Modelling the Efficiency of Local Versus Central Provision of Intravenous Thrombolysis After Acute Ischemic Stroke
Background and Purpose—Prehospital redirection of stroke patients to a regional center is used as a strategy to maximize the provision of intravenous thrombolysis. We developed a model to quantify the benefit of redirection away from local services that were already providing thrombolysis.
Methods—A microsimulation using hospital and ambulance data from consecutive emergency admissions to 10 local acute stroke units estimated the effect of redirection to 2 regional neuroscience centers. Modeled outcomes reflected additional journey time and accuracy of stroke identification in the prehospital phase, and the relative efficiency of patient selection and door-needle time for each local site compared with the nearest regional neuroscience center.
Results—Thrombolysis was received by 223/1884 emergency admissions. Based on observed site performance, 68 additional patients would have been treated after theoretical redirection of 1269 true positive cases and 363 stroke mimics to the neuroscience center. Over 5 years redirection of this cohort generated 12.6 quality-adjusted life years at a marginal cost of £6730 ($10 320, €8347). The average additional cost of a quality-adjusted life year gain was £534 ($819, €673).
Conclusions—Under these specific circumstances, redirection would have improved outcomes from thrombolysis at little additional cost.
Intravenous thrombolysis reduces the dependency and economic burden resulting from stroke.1,2 Prehospital redirection of patients to regional centers is an effective strategy for treatment provision, but there is no consensus on the circumstances that favor central versus local service configurations, particularly after the development of remote specialist assessment.3 Economic aspects of thrombolytic therapy have been examined4,5 but not in the context of the underlying service model. In a region where thrombolysis service provision was already established at local units, we used a prospective cohort to model the health outcome change and marginal costs resulting from theoretical prehospital redirection by emergency ambulance to two existing neuroscience centers (NC).
North East England is an area of 3300 square miles with a population of 2.6 million served by 12 acute stroke units (ASU), including 2 NC housing neurosurgical and neuroradiological teams in the northern and southern most cities. The mean distance from ASU to the nearest NC is 14.1 miles, range 5 to 22 (22.7 km, range 8–35). All sites offer thrombolysis 24 hrs without intentional prehospital redirection across organizational boundaries. Most patients are reviewed initially in emergency departments (ED) with secondary referral to stroke teams, although both NC offer direct admission to ASU during office hours according to agreed ambulance service protocols. Patients who do not have a diagnosis of stroke after initial assessment by the stroke team are referred back to the ED with advice, or transferred to internal medicine if symptoms persist. For 8 sites the out of hours specialist assessment is performed remotely. Collaboration between sites to ensure continuity of stroke specialist availability has created 6 independent thrombolysis services, each consisting of 1 to 3 ASU.
A prospective regional database recorded details of prehospital assessment, thrombolysis treatment, discharge outcome, and resource use for consecutive stroke admissions between October 1, 2010, and September 30, 2011. Diagnosis was confirmed by a stroke specialist within 24 hours of admission and by discharge ICD10 code (G46, I61, I62, I63, or I64). Records from the single regional ambulance provider were searched for confirmation of emergency ambulance status, Face Arm Speech Test (FAST),6 incident postcode, and journey time to the admitting ASU. Local data protection permission was granted by each National Health Service (NHS) organization. Patient identifiable data were not exported. Ethical approval was not required.
Health Outcome Model
The model was a simulation of thrombolysis status and onset to treatment time (OTT) for patients who arrived by emergency ambulance. The primary health outcome was independence at discharge, defined as modified Rankin Scale (mRS) 0–2.7 Geographic Information System (GIS) software calculated a new travel time from each incident location to the nearest NC according to the shortest journey duration. Thrombolytic treatment was considered possible if the combined hospital arrival time and observed median door to needle time (DTN) for the NC was <4.5 hours since stroke onset.1 Patients admitted to NC as their local unit did not change their thrombolysis status in the model.
The probability of a person remaining independent at discharge (ie, mRS 0–2) was calculated according to treatment status, OTT if thrombolysis given, age, and sex using results from the treatment and control arms of pooled thrombolysis trials8,9 calibrated by the UK Safe Implementation of Thrombolysis in Stroke (SITS) database.10 The model prioritized which redirected admissions would receive thrombolysis according to characteristics favoring a better outcome in SITS-UK (ie, shorter OTT, younger age, lower stroke severity, greater prestroke independence, and female sex). When the NC treatment rate was achieved, no further redirected patients were treated. The resulting output reflected a theoretical best case scenario within the observed NC treatment rate. Because of an absence of reference data, health status at discharge was assumed to be the same as 3 months after stroke. When it was not possible to locate an ambulance service record, a FAST result was imputed based on regression analysis of demographic associations within the available data. It was assumed that all FAST true positive cases were redirected and all FAST false-negative cases were admitted to a local site where thrombolysis assessment would be unavailable. The Figure summarizes the modeling process.
The economic model was a microsimulation, using estimated probabilities for changed dependency status and related costs in the 2 NC configuration compared with observed outcomes for each patient. Redirection resource implications were prehospital (extra ambulance journeys to and from NC), within hospital (NC and local bed days, thrombolysis treatment), and postdischarge (institutional care need, home carer visits). For redirected patients, a back-ward step wise regression model using predictor variables from the regional database estimated changes in length of stay (6.640*probability of mRS<3), institutionalization (ey/(1+ey) where y=−3.35+0.032*age −1.032*probability of mRS>2) and home carer visits (ey/(1+ey) where y=4.612*probability of mRS>2+1.3*total anterior circulation stroke + 0.034*age + 0.694*thrombolysis treatment).
To estimate the number of FAST false-positive stroke mimics who may have been redirected, we applied a true:false FAST positive ratio of 3.5:1 derived from descriptions of FAST validation within the same geographical region.6,11 The additional cost for each stroke mimic admission was set at 1 additional bed day plus the return journey by nonurgent hospital transport. An additional journey cost was also included for any stroke patient requiring repatriation to a local ASU because their length of stay was >48 hours. All costs (Table 1) were inflated to 2010/11 prices using the Hospital and Community Health Services Index.12
A Markov component was used to estimate the impact of service reorganization on quality-adjusted life years (QALY) over 5 years. The model assumed utility values of 0.85 and 0.27 for independence and dependence, respectively.14 All-cause mortality rates were taken from 2011 Life Tables for the United Kingdom.15 A 10% increased risk of death was assumed for the dependent population, and the probability of independence becoming dependent for any reason was estimated at 15% p.a.16
Details were collected for 1884 patients (mean age 76 years; 47% male) admitted by emergency ambulance with a confirmed acute stroke diagnosis including 223 (12%) who received intravenous thrombolysis. The mean number of consecutive months contributing data by regional ASU was 10.8 (range, 7–12). Table 2 shows the contribution per thrombolysis service and ASU. Table 3 shows the results of applying the model to the 1547 patients who were admitted to ASU other than the NC. Assuming that all were redirected to the nearest NC, there were 1238 additional admissions to NC1 and 309 to NC2, reflecting the geographical distribution of sites and lower data contribution from ASUs nearer to NC2. The median journey time was 17 minutes longer, but because of relative differences in treatment rates and DTN times between local ASU and NC, a predicted additional 97 patients received thrombolysis (a relative increase of 54% above reported activity). The net change in thrombolysis activity was not uniformly distributed across the region because of the difference in treatment rates between the 2 NC and variable efficiency across ASUs (ie, redirection to NC1 resulted in a net gain of 103 treatments, redirection to NC2 resulted in 6 fewer treatments). The total extra ambulance journey time was 440 hours but 59 hours between onset and treatment were saved for thrombolysis patients because of quicker DTN times at the NC.
In the whole cohort, 882 patients were documented as FAST positive, and 124 were documented as FAST negative. FAST status was imputed for 878 because of failure of documentation or missing paramedic records using a backwards step-wise logistic regression that showed statistically significant associations between FAST false-negative status, increasing age (coefficient 0.025 [SE 0.01] per year; P<0.01), and male sex (coefficient 0.4 [SE 0.2]; P<0.05). As a consequence, 387 patients outside of NC had a FAST positive status giving a total of 1269/1547 patients who had an observed or imputed FAST positive status (ie, a sensitivity of 82%). We estimated that 363 patients with symptoms mimicking a stroke would be redirected. Table 3 shows the additional impact of excluding from the model those 278 patients who would still have been admitted locally because they were incorrectly judged to be FAST negative. Twenty-nine fewer patients were treated, but the net increase of 68 treatments was 37% above observed activity, reflecting a regional gain of 7 good outcomes.
Resource Usage and QALY
Table 4 describes the modeled change in costs per 100 stroke admissions before and after adjustment by FAST status according to the range of treatment benefits associated with least, average, and most favorable outcome characteristics within the UK SITS register. A negative value indicates money saved. Stroke mimics were responsible for 442 and 363 additional bed days before and after FAST sensitivity adjustment, which was offset by an average of 77 bed days saved by each good thrombolysis outcome. After subtraction of the predicted social care saving from the expenditure of thrombolysis treatments and extra ambulance journeys (redirection and repatriation if required), the pre- and post-FAST annual service cost ranged from £11 266 ($17 275, €13 970) to £58 121 ($81 923, €72 070) and £3222 ($4941, €3995) to £50 962 ($78 145, €63 193) for 97 and 68 additional thrombolysis treatments, respectively. Each additional good outcome had a mean value of £2647 ($4059, €3282) without FAST adjustment and £2584 ($3962, €3204) after FAST adjustment.
Over 5 years, assuming that relevant costs remained constant and resources proportional to activity were reallocated from local ASU to NCs, redirection generated a median of 12.6 QALYs at a cost of £6730 ($10 320, €8345; ie, the cost of a single QALY gain was £534 [$819, €662]).
To examine assumptions in the modeling process we performed additional sensitivity analyses. The distribution of the missing FAST data were not associated with any demographic or clinical characteristics within the observed cohort. Without redirection, no difference in predicted treatment outcome was found when patients with an observed or imputed FAST positive status were selected preferentially for thrombolysis treatment at their local site (ie, FAST imputation did not dominate other characteristics of the group selected for thrombolysis such as onset to treatment time). For men and women separately the effect of age on imputed FAST status was identical and exclusion of sex as a cofactor reduced the number of patients receiving thrombolysis at the hub by 10%. After arrival at NC, the model assumed that clinical teams would prioritize patients likely to have the greatest benefit from thrombolysis until the observed treatment rate was achieved. If patients were selected according to 75th, 50th, and 25th centile levels of benefit there was an equivalent reduction of 1.4, 2.3, and 3.0 independent people, respectively. Table 5 shows additional analysis for QALY over 5 years according to the interquartile distribution of clinical benefits from thrombolysis within the UK SITS database.
In a region with variable provision of intravenous thrombolysis at local ASU, a model reflecting changes in travel time to 2 NC, differential treatment efficiency between units, and prehospital identification performance estimated that the number of emergency ambulance patients treated could be increased by more than one third. The average incremental cost per QALY over 5 years was well below commonly accepted thresholds for society’s willingness to pay.17 However, the overall benefit was modest, as many patients were never eligible for thrombolysis or moved between units with similar efficiency. Redirection model performance might be improved by focusing on the movement of patients between the least and most efficient units only, selective prehospital redirection according to characteristics which increase the likelihood of thrombolytic therapy, or considering the additional impact of other reperfusion interventions.
As far as we are aware, this has been the first attempt to quantify the gain of transferring thrombolysis provision between active sites based on individual patient characteristics and observed prehospital stroke identification rates. It is important to recognize that only specific marginal costs were considered, having assumed that the existing resources to provide care for up to 48 hours were redistributed from local ASU to NC. This movement of existing resources would be the dominant financial consequence from introduction of a redirection service. As pressure on healthcare resources grows, this modeling approach will become more important for optimizing provision of existing treatments.18 Future studies should also consider economy of scale as a single large site may require fewer resources to process the same number of admissions as multiple small sites.
Although settings and methodologies vary, previous economic analyses of intravenous thrombolysis report that treatment costs are offset by longer term care savings. Using data from the National Institute of Neurological Diseases and Stroke (NINDS) recombinant tissue plasminogen activator Stroke Trial and considering typical treatment, hospitalization, and social care costs, Fagan modeled a saving of $600 per treated patient at 1 year but with CI from −$3481 to $2004 USD:1996.19 Also using NINDS data, Sinclair reported a savings estimate of $3800 CAD:1999 per tPA-treated patient over a lifetime (maximum 30 years).20 These models produced absolute estimates based on trial data for patients treated <3 hours without considering relative service performance differences or the impact of prehospital identification. By applying pooled trial data for treatment up to 6 hours to a detailed model using an NHS perspective, Sandercock suggested that compared with standard care, there was a 78% probability of a gain in quality-adjusted survival during the first year with an incremental cost per additional QALY gained of £13 581 (GBP:price year not stated).5 Although there were no costs recovered from treatment during the first 2 years, this delay was likely to reflect more conservative estimates for the effectiveness of thrombolysis and patient valuation of dependency. Our simpler model suggests that social care savings occur early and that over 5 years the QALY gain would be an affordable addition to existing service costs.
It is important to recognize that our model output mainly reflected relative differences in thrombolysis service performance and other combinations of sites could be considered which are equally or more favorable, although not necessarily practical. In this specific region additional prehospital journey time did not compromise redirection, whereas for other settings remote assessment might be the only feasible mechanism for an efficient service.21 The FAST prehospital identification rate was consistent with previous descriptions.6,11 In addition to FAST false-negative cases, a further 20% to 30% of stroke patients arrive by nonemergency ambulance and private transport. Depending on the quality of local care, regional geography, and volume of patients at each site, a decision would be required whether such patients should still be redirected or admitted to a smaller scale local ASU with an option for 24-hour remote specialist review from a regional center.
In the model we did not distinguish between health states other than mRS 0 to 2 or above, and assumed that the rate of loss of independence remained constant during subsequent years. The use of median DTN at each NC to determine whether redirected patients arrived with a potential OTT <4.5 hours assumed that there was no variation in DTN because of service reconfiguration or patient characteristics (eg, the NC would not provide a quicker or slower DTN in response to the shift toward a larger number of patients arriving later from further away). Quantifying such a parameter would require an observational study following service reconfiguration. We did not capture original data for a full year of emergency admissions, but all sites contributed for >6 months. As the economic analysis compared a relative change within a single cohort, the impact of data capture bias was partially neutralized. The collection of ambulance service data were particularly challenging, but we are not aware of any previous linkage of individual prehospital and hospital records for stroke patients on this scale.
Although many patients would not experience a change in health outcome, a model based on the observed provision of intravenous thrombolysis showed an increase in good outcomes at little additional cost after redirection of stroke patients from 10 local ASU to 2 regional centers.
We acknowledge the audit facilitators, clinical stroke teams, and the North of England Cardiovascular Network for their support during data collection.
Sources of Funding
This article presents independent research commissioned by the National Institute for Health Research (NIHR) under its Program Grants for Applied Research scheme (RP-PG-0606-1241). The views expressed in this publication are those of the author(s) and not necessarily those of the National Health Service, the NIHR, or the Department of Health.
Dr Ford’s institution has received research grants from Boehringer Ingelheim (manufacturer of Alteplase) and honoraria from Lundbeck for stroke-related activities. Dr Ford has also received personal remuneration for educational and advisory work from Boehringer Ingelheim and Lundbeck. The other authors report no conflicts.
- Received March 7, 2013.
- Accepted July 26, 2013.
- © 2013 American Heart Association, Inc.
- Wardlaw JM,
- Murray V,
- Berge E,
- del Zoppo GJ
- 2.↵National Audit Office. Reducing Brain Damage: Faster Access to Better Stroke Care. Report by the Comptroller and Auditor General, Department of Health. London: National Audit Office;2005.
- 4.↵NICE. Technology Appraisal 122, 2007. National Institute for Health and Clinical Excellence. Ischaemic Stroke (acute) - alteplase. http://guidance.nice.org.uk/TA122. Accessed November 23, 2012.
- Sandercock P,
- Berge E,
- Dennis M,
- Forbes J,
- Hand P,
- Kwan J,
- et al
- Harbison J,
- Hossain O,
- Jenkinson D,
- Davis J,
- Louw SJ,
- Ford GA
- Bonita R,
- Beaglehole R
- Kent DM,
- Selker HP,
- Ruthazer R,
- Bluhmki E,
- Hacke W
- Flynn D,
- McMeekin P,
- Ford GA,
- Rodgers H,
- Thomson RG
- Lees KR,
- Ford GA,
- Muir KW,
- Ahmed N,
- Dyker AG,
- Atula S,
- et al
- Curtis L
- 13.National Audit Office (2010) Progress in Improving Stroke Care: Modelling Paper. http://www.nao.org.uk/publications/0910/stroke.aspx. Accessed November 23, 2012.
- Kim AS,
- Nguyen-Huynh M,
- Johnston SC
- 15.↵The Office of National Statistics. http://www.statistics.gov.uk/hub/population/deaths/mortality-rates/index.html. Accessed November 23, 2012.
- Slot KB,
- Berge E,
- Dorman P,
- Lewis S,
- Dennis M,
- Sandercock P
- 17.↵National Institute for Health and Clinical Excellence. Guide to the Methods of Technology Appraisal June 2008. ISBN: 1-84629-741-9.
- Mandelblatt J,
- Schechter C,
- Levy D,
- Zauber A,
- Chang Y,
- Etzioni R
- Audebert HJ,
- Kukla C,
- Vatankhah B,
- Gotzler B,
- Schenkel J,
- Hofer S,
- et al