Introduction
Almost half of older people admitted to hospital for acute illnesses, injuries or surgery will experience hospital-associated complications of delirium, functional decline, new incontinence, falls or pressure injuries [
1]. These complications are more common in frailer people, often co-exist, and contribute to longer hospital stays, greater care needs, and higher mortality [
1-
4]. Consistent application of key age-friendly care principles including early regular mobility, adequate nutrition and hydration, and meaningful cognitive and social activities reduces delirium and may reduce other complications [
5-
7]. However, care practices supporting these key principles require cooperation and negotiation between healthcare staff from different disciplines and changes in practices and policies at individual, team and system level in response to multiple and varying local and organisational barriers [
8-
10]. Implementing age-friendly principles is thus a complex healthcare intervention [
11,
12]. Design and reporting of complex interventions benefit from use of programme theory to explain core components and mechanisms (how the intervention is hypothesised to work, often articulated in a logic model), and implementation theory to describe the relationship between the intervention and the setting(s) (how it is implemented) [
12-
14].
Eat Walk Engage is a ward-based programme developed to improve multidisciplinary team delivery of age-friendly care principles (adequate nutrition and hydration, early regular mobility, and meaningful cognitive and social activities) to reduce hospital-associated complications in older inpatients [
15]. It was developed, piloted and refined in a medical [
16] and surgical ward [
17] in a metropolitan teaching hospital. The logic model (Table
1) describes the hypothesised relationship between core components, activities, goals related to the principles, and clinical outcomes. Because facilitation was identified as a central component of pilot success [
16], refinement and implementation of the program has been underpinned by the integrated Promoting Action on Research Implementation in Health Services (i-PARIHS) framework [
18], which proposes that
facilitation is the central activating mechanism for getting
innovation (in this case, age-friendly care principles) into practice within the
context (acute care wards) by engaging with the
recipients (older patients and the multidisciplinary staff caring for them). The core components of Eat Walk Engage are a trained novice site facilitator, a local multidisciplinary work group, structured measures including older patient interviews and structured observations of care practices related to programme goals, and a trained multi-professional assistant. Informed by the i-PARIHS Facilitation Guide and facilitator’s toolkit [
19] and supported by an experienced external facilitator, the site facilitator assesses the local context, measures current practice, and helps the work group to collaborate to initiate iterative cycles of improvement targeting the programme goals, including delegation of appropriate tasks to the multi-professional assistant [
15]. Improvements may include changes at individual patient or provider level (e.g. ensuring a patient is wearing spectacles), team level (e.g. creating and using a cognitive kit of puzzles, games, craft resources and fiddle objects) or system level (e.g. procuring ward signage and calendar clocks to support orientation). Improvements are informed by older patients’ reported experience and suggestions, tailored to observed care practices and context and monitored by repeated measurement.
Table 1
Eat Walk Engage logic model (articulating the programme theory) with the corresponding measures reported in the implementation evaluation below, based on guidance for process evaluation of complex interventions in health care [
13]. Patient outcomes are reported in the primary evaluation of the CHERISH cluster randomized trial [
21]
Inputs (core components) | Intervention activities | Programme goals | Patient outcomes |
• Eat Walk Engage site facilitator trained and mentored by expert facilitator • Ward-based multidisciplinary work group • Structured interviews and care process measures • Trained Eat Walk Engage assistant | Under guidance of the site facilitator, the work group: • identifies local barriers and enablers • develops shared improvement goals pertaining to the key principles • clarifies team roles and identifies opportunities for improvement • initiates small cycle improvements with re-evaluation | Improved team prioritisation Higher proportion of older patients achieve: • Early and adequate nutrition and hydration • Early mobility and independence • Meaningful cognitive and social engagement | Reduced hospital-associated complications leading to: • Reduced length of stay • Reduced facility discharge • Reduced 6-month readmission and mortality |
Implementation | Context and implementation | Mechanisms of impact | |
• Staff recruitment and training • Work group meeting frequency and attendance • Completion of interviews and audits | • Context mapping • Implementation strategies • Improvements | • Patient interviews • Behavioural mapping • Mealtime audits | |
The Collaborative for Hospitalized Elders: Reducing the Impact of Stays in Hospital (CHERISH) trial [
20] was a hybrid implementation-effectiveness trial evaluating Eat Walk Engage in four Queensland hospitals from January 2016 to May 2017. Effectiveness was evaluated using a cluster randomised trial design between October 2016 and April 2017, and demonstrated a significant reduction in hospital-associated delirium [
21], consistent with other trials of multicomponent non-pharmacological interventions [
5]. There were no significant changes in functional decline, incontinence, falls or pressure injuries, and there were promising but inconclusive findings for clinical outcomes (length of stay, facility discharge, and 6 month death and readmission) [
21].
The aim of this implementation evaluation informed by the UK Medical Research Council guidance for evaluating complex interventions in healthcare [
12,
13] and our logic model (Table
1) was to understand how Eat Walk Engage worked by
-
describing context, implementation and improvements in each site;
-
describing and analysing changes in process measures that are postulated mechanisms of impact leading to the observed patient outcomes;
-
describing how facilitation supported tailored implementation as hypothesised by the underpinning i-PARIHS implementation framework; and
-
proposing how key contextual features of different sites may have contributed to implementation success or failure [
20]. This will help to build a more advanced understanding of how and why the programme works to inform spread and scale-up.
Methods
Setting
This study was initiated by hospital-based clinician researchers through a collaborative partnership with the Queensland Government, Queensland University of Technology, geriatric and health services research academics, and two Hospital and Health Services (HHS) in Queensland, Australia. The Queensland government funds universal free hospital care through public hospitals organised into 16 HHS. Senior executive leaders from one metropolitan and two regional HHS were invited as industry funding partners. One metropolitan and one regional HHS supported participation of four hospitals (two inner metropolitan, one outer metropolitan, one regional). Each hospital nominated two acute care wards where more than half of inpatients were aged over 65 years, the program was aligned with hospital priorities, and there was nurse unit manager agreement. One ward at each hospital was randomised to implement Eat Walk Engage, providing four intervention wards. Implementation commenced in January 2016.
Participants
Participants for the implementation evaluation included inpatients on the four intervention wards before (October -December 2015) and after (March–May 2017) implementation, and multidisciplinary work group members and facilitators.
Implementing Eat Walk Engage
Implementation drew on facilitation as experiential learning, led by two experienced facilitators who recruited, trained and mentored novice site facilitators using concepts and tools outlined in the i-PARIHS Facilitation Guide [
19,
22]. The experienced facilitators were senior clinician researchers who had designed, implemented and evaluated the pilot Eat Walk Engage programme. They were supported by experts in implementation science and practice not involved in delivery of the programme, including an expert consumer advocate. They worked with local clinical leaders at each site to recruit site facilitators, who were mid-career nurses or allied health professionals selected through a competitive recruitment process focussing on skills and attributes of facilitation (e.g. self-awareness, communication, interpersonal and assessment skills) [
19,
23,
24]. The experienced facilitators provided four half-day initial group training sessions for site facilitators (October to December 2015) including didactic and interactive content based on the i-PARIHS Facilitation Guide [
19], evidence for age-friendly care principles and the prevention and management of hospital-associated complications, and provision of key readings. Mentoring included monthly half-day face-to-face peer group meetings (January 2016 to May 2017), with telephone and email support available between meetings, supporting debriefing, reflection on practice and shared learning [
25]. The experienced facilitators visited each site 3–6 times before and during implementation, meeting key stakeholders and participating in local work group meetings. Project funding supported 24 h per week experienced facilitator time to support project management and external facilitation across all sites, and 16 h per week site facilitator plus 20 h per week multi-professional assistant for each implementation ward.
Site facilitators engaged with the nurse unit manager and key staff on each intervention ward to form a work group focussed on improving care of older people. The work group included champions, opinion leaders and/or key roles for improving care of older people (e.g. nursing staff, nurse educator or clinical facilitator, physiotherapist, dietitian, occupational therapist, medical staff, patient support staff). Site facilitators scheduled monthly meetings and helped work group members reflect on evidence, tacit knowledge, older patient interviews and structured observations of care practices (see below). They supported the work group to identify and prioritise areas for improvement, clarify roles and relationships, and trial cycles of improvement aligned with programme goals, agreed team priorities and available resources.
Multi-professional assistants were allied health or nursing assistants selected by interview who received two weeks of training in care of older people, including a work instruction manual, face-to-face training with local allied health professionals (e.g. physiotherapist, speech pathologist) and work shadowing with an experienced assistant. Any clinician could delegate tasks aligned with programme goals (e.g. setting patients up for a meal, supporting supervised mobility or exercise activities, providing assisted listening devices). Assistants could also assist with ward-level tasks related to environment and resources (e.g. updating orientation boards).
Evaluating the implementation
The implementation evaluation was a prospective, multi-site evaluation using a theory-based research perspective to describe the context, implementation and mechanisms at each site [
12]. The evaluation was guided by the UK Medical Research Council guidance for process evaluation of complex interventions [
13]. The intervention and causal assumptions were articulated in our logic model (Table
1). The i-PARIHS framework [
18] informed implementation (with a central emphasis on facilitation) and context assessment. Pre-intervention measures were collected by the site facilitators in October-December 2015 before implementation commenced and presented to the local work group to prioritise and inform improvements. Post-intervention measures were repeated in March–May 2017 by trained staff not involved in implementation on that ward. Pre-intervention context and process measures (patient interviews and structured observations of care practices, described in detail below) informed and guided local implementation and improvements, and also provided a baseline for comparison with the post-intervention process measures.
Context was assessed by each site facilitator using a spreadsheet tool based on the i-PARIHS facilitation checklist [
19], focussing on determinants of implementation at the level of intervention, recipients and inner (ward) context. Each site facilitator synthesised their observations from patient interviews and structured observations (described below), group and individual discussions with local staff members and informal observations of ward practices and culture, to identify and rate barriers and enablers, supported by reflective discussions with the experienced facilitators and peer site facilitators. Contextual determinants were scored from -2 (major barrier) to + 2 (major enabler) with brief explanatory comments [
22]. Scores and comments were mapped and reviewed pre-implementation, mid-implementation and after implementation, and compiled into a summary table by the experienced facilitators.
Implementation described adherence to the core intervention components (including staff recruitment and training, work group attendance and completion of interviews and structured observations of care practices); the
implementation strategies used by facilitators to enable change; and the resulting individual, team and system-level
improvements addressing programme goals. Data were obtained from minutes and field notes maintained by the experienced facilitators during regular group face-to-face meetings with the site facilitators and electronic minutes maintained by site facilitators during their local work group meetings. The two experienced facilitators independently extracted and analysed information using directed content analysis [
26], immersing in the textual data and applying codes derived from theory and the pre-defined program logic. Implementation strategies used by facilitators were coded deductively using the Expert Recommendations for Implementing Change (ERIC) taxonomy [
27]. Improvements were coded based on the programme goals (‘Eat’, ‘Walk’, ‘Engage’ and team communication, Table
1) and summarised for each ward, including the improvement, team member(s) responsible, the barrier or enabler being addressed and the level of intervention (individual, team or system). The consensus summaries were discussed within the broader research team which included two site facilitators and four external researchers.
Mechanisms of impact were evaluated by process measures before and after the intervention was implemented, including older patient interviews and structured observations of care practices aligned with the programme logic model (Table
1). Site facilitators received group and individual training from the experienced facilitators including supervised practice of all care practice measures. Semi-structured patient interviews were conducted with a purposeful sample of inpatients likely to benefit from this ward-based intervention [
20], aiming to sample approximately ten consenting inpatients aged 65 years or more with a length of stay of 3 days or more on each ward, excluding patients with critical illness, severe cognitive impairment or at end of life who have unique considerations in their care. Questions included perceived importance of each key principle (mobility, nutrition and meaningful engagement) during acute hospitalisation, using a 4-item response scale (very important, somewhat important, not really important, important but unable to do) and whether participants had received recommendations from their health care team about each principle, reflecting team communication. Responses to open-ended questions about barriers, enablers and suggested improvements were summarised for feedback to the work group at each site and have been reported in detail elsewhere [
8].
Cross-sectional structured mealtime observations were undertaken for all ward inpatients (excluding patients who were fasting, receiving enteral or parenteral nutrition, off the ward at mealtimes or receiving end of life care) at three mealtimes (breakfast, lunch and dinner) during a 1–2 week period. The observer noted whether each patient was sitting up when the meal arrived, had the tray table in reach, and was interrupted during the meal. They recorded whether patients required assistance with meal set-up or eating, and whether they received this assistance within 10 minutes. We evaluated the percentage of mealtimes where patients were sitting in a chair and received timely assistance, as these factors are associated with increased nutritional intake [
28].
Structured observations of patients’ physical, cognitive and social activities were conducted using activity mapping [
29,
30], systematically sampling patient activities on one weekday on each ward. All patients in a room (except those receiving end of life care) were observed for two minutes before moving to the next room in continuous sequential observations between 8am and 4 pm, sampling each patient every 20–30 min depending on ward size and room configuration. The highest level of activity during each observation period was recorded using a hierarchical tool encompassing location, physical activity, cognitive/social activity, and company. We averaged the percentage of observations at each level for patients with at least 4 h of observation (excluding patients who were discharged early, admitted late or off the ward for more than half the day) [
29,
30]. We evaluated the percentage of observations where the patient was standing or walking and the percentage where they were engaged in cognitive or social activities (e.g. talking to others, reading, watching television) aligned with programme goals. Increased physical and cognitive activities may reduce delirium and functional decline in hospital [
5,
7].
Analysis
Context measures were summarised for each ward and over time in a table to allow visual comparison. Implementation strategies were summarised in a table using the ERIC taxonomy [
27] and describing the level of facilitator responsible (experienced vs novice) for implementing each strategy, while improvements were summarised by site, focussing on the persons responsible for each strategy and the barrier being addressed to illustrate how approaches were tailored to local context. Care practice measures capturing the program goals (Table
1) were summarised descriptively pre- and post-implementation at ward level in tables and graphs to illustrate site-level changes, and averages across implementation wards were compared before and after implementing the intervention to test program-level process improvement changes and understand their potential impact on program-level outcomes as reported in the effectiveness evaluation. The researcher team triangulated contextual information, improvement strategy ‘dose’ and observed care practice changes within the case studies of each ward through drafting and discussion of findings to develop consensus propositions about the variable implementation success across sites. Measures to enhance trustworthiness of analysis included use of established well-defined coding frameworks for context assessment and implementation strategies and dual coding of implementation and improvement strategies (dependability); data collection triangulation, researcher triangulation and regular peer debriefing regarding methods and findings with a team which included members internal and external to the implementation team (credibility); and thick description of context (transferability) [
31].
Discussion
This paper describes implementation of Eat Walk Engage on medical and surgical wards in four hospitals. The process evaluation illustrates the complexity of characterising, implementing, and evaluating flexible multi-component health care interventions in varying contexts [
12]. Core components were clearly specified but implementation strategies and improvements were adapted through active facilitation, informed iteratively by older persons’ perceptions and suggestions, observations of existing care practices, and dynamic features of each site’s context, in keeping with emergence properties of complex systems, which can be difficult to capture in programme theory [
32]. The multiple levels of leadership and agency illustrated in this evaluation (including experienced and novice site facilitators, work group members, assistants, other staff and families) provide an example of how differentiating implementation and improvement strategies, or implementers and participants, may be somewhat artificial in complex systems [
13]. There was substantial variation in number and type of improvements between sites, and the site with lowest fidelity to the core components had fewer improvements and less care practice changes, a ‘dose effect’ which supports our logic model.
The evaluation demonstrated improvements in many care practices within 15 months of implementation. Older patients recalled more recommendations about the key principles, suggesting improved team communication. Structured observations of care practices demonstrated improvements in mealtime care and participation in social and cognitive activities, but not consistent improvements in mobility. Mobility-related improvement strategies were observed less consistently across sites than nutrition and engagement strategies, which might help explain this difference. The facilitators’ backgrounds (dietetics, occupational therapy and nursing) may have meant that they were more confident or comfortable supporting nutrition or engagement interventions, or the well-recognised multi-level barriers to in-hospital mobility [
10,
33] may have made changes in this domain more difficult to achieve. The observed process improvements are plausible mediators of the significant reduction in delirium demonstrated in the effectiveness trial [
21]. A recent systematic review of non-pharmacological interventions for delirium prevention highlighted the importance of re-orientation, cognitive stimulation, and attention to nutrition and hydration [
5]. In contrast, inconsistent mobility improvements could explain why hospital-associated disability and incontinence were not significantly reduced [
34,
35]. A longer intervention period or earlier prioritisation of mobility improvement strategies might increase the impact of this program on these outcomes.
The study supports the i-PARIHS implementation theory, illustrating how experienced facilitators supported novices to consciously adapt implementation strategies and improvements within varied and dynamic contexts. Facilitation is both a role and a set of activities [
36]. The study funded a dedicated site facilitator within each ward, whose knowledge and skills were supported by training, mentoring, and opportunities for peer reflection [
19,
22,
37]. Detailed description of implementation strategies and improvements makes the facilitation process visible [
38], illustrating the diverse activities required of a facilitator [
36,
37]. This case study of experiential learning adds to facilitation theory [
39] by describing distinct roles of the experienced and novice facilitators. The experienced facilitators worked as an external-internal facilitator and boundary spanner to build organisational capacity [
38,
40] while the site facilitator worked as a clinical practice facilitator to support the clinical teams and assistant to trial specific improvements at individual and team level. A previous theory-based implementation evaluation of a delirium prevention program highlighted the importance of a dedicated facilitator to support iterative practice change in a dynamic environment [
41]. However our analysis also describes the important role of the experienced facilitator, who not only trained and supported site facilitators [
23,
25], but also engaged in their own facilitation roles including understanding and engaging key stakeholders within the organisational and outer context (beyond the influence of the novice facilitator), actively supporting data collection and reporting, and enabling networking between sites and stakeholders. Our findings suggest these roles need to be recognised and resourced for successful scale and spread of this complex intervention [
41].
Our longitudinal observations also illustrate how context influenced and was influenced by implementation [
42]. An adverse and deteriorating context at site D was associated with challenges to adoption of the core components (i.e. fidelity), fewer team-led improvements and limited change in care practices. Site executive leaders had committed to the trial before relocating to a newly built facility, which occurred in early 2017 as the trial was finishing. However, this major competing priority created impacts on staff recruitment, leadership and staff morale which could not be mitigated by the facilitators or local staff, and hampered success, similar to experiences reported by other investigators [
41]. Sites B and C had adverse initial features in their inner organisational context, but team communication, culture and infrastructure became more positive as site facilitators built trust and teamwork, encouraged reflection on care practices, and supported iterative improvements. Combining local tacit knowledge with shared reflection on practice to create tension for change empowers multi-level, distributed leadership which may be more effective than traditional hierarchical leadership within complex healthcare systems [
43,
44]. Sustaining this complex program will require continuing skilled facilitation and regular reassessment of context, to allow dynamic adaptation to changes in personnel, resources, leadership and organisational priorities [
45]. Although our research design was pragmatic, implementing and sustaining the program outside of a research agreement may raise new challenges in engaging and sustaining organisational leadership and visibility.
Strengths of this study are a clear logic model and prospective use of an appropriate implementation theory. Implementation was guided by older patients’ experience and local context as well as published evidence, and ward-level observations provided meaningful measurement of programme goals both as an opportunity for improvement and a measure of progress. Rich evaluation using multiple data sources captured critical concepts of context, implementation and mechanisms; an accompanying in-depth qualitative evaluation of the facilitator’s journey will augment the evaluation further. We acknowledge potential limitations, including the challenges inherent in program developers and facilitators being involved as investigators [
13]. The evaluation group included the two experienced facilitators and two novice site facilitators, but also included senior investigators not involved in the implementation and a consumer representative, and collaborative implementation group meetings for review of analysis methods and emerging findings supported researcher triangulation to enhance credibility as well as interpersonal, methodological and contextual reflexivity to reduce potential bias [
46]. The facilitator-investigator role provides valuable understanding of the breadth of implementation strategies and improvements, which could otherwise remain invisible and lead to challenges with sustainability [
47]. Our context assessments focussed most closely on the inner organisational level because the ward is the level of the intervention and site facilitator influence; while we recognised the major impact of specific outer organisational factors in Site D, there may have been other under-recognised outer organisational factors such as leadership, culture and cosmopolitanism that influenced implementation in other sites. We did not formally test inter-rater reliability for the process measures, and there is not yet sufficient published data to establish meaningful change in these measures. Implementation strategies and improvements were obtained from content analysis of project documents and field notes, which may have led to incomplete recording of strategies and may reduce dependability. This could be mitigated in future studies by designing tools for prospective, real time tracking of how implementation strategies and improvements are tailored, which has been recognised as a missing enabler for operationalising i-PARIHS and other implementation frameworks in practice [
22].
Acknowledgements
We acknowledge the facilitators, multi-professional assistants and multidisciplinary staff on the wards who engaged in this new program, and the clinical and executive leads who supported the study. We thank the older people who participated in the program, providing their feedback and sharing their ideas for improvement. We also acknowledge CHERISH investigators Merrilyn Banks, Adrian Barnett, John Endacott, Nicholas Graves, Theresa Green, Ruth Hubbard, Sharon Inouye, Sue Kurrle, Wen Kwang Lim, Phillip Masel, Shaun Pandy who supported design and conduct of the CHERISH cluster trial.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.