Clinically, thrombectomy is intuitively attractive due to the demonstrated high rates of good clinical outcome with a low number needed to treat [5]. For the hospital, however, the first treatment is associated with an initial high cost to cover the thrombectomy procedure. This study looks at the immediate hospital cost associated with endovascular treatment, focusing on the first hospital episode and first 3-months post-event in patients with large vessel occlusion as compared to conservative medical therapy. Previous studies taking into consideration an NHS economic perspective were based on simulated Markov models with the data originating from different countries [27, 28]. To our knowledge, this is the only study to date analysing matched pair patient-level costs of thrombectomy in an NHS Trust for the admitting hospital as compared to conventional treatment. As there is level 1 evidence for endovascular stroke treatment in patients with LVO, primary randomisation was not appropriate on the basis of ethical concerns [29, 30].
Thrombectomy is known to be associated with a shift towards better outcomes across the entire spectrum of disability [5], which was also demonstrated in our patients, where clinical outcome was significantly improved in the thrombectomy group with a greater number of patients with functional independence (56% vs 8% p = 0.001), fewer serious adverse events (n = 30 vs 86) and fewer additional investigations (4.5 vs 8.9 per patient; p = 0.0053; n = 112 vs 213).
Thrombectomy patients could be discharged significantly earlier (mean 44 vs 80 days; p = 0.006) and were more likely to be discharged home (60% vs 28%). This is in line with Campbell et al. who reported on 70 patients from the EXTEND-IA study (35 in each arm, mean age 69, median NIHSS 15) where in the first 90 days thrombectomy patients spent significantly more time at home (median 73 days vs 15 days; p = 0.001). LoS in an acute stroke unit was reduced from mean 12 (control group) to 8 days (endovascular group), p = 0.04. Interestingly, there was no increase in intensive care time (p = 0.51) [31], whilst in our study we found a significant decrease in ITU/HDU time (mean 1.8 vs 6.9; p = 0.0306) if patients received thrombectomy. Mean LoS was significantly shorter both when including and excluding deaths.
In EXTEND-IA patients, rehabilitation LoS was reduced in the endovascular group (mean 33 vs 14 days), p = 0.03 [31]. The same was observed in our patient cohort. Patients in the control group required significantly more rehabilitation time (p = 0.0002) which accounted for £4982 of additional rehabilitation cost per patient. A recent publication by the Council of Deans for Health has highlighted that staff shortages in rehabilitation are putting health and social care services under pressure, with England currently facing one of its most profound and sustained workforce crises in decades [32].
Approximating 1599 total hours worked by an Allied Health Professional per year, this equates to one post made available to support the shortage of healthcare staff for every 13 patients managed with thrombectomy.
Several studies have demonstrated cost-effectiveness and, in many cases, cost savings with endovascular thrombectomy [33,34,35,36,37]. A recent study estimated costs for patients receiving endovascular therapy at different time points in an NHS setting but did not compare to a control group [38].
In the EXTEND-IA study, modelled life expectancy was calculated to increase by more than 4 years in the thrombectomy group with a significantly reduced loss of disability-adjusted life years and a clear gain of 4.4 quality-adjusted life years, translating to 90-day inpatient cost savings of US $14,880 [31]. Based on simulation modelling of 90-day mRS scores, Campbell et al. predicted a sustained and statistically significant mortality benefit up to 15 years post-treatment with associated benefits in DALYs lost and QALYs gained [31].
The THRACE trial described a probability of cost-effectiveness of additional thrombectomy treatment of 84.1% for cases with an averted disability and 92.2% regarding quality-adjusted life years. Additional costs per patient with averted disability were approximately 50% below the willingness to pay threshold [37]. Further analysis showed similar results with an overall cost-saving, even when considering the initially higher treatment costs [33, 36]. Previous studies have also reported on the long-term health economic benefit when patients with large vessel occlusion (LVO) are treated with thrombectomy [34, 35], demonstrating cost-effectiveness for all subgroups of patients undergoing mechanical thrombectomy, except for those with ASPECTS < 5 and M2 occlusion-where data has so far been scarce [39]. A recent meta-analysis based on 23 studies concluded that the addition of mechanical thrombectomy is cost-saving for a patient between 50 and 79 years and cost-effective for patients between 80 and 100 years [40]. Menon et al. found that even patients with proximal M2 segment middle cerebral artery (MCA) occlusion benefitted from mechanical thrombectomy [41].
In our study, cost savings to the admitting/treating hospital were calculated at £17,221 per patient by taking only the difference between care bed days, additional investigations and rehabilitation. Importantly, no reimbursement for devices or the thrombectomy procedure was added. Fixed costs as per NHS reference costs were used, based on the number of speciality bed hours, (ITU, High Dependency Unit, Hyperacute Stroke Unit) and other standard ward bed days [15]. As patients with similar number of bed days required different intensity of management, we captured imaging, nursing and allied health interventions which otherwise may lead to variation in true patient costs. Additional costs for treatment of serious adverse events such as pulmonary embolism, sepsis, pressure ulcer, urinary tract infection, decompressive surgery, nosocomial infections and their medication were not included as they are reflected in the LoS and subject to the local payment system (i.e.block contract).
In the thrombectomy group, only 8 patients (32%) had hospital stays exceeding the trim-point, resulting in an additional 255 bed days added to the inpatient cost, versus more than twice as many (n = 17; 68%) in the control group, adding 959 additional bed days. Benefits of a shorter LoS include patient well-being with a lower risk of hospital-acquired infections and an increase in hospital capacity for new admissions and increased availability of ITU beds.
A strength of this study is its analytical approach, minimising bias due to different covariates. If covariates differ between groups, the results of regression analysis alone can be misleading [8], which is why we used coarsened exact matching (CEM), neglecting outcome and, as a next step performed adjusted regression analysis to account for the remaining bias in co-variants, again neglecting outcome [8]. This two-step approach is reported to be less prone to model misspecification and even more robust than results based on the full, unmatched data set [9, 42, 43]. Another strength is that data was collected for each patient, capturing individual levels of required care, physiotherapy, speech, language and occupational therapy.
A limitation of our study is its retrospective analysis and the relatively small sample size which, nevertheless clearly demonstrated significance on a 95% confidence interval. Another limitation is that any bias of omitted covariates cannot be completely eliminated. Also, we did not perform a baseline severity-adjusted endpoint analysis as previously suggested by Saver et al. [44]. However, pre-mRS was matched, and there was no significant difference between groups.
We couldn't assess long-term community care costs or calculate QALY's to compare the costs with the willingness to pay threshold as outcome past 90 days was not assessed. However, given the significantly greater level of disability in the control group, the costs of care beyond 90 days is expected to remain greater than in the endovascular group.
The cost of providing secondary ambulance transfers in cases where patients needed to be transported from a general hospital to the thrombectomy-capable hospital was not included in the calculated as it did not affect the thrombectomy-capable hospital. It would, however, provide an additional argument for establishing more thrombectomy-capable hospitals.