Short-term outcomes and hospitalization cost associated with flow diversion versus coil embolization for unruptured aneurysms: A national assessment

Abstract

Objective:

The use of flow diversion (FD) has rapidly increased over the last decade, particularly for treatment of complex aneurysms not amenable to conventional coil embolization (CE). We aimed to compare national outcomes and healthcare utilization associated with FD and CE of unruptured aneurysms.

Methods:

The National Inpatient Sample (2019–2022) was used to identify patients with unruptured intracranial aneurysms who underwent CE or FD (patients undergoing both FD+CE were classified as FD). Pediatric patients, non-elective admissions, and patients with subarachnoid hemorrhage were excluded. Variables included sociodemographics (e.g., age, race, gender), hospital factors (e.g., size, ownership, teaching status, location), and basic clinical variables in bivariate and multivariable regression. Outcomes included in-hospital mortality, stroke, discharge disposition (favorable/unfavorable), length of stay, and total cost of hospitalization.

Results:

7370 patients were identified, of which 4280 were CE and 3090 were FD. Given the elective nature of intervention, rates of mortality (0.19 %), unfavorable discharge disposition (2.77 %), and stroke (0.83 %) were low. On multivariable analysis, use of flow diversion was not associated with unfavorable discharge (OR 0.80, P = 0.211) or stroke (OR 0.91, P = 0.753). FD trended toward, but did not reach, statistical significance for elevated length of stay (IRR 1.04, P = 0.150); however, it did lead to a significantly higher overall cost (ß=$1260.98, P = 0.049)

Conclusion:

Nationally, short-term outcomes are similar between FD and CE, although use of FD does imbue a mildly higher hospitalization cost. Further work is needed to characterize large-scale, long-term outcome differences, particularly as FD use increases for more complex aneurysms not amenable to CE.

1. Introduction

Flow diversion (FD) has significantly altered the landscape of endovascular aneurysm treatment since its introduction to the market. Whereas traditional coil embolization (CE) is based on the premise of restricting blood flow to an aneurysm via direct filling of the aneurysm, FD focuses on healing the diseased segment of the parent vessel. By placing a flow diverting stent over the diseased vessel segment and aneurysm neck, the aneurysm is treated via a combination of hemodynamic changes, aneurysmal thrombosis, and subsequent endothelialization [1–3]. Outcomes of FD have been very promising, with a 5-year aneurysm occlusion rate of up to 96 % and a retreatment rate of only 5 % [4].

Flow-diverting stents were initially approved for treatment of large/giant wide-necked aneurysms of the internal carotid artery (ICA) in 2011; however, utilization rates saw a significant increase beginning in 2019 [3,5]. In addition, off-label usage has become increasingly common, particularly for aneurysms that are difficult to treat via traditional open or endovascular means (e.g., fusiform aneurysms, posterior circulation aneurysms) [1,5]. Despite the need for concurrent dual antiplatelet therapy (DAPT), FD has even been used for treatment of certain ruptured aneurysms, albeit less frequently [5]. This rise in popularity has outpaced the availability of literature on such off-label indications; as a result, further studies are needed to assess the true impact and utility of FD on a large scale.

While there is definitive utility, FD is not free of downsides. One of the most feared complications is in-stent stenosis, which occurs at a rate of approximately 4.8 % [4]. This, in turn, can lead to ischemic stroke, which is thought to occur at a rate of 2–5.8 % [6,7]. To lower the rate of in-stent stenosis and stroke, effectively all FD patients are placed on DAPT (generally followed by aspirin monotherapy after an initiation period) [8,9]. This can raise the rate of cerebral or systemic hemorrhagic complications. While this does not decrease the utility of flow diversion, it must be weighed against the benefit of placing a flow diverter over traditional CE when given the choice. As with any new device, cost should also be considered when determining appropriate application.

The purpose of this study is to compare FD to CE for unruptured intracranial aneurysms on a national level by using a large-scale, administrative database to assess clinical/functional outcomes and healthcare utilization associated with each procedure. While FD has been the subject of a number of prior studies, most are either single-center or small multi-center investigations, and there is minimal literature that uses large databases. Mirpuri et al. recently leveraged the MARINER database to gather a large number of FD patients; however, the purpose of this study was to assess trends in utilization and not clinical outcomes [5]. McDonald et al. used the Premier Perspective database; however, this was done approximately 10 years ago and only gathered 279 patients [10]. Finally, national databases/records were used by both Kim et al. (South Korea) and Gory et al. (France), but both studies still had notably small sample sizes (169 and 408 patients, respectively) [11,12]. Thus, the use of a larger administrative database in this study would allow for assessment of patient- and hospital-level variables that impact outcomes on a broader scale. Specifically, the aims of this study are to compare FD and CE with regards to (1) periprocedural stroke, (2) mortality, (3) length of stay, (4) discharge disposition, and (5) cost.

2. Materials and methods

2.1. Patient selection

Patients in this study were identified from the National Inpatient Sample (NIS), a national administrative databased published by the Agency for Healthcare Research and quality (AHRQ) as part of the Healthcare Cost and Utilization Project (HCUP). The NIS uses discharge-level data on approximately 7 million hospitalizations in a given year. Given the completely deidentified nature of this database, ethics committee/IRB approval were not required for this study.

International Classification of Diseases (ICD) codes were used to select for both diagnostic (ICD-10-CM) and procedural (ICD-10-PCS) inclusion and exclusion criteria. Inclusion criteria selected for patients with (1) a diagnostic code for unruptured cerebral aneurysm, (2) elective hospital admission, and (3) endovascular treatment of their aneurysm. Beginning in 2019, specific ICD-10-PCS codes were added to denote the use of flow diversion; as such, the NIS was queried from 2019 to 2022). Patients were stratified into either coil embolization (CE) or flow diversion (FD) groups. If patients had procedural codes for both CE and FD (i.e., they underwent coil embolization with adjunctive flow diverter placement), they were included with the FD group. All patients with a code for CE and not FD, including both traditional CE and stent-assisted CE (without use of a flow diverting stent), were included under the CE group. Pediatric patients (less than 18 years of age), those with missing outcome information, patients with other cerebrovascular malformations, and patients with intracranial hemorrhage (either non-traumatic or traumatic) were excluded.

2.2. Data collection

Collected sociodemographic information from NIS included age (binned into quartiles), sex, race, urban/rural (based on National Center for Health Statistics classification), income quartile (based on zip code), and primary insurer. Hospital characteristics included size, ownership status, teaching status, and US census region/division. Lastly, clinical/comorbidity variables include the Elixhauser Comorbidity Index (ECI; calculated from ICD-10-CM codes) and All Patient Refined Diagnosis Related Groups (APR-DRG) severity of illness/risk of mortality metrics.

Five outcomes were included in this study. Length of stay (LOS) was reported in days in the NIS and was analyzed as a continuous variable. Mortality and stroke are both binary variables; presence of a stroke was determined via ICD-10-CM coding. Given the elective nature of the above procedures, presence of a stroke was presumed to be peri-/post-procedural (as opposed to pre-existing). Discharge disposition is classified as either favorable (routine discharge, home with home health care, or transfer to short-term hospital) or unfavorable (other transfers, such as skilled nursing facilities, or death). Finally, cost (reported in USD) is calculated using a combination of the total hospital charges for an admission and a cost-to-charge ratio (CCR) that is unique to each hospital; thus, it is an estimated amount of the total cost based on individual hospital billing practices.

2.3. Statistical analysis

We first conducted descriptive analysis to compare variables between CE and FD. Mean and standard deviation (SD) with t-test was reported for continuous variables. Count and frequency/percentage (%) with the Chi-squared test was reported for categorical variables. Three types of multivariable regression, with generalized estimation equations (GEE) adjusting hospital clustering, were then used to assess the association between treatment groups and outcomes. Logistic regression was used for discharge disposition and stroke. Age was used as quartiles as it violated the linearity assumption. The Hosmer-Lemeshow goodness-of-fit was used to assess the model’s overall fit. Negative binomial regression was used for LOS accounting for variance inflation. Linear regression was used for cost. Linearity assumptions were checked by residual plots. All models started with all covariates mentioned in the data collection section and then underwent a variable selection process. The final model excluded variables not acting as a confounder or p > 0.05. A confounder was defined as a treatment group estimation change of more than 10 %. With this process, each outcome model had a different set of covariates. The significance level was set to 0.05 (2-sided). All analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA.).

3. Results

Overall, 7370 patients met inclusion/exclusion criteria, which included 4280 CE (58.07 %) and 3090 FD (41.93 %) (Fig. 1). Patients were relatively equally distributed between age quartiles; however, FD was noted to have a higher proportion of young patients (over half of patients under 53 were treated with FD, whereas only 37.1 % of patients ≥70 were treated with FD). Female patients comprised the majority of the cohort (n = 5803; 78.7 %); this was particularly pronounced among the FD cohort (n = 2600; 84.1 %). Multiple differences were also noted among disease severity/comorbidity metrics. Patients with a minor APR-DRG risk of mortality underwent FD more frequently those with a moderate risk or above. Similarly, higher ECI was associated with a significantly higher preponderance of CE (Table 1).

Fig. 1.

Flowchart detailing patient selection process with associated inclusion/exclusion criteria.

Table 1.

Sociodemographic variables and univariable analysis for the entire cohort, stratified by treatment type (coil embolization vs. flow diversion).

	Total Cohort		Treatment Type				P
			Coil Only		Flow Diversion
	Mean	SD	Mean	SD	Mean	SD
LOS (days)	1.59	2.24	1.58	2.35	1.61	2.06	0.54
Cost (USD)	34362	20508	34031	20444	34820	20590	0.11
	N	%	N	%	N	%	P
Age							< 0.001
< 53	1835	24.9	914	21.4	921	29.8
53–62	1765	24.0	1038	24.3	727	23.5
62–70	1873	25.4	1130	26.4	743	24.1
≥ 70	1897	25.7	1198	28.0	699	22.6
Gender							< 0.001
Male	1567	21.3	1077	25.2	490	15.9
Female	5803	78.7	3203	74.8	2600	84.1
Race							< 0.001
White	4793	65.0	2902	67.8	1891	61.2
Black	965	13.1	513	12.0	452	14.6
Hispanic	923	12.5	489	11.4	434	14.1
Asian or Pacific Islander	251	3.4	125	2.9	126	4.1
Other	251	3.4	141	3.3	110	3.6
Missing	187	2.5	110	2.6	77	2.5
Insurance							< 0.001
Medicare	3144	42.7	1936	45.2	1208	39.1
Medicaid	1022	13.9	566	13.2	456	14.8
Private insurance	2779	37.7	1523	35.6	1256	40.7
Self-pay	140	1.9	78	1.8	62	2.0
Other	276	3.7	171	4.0	105	3.4
Missing	DS*	DS*	DS*	DS*	DS*	DS*
APR-DRG: Risk of Mortality Subclass							< 0.001
Minor	6055	82.2	3458	80.8	2597	84.1
Moderate	943	12.8	610	14.3	333	10.8
Major	261	3.5	149	3.5	112	3.6
Extreme	111	1.5	63	1.5	48	1.6
APR-DRG: Severity of Illness Subclass							0.23
Minor	4865	66.0	2803	65.5	2062	66.7
Moderate	2024	27.5	1191	27.8	833	27.0
Major	372	5.1	229	5.4	143	4.6
Extreme	109	1.5	57	1.3	52	1.7
Elixhauser							< 0.001
0	1090	14.8	586	13.7	504	16.3
1	2047	27.8	1141	26.7	906	29.3
2	1854	25.2	1088	25.4	766	24.8
≥3	2379	32.3	1465	34.2	914	29.6
Patient Location: NCHS Urban-Rural Code							0.004
Central of metro ≥ 1 million	2075	28.2	1168	27.3	907	29.4
Fringe of metro ≥ 1 million	1946	26.4	1103	25.8	843	27.3
Metro 250,000–999,999	1695	23.0	975	22.8	720	23.3
Metro 50,000–249,999	611	8.3	384	9.0	227	7.4
Micropolitan	593	8.1	367	8.6	226	7.3
Not metropolitan or micropolitan	435	5.9	272	6.4	163	5.3
Missing	15	0.2	11	0.3	DS*	DS*
Income							0.42
1st quartile	1804	24.5	1042	24.4	762	24.7
2nd quartile	1845	25.0	1050	24.5	795	25.7
3rd quartile	1877	25.5	1117	26.1	760	24.6
4th quartile	1741	23.6	1002	23.4	739	23.9
Missing	103	1.4	69	1.6	34	1.1
Bed Size of hospital							0.06
Small	382	5.2	232	5.4	150	4.9
Medium	1228	16.7	744	17.4	484	15.7
Large	5760	78.2	3304	77.2	2456	79.5
Control/Ownership of Hospital							0.85
Government	1043	14.2	607	14.2	436	14.1
Private, not-profit	5843	79.3	3398	79.4	2445	79.1
Private, invest-own	484	6.6	275	6.4	209	6.8
Census Division of Hospital							< 0.001
New England	466	6.3	263	6.1	203	6.6
Middle Atlantic	1203	16.3	637	14.9	566	18.3
East North Central	1208	16.4	746	17.4	462	15.0
West North Central	459	6.2	308	7.2	151	4.9
South Atlantic	1492	20.2	818	19.1	674	21.8
East South Central	520	7.1	303	7.1	217	7.0
West South Central	597	8.1	352	8.2	245	7.9
Mountain	483	6.6	240	5.6	243	7.9
Pacific	942	12.8	613	14.3	329	10.7
Region of Hospital							< 0.001
Northeast	1669	22.7	900	21.0	769	24.9
Midwest	1667	22.6	1054	24.6	613	19.8
South	2609	35.4	1473	34.4	1136	36.8
West	1425	19.3	853	19.9	572	18.5
Stroke							0.88
No	7309	99.2	4244	99.2	3065	99.2
Yes	61	0.8	36	0.8	25	0.8
Died during Hospitalization							0.94
No	7356	99.8	4272	99.8	3084	99.8
Yes	14	0.2	DS*	DS*	DS*	DS*
Discharge Disposition							0.07
Favorable	7166	97.2	4149	96.9	3017	97.6
Unfavorable	204	2.8	131	3.1	73	2.4

When directly comparing outcomes of CE vs. FD, there was no significant difference in LOS (1.58 vs. 1.61 days, P = 0.54) or cost ($34,031 vs. $34,820, P = 0.11). Overall rates of stroke (N = 61; 0.8 %), inpatient mortality (N = 14; 0.2 %), and unfavorable discharge (N = 204; 2.8 %) were low and not significantly different between the two groups. Descriptive statistics and univariable comparisons are provided in Table 1.

Results of multivariable analysis are shown in Tables 2–5. Advanced age was significantly associated with unfavorable discharge, the effect of which increased in older age groups (OR 3.94, P < 0.001 for an age of ≥70 years). APR-DRG severity of illness was also significantly associated with unfavorable discharge, with an OR of 142.62 (P < 0.001) in the extreme category. Unexpectedly, however, ECI did not appear to be associated with discharge disposition. When controlling for other factors, FD was not related to discharge disposition (OR 0.80, P = 0.21) (Table 2).

Table 2.

Multivariable analysis assessing associations between patient characteristics and unfavorable discharge.

Variable	Category	OR	95 % CI		P
Treatment Group	CE	-
	FD	0.80	0.56	1.14	0.21
Age	< 53	-
	53–62	2.36	1.28	4.34	0.006
	62–70	2.99	1.66	5.41	< 0.001
	≥ 70	3.94	2.20	7.05	< 0.001
APR-DRG Severity of Illness	Minor	-
	Moderate	4.34	2.60	7.23	< 0.001
	Major	23.91	14.05	40.68	< 0.001
	Extreme	142.62	81.09	250.85	< 0.001
Elixhauser	0	-
	1	1.10	0.45	2.66	0.84
	2	1.17	0.48	2.84	0.74
	≥ 3	2.12	0.94	4.80	0.07

DS* = data suppressed (required per NIS policy for cells with a patient count of ≤10

Table 5.

Multivariable analysis assessing associations between patient and hospital characteristics and estimated cost.

Variable	Category	β	95 % CI		P
Treatment group	Coil	-
	Flow Diversion	1261	6	2516	0.049
Age	<53	-
	53–62	642	−610	1894	0.31
	62–70	1377	127	2627	0.031
	≥70	1816	557	3076	0.005
Income	1st quartile	-
	2nd quartile	1160	−57	2376	0.06
	3rd quartile	2264	1003	3526	<.001
	4th quartile	4049	2487	5611	<.001
Census Division of Hospital	New England	-
	Middle Atlantic	−2480	−7178	2218	0.30
	East North Central	−3153	−6962	656	0.10
	West North Central	−4277	−8281	−273	0.036
	South Atlantic	−708	−4309	2893	0.70
	East South Central	612	−4212	5436	0.80
	West South Central	−1608	−5382	2167	0.40
	Mountain	−2765	−7122	1592	0.21
	Pacific	4972	999	8946	0.014
Bed Size of Hospital	Large	-
	Medium	1253	−850	3355	0.24
	Small	3766	742	6791	0.015
Control/Ownership of Hospital	Government	-
	Private, not-profit	−926	−3321	1469	0.45
	Private, invest-own	−8492	−11751	−5233	<.001
APR-DRG Severity of Illness	Minor	-
	Moderate	2271	1301	3242	<.001
	Major	12672	9908	15436	<.001
	Extreme	53603	41892	65314	<.001
Elixhauser	0	-
	1	808	−443	2059	0.21
	2	1056	−178	2289	0.09
	≥3	2418	992	3845	<.001

Similarly, only age and APR-DRG severity of illness were related to stroke rates. That said, an increase in stroke rates was only seen in the 53–62 age group. Female sex trended toward an increase in stroke rate but similarly did not reach statistical significance (OR 2.09, P = 0.07). FD had no significant relationship with stroke rates (OR 0.91, P = 0.75) (Table 3).

Table 3.

Multivariable analysis assessing associations between patient characteristics and stroke.

Variable	Category	OR	95 % CI		P
Treatment Group	CE	-
	FD	0.91	0.49	1.67	0.75
Age	<53	-
	53–62	3.34	1.21	9.23	0.020
	62–70	2.14	0.74	6.18	0.16
	≥70	2.48	0.95	6.48	0.06
Gender	Male	-
	Female	2.09	0.93	4.68	0.07
APR-DRG Severity of Illness	Minor	-
	Moderate	10.75	2.29	50.54	0.003
	Major	217.73	53.75	881.94	<0.001
	Extreme	519.16	126.04	2138.45	<0.001
Census Division of Hospital	New England	-
	Middle Atlantic	1.98	0.45	8.74	0.37
	East North Central	2.07	0.46	9.32	0.34
	West North Central	1.84	0.33	10.33	0.49
	South Atlantic	1.01	0.22	4.65	0.99
	East South Central	0.36	0.03	3.82	0.39
	West South Central	0.48	0.06	3.57	0.47
	Mountain	1.04	0.18	6.14	0.97
	Pacific	2.99	0.66	13.52	0.15

Multivariable analysis for length of stay is shown in Table 4. Multiple sociodemographic variables, including both advanced age (≥62) and non-white race were associated with increased length of stay, as was private/for-profit hospital status (IRR 1.32, P = 0.002). Both increased APR-DRG severity of illness and ECI were associated with increased LOS. FD trended toward an increase in LOS over CE; however, this did not reach statistical significance (IRR 1.04, P = 0.15).

Table 4.

Multivariable analysis assessing associations between patient and hospital characteristics and length of stay.

Variable	Category	IRR	95 % CI		P
Treatment group	Coil	-
	Flow Diversion	1.04	0.99	1.09	0.15
Age	<53	-
	53–62	1.06	1.00	1.12	0.06
	62–70	1.08	1.01	1.15	0.016
	≥70	1.15	1.08	1.22	<0.001
Race	White
	Black	1.10	1.02	1.19	0.017
	Hispanic	1.14	1.07	1.22	<0.001
	Asian or Pacific Islander	1.18	1.06	1.32	0.004
	Other	1.11	1.01	1.23	0.039
Census Division of Hospital	New England	-
	Middle Atlantic	1.06	0.95	1.17	0.30
	East North Central	1.06	0.96	1.17	0.23
	West North Central	0.94	0.84	1.06	0.30
	South Atlantic	0.98	0.89	1.08	0.66
	East South Central	1.09	0.92	1.30	0.30
	West South Central	1.07	0.95	1.21	0.28
	Mountain	0.93	0.82	1.06	0.29
	Pacific	0.95	0.86	1.05	0.32
Patient Location: NCHS Urban-Rural Code	Central of metro ≥1 million	-
	Fringe of metro ≥1 million	1.08	1.02	1.14	0.014
	Metro 250,000–999,999	1.05	0.97	1.13	0.22
	Metro 50,000–249,999	0.99	0.90	1.09	0.82
	Micropolitan	0.96	0.87	1.05	0.38
	Not metropolitan or micropolitan	0.96	0.87	1.06	0.39
Bed Size of hospital	Large	-
	Medium	1.01	0.94	1.09	0.73
	Small	0.87	0.77	0.98	0.017
Control/Ownership of Hospital	Government	-
	Private, not-profit	0.99	0.92	1.06	0.75
	Private, invest-own	1.32	1.11	1.56	0.002
APR-DRG Severity of Illness	Minor	-
	Moderate	1.27	1.22	1.33	<0.001
	Major	2.79	2.48	3.14	<0.001
	Extreme	8.21	6.87	9.82	<0.001
Elixhauser	0	-
	1	0.95	0.90	1.01	0.09
	2	1.03	0.97	1.10	0.39
	≥3	1.14	1.07	1.22	<0.001

Finally, variables associated with total cost are shown in Table 5. FD was associated with a modest but significant elevation in cost (ß $1261, P = 0.049). Advanced age and higher income quartiles were also associated with increased cost, as was the Pacific US census division (ß 4972, P = 0.014). Private/for-profit hospital status was associated with significantly lower cost (ß −8492, P < 0.001). Both higher APR-DRG severity of illness and ECI were associated with increased cost.

4. Discussion

This study compared outcomes and healthcare utilization for FD and CE using the NIS, one of the largest existing administrative databases. Prior to 2019, there were no specific ICD-10-PCS codes to differentiate FD from other forms of endovascular embolization; as such, this is the first study to the authors’ knowledge to analyze outcomes on FD for unruptured aneurysms in the NIS.

Overall, FD accounted for 41.9 % of all endovascular procedures identified in this study. While FD was first approved in 2011, utilization first began to rise in the mid 2010s and saw a rapid upswing beginning in 2019. According to Mirpuri et al., flow diversion accounted for 34.1 % of all unruptured aneurysm treatments (including FD, CE, and clipping), up from 12.1 % in 2019. When considering only endovascular treatments, FD accounted for 39.7 % of unruptured aneurysm treatments in 2020 [5]. Thus, the rate in this study is in range of, but slightly higher than, the rate reported from 2020. This would suggest that, while growth rates may have slowed, the utilization of FD is still increasing.

Demographically, FD had a higher preponderance of female patients than CE (84.1 % vs. 74.8 %, P < 0.001). It is not surprising that female patients make up a majority of the total study population; female sex has been repeatedly shown to be a risk factor for aneurysm development (with ORs ranging from 1.81 to 3.83 compared to males) [13–15]. This relationship is likely multifactorial and complex but is felt to be largely attributed to hormonal effects and estrogen deficiency following menopause [14,15]. That said, this does not necessarily explain why female patients would have higher rates of FD. FD was also used more frequently in younger patients (53.3 % under 62 for FD vs. 45.6 % for CE, P < 0.002), which has been previously reported [16]. There are multiple potential causes of this. In general, aneurysms in young patients (especially pediatric patients, although they were excluded from this study) are more commonly larger, fusiform, and located in the posterior circulation – all features that make them more amenable to flow diversion over coil embolization alone [17–19]. Additionally, the risks of placing patients on antiplatelet therapy may be higher in elderly patients [20]. It is important to note that the primary goal of this study was to assess outcomes of FD against CE; investigation into which sociodemographic factors favor one treatment type over the other would require further investigation.

The rate of documented ischemic stroke was low in this study for both FD and CE (0.81 % vs. 0.84 %, P = 0.88) and did not differ significantly between the two groups. The only variables found to be associated with stroke upon multivariable analysis were age, which has been noted in a number of prior studies, and APR-DRG severity of illness. Of note, the stroke rate noted in this study is lower than prior investigations, which generally report a range of approximately 2.0–5.8 % [6,7]. While this could be due to improved surgeon comfort with flow diversion over time, it is important to note that strokes in this study may either (1) not be coded appropriately or (2) occur in a delayed fashion, and thus not be captured during the index hospitalization.

There is minimal prior literature on post-FD stroke rates that occur on an inpatient basis, as most occur in a delayed fashion. In one study with a stroke rate of 5.8 %, the actual procedural stroke rate was approximately 1.38 %, which would be closer in line with our finding [6]. A few studies have assessed the rates of acute/intraprocedural thrombosis. Hohenstatt et al. reported an intraprocedural thrombosis rate of 7.5 %; however, most resolved with medical management and only 1.2 % had re-thrombosis [21]. Adeeb et al. identified an intraprocedural thrombosis rate of 4.6 %, but only 0.9 % of these had post-procedural strokes [22]. Finally, Townsend et al. reported an acute post-procedural thrombosis rate of 1.3 %, yet even these had an average time to presentation of 5.5 days, which is generally well past a patient’s estimated discharge date following uncomplicated FD [23]. Thus, while there does not appear to be a difference in stroke rates between FD and CE in this study, the overall rate in either group may be under-reported due to inherent methodological limitations. Of note, post-FD stroke rates are also dependent on aneurysm location and morphology (e.g., fusiform morphology and size >15 mm are associated with stroke); such details are out of the scope of the NIS database [7,24].

Healthcare utilization metrics were also similar between the two groups. FD trended toward, but did not reach, statistical significance for increased LOS (IRR 1.04, P = 0.15). Prior reports have been variable in this regard, ranging from decreased to increased LOS after FD. It is likely that, with additional patients in this study, the LOS would be higher for the FD group. There could be a number of reasons for this; however, in the absence of any complication or outcomes difference, it may be related to tailoring of DAPT therapy prior to discharge, such as for clopidogrel non-responders (accounting for 4–30 % of patients based on prior literature) [25]. For example, if a patient is found to be a non-responder in the peri-operative setting, their hospital stay may be extended while they are converted to another agent.

The only variable significantly different between the two groups was cost, with FD reaching statistical significance for an increase in overall cost (ß 1261, P = 0.049). Cost has been the subject of multiple prior studies comparing FD to conventional endovascular treatments [26–28]. The up-front cost of FD vs. CE is largely based on the size of the aneurysm, given that the largest driver of cost are supplies [29]. For example, a large/giant aneurysm may still only require one flow diverting stent but would require a significantly larger number of coils than a small aneurysm. Initial studies suggested that price parity is reached at approximately 12 mm, above which FD is more cost effective [26,27].

However, it is also important to consider long-term costs, which are not captured in this study given NIS limitations. The most obvious long-term cost associated with FD and not CE is the need for prolonged anti-platelet therapy. This would not only add the cost of the medications themselves but also could put patients at higher risk of hemorrhagic complications (whether neurologic or non-neurologic), thus raising the chance of incurring future hospitalization costs. That said, retreatment rates for FD are generally lower than CE. As such, this leads to higher potential long-term costs of CE when considering the possible need for additional procedures/retreatment [28,30]. Lastly, there are other considerations – such as follow-up imaging regimens, etc. – that may differ between the groups and further impact cost. As such, while FD was marginally more expensive in this study, this finding should be taken in context, as it could be reversed with the availability of longitudinal data given lower retreatment rates. For a full assessment of cost, subsequent studies should ideally track patients over multiple years to fully assess long-term cost differences, likely in the form of a multicenter investigation.

It is important to note that ICD-10 coding does not allow for assessment of aneurysm-specific characteristics, such as morphology, size, or location. There are notable differences between aneurysms that may be more amenable to simple CE (e.g. narrow-necked bifurcation aneurysms) and those in which FD is either preferable or even necessary (such as giant carotid artery or fusiform aneurysms). As such, there may be inherent differences in aneurysm characteristics between the two groups that are unable to be controlled for. Thus, this analysis likely does not represent a direct, pairwise comparison of CE and FD, and it is possible that some aneurysms in the FD cohort may represent more complex aneurysms that are not amenable to CE. This could, in turn, have implications on interpretation of the above data – for example, while FD is associated with a mild increase in cost, this may not be an important consideration if CE is not a reasonable option for a given aneurysm.

This study is not without limitations. The largest limitation is that, while the NIS contains a large number of patients, it does not allow for tracking of patients across multiple hospitalizations. As such, long-term outcomes (e.g., strokes from chronic in-stent stenosis, aneurysm occlusion and recanalization rates, etc.), as well as longitudinal cost, are not able to be calculated. The NIS also does not track which cases use FD in an off-label fashion. As with any administrative database study, there are likely coding errors or discrepancies that are inherently impossible to identify. Some centers may perform some of these procedures on an outpatient basis; as such, these would not be included in the NIS, introducing a possible bias. The NIS is inherently limited to data from the United States and thus may not be generalizable to other regions/countries. Lastly, while each hospital has a unique cost-to-charge ratio, costs are estimated from hospital charges and not an exact reflection of true cost.

5. Conclusion

This study uses the NIS, a large, nation-wide, administrative database to assess both clinical outcomes and healthcare utilization metrics associated with flow diversion versus coil embolization for the treatment of unruptured intracranial aneurysms. On a hospitalization-level, stroke, mortality, and unfavorable discharge rates were similar between the two groups; there was also no significant difference in length of stay. Hospitalization cost was higher for flow diversion than for coil embolization. Additional work is still needed to further delineate long-term implications on a national level.

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.clineuro.2025.109236.

Data Availability

All data was obtained from the National Inpatient Sample, which is available online via the Healthcare Cost and Utilization Project.