Cost-Effectiveness of Hub-and-Spoke Telestroke Networks for the Management of Acute Ischemic Stroke From the Hospitals’ Perspectives

Introduction

Intravenous thrombolysis reduces long-term disability and is cost-effective in acute ischemic stroke (AIS).^1–3 Unfortunately, intravenous thrombolysis is underused with rates <5%,⁴ attributed in part to a lack of stroke specialists. In rural areas, use of intravenous thrombolysis is further restricted, exacerbated by the dearth of stroke experts and lack of access to endovascular revascularization therapies.^5,6 Hub-and-spoke telestroke networks can overcome these geographic barriers to acute stroke care, enhance stroke diagnosis, increase intravenous thrombolysis administration rates, and improve long-term outcomes.^7–11 In the past decade, telestroke networks have become widespread and represent an expanding model of stroke care, particularly in rural and small hospitals.¹²

However, telestroke networks are associated with significant upfront costs, and the absence of cost-effectiveness data has hindered further dissemination. Different stakeholders, including government and nongovernment insurers, hospital administrators, and health practitioners, may be interested in the costs and benefits of these establishments to better inform their decisions. A previous study showed that telestroke networks were cost-effective in the long term from a societal perspective.¹³ To date, cost-effectiveness of telestroke networks from individual hospitals’ perspectives has not been assessed. Such information would be valuable because, to a large extent, decisions on setting up (in the case of a hub hospital) or joining (in the case of a spoke hospital) a telestroke network are made by individual hospitals. Using data from 2 existing telestroke networks, we developed a decision model to assess the cost-effectiveness of establishing a hub-and-spoke telestroke referral network from the perspectives of the hub, the spoke, and the network. Our goal was to determine whether the costs and benefits of a telestroke network support its implementation from these 3 key perspectives and to evaluate the sensitivity of these results across a range of plausible variations in model inputs.

Methods

Model Overview

A decision analytic model (Figure 1) was developed to compare the cost and effectiveness of treating AIS patients with and without a telestroke network from the perspectives of an entire network, a hub hospital, and a spoke hospital. The model included important branching points from the time a patient presented to an emergency department, either in a network or not in a network, through hospital discharge, including intravenous thrombolysis, transfer from spoke to hub, and endovascular stroke therapy. Outcomes were measured by discharge destination (ie, home discharges, rehabilitation/nursing home discharges, and in-hospital death). A 5-year time horizon was used for the base case.

• Download figure
• Download PowerPoint

The following assumptions were made in model estimation:

• Improvements in outcomes only resulted from increases in the proportions of AIS patients receiving intravenous thrombolysis and endovascular stroke therapies among patients who presented to the emergency department in a spoke hospital.

• All patients (with or without intravenous thrombolysis) in spoke hospitals could be transferred to the hub hospital, where they could receive further treatment with endovascular stroke therapy.

• Patients transferred from spokes to hub would not be initiated on intravenous thrombolysis in the hub hospital.

• Intravenous thrombolysis had no impact on in-hospital mortality.

• Both hub and spoke hospitals had the capacity to administer intravenous thrombolysis or to treat additional AIS patients without requiring additional resources (such as additional beds, new stroke specialists) except for the resources required for telemedicine.

Model Inputs

Model inputs included network characteristics, cost inputs, and effectiveness inputs.

Network Characteristics

In the base case, the network was assumed to consist of 1 hub hospital and 7 spoke hospitals, the average number of hospitals in the US telestroke networks.¹² Other network parameters, including number of AIS patients, rates of teleconsultations, spoke-to-hub transfers, intravenous thrombolysis, and endovascular stroke therapies, were obtained from unpublished data from the Georgia Health Sciences University and the Mayo Clinic telestroke networks. The mean values were used in the base case (Table 1). Additional details on the characteristics of these 2 networks are given in Table I in the online-only Data Supplement.

Cost Inputs

Cost inputs, consisting of telestroke setup and maintenance costs, AIS treatment costs, and reimbursements, were obtained from literature and publicly available data. Because the model assumed that all hospitals had the capacity to treat AIS patients without additional resources, marginal costs, which accounted for 31% of the total costs (the proportion of nonlabor costs), were used in the model inputs.¹⁴ All costs were adjusted to 2011 US dollars with the Consumer Price Index for medical services (Table 2).^15–26

Effectiveness Inputs

Effectiveness inputs included discharge dispositions for each patient group defined by treatment with intravenous thrombolysis, endovascular stroke therapy, and onset-to-treatment time (Table 3). Because these data were not directly available from the literature, these inputs were estimated from existing clinical trials.

Discharge dispositions associated with intravenous thrombolysis treatment were estimated from the following data: the association between modified Rankin scale (mRS) outcomes at 3 months and discharge destination from the National Institute of Neurological Disorders and Stroke (NINDS) trial³ and mRS at 3 months for different stroke onset-to-treatment time categories (ie, 0–90, 91–180, and 181–270 minutes) from a meta-analysis.²⁷ In-hospital mortality, estimated to be 10.3%, was obtained from a study using the National Inpatient Sample²⁸ and applied to patients with and without intravenous thrombolysis. The rates of home and rehabilitation/nursing home discharges were then redistributed on the basis of this mortality rate (Table II in the online-only Data Supplement). Percentages of patients with mRS score of 0 to 1 at 3 months for each onset-to-treatment time category among those with and without intravenous thrombolysis were obtained from a meta-analysis (Table III in the online-only Data Supplement).²⁷ These numbers were then used to estimate the percentages of patients discharged to home by use of the ratio between the derived rate of home discharges and the rate of mRS score of 0 to 1 at 3 months in each treatment group from the NINDS trial.³

Similarly, discharge dispositions associated with endovascular stroke therapies were estimated. mRS scores of 0 to 2 at 3 months among patients with endovascular stroke therapy were obtained from the Mechanical Embolus Removal in Cerebral Ischemia (Multi-MERCI) trial.²⁹ Because the Multi-MERCI trial did not have a control arm, mRS score of 0 to 2 at 3 months among patients without endovascular stroke therapy were estimated from the Multi-MERCI trial and the ratio between the active and control arms in the Prolyse in Acute Cerebral Thromboembolism II (PROACT II) trial,³⁰ the only trial reporting the relative efficacy of an intra-arterial intervention versus non-intra-arterial intervention. mRS score of 6 (death) at 3 months among patients with endovascular stroke therapy was obtained from the Multi-MERCI trial and assumed to be the same for patients without endovascular stroke therapy in the base case. In the sensitivity analysis, an alternative method was also used to estimate mRS at 3 months. Specifically, mRS score at 3 months was also estimated from the recanalization rates for endovascular stroke therapy reported in the Multi-MERCI trial and the recanalization rates for nonendovascular stroke therapy estimated with the methods described by Patil et al.³¹

Model Outputs

Model outputs included total incremental costs and incremental effectiveness as measured by incremental numbers of home discharges, inpatient rehabilitation/nursing home discharges, and in-hospital deaths. All outputs were estimated from each of the 3 perspectives. Costs were discounted at 3% annually.

Incremental effectiveness from a spoke hospital perspective was estimated based on the final discharge destinations of patients presenting to the emergency department in that hospital, regardless of whether they were transferred to the hub. Incremental effectiveness from the hub hospital perspective was estimated only among patients transferred from spoke hospitals because those presenting to the emergency department in the hub hospital would receive the same treatments with and without a network. Incremental effectiveness from the network perspective was estimated among AIS patients from all spoke hospitals in the network.

As intermediate outcomes, the incremental numbers of AIS patients treated with intravenous thrombolysis and endovascular stroke therapy and admitted to each hospital were also reported to demonstrate the impact of a telestroke network on AIS care in different hospitals.

Sensitivity Analyses

Extensive 1-way sensitivity analyses were performed by varying 1 model input at a time while holding other model inputs at the base-case value. Model inputs included in the sensitivity analysis were (1) network characteristics, including increasing the number of spoke hospitals up to 40; (2) setup and maintenance costs of telestroke systems; (3) marginal costs for treating AIS at the hub or spoke; and (4) discharge dispositions associated with endovascular stroke therapy. The majority of inputs were varied within ±25% of the base-case value. Rates of intravenous thrombolysis and endovascular stroke therapy without the network were assumed to be 10% and 50% of the network values, respectively. Marginal costs for treating AIS were set at 25%, 50%, and 75% of the total costs, respectively. Two-way sensitivity analyses were performed by varying both spoke-to-hub transfer rate (from 0%–100%) and endovascular stroke therapy rate among transferred patients (at 25% and 50% of the base-case value).

Results

The base-case analysis represented a network with 1 hub and 7 spoke hospitals, with a total of 1112 unique AIS patients presenting to the emergency departments in the network hospitals per year. With the telestroke network in place, the model predicted that about 114 fewer AIS patients would be admitted to the hub hospital each year, whereas approximately 16 more patients would be admitted to each spoke hospital compared with a no network setting (Table 4). The model predicted that about 45 more patients would be expected to be treated with intravenous thrombolysis and 20 more with endovascular stroke therapy in a telestroke network per year. From the network perspective (ie, hub and spoke hospitals taken together), an estimated average cost saving of $358 435 per year could be achieved with a telestroke network versus without a network during the first 5 years. The cost savings estimated by the model increased over time for patients treated during the first year, from $234 836 at the end of 1 year to $393 712 at the end of 5 years. The hub would bear positive costs of $405 121 per year, but each spoke would save $109 080 per year. With cost-sharing arrangements between the hub and spoke hospitals, this analysis suggests that each hospital could achieve equal cost savings of $44 804 per year during a 5-year time horizon.

Compared with a no network setting, the model estimated that 6.1 more patients would be discharged home per year. Because the model assumed that intravenous thrombolysis did not affect mortality, the same number of decreases in discharges to rehabilitation facilities or nursing homes was expected in the telestroke network (Table 4). From the perspectives of the hub and each spoke, 4.6 and 0.9 additional patients, respectively, were expected to be discharged home each year.

The 1-way sensitivity analysis showed that the results were robust overall. The network achieved cost savings in the model in all scenarios except when the number of spoke hospitals was reduced to 1 (incremental costs of $24 202 per year) and the marginal treatment costs for each spoke hospital were assumed to be 75% of the total treatment costs (incremental costs of $104 534 per year). The cost savings to the network were estimated to improve with increasing numbers of spokes. When the number of spoke hospitals in each network was varied from 0 up to 40, the incremental costs per year over a 5-year time horizon ranged from $87 974 (0 spokes) to a network savings of about $2 400 000 (Figure 2). In all other model scenarios, the estimated cost savings ranged from $159 718 to $1 359 500 per year from the network perspective, and the estimated incremental effectiveness ranged from 3.83 to 7.63 additional home discharges per year. With the use of the alternative method of estimating discharge dispositions associated with endovascular stroke therapy, about 0.98 in-hospital deaths could be avoided per year in the network. When transfer rate increased from 0% to 100%, the estimated network cost savings were reduced from $555 818 to $5588 per year; incremental costs decreased substantially for the hub hospital but increased for each spoke hospital (Figure 3). The 2-way sensitivity analysis predicted that with 50% and 75% reductions in the rate of endovascular stroke therapy, the network could achieve cost savings if the spoke-to-hub transfer rate was <62% and <43%, respectively.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Discussion

The results of this study suggest that a telestroke network may be an effective way to extend the reach of stroke specialists to remote areas and thus to improve the overall quality of care for AIS patients. Despite strong evidence to support its applications in acute stroke treatment,³² no studies have been conducted to help hospital administrators evaluate the impact of such interventions on their hospitals or hospital systems. Our modeling estimated that compared with no network, a telestroke system of 1 hub and 7 spoke hospitals may result in more intravenous thrombolysis, more endovascular stroke therapies, more stroke patients discharged to home, and despite upfront and maintenance expenses, greater cost savings for the entire network. Under the base-case scenario, the model predicted that a minimum of 2 spoke hospitals is required for the network to break even in terms of cost; however, with more spokes, the network was estimated to achieve additional cost savings. Moreover, the telestroke network strategy remained cost-effective across a wide range of uncertainties in model inputs.

Hospital costs were particularly sensitive to transfer rates from spoke hospitals to the hub. Telestroke remains cost-effective across transfer rates from 0% to 100% in the base-case model. However, as the transfer rate increases, the model suggests that the cost savings for the network and individual spokes diminish, whereas it increases for the hub hospital. From the spoke hospital perspective, there is a financial incentive to keep patients and receive the higher diagnosis-related group reserved for ischemic stroke with thrombolysis. Conversely, hub hospitals collect greater inpatient reimbursement for the additional endovascular stroke therapies performed on transfer patients. Simply put, the cost savings from each hospital’s perspective (hub or spoke) is greatest when the patient is discharged from their hospital. To optimize cost-effectiveness across the network, transfers should be limited to patients who are candidates for endovascular stroke therapy, hemicraniectomy, or other surgical or endovascular surgical interventions. In reality, transfers between hospitals occur for multiple reasons, reflecting patients’ health needs, available resources, and the expertise of hospitals. Small, rural hospitals often lack neurologists, interventionalists, neurosurgeons, and intensivists and may be incapable of managing complex stroke cases.

Contrary to the common perception that a telestroke referral network poses a substantial financial burden on hospitals, our study revealed that it is likely to be a cost-saving strategy from the hospitals’ perspectives while also improving patient outcomes in terms of home discharges. To improve its generalizability, our model included the costs of maintaining round-the-clock coverage with stroke experts from the hub. Which hospitals, then, should bear the cost burden of a telestroke network? The results of this economic research have implications on the assignment of financial responsibility between hub-and-spoke partners. In a network that is fundamentally designed to transfer patients from spoke to hub, the hub hospital should subsidize the costs of the telestroke, whereas in a network designed predominantly to aid spoke hospitals’ capability to effectively maintain more AIS patients, the spoke hospitals should finance the system. In the base-case model, targeting a spoke to hub transfer rate of ≈30% resulted in an economic benefit for the hub, the spokes, and the telestroke network overall.

To the best of our knowledge, our study is the first to evaluate the cost-effectiveness of a hub-and-spoke telestroke network from hospitals’ perspectives. Previous studies have shown that telestroke networks are cost-effective for AIS from the societal perspective in the long term. Nelson et al estimated that a telestroke network had an incremental cost-effectiveness ratio of $2400 per quality-adjusted life-year compared with no network over a lifetime horizon.¹³ Both our study and Nelson et al used data inputs from Arizona telestroke networks. However, although Nelson et al combined data from the Stroke Telemedicine for Arizona Rural Residents network with those of the University of Utah network, our study combined data from the Mayo Clinic Telestroke network in Arizona (independent from Stroke Telemedicine for Arizona Rural Residents) with those of the Georgia Health Sciences University network. In both analyses, the number of hospitals in the telestroke system base-case scenario was also similar (8 spokes in the study by Nelson et al and 7 spokes in the present study). However, there were significant differences in study design. Our analysis focused on cost-effectiveness from the hospital perspective, whereas the analysis by Nelson et al was from the societal perspective. Furthermore, our decision model included the possibility of endovascular therapy in addition to intravenous thrombolysis. Cost inputs in our study included those associated with dedicated network program managers and personnel; higher estimates of inpatient care, interhospital transfer, rehabilitation, long-term care, and caregiver costs; and a wider range of spoke-to-hub transfer rates. The costs in the Nelson et al study were converted to 2008 US dollars, whereas ours were study converted to 2011 US dollars. Another study in Denmark found that a telestroke system was associated with cost savings and greater effectiveness from 2 to 30 years. Ehlers et al³³ analyzed the use of telestroke from a national perspective within the context of a state-run universal healthcare system in which 5 regional stroke hubs would be linked to a single spoke. Similar to the Nelson et al study, this article examined telestroke from a societal viewpoint, comparing the long-term costs of stroke patients (both inpatient hospitalization and postdischarge health care) with improvements in quality-adjusted life-years. In contrast, this analysis focuses on the initial prehospital and in-hospital costs, and outcome was assessed in terms of discharge status. Taken together, these studies suggest that telestroke networks can be a cost-effective strategy and should be considered by decision makers to improve the quality of stroke care.

Our results should be assessed in the context of the limitations and assumptions of our model. First, the model assumed that the benefits of telestroke were limited to the increased use of intravenous thrombolysis and endovascular stroke therapies in AIS patients. Previous studies have shown that patients would also benefit from improved subacute care and rehabilitation through telestroke systems.³² Additionally, a telemedicine system can improve quality of care for patients with other types of stroke or even other diseases. If these benefits are taken into account, a telestroke network may be more cost-effective. Second, we assumed that hubs and spokes have the personnel and resources to manage additional stroke patients who could arise from network development. When the marginal costs to care for additional stroke patients are increased by 75% of the total treatment costs at spoke hospitals, the network is estimated to have positive costs, and there was an estimated incremental cost of $17 117 per discharge home from the network perspective. Third, the study is limited by the data that can be obtained in the existing literature and public domains. For example, without randomized, controlled trials, assumptions were made in the estimation of the effectiveness of endovascular stroke therapies. Importantly, however, telestroke remained cost-effective in the model even with significant reductions in the number of endovascular stroke therapies performed at the hub, as long as the transfer rate was kept low. For example, with a 50% reduction in the rate of endovascular stroke therapies, the model suggests that a network would maintain cost savings if the transfer rate was <62%; with a 75% reduction in the rate of endovascular stroke therapies, the network would maintain cost savings as long as the transfer rate was <43%. In addition, discharge destinations were not directly reported in the clinical trials and thus had to be derived from multiple data sources. Given these uncertainties, conservative assumptions were made for the base case. Fourth, most model inputs related to the network characteristics were estimated from 2 telestroke networks, which may not be representative of the broad range of network models. Finally, we did not consider the cost-effectiveness of a telephone-only link from spoke to hub. Although this strategy may be used to facilitate intavenous thrombolysis administration for stroke,³⁴ compared with telestroke, the correct treatment decision for thrombolysis is made significantly less frequently.^9,35 Furthermore, when considering whether to transfer a patient from a spoke hospital for endovascular stroke therapy, visual inspection of the patient to aid in quantification of the National Institutes of Health Stroke Scale and review of the baseline neuroimaging may be invaluable but cannot be achieved in a telephone-only network.^36,37

Conclusion

This study suggests that a telestroke network may increase the number of patients discharged home and reduce the costs borne by the network hospitals. The economic implications of a telestroke network vary for individual participating hospitals, depending on available resources and spoke-to-hub transfer rate, and should be considered when a network is being developed.

Source of Funding

This study received funding from Genentech, Inc.

Acknowledgments

The authors had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Footnotes