Background

Worldwide, stroke is the second leading cause of death and the most common cause of permanent disability, resulting in huge societal and economic burdens related to long-term care, rehabilitation, and productivity loss [1, 2]. Acute ischemic stroke (AIS) represents the majority of stroke patients, and reperfusion treatments like intravenous thrombolysis (IVT) and endovascular thrombectomy (EVT) have shown to be effective in improving functional outcomes. Both treatments are highly time-dependent, with IVT effective up to 4.5 h after symptom onset and EVT within 6 h [3, 4]. For selected patients suspected of a large vessel occlusion (LVO), EVT has shown to be effective even up to 24 h [5]. Importantly, shorter time from onset to treatment (OTT) with IVT or EVT is associated with better functional outcomes [6,7,8].

Due to the time dependency of reperfusion treatments, multiple strategies or interventions have been proposed to reduce time delays along the acute stroke pathway. Examples include educational interventions to create awareness among citizens to contact emergency services immediately following symptom onset [9], a mobile stroke unit that brings IVT to the patient instead of transporting the patients to the nearest IVT capable facility [9], telemedicine solutions for expert opinion at a distance [10], and workflow improvements including inter-hospital patients transfer, teamwork and feedback on performance [9, 11]. In addition, direct transport of acute stroke patients suspected of LVO from the onset scene to a comprehensive stroke center based on prehospital triage instruments [12] is a promising alternative organizational model, which could decrease the OTT time for patients eligible for EVT [13].

While several strategies have been developed to reduce time to reperfusion treatments, less is known about their cost-effectiveness. As stroke incidence and its burden on society are expected to increase [14], evidence generated by economic evaluations of these strategies will support policymakers, clinicians, and other stakeholders in deciding how to allocate scarce resources whilst optimizing clinical outcomes for patients. Therefore, the aim of this study is to systematically review the cost-effectiveness of strategies directed at reducing time to reperfusion treatments for AIS patients.

Methods

Search strategy and study selection

This systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [15]. The search strategy was constructed by the first two authors (C.P.N. and W.J.M). Three electronic databases (EMBASE, MEDLINE/ Pubmed, and Web of Science) were searched to identify relevant articles published between January 2010 to January 2022. The search strategy was based on the PICOS format: The population (P) were ‘stroke’ patients, the intervention (I) ‘EVT or IVT’ and ‘reducing time-to-treatment’, and the outcome (O) was ‘incremental cost-effectiveness ratio’ (ICER). Comparators (C) and study design (S) were not included to maximize records retrieved. Our eligibility criteria were: (1) stroke patients treated with IVT and/or EVT, (2) full economic evaluation was performed (i.e., cost-effectiveness, cost-utility, or cost–benefit analysis), and (3) strategies or interventions that aim to reduce time to treatment or increasing the proportion of eligible patients for treatment due to time delay reduction. Papers were excluded if: (1) articles published before 2010; (2) articles not written in English, and (3) reviews, protocol papers, conference abstracts, letters to editors, or case reports. All identified records from three databases were stored in Endnote® X8 software [16]. Details of the search strategy are reported in the Supplementary Information (Appendix 1). In addition, a manual cross-reference of selected papers was performed to include more relevant papers.

Study selection, data extraction, and quality assessment were performed independently by two reviewers (C.P.N. and W.J.M.). Disagreements were discussed between reviewers until a consensus was reached. Discrepancies were resolved by consulting a third reviewer (M.M.H.L.).

Data extraction and analysis

The following information was extracted from the selected papers independently by two reviewers (C.P.N. and W.J.M): country of origin, strategies including intervention, comparator, type of economic evaluation, effectiveness measurement, model type (i.e. Markov model, decision tree or other), economic perspective, time horizon, discount rate, reference year of cost, outcomes including incremental costs, incremental effectiveness and an ICER. An economic evaluation was stratified by type of strategy or intervention and compared with the same effectiveness measurement. An ICER was calculated by dividing the different costs by the different effectiveness of two strategies. The strategy or intervention was considered cost-effective if an ICER was less than the willingness-to-pay threshold mentioned in the study. In case specific willingness-to-pay thresholds were not stated in the study, the recommended threshold of one Gross Domestic Product (GDP) per capita was used according to the World Health Organization for cost-effectiveness studies [17]. The strategy or intervention was dominant if the strategy resulted in higher effectiveness (i.e., QALY) and less costs than the comparator. In case data were insufficient or unclear for data analysis, additional information was requested by contacting the corresponding author of individual papers.

Quality assessment

For quality assessment, the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) guideline [18] was used, scoring a total of 24 criteria. Papers were classified into three categories: high (75.0% or more), medium (from 50.0% to 74.9%) and low quality (less than 50.0%) [19, 20].

Results

Study selection

The initial literature search resulted in 1,790 records (Fig. 1). After removing duplicates, 1,294 articles were screened on title and abstract independently by both reviewers, and 36 articles were considered relevant for the full-text review. Of these articles, 18 articles were excluded because of missing outcome measures (n = 1), missing time to treatment variables (n = 13), missing full economic evaluation (n = 2) and data duplication between articles (n = 2). Two additional articles were identified when an additional search in Pubmed, EMBASE, and Web of Science was conducted until January 9th 2022. Ultimately, 20 studies were included.

Fig. 1
figure 1

PRISMA flow chart

Full size image

Study characteristics

The majority of economic evaluations were conducted in high-income countries, and only one study was performed in an upper-middle-income country (China) [21]. Five studies were based in the United States [22,23,24,25,26], one study in Canada [27], 10 studies in Europe [28,29,30,31,32,33,34,35,36,37], one study in Japan [38], one study in Singapore [39], and one study in Australia [40]. The type of interventions varied across studies, including educational interventions [23, 34, 39], organizational models [22, 24,25,26,27, 30,31,32, 36, 38, 40], healthcare delivery infrastructure [29, 33, 35], and workflow improvement [28, 34]. An overview of all study characteristics is provided in Table 1.

Table 1 Study characteristics
Full size table

The majority of studies applied cost-utility analysis (n = 14) with quality-adjusted life year (QALY) (n = 13) [21,22,23, 26,27,28,29,30, 32, 34, 36, 38, 39] and disability-adjusted life year (DALY) (n = 1) [40] as the primary outcome of effectiveness. Two studies reported cost-effectiveness based on the IVT rate [29, 30], one study based on the number of home discharges [20], one study based on three outcome variables, including the IVT rate within three hours from symptom onset, OTT time, and door-to-needle time [25], and one study based on door-to-needle time, and death averted [37]. One study analyzed the cost–benefit of the used strategy [31].

Most economic evaluations conducted a model-based approach, while two studies were trial-based analyses [32, 37]. Simulation modeling was applied in six studies [25, 29, 30, 33, 38, 40], decision tree analytic modeling was used in four studies [24, 26, 34, 35], the use of a decision tree and a Markov model was adopted in four studies [21, 27, 28, 36], and two studies used a Markov model [22, 23]. Additionally, one study used a population-level systems dynamics model [39], and one study did not report a clear model [31].

Economic evaluations were analyzed from various perspectives containing a healthcare payer or payer perspective (n = 7) [26, 27, 29, 32, 36, 37, 39], a hospital or healthcare provider perspective (n = 3) [24, 35, 40], a national health service viewpoint (n = 2) [28, 34], a societal perspective (n = 3) [21,22,23], a policy maker viewpoint (n = 1) [33] and a government perspective (n = 1) [38]. Hence, most of studies included medical service costs, such as ambulance transportation, hospitalization, outpatient visits, rehabilitation, and long-term care. However, only two studies with a societal perspective reported indirect costs (productivity loss) [21, 23]. In three studies, no perspective was mentioned. One study included health care costs (ambulance service, treatment costs, bed day, nursing home, residential care home, assisted living facility, and carer visiting) [30], and two studies considered intervention costs and costs of different functional status after stroke [25, 31].

Time horizons ranged from 90 days to a lifetime. Seven studies used a 5-year horizon [23, 24, 29, 30, 32, 37, 40], and six studies employed a lifetime perspective [22, 25, 27, 28, 34, 36]. A ninety-day horizon was used in one study [26]. Discount rates were reported in 13 studies ranging from 1.5% to 5.0% [21,22,23,24,25, 27, 28, 32, 34, 36, 38,39,40]. Seven studies did not report any discount rate. Of those, the discount rate was not applicable for three studies due to a short time horizon of 90 days or 12 months [26, 31, 35].

Quality of studies

Total CHEERS scores, along with percentage scores of included studies, are presented in Table 2. Fourteen studies were assessed as high quality (ranging from 79.2% to 93.8%) and six studies as medium quality (ranging from 66.7% to 72.9%). There were no studies that were assessed as low quality. The assessment on each item of the CHEERS statement is described in the Supplementary Information (Fig. S1).

Table 2 Reporting quality assessment
Full size table

Results of economic evaluation

Included studies were categorized into four main groups based on the type of interventions: (1) educational interventions, (2) organizational models, (3) healthcare delivery infrastructure, and (4) workflow improvements (see Table 3).

Table 3 Study results
Full size table

Educational interventions

Three studies assessed the cost-utility of educational interventions to reduce time delays to treatment in AIS patients [23, 34, 39]. Topics varied across studies, including public campaigns, interactive interventions for patients (i.e. educational materials, medical alert bracelets, and in-hospital interactive group sessions), and training staff of general practitioner offices. Irrespective of the type of intervention, all educational interventions were cost-effective as their ICERs were lower than the willingness-to-pay thresholds. One study even demonstrated that educational interventions for both health staff and patients were dominant compared to current practice (higher QALY gained (1.14- 2.26 QALYs) and saving $16,153- $32,305) [34].

Organizational models

Over half of the selected studies (12/20) performed economic evaluations on organizational models. Different strategies were considered, including studies on telemedicine solutions between stroke centers and community hospitals [21, 22, 24,25,26], in-ambulance telemedicine [36], mobile stroke units [31, 32, 40], a combination of mothership model and drip-and-ship model with alternative transportation modes (ground or air) [27], prehospital redirection of patients to regional IVT center instead of local stroke units [30], and a mobile interventionist [38]. Most of these studies (11 out of 12) reported that interventions were cost-effective or dominant (higher QALY gained and saving costs) [22, 24,25,26,27, 30,31,32, 36, 40]. However, one Japanese study reporting on the use of a mobile interventionist showed that this approach was only cost-effective in a specific region (i.e. Kamikawachubu area in Hokkaido). In Hokkaido, the ICER of mobile interventionists was higher than the threshold of $48,146 in other areas [38].

Healthcare delivery infrastructure

Three economic evaluations reported interventions related to the regional healthcare delivery infrastructure for acute stroke care, such as introducing new EVT centers [29], upgrading hospitals to IVT-capable stroke care centers [33, 35] and centralization of acute stroke care treatment [33]. Effectiveness was measured as the percentage of the IVT rate in two studies. Within these studies, the ICER ranged from $56 to $660 per one percentage increase in the IVT rate [33, 35]. One study showed that adding new EVT-capable centers in their region gained an additional 213 QALYs and saved £2,870,000 compared to the current situation [29].

Workflow improvement

Three studies assessed economic evaluations of workflow improvements [28, 34, 37]. Implementing a quality improvement project, including streamlining stroke care pathway and simulation-based training, produced an additional $29 per minute door-to-needle time reduction and $10,543 per death averted [37]. One study evaluated inter-hospital transfer by helicopter emergency medical service compared to ground emergency medical service, which appeared cost-effective at the threshold of £30,000 per QALY [28]. Another study demonstrated that an immediate computed tomography (CT) scan strategy, in which the CT scanner was moved closer to the emergency department, was dominant when compared to current practice [34].

The ICERs of studies reporting QALYs (n = 13) [21,22,23, 26,27,28,29,30, 32, 34, 36, 38, 39] were assessed as a proportion of country GDP per capita (Fig. 2). Cross-study analysis indicated that the strategies of educational interventions, organizational models, healthcare delivery infrastructure, and workflow improvement were highly cost-effective with the ICER less than one times GDP per capita (ranging from -1.58 to 0.85 times GDP per capita) in 11 studies. Stevens et al. showed that the educational interventions in the US exceeded one times GDP (1.05 to 1.49 times GDP per capita). However, organizational models in Japan were not considered cost-effective in most of regions with the ICERs ranging from 2.25 to 40.89 times GDP per capita.

Fig. 2
figure 2

ICERs compared to country GDP per capita

Full size image

Three studies had more than 1 intervention (3 interventions in Bayer’s study, 2 interventions in Stevens’ study, 3 interventions in Ramos’ study); 1 intervention in 7 regions in Morii’s study. ICER: incremental cost-effectiveness ratio.

The majority of included studies (18/20) performed sensitivity analysis. Of these studies, probabilistic sensitivity analysis was conducted in seven studies. More information on the sensitivity analyses is summarized in the Supplementary Information (Table 1).

Discussion

Principal findings

To our knowledge, this is the first systematic review on the cost-effectiveness of strategies aimed to reduce OTT time for AIS patients. Twenty studies met the eligibility criteria and were included in this systematic review. Based on factors associated with time reductions along the acute stroke care pathway [9, 11, 44], we categorized strategies into four groups: educational interventions, organizational models, healthcare delivery infrastructure and workflow improvements. The results of this review demonstrate that all intervention types may be cost-effective or even dominant when compared with current stroke care practices [21,22,23,24,25,26,27,28,29, 31, 34, 36, 39, 40]. Two studies in ‘organizational model’ category [30, 32], in which the willingness to pay threshold was not stated, were deemed cost-effective if the threshold of one GDP per capita was applied [17]. Accordingly, these strategies should be considered as interventions to optimize acute stroke care systems, as they improve functional outcomes for AIS patients and reduce long-term costs of disability after stroke. However, some improvement approaches might be region-specific, as Morii et al. [38] found that a mobile interventionist strategy was cost-effective in only one area (i.e. Kamikawahokubu) while not feasible in other areas due to the close distance between the hub and spoke centers and a relatively low incidence of LVO patients. We could not conclude whether interventions were cost-effective for two studies [33, 35] in the category ‘healthcare delivery infrastructure’ and for one study [37] in the category ‘workflow improvement’ because the primary outcome was not measured in QALYs. In addition, cross-study analysis of studies reporting QALYs showed that all intervention types were highly cost-effective in most of countries (11/13 studies).

In our review, 70% of included studies were classified as high quality based on the CHEERS score. The majority of the studies referred to a cost-utility analysis, in which the study design (i.e. the perspective, time horizon, and models) differed across studies. However, most of studies (16/20) showed that strategies aimed at reducing the OTT time were cost-effective. Although strategies were categorized into four groups, classifying articles was not straightforward because various strategies were simultaneously considered in economic evaluations. For example, Penaloza-Ramos et al. conducted health economic evaluations of seven strategies simultaneously, including health staff training (in the ‘educational intervention’ category) and timely CT scan (in the ‘workflow improvement’ category) [34]. Bayer et al. performed the cost-effectiveness analysis of combining a public information campaign with five other interventions, such as IVT and EVT treatment in the acute stroke unit, out-of-hospital rehabilitation, and secondary prevention. The results showed that combining interventions was the most cost-effective strategy [39].

Policy relevance

Incorporating the cost-effectiveness evidence, apart from the current set-up of regional stroke services, other local factors need to be taken into account before intervening in regional stroke care pathways, such as the incidence of stroke, population density, and geographic location of stroke centers. From a variety of possible improvement strategies, policymakers and other stakeholders will have to identify the ones most favorable in terms of effects and cost implications. For example, the mobile interventionist strategy is cost-effective in the specific area with a high incidence of LVO, more than three-hour travel for patients to a hub (intervention) center, and within one-hour travel for interventionists [38]. Furthermore, Yan et al. identified the optimal strategy in delivering EVT, as the combination of both mothership model and drip-and-ship model based on the geographical location of stroke onset and stroke centers [27]. As such, it can be inferred that multiple organizational models need to be considered in stroke management according to local characteristics.

Strengths and limitations

The strength of our systematic review is in its wide scope, providing a broad range of various strategies to reduce time delay compared with focusing on one or more specific interventions. Additionally, all types of full economic evaluations, with both trial- and model-based analysis, were included in this review. Furthermore, it provides an informative synthesis of the cost-effectiveness evidence for all strategies, enabling better-informed decisions in stroke care, especially in the context of constrained health resources and the ageing population.

Nevertheless, several limitations in the review should be noted. Firstly, direct comparison between strategies was not possible due to heterogeneity across studies, such as setting, effectiveness outcomes, and comparators. Therefore, a ranking and assessment could not be offered. The latter would be difficult, as generic weighing and balancing are impractical without taking region-specific characteristics of the local setting into account. Instead, we offer a narrative synthesis which will help local policymakers to consider their individual options. Another limitation is to include only English papers in the present systematic review. However, no language restriction was applied in our search strategy, and screening title and abstract revealed that non-English articles also met other exclusion criteria. Finally, the results of this review were based on published studies only, thus potentially leading to publication bias.

Future perspective and recommendation

Our study demonstrates the cost-effectiveness of various strategies aiming at reducing OTT, but other organizational strategies are currently lacking a cost-effectiveness perspective. For example, the use of prehospital triage tools to transfer stroke patients directly to the appropriate target hospitals [45] or workflow improvements in which suspected LVO patients are directly transferred to the angiography suite [46] may be efficient solutions to further reduce time delays to EVT initiation. This emphasizes the need of economic evaluations for these strategies.

There is also a notable gap persisting in the cost-effectiveness analysis of strategies reducing OTT time in low- or middle-income countries. In such settings policymakers and health care providers are possibly in the phase of building infrastructure without the opportunity to consider alternative organizational set-ups (i.e. mobile stroke unit, mobile interventionist) [47]. Although stroke remains a major cause of death and disability worldwide, stroke incidence and mortality rates recently have shown to decline in high-income countries due to improvements in primary and secondary prevention as well as in acute reperfusion treatments [48]. However, the trend in low- and middle-income countries is the opposite, resulting in a significant stroke burden in these regions [49]. Given the concerning shift in stroke epidemiology to low- or middle-income countries, important choices lie ahead, even when considering the very basic infrastructure.

Current economic evaluations use healthcare or payer perspectives that do not include indirect costs of post-stroke, such as productivity loss and informal caregiving for stroke patients. However, indirect costs of post-stroke account for approximately 33% of the total economic burden of stroke [50]. There is a remarkable increase worldwide in stroke incidence among younger groups (i.e. less than 65 years old) [51], which is likely to increase the indirect cost of stroke burden. Hence, future studies need to take into account the lack of evidence of societal perspective, and in low-and middle-income countries.

Conclusions

The results of this systematic review show that reported strategies reducing time delay in stroke care services are mostly cost-effective across different settings. While the findings from this review provide mainly positive results, local characteristics and background of regional health care systems should be taken into account.

All authors were involved with the conception and study design. Chi Phuong Nguyen and Willemijn J Maas performed screening records, data extraction, and data analysis. Maarten M H Lahr was the third reviewer. Chi Phuong Nguyen outlined the manuscript. Chi Phuong Nguyen and Willemijn J Maas drafted the manuscript. All authors critically reviewed the entire manuscript and approved the final submission.