The present analysis reveals short-protocol MRI as a cost-effective strategy in emergency patients with suspected intracranial pathology and negative non-contrast head CT presenting with inconclusive neurological symptoms, being dominant even from an economic perspective. In this setting, the cost-savings due to a lower rate of major strokes and gains in effectiveness by application of preventive treatment may by far outweigh the additional costs of supplemental short-protocol MRI subsequent to negative head CT in the emergency situation.
Major strokes are preceded by minor strokes or transient ischemic attacks (TIA) in 15–30%, and about 40% of recurrent major strokes occur within 7 days and about 20% within 24 h after the initial minor stroke or TIA [6, 7, 30]. Urgent initiation of secondary prophylactic treatment after a minor stroke or TIA was demonstrated to prevent 80 to 90% of recurrent major strokes [9,10,11]. The importance of a rapid treatment start has been demonstrated in detail in the EXPRESS study with a rate of recurrent stroke of 2.1% in patients receiving treatment within 1 day of the index event compared to a rate 10.3% in patients receiving treatment within 3 days of the index event [11]. These data clearly point to the pivotal role of an early detection of minor strokes, a current diagnostic gap that may be closed with MRI in the acute setting [14,15,16].
Major stroke is of utmost clinical and economical importance, as it is associated with a very high economic burden for healthcare systems. About 40% of all stroke patients suffer moderate to severe disabilities and need special care; about 10% depend on long-term care facilities [4]. Acute stroke care accounts for about one half of total direct medical costs within the first 12 months following ischemic stroke, and post-stroke care significantly contributes to the large expenses. Rehabilitation services and facilities represent the most expensive factors [4, 5]. In the acute setting, the length of stay in hospital is the most important cost driver [35]. Minor stroke or TIA patients have a significantly shorter length of stay in hospital, less patients have to be admitted to the intensive care unit, and less patients are discharged to skilled nursing facilities with consecutive high expenses [3, 5].
In this context, our results indicate that additional investment in short-protocol MRI examinations in the emergency setting does not only increase the detection rate of minor strokes, but can be regarded as a highly cost-effective, dominant strategy by preventing major strokes with consecutively high costs. In our analysis, we found overall higher costs for patients without additional short-protocol MRI compared to patients undergoing additional MRI. Further, the cumulative calculated effectiveness in the CT-only group was lower compared to the additional short-protocol MRI group. In our deterministic sensitivity analysis, additional short-protocol MRI remained the dominant strategy in the ranges investigated, and even when assuming a relatively high cost of the ultrafast MRI of $500, it remained the cost-effective strategy. Transition probabilities, detection rates, and quality of life after a major stroke had a higher influence on the cost-effectiveness of additional short-protocol MRI than costs and sensitivity of MRI itself.
Our study adds to the field as it demonstrated the cost-effectiveness of supplemental short-protocol MRI subsequent to negative head CT in neurological emergency patients. In many related fields of research, the additional utilization of procedures and diagnostic tests has been shown to be cost-effective: Despite high costs in the acute setting, interventional treatment has been shown to be cost-effective in major stroke patients, even in patients with a limited overall life expectancy [26, 36]. Furthermore, a transition from a time window to a tissue window for stroke patients presenting with unknown onset time for endovascular treatment and intravenous thrombolysis has taken place [37,38,39]. In this context, higher costs of MRI have been found to be cost-effective for major stroke patients with unknown onset time in order to allow treatment beyond traditional time windows in selected patients with salvageable brain tissue, thereby improving functional outcome after major stroke [40]. The cost-effectiveness of secondary prophylaxis, e.g., with antiplatelet therapy, has also been demonstrated in previous studies [41,42,43].
In future studies and cost-effectiveness analyses, initial imaging with short-protocol or standard-length MRI instead of CT for selected patients should be investigated, as this may lead to an optimized diagnosis and treatment planning in patients with minor stroke symptoms.
Limitations
Yet, results of this cost-effectiveness evaluation have to be interpreted in the context of certain limitations:
First, input parameters are key factors for the modeling of cost-effectiveness evaluations. Still, these input parameters are derived from published literature and their data basis can vary between input variables. Further, the yearly detection rate of minor strokes or TIA had to be assumed since data on this parameter were not available and the number of patients with mild and unspecific symptoms and missed minor strokes discharged from the emergency department without secondary prophylaxis is not known and therefore had to be estimated. On the other hand, deterministic sensitivity analysis with wide ranges for input parameters allowed to model a variety of scenarios with all of them remaining cost-effective for additional short-protocol MRI in patients with neurological symptoms and inconclusive non-contrast head CT.
Second, every cost-effectiveness modeling has limitations due to the model structure. Although the Markov model applied takes into account several clinical situations, there may be special cases not fully reflected in the model. However, due to a tradeoff between model complexity and availability of (published) data, the authors believe that the current modeling approach may represent an acceptable middle ground in this respect.
Third, post-stroke care costs used by Earnshaw et al and Kunz et al are based on the stroke treatment economic model from 1996 [25, 26]. However, detailed and more recent data on post-stroke care costs from the USA for this particular case were not available. Acute stroke care costs increased disproportionately due to novel medical and mechanical treatment options. For these costs, more recent data were available.
Fourth, the rate of minor strokes in our cohort in selected emergency patients presenting with inconclusive neurological symptoms was based on study results from a prospective single-center diagnostic accuracy study. Minor stroke rates might differ slightly in large-scale multi-centric studies in this patient subpopulation. However, short-protocol MRI remained the dominant strategy in wide ranges for this input parameter assessed with our deterministic sensitivity analysis.
Fifth, our analysis was based on the US healthcare system. Cost-effectiveness might differ substantially between countries, and our results may not be transposed to other health care systems without adjustments and modifications.