and promote reflection on their performance 2122. For example, when staff members interact, they develop networks 52. These new networks of connections allow local changes in behavior to result in system-wide change. When staff members interact they exchange information, which generates new understanding and knowledge 5354. With this knowledge, staff members learn, change behaviors 54, and become capable of accomplishing something new. Finally, cognitive diversity, the use of multiple perspectives to make sense of information 5054, arises from interaction among people. The more diverse the individuals (e.g., varying roles, education, social or cultural backgrounds, age cohorts 52, and external collaborations 53) the richer the interpretation of data, the more appropriate the decision making, and the more effective the action planning 5051.

We propose that systematic use of these local interaction strategies to create relationship networks and channels of communication for learning together, exchanging information, and solving problems, is a prerequisite to the ability to effectively implement a fall reduction program. Based on complexity science theory, if we achieve expected changes in staff interactions, we will observe changes in measures of communication, participation in decision-making, relational coordination, psychological safety, and safety culture (Figure 1). These measures in turn are expected to be related to more effective fall risk factor reduction measures.

CONNECT is expected to work in combination with QI programs because CONNECT creates processes for group learning and implementation of evidence-based content introduced by the QI program. FALLS will include content on evidence-based practices found to reduce falls in efficacy trials 234. Modifiable fall risk factors, suggested by clinical practice guidelines and Agency for Healthcare Research and Quality's fall management program are: orthostatic hypotension 5556; sensory impairment 5758; footwear; gait and assistive devices 4567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859; toileting needs 60; environmental problems 61; fall-related medications 6263; and vitamin D. 64656667 CONNECT is an important companion for QI interventions such as FALLS because it creates relationship networks and communication channels for learning, information exchange, and problem solving.

A major challenge for nursing home research is the potential for staff turnover. Using a successful strategy from our prior studies, we will secure a written commitment from the nursing home administrator, director of nursing, and if relevant, a corporate representative, that the study will continue even if one or more top administrators leave. We also have designed this study to be robust to staff turnover by incorporating the CONNECT in-class learning sessions into the nursing home's orientation for new staff. Exploratory analyses will determine whether staff turnover affects the fall-related processes or fall rate measures. Another challenge for nursing home research is designing approaches that are appropriate and acceptable for all levels of staff, regardless of education and socio-economic background. We use storytelling, which is an efficient yet high-impact method of conveying information, infused with relevant nursing home cultural norms, values, and beliefs 72. Because storytelling and role play are based on descriptive verse, they may attenuate learning barriers associated with low literacy and English as a second language 73.

CONNECT will be implemented over 12 weeks, followed by FALLS for an additional 12 weeks. The FALLS intervention is a modification of interventions previously tested by the PI 769 and is based on the Falls Management Program developed by the Agency for Healthcare Research and Quality 5974. Details of the intervention components, rationale, participants, and time required for CONNECT and FALLS are in Tables 1 75 and 2 75 respectively. All aspects of the interventions are applied to cluster level; even aspects that are delivered to individuals, address cluster level interactions. The complete text of the CONNECT and FALLS protocols are available on request to the authors.

Our treatment fidelity protocols use the National Institutes of Health Behavior Change Consortium's 76 model of treatment fidelity.

To ensure design fidelity, we standardized the CONNECT and FALLS protocols to a specified dose in terms of number, frequency, and length of contact.

CONNECT and FALLS will be delivered by different research interventionists trained separately to minimize contamination. The protocol specifies training content, structured practice, and role-play exercises to ensure that interventionists' skills meet established standards.

To ensure that CONNECT and FALLS are delivered as intended, a research team member will observe the interventionists on a random schedule, completing standardized checklists. The interventionists and PIs will discuss the results and problem-solve barriers to adherence and repeat concepts and role-play as needed. We will track participants that complete study components. For CONNECT, we will use: contact summary sheets for each visit to a research site; databases for interventionists to record contacts with participants; and sign-in sheets to document participation in sessions. For FALLS, we will use: contact sheets to record each contact between interventionists and the Fall Team; sign-in sheets to document participation in post-fall problem-solving sessions; and databases to track completion of educational modules via requests for continuing education credit or certificate of completion.

For CONNECT, participants' self-monitoring of local interactions will provide a measure of adherence and behavior change. The class sessions will include discussion and practice during which skills can be systematically assessed. For FALLS, participants will complete post-tests in the educational modules.

Researchers will systematically assess enactment when they shadow the in-house facilitators to observe how they practice mentoring behaviors. The researchers will also assess and record enactment by participants during structured mentoring. Finally, they will assess enactment by observing at least two orientation sessions in which the in-house facilitator delivers the in-class session to new employees. Fall risk reduction indictors will be used to measure enactment of the FALLS intervention.

Because of the nature of the intervention, it is not possible to mask the study assignment from the subjects or the research interventionists. However, the outcomes assessment of falls quality indicators will be completed by independent nurses employed by the state Quality Improvement Organization who will be blinded to study assignment. Success of blinding will be evaluated by asking these nurses to state which intervention they believe that the nursing home received.

The study hypotheses pertain to the cluster level of the nursing homes. Hierarchical linear modeling with Glimmix was used to account for clustering in this study. This procedure is useful when there are multi-level, nested sources of variability such as patients and staff clustering within nursing homes. The Glimmix procedure analyzes both individual and group level trajectories of change over time. Our hypotheses (H) to address the study aims and related analysis are listed below.

We used algorithms developed to estimate power for longitudinal models based on the formulae of Jung and Ahn 96, a type I error rate of 0.05 (two-tailed), and a 15% rate of attrition for the staff samples. Hierarchical linear modeling with SAS PROC Glimmix was used to account for clustering in this study. This procedure is useful when there are multi-level, nested sources of variability such as patient and staff clustering within nursing homes. The Glimmix procedure analyzes both individual and group level trajectories of change over time, and can be used to estimate models where persons within clusters are changing over time. Maximum cluster size was limited by the pool of resident fallers and number of staff in each facility. The power analysis algorithms used to determine cluster size take clustering at the individual level into account. In the analyses, we will treat cluster as a fixed rather than as a random effect. This approach will more adequately control on potential confounders at the level of the nursing home. With cluster analyzed as a fixed effect, inter-cluster correlations are not needed to calculate power.

For aim one, we will have 80% power to detect a 15% difference in risk factor assessment and intervention scores, which is considered to be the minimally clinically significant improvement in falls care practice. For aim two, the resident sample will provide 80% power to detect a 23% difference in the fall rate due to intervention, and a 23% difference in the probability of a recurrent fall. Because this is a real world effectiveness study, this change in fall rate is slightly smaller than that seen in a randomized controlled trial of multifactorial risk factor reduction, but still clinically meaningful. For the continuous outcomes in aim three, we will have 80% power to detect standardized differences of 0.21, a magnitude considered small in the statistical literature 97. As we have a single primary outcome and several additional outcomes that are exploratory, we will not adjust our significance tests for multiple tests.

Item-specific missing data on potential covariates will be handled with maximum-likelihood and multiple imputation techniques 98. Missing values can be imputed with SAS PROC MI using the Markov chain Monte Carlo algorithm, which can be used with complex missing data patterns as well as for continuous, ordinal, and dichotomous measures 99. We expect 15% attrition on our dependent variables across waves. Recent work by Chang et al. 100 shows how shared parameter models 101102 can be used to address potential bias due to a failure to meet the missing at random assumption. These models will be operationalized where necessary.

We were funded (56NR003178) to conduct a pilot study of the CONNECT + FALLS intervention in two nursing homes and FALLS in two nursing homes 31. We have now been funded (R01NR003178) to conduct this study in 16 additional nursing homes to allow for a full test of the intervention. We will begin nursing home recruitment in fall of 2011 and will stagger recruitment and interventions over four years (completed in 2015). Follow-up data collection and analyses will be completed in 2016.