A total of 2181 articles were retrieved, of which 863 were duplicates. Overall, 1,259 articles were excluded following title and abstract screening, and 37 were excluded following the full-text screen (Appendix 2); thus, 22 reviews were included in this overview (Cohen’s κ = 0.65; substantial agreement) (Fig. 1).
Most of the included reviews were systematic [20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39], while two were deemed eligible for inclusion due to their systematic approach on synthesising and reporting the data [40, 41]. Most of the included reviews (n = 15/22) reported a meta-analysis [20,21,22,23,24,25,26,27,28,29,30,31,32,33, 40], while the rest narratively synthesised the findings [34,35,36,37,38,39, 41].
Characteristics of the included reviews
Included reviews were published between 2018 and 2023. The reviews were from Canada (n = 9), UK (n = 4), USA (n = 2), China (n = 3), Germany (n = 1), Sweden (n = 1), Ireland (n = 1), and the Netherlands (n = 1). All reviews used more than two databases for their searches, with Medline, EMBASE, and Cochrane Central Register of Controlled Trials and Cochrane Library being the most common searched databases. Most reviews included up to 50 primary studies, while three reviews included more than 100 studies [30, 31, 41]. Six reviews were focussed on specific patient groups [21, 26, 27, 34, 37, 40], while the rest were focussed on different types of patients at pre- to post-discharge [20, 22,23,24,25, 28,29,30,31,32,33, 35, 36, 38, 39, 41]. All reviews evaluated the impact of interventions on healthcare- and patient-related outcomes. Most reviews focussed on transitional care interventions (TCIs) with the most common being telehealth-based interventions. The most frequently assessed outcomes across reviews were ED/hospital readmissions, ED visits, LOS at hospital/ED, and mortality. Reviews characteristics table can be found in Appendix 3.
Primary study overlap
Overall, 22 reviews reviewing 570 unique primary studies were included in the overlap assessment (Appendix 4). Overlap of included primary studies across reviews was low, with an overall corrected cover area (CCA) of 1.1%.
Risk of bias in systematic reviews and primary studies
Overall, this sample of reviews was assessed as being of only low quality according to AMSTAR 2 criteria. Downgrades for risk of bias mostly occurred in the following AMSTAR 2 domains: (i) explanation of included studies’ design, (ii) duplicate data extraction, (iii) list of excluded studies with justification, and (iv) role of funding sources (Fig. 2). In total, 12 reviews had an overall confidence rating of low (54.5%), while seven and three reviews had an overall rating of critically low (31.8%) and moderate (13.6%), respectively. Of the critical domains, 17 out of 22 reviews had a pre-established protocol, all reviews reported a generally satisfactory literature search, while 16 out of 22 did not provide a justification for the excluded studies. Most reviews generally used satisfactory techniques to assess RoB in the included primary studies (n = 21), while all the reviews with a meta-analysis used appropriate statistical methods. Most reviews considered RoB assessments in the interpretation of the results (n = 20/22), while publication bias was not assessed in six reviews. Confidence ratings graph and table with the critical appraisal ratings by study can be found in Appendix 4.
Stacked bar plot of the distribution of ratings as percentages by AMSTAR 2 domain
Overall, 155/353 (43.9%) unique primary studies included in the meta-analytic reviews of this overview were rated to be of high-risk, 98 (27.75%) were rated to be of low-risk, 46 (13.1%) were rated to be of unclear risk, while 38 (10.7%), 8 (2.3%), and 8 (2.3%) of primary studies were judged to be of some, serious, and critical concerns respectively (Appendix 4). The most affected RoB domains across the primary studies were the randomisation process, allocation concealment and blinding procedures.
Summary of results
This overview synthesised evidence from 22 reviews, encompassing 570 unique primary studies, on the effectiveness of intermediate care interventions. The main outcomes assessed were LOS, ED visits, hospital readmissions, mortality, QoL, and costs. Meta-analyses were reported for LOS (6 reviews), ED visits (9 reviews), and readmissions (12 reviews). Mortality was examined in 13 studies, with 8 reporting meta-analyses. QoL was assessed in 8 studies, 4 of which included meta-analyses. Cost-effectiveness was evaluated in 6 studies.
Length of stay (LOS)
Overall, data regarding the effects of intermediate care interventions on LOS was drawn from 12 reviews [22, 25,26,27, 30, 31, 36,37,38,39,40,41] with six of them reporting meta-analyses [22, 25, 26, 30, 31, 40] (Fig. 3). Meta-analyses included from four [25] to 12 reviews [30, 31]. Meta-analytic estimates showed that high-complexity TCIs (indicated by the number of components and discharge stages) (SMD = -3 [95%CI: -3.61, -2.39], p < 0.05; SMD = -0.2 [95%CI: -0.38, -0.03], p < 0.5) [22, 31], VWs (MD = -1.94 [95%CI: -3.28, -0.6], p < 0.05) [25], PERTs (MD = -1.61 [95%CI: -3.21, -0.02], p < 0.05; MD = -1.79 [95%CI: -3.29, -0.28], p < 0.05) [40], and nurse-led TCIs (MD = -2.37 [95%CI: -3.16, -1.58], p < 0.05) [26] were effective in reducing EDs/hospital/ICU LOS, and LOS readmission following discharge. The remaining studies found that nurse-led TCIs were associated with reduced LOS [27, 36, 37], while inconsistent effects were observed regarding the impact of ED-based interventions, rehabilitation, and palliative care support [38, 39, 41].
Meta-analytic estimates regarding the effectiveness of intermediate care interventions in reducing LOS*. Note. HC: high-complexity; ICU: intensive care unit; LC: low-complexity; MC: medium-complexity; MD: mean difference; NL: nurse-led; PERTs: pulmonary embolism response teams; SMD: standardised mean difference; TCIs: transitional care interventions. *Values < 0 indicate shorter LOS at ED/hospital favouring the intermediate care interventions
ED visits
Overall, data regarding the effects of intermediate care interventions on ED visits was drawn from 13 studies [20,21,22,23,24,25,26,27, 31,32,33, 35, 39] with nine of them reporting meta-analyses [20,21,22,23, 25,26,27, 31, 33] (Fig. 4). Meta-analyses included from two [21] to 41 studies [31]. Meta-analysis estimates showed that telemonitoring (RR = 0.62 [95%CI: 0.42, 0.94], p < 0.05) [21], PFC (OR = 0.56 [95%CI: 0.34, 0.95], p < 0.05) [23], VWs (RR = 0.83 [95%CI: 0.7, 0.98], p < 0.05) [25], and low-complexity TCIs (OR = 0.68 [95%CI: 0.48, 0.96], p < 0.05) [31] were effective in reducing ED visits within 30 days post-discharge. Also, ED-based TCIs were found to be effective in reducing outpatient follow-up rates (OR = 1.79 [95%CI: 1.43, 2.24], p < 0.05) [20]. Two reviews found that ED-based interventions were associated with reductions in ED visits [24, 35], while inconsistent findings – in terms of the direction and precision of the effects – regarding the effectiveness of TCIs and ED-based interventions were observed in two reviews [32, 39].
Meta-analytic estimates regarding the effectiveness of intermediate care interventions in reducing the number of (re)-visits at ED*. Note. CI: confidence interval; COPD: chronic obstructive pulmonary disease; HF: heart failure; LC: low-complexity; NL: nurse-led; NLPD: nurse-led peri-discharge; OR: odds ratio; PFC: Person- and family-centred care; RR: risk ratio; TCIs: transitional care interventions; VW: virtual ward. *Values < 1 indicate lower probability for subsequent ED visits, favouring the intermediate care interventions
Readmissions
Overall, data regarding the effects of intermediate care interventions on ED/hospital readmissions was drawn from 20 studies [20,21,22,23,24,25,26,27,28,29,30,31,32,33, 35,36,37,38,39, 41] with 12 of them reporting meta-analyses [20,21,22,23, 25,26,27,28,29, 31,32,33] (Appendix 5). Meta-analyses included from four [23] to 73 studies [31]. Apart from three reviews [20, 29, 33], meta-analysis estimates showed that intermediate care interventions, namely telephone-based (RR = 0.72 [95%CI: 0.63, 0.81], p < 0.05) [21] and multi-component TCIs (RR = 0.63 [95%CI: 0.45, 0.88], p < 0.05; OR = 0.75 [95%CI: 0.62, 0.91], p < 0.05) [21, 32], PFC (OR = 0.56 [95%CI: 0.34, 0.95], p < 0.05) [23], VWs (RR = 0.91 [95%CI: 0.85, 0.98], p < 0.05) [25], nurse-led TCIs (RR = 0.91 [95%CI: 0.82, 0.99], p < 0.05; RR = 0.89, [95%CI: 0.82, 0.97], p < 0.05; RR = 0.78 [95%CI: 0.68, 0.89], p < 0.05) [26, 27], TCIs of low- and middle-complexity (OR = 0.78 [95%CI: 0.66, 0.92], p < 0.05; OR = 0.82 [95%CI: 0.68, 0.97], p < 0.05) [31], critical care outreach (RR = 0.64 [95%CI: 0.42, 0.99], p < 0.05; RR = 0.62 [95%CI: 0.43, 0.9], p < 0.05) [28] and rapid response teams (RR = 0.76 [95%CI: 0.67, 0.87], p < 0.05) [28], were effective in reducing the hospital readmissions within 30 days post-discharge [21,22,23, 25,26,27,28,29, 31, 32], with these effects extending to >30 days post-discharge [21, 31, 32]. Also, the implementation of TCIs was found to be associated with a reduction in hospital/ED readmissions (OR = 1.66 [95%CI: 1.18, 2.35], p < 0.05) [22]. The remaining reviews found that ED-based, primary-care and home-based interventions, outpatient telehealth consultations, ED-based TCIs and advanced nursing care interventions, and multi-disciplinary coaching interventions were associated with reductions in hospital readmissions [24, 30, 35, 36, 39]. Inconsistent findings regarding the effectiveness of TCIs and palliative care interventions were observed in two reviews [37, 38].
Mortality, quality of life (QoL) and costs
Overall, data regarding the effects of intermediate care interventions on mortality was drawn from 13 studies [21, 22, 24, 25, 28,29,30, 33, 36,37,38,39,40,41] with eight studies reporting meta-analyses [21, 22, 25, 28,29,30, 33, 40] (Appendix 5). Meta-analysis estimates showed that telemonitoring (RR = 0.62 [95%CI: 0.42, 0.94], p < 0.05) [21] and VWs (RR = 0.9 [95%CI: 0.82, 0.97], p < 0.05) [25] were associated with lower mortality rates. No evidence regarding the effects of home-based, advanced nursing and palliative care and TCIs in reducing mortality were detected [24, 36,37,38, 41]. Overall, data regarding the effects of intermediate care interventions on patients’ QoL was drawn from eight studies [21, 22, 25, 31, 34, 37, 38, 41] with four of them reporting meta-analyses [21, 22, 25, 31]. Meta-analysis estimates showed no evidence regarding the effectiveness of TCIs and VWs in increasing QoL across patients [21, 22, 25, 31]. The remaining reviews found no evidence regarding the effects of PFC, TCIs, palliative care interventions [34, 37, 38], while education coaching interventions were found to be generally effective [41]. Overall, six studies assessed the cost related to the implementation of intermediate care interventions [20, 24, 30, 36, 37, 41]. Inconsistent findings – in terms of the magnitude and precision – regarding the cost-effectiveness of intermediate care interventions were found, while primary care interventions (MD=-$4119 [95%CI: -$7935, -$302], p < 0.05) [24] and TCIs with assisted technology tools seemed to be cost-effective [36, 37, 41].
Assessment of reporting bias and certainty of evidence
All reviews were generally well-reported on their pre-defined analysis and used appropriate methods of analysing the data. Most of the primary studies included in the reviews were deemed to be of high risk of bias at least in one domain. In four out of six reviews reporting meta-analysis of LOS data, the level of heterogeneity was higher than 75%. In three out of nine reviews reporting meta-analysis of ED-visits data, the level of heterogeneity was over 70%. In three out of 12 reviews reporting meta-analysis of readmissions data, the level of heterogeneity was over 70%. In three out of eight reviews reporting meta-analysis of mortality data, the level of heterogeneity was over 75%.
No Grading of Recommendations Assessment, Development and Evaluation (GRADE) assessment was provided for LOS data. Three out of nine reviews that meta-analysed ED-visits data reported GRADE certainty of evidence assessments. In two reviews the certainty of evidence was judged to be low [20, 21], while in one review the certainty of evidence was judged to be high [33]. Four out of 12 reviews reporting meta-analyses of readmissions data provided GRADE certainty of evidence assessments. The certainty of evidence was judged to be low [20], moderate [21, 33], and very low [28]. Four out of eight studies reporting meta-analysis on mortality data provided certainty of evidence assessments with those assessments ranging from very low [28] to moderate [21, 30, 33].
link