Findings and interpretation
This is the first meta-analysis of economic evaluations of integrated care across different clinical and care areas as well as types of integration. The results indicate that integrated care was associated with lower costs and improved outcomes compared with usual care, especially in studies with a follow-up period over a year. This may reflect the need for a sufficiently long follow-up period for effects to emerge, especially if there is a learning period following implementation [
85]. In addition, studies with an extended follow-up period are more likely to capture long-term reductions in cost that may negate and surpass the initial investment in developing and implementing integrated care [
86].
Moreover, our results raise questions about whether the long-term impact of integrated care has been captured in the economic evaluations. Decision analytic modeling that extrapolates costs and outcomes beyond study follow-up is recommended, especially when benefits of deliberating treatment plans for chronic diseases may be ongoing [
28]. Nevertheless, it was performed in only one study [
43]. Similarly, integrated care inherently intersects care boundaries and impacts a broad range of costs and outcomes within and beyond the healthcare system. However, only about a third of the economic evaluations included an (sensitivity) analysis taking the societal perspective, potentially missing more widespread cost savings of integrated care interventions related to costs in other sectors, informal care-giving, and productivity [
87]. Taking the societal perspective in economic evaluations of integrated care is demanding and requires more complex and costly data collection. Although a health payer perspective remains the approach recommended by bodies such as the National Institute for Health and Care Excellence (NICE), this may jeopardize the quality of evidence about the cost-effectiveness of integrated care.
The pooled results of observational economic evaluations showed both significant reduction in costs and improvement in outcomes. However, this evidence was not found in studies with experimental designs. This contrast in the findings by study design highlights a well-documented trade-off between attributability and practicability [
88]. In our review, several observational economic evaluations barely took any measures to mitigate for treatment contamination or selection bias, thereby jeopardizing causal inference [
36,
52,
55,
64,
68,
69,
71,
74]. Although experimental designs are the gold standard for robust causal inference, their adoption in evaluating integrated care has been criticized due to their rigidness and low generalizability [
89,
90].
Studies from Europe and Australia/Asia were significant in both costs and outcomes; whereas, studies from North America showed no significant effects. The reasons for this are unclear but could be owing to differences in healthcare systems [
91] or the stage of implementation of integrated care interventions. North America is at a more advanced stage; it is possible, therefore, that the studies implemented are broader and involve larger populations, which may dilute the effects relative to smaller studies elsewhere where integrated care is still at a nascent stage [
18].
Among types of intervention, disease management interventions alone showed significant decreases in costs and improvements in outcomes. This is similar to the findings of previous meta-analyses of disease management programs on single chronic conditions [
20,
92,
93] and may mean that integrated care interventions are implemented more easily within single disease areas. Indeed, many initiatives around the world have started integrating services within single chronic conditions as a first step towards a wider integration [
17]. Disease management programs have been long implemented in North America and Europe and certain levels of efficiency may have been achieved due to experience and productivity [
94]. However, an important challenge remains, since disease management programs may not meet the needs of a patient with multiple health problems with complex needs [
95].
Quantity and quality of economic evaluations
Substantial investment into the implementation of integrated care is occurring on a global scale [
96]. Despite this, only 34 economic evaluations of integrated care were identified in this review that had sufficient reported costs and outcomes to be included in the meta-analysis. Of these studies, only 19 (56%) had a quality score over 70%—a score above which is generally given to a study of “fair standard” [
97]. The relatively low number of economic evaluations and their moderate methodological quality may stem from two reasons. First, economic evaluations are increasingly piggy backing effectiveness assessments of integrated care and are subject to insufficient communication between health economists and clinical/health service researchers [
88]. Hence, regarding economic evaluations as an “afterthought” may be contributing to such a remarkably low number of suitable studies for meta-analysis. Second, integrated care interventions, like many complex interventions, are frequently not subject to extensive health technology assessment (HTA) as part of a reimbursement process at the national level. As a result, the cost-effectiveness of integrated care may receive less scrutiny than other health interventions (e.g., pharmaceuticals and clinical technologies) traditionally subject to HTA.
Limitations
First, although the search strategy deployed aimed to include all studies broadly fitting the pre-set definition of integrated care, this review may have missed studies characterized by an alternative approach not covered by the search terms. Therefore, this review and meta-analysis may exclude some economic evaluations of interventions that could broadly fall under the integrate care umbrella term without explicitly fitting our working definition. However, it is expected that this may be the case for only a few studies as we have used broad concepts of integrated care in our search strategy. Second, despite the use of a binary system to review and assign quality scores to each economic evaluation based on the CHEERS-adapted checklist, there was opportunity for subjectivity which may have biased the scoring. Finally, due to the lack of reported standard error/ deviation of mean costs and outcomes, the heterogeneity across the studies reflected in the
I2 statistic was based on the study quality rather than the precision of the mean [
98]. Therefore, the statistical significance of the meta-analysis results should be interpreted with caution.
Policy and research implications
Our findings support the reorientation of healthcare systems towards integration of care to help policy makers to meet increased demand for health and social care within tight budgets. However, with such sparse economic evaluations in integrated care, there is insufficient evidence about the factors that determine the cost-effectiveness of integrated care, such as models of care integration, implementation process, and target population. Efficiency of research is being streamlined in most other health innovations (e.g., pharmaceuticals and medical technologies) by including them in reimbursement processes with cost-effectiveness as an explicit criterion for market access [
99]. Similar efforts should be made in assessing the cost-effectiveness of integrated care. Measures could be taken to standardize description of intervention and comparator, reporting of methods and results; apply appropriate follow-up periods and decision-analytic models; address bias; and deploy explicit decision criteria when value-for-money is uncertain. Such directed expansion of health economics towards the evaluation of integrated care is necessary to ensure decisions surrounding the implementation of integrated healthcare delivery are likely to benefit, rather than hinder, aims to meet increasing demands on tightening budgets.