SCOPE was a pragmatic controlled trial with each randomly selected LTC home identifying a care unit to participate in the intervention. Care aides and residents of these units formed the units of analysis. Control (usual care) units in non-intervention LTC homes were matched to intervention units post hoc (see below).
Setting
This study was part of a larger research program examining modifiable contextual factors that influence implementation and improvement efforts in LTC homes in British Columbia, Alberta, and Manitoba: the Translating Research in Elder Care (TREC) research program. TREC’s aim is to improve the quality of care and quality of life for LTC home residents, and the quality of work life for the staff who care for them (66). It operates at the level of the clinical microsystem (resident care units) where quality is created (67, 68). The overall TREC cohort includes 94 urban homes and was created using a stratified (owner-operator model, size, region) random sample (66). SCOPE homes were selected using a stratified random sample (66) of TREC homes in Alberta and British Columbia.
Outcomes and Measures
The primary outcome measure for this study aimed at improving use of best practices for resident care was care aide-reported conceptual use of best practices, defined as the cognitive, reflective use of research (best practices) where the knowledge may change one’s opinion or mindset about a specific practice area but not necessarily one’s direct actions. This scale asks about how often on a typical workday best practice knowledge helped with conceptual thinking about resident care, for example by making sense of things related to resident care. It is an indirect application of research (69-72) measured using the five-item CRU scale. The CRU scale has demonstrated acceptability, reliability, and validity (73-75).
Secondary outcomes
Secondary outcomes included validated measures of care aide-reported outcomes on work engagement, job satisfaction and burnout; and resident outcomes on clinical indicators for pain, mobility and responsive behaviours collected as part of the Resident Assessment Instrument – Minimum Data Set (version 2.0)(RAI-MDS)(76). All outcomes were collected at baseline and at the end of the SCOPE trial. Measures of implementation fidelity, measured at the SCOPE team level in a concurrent process evaluation [currently under review with this journal as a companion paper to this manuscript], were also collected. Only fidelity enactment data are incorporated into the main trial analysis reported here. Full details are shown in Table 1.
Table 1
Care aide outcomes: Collected at baseline and end of study, as part of the TREC care aide survey | burnout, using the Maslach Burnout Inventory (83), for which adequate reliability (84) and validity are established (84–86); job satisfaction, (87, 88) work engagement (89); psychological empowerment, (90); organizational citizenship behaviours directed at the organization (91). | The work engagement, psychological empowerment, and organizational citizenship behaviour measures were adapted and validated for use with the care aide population (92). |
resident outcomes: all obtained from quarterly RAI-MDS 2.0 reports (93) included | (1) physical functioning (Activities of Living – Hierarchy [ADL-H] scale score (94) (2) responsive behaviours (Aggressive Behaviour Scale [ABS] score of 2+) (95) (3) a pain score based on observable indicators of pain, which was developed and validated by TREC researchers to overcome the issue of under-detection of pain in residents with dementia (96, 97). | |
implementation fidelity: measured at the level of the SCOPE team | fidelity enactment (intervention participants actual performance of intervention skills/implementation of the core intervention components in the intended situation) was assessed on a four-point scale by a panel of investigators at the final celebration learning congress. The panel rated each team’s actual implementation of SCOPE activities: defining aims, generating change ideas, using PDSA cycles and measurement to test changes, modifying unsuccessful changes, spreading successful changes to residents and staff across the unit. The single-item global fidelity enactment measure was developed in a previous study (98) and adapted for use in the SCOPE pilot study (44, 99). |
Sample size and power calculation
The primary outcome measure was change in Conceptual Research Use (CRU), from baseline to post intervention, compared between intervention and control (usual care) units. Initial modelling was based on unit aggregate expected change in the primary outcome, dictating a sample size of 34 units to be matched to usual care units, but was replaced by a care aide level analytical model, deviating from the original published trial protocol (NCT03426072). Thus, for an effect size of d=0.22 (based on a mean difference of 0.11 in the CRU score between the intervention and control group at follow up and a standard deviation of 0.5 in both groups, informed by our pilot data (40)), the required sample size was 652 CA surveys, n=326 in each study group (based on a two-tailed test for independent study groups, at 80% power, with an alpha of 0.05). Taking into account possible clustering effects, we multiplied this required sample size by a variance inflation factor (VIF=1+[cluster size – 1]*intra-cluster correlation). Based on previous TREC data, we assumed a cluster size of 15 care aide surveys per unit and an intracluster coefficient of 0.01. Therefore, our required sample size was 652*1.14=744 care aides (n=372 per study group) or 50 care units (25 in each study group, each providing an average number of 15 care aide surveys).
Sampling
To be eligible to participate, LTC homes had to, a) be a part of the TREC cohort in Alberta and British Columbia, b) have units comprising general nursing care for older adults, rather than those co-managed with acute care, c) have the majority of residents over the age of 65, d) have more than 35 beds in total, e) be geographically located within 100km of either Edmonton: Edmonton Health Zone (EH) or Calgary: Calgary Health Zone (CH) in Alberta (AB), or Kelowna: Interior Health (IH), or New Westminster: Fraser Health (FH) in British Columbia (BC), f) use the Resident Assessment Instrument- Minimum Data Set 2-0 (RAI-MDS) to gather resident level care indicators and, g) have 8 or more care aide responses to the baseline trial data collection survey.
Eligible LTC homes were stratified by region (EH, CH, IH, FH), owner operator model (for profit, not for profit), and size (small: <80 beds, medium: eight-120 beds, large: >120 beds), and randomly selected for participation. Based upon feedback from decision-makers and LTC home administrators, it was decided that randomization to intervention or control at the outset would not be feasible because of the likelihood of bias favouring refusal to participate as an inactive ’control’ site. Thus, random selection was undertaken only for intervention sites, with replacement for refusals. Once the number of LTC homes within the same stratum was exhausted, a replacement home was randomly selected from the remaining homes in that region.
Because of the limited number of eligible LTC homes in the cohort, homes which declined to participate were returned to the main TREC cohort to act as usual care (control) comparators. After removing ineligible units (those who did not participate in both the baseline and follow up data collections and those with fewer than eight care aide responses to the TREC care aide survey) to ensure stability of measures at either the baseline or follow up data collections, we randomly matched a control unit to each intervention unit, based on the unit type: general long-term care versus dementia care unit, number of beds on unit, facility size category (small: <80 beds, medium: eight-120 beds, large: >120 beds), ownership model (for-profit, not-for-profit), and region.
Directors of care in charge of each selected home were invited to participate, provided with information about the study, and included if they consented to participate. Participating homes were provided with $3000 as partial compensation for the time and resources required to participate. Following discussion of trial requirements, Directors of Care were given the task of identifying one care unit within their home to participate in the intervention and to identify staff as members of their SCOPE team.
Participants
SCOPE teams were composed of four to seven members, at least two of which were care aides. Each team was either led by a care aide or co-led by two care aides. Other team members consisted of unit-based care aides and/or professional staff (e.g., registered nurse, physiotherapist, occupational therapist, recreation therapist). A team sponsor (usually a unit-level clinical nurse manager) was responsible for supporting day-to-day project activities. A senior sponsor, normally at the facility Director of Care or the care manager level in large units, agreed to actively support each team, removing barriers to change, and supporting time spent on quality improvement.
Analysis
SAS® 9.4 (SAS Institute Inc., Cary, NC, USA) was used for all statistical analyses. Using descriptive statistics, baseline characteristics of LTC homes, care units, care aides and residents were compared between study arms. To assess intervention effectiveness, mixed effects regression models were used (77, 78). All models were adjusted for sampling strata, baseline differences of the outcome variables, care aide characteristics (sex, age, English as first language [yes/no]), and care unit staffing (total care hours per resident day and percentage of total hours per resident day provided by care aides). A unit-level random intercept was added to account for dependencies of responses provided by care aides on the same care unit. Similar models were used to assess the impact of the intervention on resident outcomes but adjusted for resident characteristics (age, sex, case mix index). Finally, to assess whether improvements in outcome scores were higher in intervention facilities with higher (above median) enactment scores, mixed effects regression models were used, adjusted for the same variables as above, and included an interaction term between the dichotomous enactment variable (high/low) and data collection (baseline/follow up).