Summary box

What is already known on this topic

  • Sustainability of knowledge translation (KT) interventions supporting implementation of chronic disease management (CDM) in older adults (> 65 years) with chronic diseases is vital to ensure long-term, high-quality patient care.

What this study adds

  • Few RCTs assessed sustainability, fidelity, and adherence of KT interventions for CDM practice for at least 1 year.

  • Sparce evidence assessing quality of life and care following sustained KT interventions present KT knowledge gaps and analytical challenges.

  • More studies providing an operational standardized measure of sustained KT interventions are necessary to explore patient outcome heterogeneity and robust conclusions regarding treatments and associated results.

Background

Evidence-based clinical interventions (i.e. early mobilisation in older adults or heart failure medications use) require tailored knowledge translation (KT or implementation) interventions (i.e. patient education or team changes) to optimise use in practice or policy. KT interventions are strategies that facilitate research uptake in practice and policy and include any action or set of actions that target factors that hinder or help someone to use a new practice or evidence-based program [1]. KT interventions are diverse and can focus on patients, caregivers, clinicians, managers and policy makers [2, 3]. Adoption of KT interventions can impact patient care and health system outcomes; however, there is a tendency to return to prior behaviours after initial interventions end [4]. Sustainability of KT interventions is defined as the continued delivery of clinical and KT intervention after its adoption is secured over a period of time (depending on the implementation context), while producing benefits for individuals and systems [5]. Failure to sustain KT interventions can lead to declining patient and health system outcomes and diminish confidence and support for future KT [6, 7].

Adults aged 65 years and older are the largest growing proportion of the global population, and many are affected by chronic diseases [8, 9]. Evidence-based clinical interventions to manage these conditions often include a combination of pharmacological and non-pharmacological interventions. However, to optimise intervention impact, their use needs to be supported at the patient, healthcare provider and health system levels via KT interventions [6]. Sustainability of KT interventions to manage chronic diseases is of paramount importance to ensure long-term, high-quality patient care and optimise health system impact consistently [10,11,12,13]. Specifically, optimal chronic disease management (CDM) in older adults requires sustained use of CDM interventions via effective KT interventions [14]. More importantly, it is expected that fostering sustainability will help reduce waste in health by facilitating their effective use. Our previous sco** review on the sustainability of KT interventions to manage CDM in adults included 62 experimental, quasi-experimental and observational studies assessing 13 different types of KT interventions [14]. Evidence showed that 56.1% of the eligible patients received a KT intervention for CDM, and even fewer maintained their use (e.g. 45.4% with diabetes mellitus, 24.7% with atrial fibrillation) over 2 years [15]. Moreover, it remains unclear which KT interventions and their individual components are most effective and sustained to optimise CDM.

The aim of this systematic review and meta-analysis was to describe sustainability of KT to implement a CDM intervention for at least 12 months by engaging 17 knowledge users, including patient partners, throughout using integrated KT. A knowledge user is defined as an individual who is likely to be able to use research results to inform their decisions about health policies, programs and practices (e.g. clinicians, managers, policy makers, patients/families and others) [16, 17]. We aimed to systematically assess the efficacy of sustainability of KT intervention for CDM end-users with comorbid conditions including older patients, their caregivers, health and policy-decision makers on healthcare outcomes (including quality of life [QOL] and quality of care [QOC]) at least 1 year after CDM intervention implementation or the termination of initial funding.

Methods

We registered our protocol with PROSPERO (CRD42018084810) and published it in an open-access journal [18]. Our systematic review follows the PRISMA 2020 [14, 18, 21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40]). Any deviations from the protocol are reported in Additional file 1: Appendix 2 [41, 42].

Knowledge user engagement

We enhanced systematic review conduct by employing an integrated KT approach [4] from project onset via established partnerships with 17 knowledge users, including one patient partner (KT), one funder (DAC), one policymaker (AE), 11 international KT researchers (BRH, IDG, JES, JM, JP, LRD, LS, PPG, RCB, WI, TVdW) and four clinicians (BH, FL, HTS, SES). The knowledge users provided input throughout the research process, including formulation of the research question, study protocol, prioritization of outcome measures and interpretation of results based on context relevance [18].

Eligibility criteria, search strategy and selection process

We included randomised controlled trials (RCTs) where the target population for the CDM intervention included patients (at least 65 years old with one or more chronic disease [22]) or their caregiver. End-users of the KT intervention to implement a CDM intervention for at least 12 months included patients aged 65 years and older with at least one chronic disease, their caregivers, clinicians (all disciplines), public health officials, health care managers and policy-makers. RCTs comparing a KT intervention versus other KT interventions or usual care were eligible.

KT interventions were classified using (1) a pre-existing taxonomy developed by the Cochrane Effective Practice and Organisation of Care (EPOC) group and (2) the behaviour change technique (BCT) taxonomy. The primary outcome was sustained implementation of a KT intervention for CDM beyond 1 year after implementation or termination of funding and which KT interventions were used (Additional file 1: Appendix 3 [14]). Secondary outcomes were health-related or disease-specific QOL and process or QOC (Additional file 1: Appendix 4).

We searched the bibliographic databases MEDLINE, EMBASE, and CENTRAL up to March 4, 2020, and developed a grey literature search strategy [21] to seek unpublished studies (Additional file 1: Appendix 5). Reviewers independently and in duplicate screened titles/abstracts in level one and similarly full-text articles in level two. Pairs of reviewers independently abstracted data from each included study. Two pairs of reviewers (ACT, CF, CS, SES) coded each KT intervention within the included studies independently using EPOC and BCT taxonomies [3, 14, 23] (Additional file 1: Appendices 5, 6 and 7 [43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59]).

Within and across study bias assessment

Pairs of reviewers appraised included studies using the EPOC risk of bias (ROB) tool independently [29]. We visually inspected small-study effects and reporting bias using the contour-enhanced funnel plot and Egger’s test when at least ten studies were available [30].

Synthesis

We performed a descriptive analysis for the primary outcome and sustainability of KT interventions and used frequencies and percentages for the a priori defined KT dimensions: sustainability, adherence and fidelity assessment.

We combined study-level data in a meta-analysis using the mean difference (MD) for continuous outcomes (i.e. QOL) and odds ratio (OR) for dichotomous outcomes (i.e. QOC) along with corresponding 95% confidence intervals (95% CI) when at least two studies were available. We performed both random-effects and fixed-effect meta-analysis models using the inverse-variance method. Under the random-effects model, we estimated the overall effect size and its 95% CI using the Hartung–Knapp–Sidik–Jonkman method to handle meta-analyses with few studies [39,40,41]. In line with recent recommendations, when the estimated heterogeneity was positive (> 0) and at least three studies were included in the meta-analysis, we prioritized the random-effects model, since we expected the studies to be methodologically and clinically different [27, 37]. When the estimated heterogeneity was zero, we prioritized the fixed-effect model since the Hartung–Knapp–Sidik–Jonkman method is considered inadequate [26, 33, 39,40,41, 60]. When two studies were included and the estimated heterogeneity was positive (> 0), we presented both fixed and random effects findings [37]. We calculated prediction intervals (PIs) for the overall effect under the random-effects model to capture the interval within which we expected the true intervention effect of a new study to fall. We used the restricted maximum likelihood method [29] to estimate the between-study variance τ2 and the Q-profile approach to calculate its 95% CI [32, 36]. We explored potential heterogeneity using predefined meta-regression, subgroup or sensitivity analyses.

Results

Study selection

Overall, 157 RCTs (973,074 participants overall [961,745 patients, 5540 caregivers and 5789 providers]) and 39 companion reports were included, after screening 15,361 citations and 3145 full-text articles (Fig. 1). Of the included studies, one was written in non-English language, that was in Chinese [61]. The 157 RCTs included 110 RCTs identified from literature search, 27 RCTs from reference scanning, 19 from other reference scanning in related reviews, protocols, and conference abstracts, and one study from contacted authors (Additional file 1: Appendix 8). Of the 157 RCTs, 51 were cluster-RCTs. A total of 66 of the 197 contacted authors responded to our emails, and 36 provided additional data for analysis.

Fig. 1
figure 1

PRISMA flow diagram for identification of eligible studies. Abbreviations: CA conference abstract, KT knowledge translation, mos months, RCT randomised controlled trial. From [157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,157,158,159,160,161,162,163,164,165,166,167,168,169,170, 172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,201]).

Table 2 Summary of patient characteristics
Full size table

Overall, 327 KT interventions were identified across all study arms (Additional file 1: Appendix 11), which focused primarily on patients only (42.8%), healthcare providers only (20.8%) and both patients and healthcare providers (11%; Table 3). Most KT interventions were single interventions (42.5%) and were not tailored to end-user type (Table 3). The KT intervention delivery method was not reported in many studies (39.1%), but when reported was frequently in-person (26.3%). Across the 157 RCTs, instruction on performing a behaviour and education targeting patients/caregivers were the most frequently reported BCT and EPOC components (Table 4).

Table 3 Summary of KT intervention characteristics across study arms
Full size table
Table 4 Summary of KT intervention behaviour change characteristics across studies
Full size table

Within-study risk of bias and across-study reporting bias

Within-study bias appraisal suggested that low ROB was present for 105 (67%) RCTs for random sequence generation, 63 (40%) RCTs for allocation concealment, 121 (77%) RCTs with incomplete outcome data and 119 (76%) RCTs with ‘other’ bias. Participant and personnel blinding and outcome assessment were judged at high ROB in 121 (77%) and 68 (43%) RCTs, respectively. Selective reporting was of unclear ROB in 74 (47%) RCTs (Additional file 1: Appendix 12 [43,44,45,46,47,48,49,50, 52,53,54,55,56,57,58,59, 61,62,63,64,65, 67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,201]).

Reporting bias assessment across studies using Egger’s test for each outcome and measurement scale separately suggested no evidence of publication bias or small-study effects (Additional file 1: Appendix 13).

Results of syntheses

Sustainability, fidelity, adherence of KT interventions for CDM practice

Overall, 157 RCTs reported on the primary outcome of sustained implementation of KT intervention for CDM practice. Of these, studies used different terms for sustainability dimensions: 14 studies were described by the authors as assessing sustainability in 25 (8.1%) interventions, 67 studies were described as assessing adherence in 115 (35.6%) interventions and 19 studies were described as assessing fidelity in 27 (8.7%) of the total 327 interventions (Figs. 2, 3 and 4, Additional file 1: Appendix 11). Of the 14 studies, five studies described adherence. Of 67 studies, five were also described by authors as assessing sustainability and 12 described as assessing fidelity. No study reported on all three dimensions. The 36.4% of the 327 KT interventions, representing most of the identified KT interventions, had a duration up to 15 months (Table 3).

Fig. 2
figure 2

Cochrane Effective Practice and Organisation of Care (EPOC) risk of bias summary results (n = 157 RCTs)

Full size image
Fig. 3
figure 3

Stacked bar plot of knowledge translation (KT) sustainability dimensions

Full size image
Fig. 4
figure 4figure 4

Forest plots for quality of life (a, b, c and d) and care outcomes (e). a Short-Form (SF) Mental Health Survey. b SF Physical Health Survey. c European Quality of Life-5 Dimensions (EQ-5D). d St George’s Respiratory Questionnaire (SGRQ) and Minnesota Living with Heart Failure Questionnaire (MLHFQ). e Quality of care. CI confidence interval, EQ5D European Quality of Life-5 Dimensions, HK Hartung–Knapp–Sidik–Jonkman method, MD mean difference, MLHFQ Minnesota Living with Heart Failure Questionnaire, OR odds ratio, SF Short-Form, SGRQ St George’s Respiratory Questionnaire, UC usual care

Full size image

Healthcare outcomes with meta-analysis—quality of life (QOL)

QOL was described in 50 studies reporting seven different measurement scales and 49 different interventions, including usual care. Below we present the results of each scale informed by at least 10 studies separately, whereas in Additional file 1: Appendix 14 [103, 137] we show the results with < 10 studies. The individual study results are reported in Additional file 1: Appendix 15 [44, 46, 49, 50, 62, 63, 65, 66, 72, 73, 77, 78, 83, 88, 94, 97, 100, 102,103,104, 111,112,113, 123, 124, 126, 128, 129, 134, 136,137,138,139, 143, 147, 151, 153, 155,156,157, 159, 161,162,163, 170, 176, 177, 180,181,182,183, 187,188,189,190,191,192, 194, 14]. In the present systematic review and in contrast to the sco** review, we assessed a more focused research question. We examined the impact of sustainable KT interventions on health outcomes, included RCTs, and performed a meta-analysis of the RCT findings. In this systematic review, we found substantial publication growth, and while most interventions were similarly intended for patients, they were not tailored for patient use. Stirman and colleagues identified 125 studies in their systematic review of public health and clinical intervention sustainability; half were quantitative studies and few reported rigorous evaluation methods [7]. The authors noted a limitation that there is insufficient intervention or outcome details to inform what interventions are effective in which contexts [7].

Two frequent KT challenges in the majority of studies included in this review are a lack of a clear definition of sustainability and the scarcity of evidence assessing QOL and QOC in KT interventions. We defined KT sustainability in this study as clinical and KT interventions continuing to be delivered beyond a certain period of time. Ideally, sustainability studies should specify whether the relevant outcomes are sustained, which is difficult to report given the short duration of grant funding. Researchers and implementers should consider other sustainability aspects, including capacity to sustain implementation. Our findings can be used by knowledge users (e.g. patients, clinicians, policy-makers) regarding the sustainability of KT interventions for CDM. Initial implementation strategies may need to be modified over-time to facilitate the intervention’s sustainability, as inducing behavioural changes in patients for extended periods of time may be difficult.

Prolonged implementation of effective clinical CDM interventions through sustainable KT interventions has the potential to optimise QOL and QOC in older adults with chronic diseases. More studies are necessary to assess the efficacy of individual KT interventions and their separate components in a network meta-analysis [18]. Future work could build on our study by addressing this research gap and relevant KT intervention costs. We anticipate that these results will help to explore sustainable KT interventions development for CDM in older adults and outline how to tailor interventions. In particular, our unique review provides a more granular look at KT intervention components and behaviour change strategies.

Strengths of our study include that we followed the Cochrane Handbook methods for systematic reviews [26]. Reviewers worked in pairs and independently for screening, data abstraction and risk of bias appraisal. We reported the results using the PRISMA 2020 statement [19]. To our knowledge, this is the first study assessing the KT intervention efficacy in a systematic review with meta-analysis of RCTs. We used novel approaches to engage knowledge users and integrate their views and values in this research [4]. We used different taxonomies (EPOC and BCT) to code KT interventions, allowing researchers to use our results to build their interventions to optimise future studies [23].

Our study has some limitations to be considered. First, due to the small number of studies, we were unable to compare the efficacy of different KT interventions. High heterogeneity might be due to varied KT interventions combined in a single group. Initially, we aimed to perform a network meta-analysis to compare multiple KT interventions and produce a ranked order of their KT sustainability efficacy; however, the available evidence did not permit this. Based on the network meta-analysis results, we planned to perform an economic analysis of the interventions identified as effective. Moving forward, we plan to update our systematic review and conduct a network meta-analysis to examine the impact of different sustained KT interventions in older adults with comorbid conditions and determine which approaches are most successful and cost-effective. We will explore how different KT intervention types are linked to CDM practice. Second, the scarcity of available data is a limitation in that many KT interventions were informed by only a few studies and patients. This could affect our ability to detect differences in effects due to reduced statistical power. Also, demographic variables that may explain heterogeneity, such as age categories, living with or without a partner, were not available in the original studies. Third, our literature search is about 3 years old and new relevant studies may be available [205]. However, institutional COVID-19 lockdowns, remote work and logistical difficulties in coordinating a geographically dispersed team have resulted in extended time taken to gather, analyze, organize and present this data—excessive financial cost and lost personnel make updating this review non-feasible at present.

Conclusions

Detailed assessment of KT intervention sustainability and understanding which are the most effective intervention components remain important research gaps. The overall efficacy of KT interventions regarding supporting a better QOL and QOC remains uncertain. Our results should be interpreted with caution due to small, imprecise and heterogeneous observed study effects with high risk of bias in participant and personnel blinding and outcome assessment. Also, KT intervention efficacy may vary depending on the intervention type, number of chronic diseases, comorbidities and participant age, among other effect modifiers. For example, the number of chronic diseases and patient comorbidities may account for varying KT intervention effect, with a tendency to observe improved KT intervention efficacy as health issues accumulated. However, it is important to note that the relationship between these factors and KT intervention efficacy is complex and requires careful interpretation. Addressing specific outcome effect modifiers can be exploited by tailoring KT interventions in future studies.