The diagnosis and treatment of cerebral AVMs is an important component of modern neurosurgical care.¹⁸ Goals of care focus on prevention of hemorrhage, seizures, and neurological deficits associated with AVMs.¹⁴ The treatment of AVMs has increased in complexity with the development of multimodal approaches¹ that involve open microsurgical excision,⁴ endovascular therapy,²¹ and radiosurgery.^2,13 Each of these procedures has associated complications, with approximately 7.1% of surgical candidates, 6.6% of endovascular candidates, and 5.1% of radiosurgical candidates facing permanent neurological deficits after treatment.²⁰ With such a dynamic treatment approach wherein each modality is associated with significant complications, constant assessment of clinical outcomes is imperative in determining which treatment options afford consistently superior outcomes and under what circumstances.

Several AVM grading schemes currently in use have shown that certain patient-level characteristics correlate with outcome after intervention and are used to guide therapeutic decisions in clinical practice. The Spetzler-Martin grading system¹⁹ and Lawton supplementary grading system¹² together assess for the following 6 factors: size, location, and eloquence for the former and age, hemorrhage, and nidal compactness for the latter. Both have been validated for their utility in predicting surgical morbidity and have proven useful in clinical practice; however, these factors have not been assessed on a broader scale.

Given the variation in institutional demographics, assessing the comparative effectiveness of care delivery across institutions is imperative in ensuring that the highest quality of care is delivered consistently. To date, studies in the neurosurgical literature have almost exclusively examined the experiences at a single institution.²⁰ These results, while helpful in delineating some of the trends, are highly variable from institution to institution and are difficult to generalize. A meta-analysis published in 2011 summarizes the results from these studies.²⁰ Unfortunately, comprehensive assessments on a truly national scale remain sparse. Examining patient outcomes and hospital charges allows us to look at the value of health care provided, defined as the ratio of outcomes to costs. As we begin to look at the comparative effectiveness of these different modalities in treating AVMs, the value of care delivered will be an essential parameter to examine. National patient databases offer an unparalleled tool for assessing the efficacy of rapidly evolving therapeutic modalities. Herein, we use the database to assess the diagnosis and management of cerebral AVMs between 2000 and 2009,⁵ examining demographics, outcomes, and value over this period. Nationwide Inpatient Sample analyses have become increasingly important by enabling rapid evaluation of national trends in neurosurgery,^{6–11,15–17} and our findings establish a baseline of care for cerebral AVMs against which future innovations in care can be compared and assessed for efficacy.

Methods

Results

Annual Discharges

A mean of 4191 patients were admitted annually with a primary diagnosis of AVM over the 10-year period from 2000 to 2009. The trend in annual admission rates is shown in Fig. 1A. Admissions ranged from 3503 in 2005 to 5209 in 2003. There was no statistically significant variation over this period of time as assessed by a year-to-year comparison.

A: Overall AVM admissions between 2001 and 2009. B: Annual mean hospital charges for patients admitted with a primary diagnosis of cerebrovascular anomaly between 2001 and 2009. C: Discharge proportion for patients admitted with a primary diagnosis of cerebrovascular anomaly between 2001 and 2009. D: Average LOS for patients admitted with a primary diagnosis of cerebrovascular anomaly between 2001 and 2009

Treatment

A mean of 1232 patients with AVMs were treated annually with open excision. The trend in the proportion of AVMs treated by open excision is shown in Fig. 2A. There was an initial decline in the overall proportion treated with excision over the first half of the decade, with increased surgical management in the latter half. Concurrently, the proportion of endovascular interventions increased rapidly over the first half of the study period from 2.6% in 2000 to 36.6% in 2005 (p < 0.05), with modest decline toward the end of the decade.

A: Proportion of AVMs treated with open surgery, endovascular treatment, radiosurgery, or observation by year. B: Proportion of AVMs treated with open excision, endovascular treatment, radiosurgery, or observation stratified by age group. C: Median total charges by year for open surgery, endovascular, radiosurgery, or observation from 2000 to 2009.

We were further able to analyze treatment trends by stratifying on patient demographic, as shown in Fig. 2B. The distribution of treatments correlates closely with age with several clear trends observed when stratified as follows: younger than 18 years, 18–44 years, and 45 years or older. Both open surgery and radiosurgery had the highest rates among the young, at 40% and 4%, respectively. These proportions decrease linearly across age strata to 32% and 2%, respectively (p > 0.05). In contrast, endovascular treatments show a countercurrent trend, increasing in frequency as age increases, from 21% in the youngest strata to 26% in the oldest (p < 0.05). There was no statistically significant correlation between patient income or sex and treatment modality. Total hospital charges also showed distinct trends, overall increasing over the course of the decade for all treatment modalities (Fig. 2C). Open surgery increased the fastest with peak median charges occurring in 2007, with subsequent regression, yielding an overall yearly average growth rate of 9.18% (p < 0.05). Endovascular treatments likewise increased at 8.63% yearly, and radiosurgery remained relatively constant given inflation at a yearly growth of 3.31%.

Comparative Effectiveness of Treatments

We evaluated the comparative effectiveness of treatments by analyzing our primary outcomes measures, in-hospital mortality, LOS, routine discharge, and total charges, while controlling for both patient and hospital factors (Table 2). We stratified by presence of subarachnoid or intracerebral hemorrhage and by age according to the supplemental grading scheme for each treatment modality, including uni- and multimodal approaches. We assessed the impact of these factors by univariate and multivariate analysis and found that several patient-level factors were significant predictors of outcome (Table 3). Figure 3 illustrates the impact of age and hemorrhage status on each of the primary outcome measures.

TABLE 2:

Outcomes of patients treated with open surgery, endovascular therapy, radiosurgery, or combination therapy with a primary diagnosis of AVM from 2000 to 2009 stratified by presentation hemorrhage status^*

Outcome	Treatment									p Value
	Op	Endo	SRS	Op + Endo	Endo + SRS	Op + SRS	Op + Endo + SRS	Op + Angio	None
unruptured
in-hospital mortality (%)	1.0	1.6	0.0	0.0	0.0	NA	NA	0.5	0.7	<0.01
mean LOS (days)	5.45 ± 0.22	2.49 ± 0.11	1.37 ± 0.15	8.36 ± 0.45	2.70 ± 1.54	NA	NA	5.80 ± 0.18	3.45 ± 0.09	<0.01
routine discharge (%)	67.5	93.4	98.9	75.9	100	NA	NA	83.8	72.9	<0.01
mean total charges ($)	57,062 ± 1,600	49,839 ± 1,301	45,393 ± 1,422	109,467 ± 4,861	49,172 ± 11,005	NA	NA	63,945 ± 2,011	22,946 ± 611	<0.01
ruptured
in-hospital mortality (%)	6.6	10.6	NA	10.3	NA	NA	NA	2.9	3.1	<0.01
mean LOS (days)	11.86 ± 1.20	10.59 ± 1.69	NA	26.11 ± 6.42	NA	NA	NA	15.78 ± 1.84	6.98 ± 0.63	<0.01
routine discharge (%)	42.5	56.7	NA	36.0	NA	NA	NA	49.0	38.7	<0.01
mean total charges ($)	105,339 ± 11,668	127,652 ± 18,269	NA	220,115 ± 26,696	NA	NA	NA	138,713 ± 13,956	55,553 ± 5,946	<0.01

* Mean values are represented as the mean ± SE. Abbreviations: Angio = angiography; Endo = endovascular treatment; NA = not available; SRS = stereotactic radiosurgery.

TABLE 3:

Results of multivariate modeling of patient-level factors as predictors of mortality, length of stay, discharge proportion, and total charges^*

Patient Characteristics	p Value
	Mortality	LOS	Total Charges	Discharge Proportion
admission source	0.0017	0.391	0.0001	0.0049
admission type	0.0001	0.0017	0.7221	0.4442
sex	0.5636	0.0655	0.5473	0.0423
elective	0.0001	0.3198	0.5946	0.0863
comorbidity index	0.1723	0.0001	0.0001	0.0001
age	0.003	0.0086	0.1173	0.0001
hemorrhage	0.0001	0.0001	0.0001	0.0001

* Values in bold face are statistically significant.

Comparison of outcomes measures, mortality (A and B), LOS (C and D), discharge proportion (E and F), and total charges (G and H), stratified by age and hemorrhage status at presentation. Angio = angiography; Endo = endovascular; Obs = observation; SRS = stereotactic radiosurgery; Surg = surgery.

Among patients in whom the AVM did not rupture, those treated radiosurgically fared best with no reported mortalities, the lowest LOS (1.37 days), and the highest discharge proportion (98.9%). By comparison, patients treated endovascularly had the highest mortality rate (1.6%), and yet had a relatively low LOS index (2.49 days) and a high discharge proportion (93.4%). Patients treated surgically tended to remain in the hospital longer (5.45 days [surgery] and 8.36 days [surgery plus endovascular treatment], p > 0.05), and had lower discharge proportions (67.5% [surgery] and 75.9% [surgery plus endovascular treatment], p > 0.05). Patients who underwent surgery and angiography had very low mortality (0.45%) and an improved discharge proportion (83.8%).

Patients experiencing AVM rupture had very different outcomes. In general, their mortality and LOS were significantly higher than for equivalent unruptured patients. Those who underwent embolization had the highest mortality (10.6% [endovascular treatment] and 10.3% [endovascular treatment plus surgery]), which is nearly 60% higher than the next highest (6.6% [surgery], p < 0.05). Surgery plus angiography had the lowest mortality rate at 2.87%. Those who underwent surgery plus endovascular treatment had the worst outcomes profile, with all indices significantly worse than those who had no treatment at all (p < 0.05 for each index). The presence of any procedure significantly increased LOS (6.98 days [no therapy] and > 10.5 days [all therapies], p > 0.05). Hemorrhage was correlated with higher charges (p > 0.05), on average doubling charges for each treatment category. There were too few radiosurgical cases to include in these analyses.

We analyzed admissions for comorbid conditions, procedures, and potential adverse events (Table 4). Hemorrhage and procedures performed were the most important factors in predicting complication rates. Although patients who suffered AVM rupture had overall higher complication rates, the change was largest for extraventricular drain usage and pulmonary complications, averaging a nearly 10-fold increase across treatment modalities. Comparing treatments, patients who underwent surgery were 4 times as likely to undergo ventriculostomy (3.74% vs 0.85% [endovascular treatment], p < 0.01). Surgically treated patients also experienced potential adverse events at much higher rates than both other groups of patients, including pulmonary, venous thromboembolism, cardiac, renal, gastrointestinal, infectious, pneumonia, and wound complications (p < 0.01). The 2 types of complications wherein surgical patients did not have significantly difference rates were for implant complications and accidental puncture/laceration.

TABLE 4:

Complications and associated procedures for patients admitted with a primary diagnosis of AVM from 2000 to 2009 by treatment modality and stratified by presentation hemorrhage status^*

Intervention	Complications & Associated Procedures Rates (%)
	EVD	Imp	Lac	Pulm	DVT	Card	Renal	GI	Infect	PNA	Wound
unruptured
no Tx	0.63	0.26	0.21	1.41	0.31	0.09	0.72	0.38	2.58	0.83	0
op	2.83	0.71	0.41	2.98	0.27	0.88	0.43	0.68	2.09	0.75	0.86
endo	0.12	0.29	0.41	1.32	0.12	0.31	0.06	0.37	1.13	0.13	0
SRS	0	0.51	0	0	0	0	0	0	0.55	0	0
op + endo	1.80	0.33	1.01	4.88	1.01	1.41	0	1.02	4.79	0.66	1.08
op + angio	2.27	0.11	0.59	3.51	0.96	0.25	0.26	0.83	1.69	1.55	0.11
ruptured
no Tx	10.49	0.92	0	5.90	1.23	0	0.46	0.84	6.24	0.41	0
op	11.70	1.63	0	12.63	0.80	0.75	0	1.78	2.45	5.11	2.50
endo	16.84	2.59	3.81	13.06	1.42	0	0	1.34	9.34	6.81	0
SRS	0	0	0	0	0	0	0	0	0	0	0
op + endo	18.02	2.34	2.55	33.19	14.20	4.67	0	6.42	15.39	9.32	2.34
op + angio	15.42	0.85	0	11.30	0.89	1.79	2.02	3.75	9.25	2.82	0.89

* Card = cardiac; DVT = deep venous thrombosis; EVD = external ventricular drainage; GI = gastrointestinal; Imp = implant; Infect = infection; Lac = laceration; PNA = pneumonia; Pulm = pulmonary; TX = treatment.

Analysis of admission type and source of admission also revealed differences (Table 5), potentially indicative of patient demographics and severity of illness. Radiosurgically treated patients were the youngest contingent at 34.1 years (p < 0.01). Those admitted through the emergency room or transferred urgently were nearly twice as likely to undergo surgery over endovascular treatment (8% vs 4%, p < 0.01) and 4 times more likely to undergo surgery than radiosurgery (8% vs 2%, p < 0.01). Patients admitted on Saturday or Sunday were also more likely to receive surgical than endovascular (5.21% vs 1.94%, p < 0.01) or radiosurgical (0.53%, p < 0.01) therapy.

TABLE 5:

Admission characteristics of patients treated with open surgery, endovascular therapy, and radiosurgery with a primary diagnosis of AVM from 2000 to 2009^*

Characteristic	Treatment			p Value
	Op	Endo	SRS
admission source (%)
routine	86	90	95	<0.01
ED	8	4	2	<0.01
admission type (%)
elective	76	78	91	<0.01
emergency	13	8	3	<0.01
weekend (%)	5.21	1.94	0.53	<0.01
mean age (yrs)	37.4 ± 0.4	40.2 ± 0.4	34.1 ± 1.3	<0.01

* Mean values are ± SE.

Discussion

The treatment of cerebral AVMs over the first decade of the 21st century has rapidly evolved with the advent of multimodality therapies, including open excision, endovascular treatment, and radiosurgery, and such advances continue to shape our treatment approach. Examining trends in parameters of care and patient outcomes is imperative in any period of rapid evolution in treatment, and enables us to assess how well new therapies meet the underlying goals of care, including safety, efficacy, and cost-effectiveness. This analysis has demonstrated several important trends.

Over this period of 9 years, the total charges for care and cost per day of care have increased for all, both averaging 5.8% increases yearly over the nominal rate of inflation. This is in line with general medical inflation as assessed by other high-resource procedures such as coronary artery bypass grafting and hip replacement surgeries (ICD-9-CM 36.11 [5.4%] and ICD-9-CM 81.51 [3.4%]). However, stratification of the data by procedure performed and hemorrhage status shows that both of these variables significantly impact costs. If we define value as the ratio of unit of outcome received per unit of cost, we find that the most resource-intensive therapy, surgery plus endovascular treatment for patients with hemorrhage, shows a distinct decrease in value received for care delivered and should therefore be critically evaluated on a case-by-case basis to determine if expected benefit is worthwhile given the high rates of poor outcome.

The comparative effectiveness of the treatment modalities can be assessed on a nationwide basis. The proportion of AVMs treated with microsurgical excision has remained relatively constant over this period of time. However, the data show tremendous and durable expansion in the proportion of AVMs treated endovascularly, either alone or in combination. There is a statistically significant decrease in the proportion of patients treated with open microsurgical excision of the AVM as patient age increases. This trend also correlates with the most dramatic decrease in overall LOS among the 45- to 64-year-old age group, suggesting that there may be a parallel between the adoption of endovascular modalities for AVMs and decrease in LOS. Microsurgical treatment remains the most common treatment for patients admitted urgently, such as those admitted over weekends or through the emergency room or emergency transfers. The prevalence of surgery among these patients is at least in part an indicator of severity of disease, such as patients who require emergency decompression. Furthermore, patients with poor access to care tend to be overrepresented in these emergency care categories. Their access to procedures typically offered on an outpatient basis, such as radiosurgery, is diminished, even if their disease could have been adequately managed through that modality.

These data suggest that the choice of therapy should be considered carefully and independently for patients regardless of rupture status (ruptured or unruptured). In both cases, surgery plus angiography is associated with the most favorable outcomes, and on a value basis is the superior choice. One surprising finding was that patients who received what many consider “maximal therapy,” consisting of surgery plus endovascular therapy, experienced the worst suite of outcomes and represent the worst value proposition. These patients had poor immediate outcomes at discharge at nearly twice the costs for patients regardless of rupture status. It is notable that embolization, either alone or in combination, was linked with substantially increased mortality risk, and for this reason should be undertaken with care and full knowledge of implications. Furthermore, embolization is associated with additional days of stay and substantial charges, and yet it adds little to improve outcome. On a larger scale, real costs related to the care of patients who underwent surgery plus endovascular treatment are likely higher given that among those who survive, fewer go home at discharge and likely incur substantial additional costs for long-term care and rehabilitation.

The data presented suggest that patient-level data may be useful when considering the choice of treatment regimens. Indeed, hemorrhage status figures prominently in our outcomes analyses, with presence of hemorrhage universally connected with poorer outcomes. Age also had a prominent effect, but the direction of the effect was less consistent. Age and hemorrhage have also been shown to correlate with surgical outcome in the Lawton supplementary grading system.¹² Other factors commonly used in grading schemes, size, location, eloquence, and nidal compactness, are not directly obtainable from such national data sets. Based on the senior author's (M.T.L.) approach to triage and treatment of AVMs, it seems likely that universus multimodal treatment may act as a surrogate for the complexity of the lesion. More complex lesions tend to require additional interventions, such as embolization, to reduce risk to the patient prior to resection. We anticipate that further assessment of these patient-level factors in more detailed institutional and multiinstitutional data sets will bear out these findings.

There are limitations inherent to the data source. It is not possible to assess the relative superiority of treatment modalities from the data analyzed. Rather, the analyses are designed to evaluate associations between treatments and certain outcome measures. It is not possible to discern causality from the results.

It is impossible to know what proportion of patients who underwent endovascular therapy were treated with the goal of cure compared with those who underwent partial embolization prior to adjuvant therapies on separate hospital admissions. Conversely, in patients who underwent multiple inpatient procedures, we cannot accurately attribute adverse outcome events to a given intervention. For instance, the data clearly demonstrate that ventriculostomy among patients undergoing surgery is 4 times more prevalent, but we cannot know if patients had more severe presentations requiring presurgical treatment or whether the need for ventriculostomy was a result of the surgical intervention. Surgery plus endovascular therapy for patients in whom the AVM has ruptured also demonstrated poor outcomes at high cost, or extremely low value, but since the data source does not allow us to easily control for severity of presenting disease, it is possible that this therapy option acts as a surrogate for extremely high-risk lesions that would have poor outcome regardless of therapy chosen. Analysis of admission type and source likewise revealed significant differences between the patient populations that could be indicative of patient demographics and the severity of their presenting condition.

Not all therapies are completed in a single hospitalization. Tracking individual patients across multiple hospital admissions could alleviate some of these limitations. However, these data have been de-identified, and only a small subset of states provide patient-level identifiers to allow researchers to track readmissions, so a nationwide analysis of questions related to subsequent admissions is not possible. In the case of radiosurgery, admission may not require admission universally, and in fact admission may be a marker of complication or severity of disease. Even if this sample is biased to be sicker than the average patient undergoing radiosurgery, in which case the patient is likely to be an outpatient, these patients still perform better than those who undergo other types of treatment in terms of outcomes. Although evaluation of a more representative sample including outpatients would require different outcomes metrics, it is anticipated that these outcomes would be more positive than those reported here, and as such the bias is acceptable.

Finally, the extent to which the care parameters we examined reflect the goals of care that patients and society have in AVM management remains uncertain. Parameters such as complete AVM obliteration, survival, functional outcomes, quality of life, and complication rates, among others, must be assessed to really examine the value of care delivered in a meaningful way. These data are unfortunately not tabulated on a national level. It is critical that we develop nationally coordinated databases to facilitate long-term study of care management in a way that will more closely reflect goals of care and enable us to truly evaluate the safety, efficacy, and cost-efficiency of care parameters in a manner that provides value and utility to patients and society. Importantly, from the neurosurgical standpoint, this may be particularly feasible in a small field such as neurosurgery, where national coordination may be somewhat easier than in larger fields.

Conclusions

Charges for treating cerebral AVMs have increased dramatically over the first decade of the 21st century: 5.8% annually over the rate of nominal inflation during that period of time. However, cost-efficiency is more important than cost of care. Although it is more difficult to ascertain, cost per unit of value derived is a more appropriate parameter. Data sources such as the NIS allow for analysis of comparative effectiveness as well as comparative cost-efficiency of multiple alternative treatments across a nationwide landscape. This analysis provides a first look at national trends in AVM management from the standpoint of safety, efficacy, and cost-efficiency of care. We demonstrate that there are differences among treatment modalities that can significantly influence patient outcomes. The surgeon's choice of therapies must be undertaken with great care and a clear understanding of anticipated benefits and their attendant risks to the patient.

Disclosure

Dr. Lawton receives royalties from Mizuo America, Inc.

Author contributions to the study and manuscript preparation include the following. Conception and design: Davies. Acquisition of data: Davies. Analysis and interpretation of data: all authors. Drafting the article: all authors. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Statistical analysis: Davies, Yanamadala. Administrative/technical/material support: Lawton. Study supervision: Lawton.