Study protocol for the translating research in elder care (TREC): building context – an organizational monitoring program in long-term care project (project one)

Rogers' representation of classical Diffusion of Innovations theory [6] is the dominant and most consistently used theory in this field. In it, Rogers describes the spread of new ideas using four main elements: the innovation, time, communication channels, and a social system. In addition to theories, a range of models addressing more focused areas of knowledge translation are also available [4, 7] (Table 1). A recent framework with similarities to Rogers' Diffusion of Innovations Theory, is the Promoting Action on Research Implementation in Health Services (PARiHS) framework [8]. Its authors argue that successful research implementation (a specialized form of knowledge translation) is a function of the interplay between evidence, context, and facilitation. They hypothesize that it is when each of these three elements is high that successful research implementation is most likely to occur [9‐11].

Rogers [6] argued that the adoption of an innovation (or research) is influenced by the interaction among three key components: the innovation, the adopter, and the environment. Investigators studying nursing services delivery have used this theory widely to frame studies of research use [12‐20]. Little work has been done on characteristics of the innovation in healthcare [21]. Until recently, research has focused largely on changing individual (the adopter) behaviour. For example, in studying physician behaviour, investigators have focused on interventions, such as academic detailing [22], educational influentials [23‐25], reminder systems [22, 26], and audit and feedback [27, 28]. While these interventions result in modest to moderate improvements in patient care, generalizability remains uncertain because of a limited understanding of the contextual, individual, and organizational factors that may influence the effectiveness of the different interventions [25, 29].

In the study of nurse (adopter) behaviour, the focus has largely been on examining individual determinants of research use, such as attitude [30‐32], age [31, 33], education [17, 33‐36], experience [31, 33], clinical area [17, 30], journals read [19, 37, 38], employment status [33], and most recently, critical thinking behaviour [39]. Less attention has been given to interventions, such as opinion leaders [34] or multidisciplinary teams [40]. In a systematic review by Estabrooks et al. [41], the most frequently studied individual determinant, and the only one with a consistently positive effect, was attitude towards research. Findings for other individual determinants were highly equivocal and most studies were characterized by serious design and methodological flaws. Further, investigators have not selected individual factors for study with the important requirement that the factor be potentially modifiable.

Numerous organizational (environmental) factors thought to influence innovation adoption in industry and health services have also been studied. Those shown to have an influence include organizational complexity [42‐46], centralization [47], size (e.g., number of beds) [20, 42, 44, 48, 49] presence of a research champion [50‐52], traditionalism [53, 54], organizational slack [42, 55], access to and amount of resources [56], constraints on time [12, 57‐67], professional autonomy [58, 68, 69] and organizational support [30, 31, 56, 68, 70, 71]. Again, investigators have generally not selected factors for study with a requirement for potential modifiability.

While there is generally a growing awareness and acceptance among researchers of the importance of organizational context (the local environment) to successful knowledge translation, and successful knowledge translation to improved patient, provider (staff), and system outcomes, astonishingly little empirical evidence supports these assumptions. Further, we know little about knowledge translation in the long-term care environment – an environment where: the quality of care is suboptimal [72] and the model of care is a nursing services delivery model where the majority of caregivers provide some level of nursing services.

In this project, we aim to investigate the impact of organizational context (giving specific attention to those factors which may be potentially modifiable) on knowledge translation and the effect of both organizational context and knowledge translation on resident, provider (staff), and system outcomes using long-term care as a naturally occurring laboratory.

We are using an extension of the PARiHS framework to frame this research project. In the PARiHS framework, the continuous interaction between context, evidence, and facilitation is hypothesized to lead to increased research implementation. This project is particularly focused on increasing understanding of the role of one of these elements, context, on promoting knowledge translation and improving outcomes. We define context as "...the environment or setting in which people receive healthcare services, or in the context of getting research evidence into practice, the environment or setting in which the proposed change is to be implemented." [[73], p. 176]. Context according to PARiHS consists of three core dimensions: culture, leadership, and evaluation. In this project, however, we take an expanded view of context to include additional modifiable elements of the work setting, such as interactions (formal and informal), social capital, resources, and organizational slack.

The purpose of this project is to establish the magnitude of the effect of organizational context on knowledge translation, and of organizational context and knowledge translation on resident, provider (staff), and system outcomes. The primary objectives of the project are:

Interviewer training for individuals conducting CAPI with healthcare aides has been undertaken to ensure standardized interviewer technique and the collection of high-quality data. Interviewer and quality control manuals have been created to facilitate data quality processes for these interviews. The interviewer manual describes the step-by-step process of conducting a CAPI interview, and the process by which the data are handled. The quality control manual outlines the characteristics of a successful interviewer and the training and process that must be undertaken before someone is deemed to be prepared to begin interviewing. Quarterly and yearly quality control and improvement processes are in place.

Data analysis is an ongoing iterative process. Data are cleaned and processed for analyses at the close of each quarter. Real-time descriptive analyses are completed more frequently to assess response rates and to ensure that interviewer variation is within expected limits. As the data set is assembled, we are performing ongoing descriptive analyses to: check for outliers and systematic biases, monitor response rates, and inform variable selection for modeling. These analyses are also being used to inform TREC project two data collection [2]. In addition, we are computing response rates and distributions (means, medians, standard deviations) for the knowledge translation measures and all of the constructs assessed in the TREC survey by provider group, unit, facility, region, and province.

After reviewing the descriptive analyses for the total data set we will undertake analysis of variance (ANOVA) and multiple comparison tests as sample size permits in order to investigate differences in knowledge translation behaviours among staff groups (i.e., healthcare aides, nurses, physicians, allied health, practice specialists, care managers) and between units, facilities, regions, and provinces. We will use similar methods to describe and assess differences in resident (e.g., falls) and provider (e.g., health status, burnout) outcomes. Additionally, differences among staff groups, units, facilities, regions, and provinces on all independent variables (e.g., ACT dimensions) will be examined with similar descriptive and ANOVA methods.

The majority of our analytical work will consist of a series of regression models, then multi-level and structural equation models, and finally, if our data permits, hierarchical structural equation models. We will estimate the knowledge translation dependent variables at the individual provider (staff) level. Staff characteristics, individual, and context variables will be the primary explanatory variables in these equations. We will then use the predictions of knowledge translation variables as independent variables in additional equations to estimate staff and resident outcomes, with the individual staff member/resident as the unit of analysis. Resident characteristics, staff characteristics, and predictions of knowledge translation variables aggregated to the unit or facility level will be the primary explanatory variables in these equations. We will perform multi-level analysis using organizational data (aggregated) at the unit level with subjects nested within each unit. We have three levels of organizational data in the survey- facility (level three), unit (level two), and individual (level one). In further analyses, we will use these data in structural equation models to explore the relationships among different outcomes and context variables, including latent variables. This may be of particular importance in analyzing the knowledge translation variables, which have qualities that are difficult to observe directly.

As a value-added function for the participating long-term care facilities we will provide them with annual facility reports approximately six weeks after we have the second wave of data collection (so that we can provide wave 1 and wave 2 comparisons). Our decision makers have informed us that because many of the long-term care facilities within their regions have limited internal data analysis capability, periodic private reports on their own data would be of value. These reports will be at the facility level and may include some de-identified unit-level feedback, but will not allow for identification of specific residents, staff and/or units. The format of the report will be the same in all facilities and have been determined in a consultative process with the facilities in the first year of the main study. In addition to the agreed upon data elements requested by the facilities, a section of the report will also emphasize variances of note for the individual facility.