A national analysis of 9655 pediatric cerebrovascular malformations: effect of hospital volume on outcomes

Full access

OBJECTIVE

Comprehensive multicenter data on the surgical treatment of pediatric cerebrovascular malformations (CVMs) in the US are lacking. The goal of this study was to identify national trends in patient demographics and assess the effect of hospital case volume on outcomes.

METHODS

Admissions for CVMs (1997–2012) were identified from the nationwide Kids’ Inpatient Database. Admissions with and without craniotomy were reviewed separately. Patients were categorized by whether they were treated at low-, medium-, or high-volume centers (< 10, 10–40, > 40 cases/year, respectively). A generalized linear model was used to evaluate the association of hospital pediatric CVM case volume and clinical variables assessing outcomes.

RESULTS

Among the 9655 patients, 1828 underwent craniotomy and 7827 did not. Patient age and race differed in the two groups, as did the rate of private medical payers. High-volume hospitals had fewer nonroutine discharges (11.2% [high] vs 16.4% [medium] vs 22.3% [low], p = 0.0001). For admissions requiring craniotomy, total charges ($106,282 [high] vs $126,215 [medium] vs $134,978 [low], p < 0.001) and complication rates (0.09% [high] vs 0.11% [medium] vs 0.16% [low], p = 0.001) were lower in high-volume centers.

CONCLUSIONS

This study revealed that further investigation may be needed regarding barriers to surgical treatment of pediatric CVMs. The authors found that surgical treatment of pediatric CVM at high-volume centers is associated with significantly fewer complications, better dispositions, and lower costs, but for noncraniotomy patients, low-volume centers had lower rates of complications and death and lower costs. These findings may support the consideration of appropriate referral of CVM patients requiring surgery or with intracranial hemorrhage toward high-volume, specialized centers.

ABBREVIATIONS AVM = arteriovenous malformation; CVM = cerebrovascular malformation; HCUP = Healthcare Cost and Utilization Project; ICD-9-CM = International Classification of Diseases, Ninth Revision, Clinical Modification; ICH = intracranial hemorrhage; KID = Kids’ Inpatient Database.

OBJECTIVE

Comprehensive multicenter data on the surgical treatment of pediatric cerebrovascular malformations (CVMs) in the US are lacking. The goal of this study was to identify national trends in patient demographics and assess the effect of hospital case volume on outcomes.

METHODS

Admissions for CVMs (1997–2012) were identified from the nationwide Kids’ Inpatient Database. Admissions with and without craniotomy were reviewed separately. Patients were categorized by whether they were treated at low-, medium-, or high-volume centers (< 10, 10–40, > 40 cases/year, respectively). A generalized linear model was used to evaluate the association of hospital pediatric CVM case volume and clinical variables assessing outcomes.

RESULTS

Among the 9655 patients, 1828 underwent craniotomy and 7827 did not. Patient age and race differed in the two groups, as did the rate of private medical payers. High-volume hospitals had fewer nonroutine discharges (11.2% [high] vs 16.4% [medium] vs 22.3% [low], p = 0.0001). For admissions requiring craniotomy, total charges ($106,282 [high] vs $126,215 [medium] vs $134,978 [low], p < 0.001) and complication rates (0.09% [high] vs 0.11% [medium] vs 0.16% [low], p = 0.001) were lower in high-volume centers.

CONCLUSIONS

This study revealed that further investigation may be needed regarding barriers to surgical treatment of pediatric CVMs. The authors found that surgical treatment of pediatric CVM at high-volume centers is associated with significantly fewer complications, better dispositions, and lower costs, but for noncraniotomy patients, low-volume centers had lower rates of complications and death and lower costs. These findings may support the consideration of appropriate referral of CVM patients requiring surgery or with intracranial hemorrhage toward high-volume, specialized centers.

In Brief

In this study, the authors utilized the Kids’ Inpatient Database (KID) to study the effect of hospital volume of pediatric cerebrovascular malformations (CVMs) on outcomes. This study is important because it identifies demographic trends and highlights the finding that children requiring craniotomy for intervention of CVM had better outcomes at high-volume centers. This finding can help direct these patients to high-volume centers in order to improve care and outcomes.

Cerebrovascular malformations (CVMs) are defined as developmental arterial and/or venous disorders of the brain, including arteriovenous malformations (AVMs), cavernous malformations and venous malformations, telangiectasias, and other anomalies.5,16 Previous pathological studies have indicated that the prevalence of CVMs is 360–4700 cases per 100,000 population.14,16 For AVMs specifically, the prevalence is between 140 and 521 cases per 100,000.16,17 Because of the relative rarity of CVMs in children, complete and accurate epidemiological data with demographics and outcomes are lacking. CVMs are associated with neurological morbidity, particularly ischemic stroke and intracranial hemorrhage (ICH), epilepsy, and potentially death. Hemorrhagic stroke, including ICH, accounts for 50% of pediatric strokes versus only 20% of adult strokes and is often secondary to ruptured AVMs and cavernous malformations.6,7 The risk of hemorrhage for untreated cerebral AVMs has been estimated at 2%–4% per year, with a mortality rate of 5%–12%18 and a 50% risk of serious neurological morbidity with each hemorrhage.11,24 Treatment can be supportive or can include operative intervention, including craniotomy for hematoma evacuation, decompressive craniectomy, or lesion excision. Resection can be done on an emergency basis or in a delayed, elective fashion.

There have been no US population–based studies regarding healthcare utilization among children with CVMs. Furthermore, data to objectively compare treatment of CVMs across institutions are lacking, although hospital and provider case volume have been demonstrated to be significant predictors of morbidity and mortality when treating other neurosurgical conditions.21

To address this current gap in knowledge, we analyzed admissions for CVMs, with and without craniotomy, from the Kids’ Inpatient Database (KID). We hypothesized that a higher institutional volume of pediatric CVMs would correlate directly with favorable financial and clinical outcomes. If true, this would inform the current debate centered on the utility of prioritizing referral of children with these disorders to high-volume centers.

Methods

Data Source

The data that support the findings of this study are available from the Kids’ Inpatient Database (https://www.hcup-us.ahrq.gov/kidoverview.jsp). The KID is a pediatric inpatient sample generated by the Healthcare Cost and Utilization Project (HCUP) triennially from 1997 to 2012. It is the largest US pediatric inpatient database available and contains more than 18 million records. Institutional review board approval is not required because all data are de-identified before distribution to researchers. Inclusion criteria for this study were age < 22 years and an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis of CVM with inpatient admission to a KID participant hospital. Given the rarity of the conditions studied, the propensity for high-level centers to treat patients until they are older, and the availability through the KID data set, age 22 years was used as the cutoff point.21 The ICD-9-CM designation 747.81 was used as a marker of CVM and encapsulates anomaly of the cerebral arteriovenous system, AVM (cerebral/congenital), cerebral AVM, congenital cerebral AVM, congenital cerebral AVM (at birth), and nonruptured congenital cerebral aneurysm. AVM is not separately coded in the ICD-9-CM; therefore, all pathologies of this type were grouped together. We also noted whether the identified patients with ICD-9-CM designation 747.81 underwent a craniotomy (ICD-9-CM 01.24) or had a diagnosis of ICH (ICD-9-CM 432.9).

Patient-level variables included age, sex, race, payer status, admission source, and admission type. Institutional variables included hospital region, location (rural or urban), teaching status (teaching or nonteaching), and number of beds. For each year, a hospital’s annual CVM admissions were calculated. Subsequently, all of the hospitals were categorized as high-, medium-, or low-volume institutions by determining cut points that divided annual total CVM caseload into thirds; these cut points were determined to be > 40, 10–40, and < 10 cases/year. This equalized numbers of admissions in each of the 3 volume categories. The methodology was adopted from a previous study.21

Outcomes

Endpoints for analysis included patient disposition, cost, and complications. Disposition was dichotomized based on previous investigations into two categories: routine and nonroutine.20 Routine discharge was defined as release to home or self-care. Nonroutine discharges included transfer to short-term hospital, skilled nursing facility, other transfer, home healthcare, or leaving against medical advice. Complications were determined from the ICD-9-CM in concordance with previously published methods,15,20,23 including wound, infectious, urinary, pulmonary, gastrointestinal, cardiovascular, systemic, intraoperative, neurological, and anesthetic complications, and any complications requiring additional procedures. The total number of complications was treated as a continuous variable, and specific complications were not analyzed individually. Cost included direct charges found within the database and was not adjusted using state-specific cost-to-charge ratios.

Statistical Analysis

Samples were weighted utilizing a weighting variable available within the database. We compared categorical variables using the chi-square test and continuous variables by 2-tailed Student t-test or one-way ANOVA. A generalized linear model was used to evaluate interactions between hospital volume, craniotomy, and ICH for various endpoints. The HCUP databases lack unique patient identifiers; therefore, the data do not allow for control of multiple measurements in a single patient.

Continuous and discrete values are represented as mean (± SEM) and percentages of total, respectively. For graphed results, means with 95% confidence intervals were analyzed. A p value < 0.05 was considered statistically significant. However, because of the limitations of p values in large administrative data sets, we further evaluated the effect size for significant results by calculating mean differences and confidence intervals for the variables of interest. All analyses were conducted using IBM SPSS (version 23.0, IBM Corp.).

Results

Demographics

A total of 9655 admissions for CVM were identified, with 7827 admissions without craniotomy and 1828 admissions with craniotomy (Table 1). White patients accounted for the largest group of CVM admissions in both nonoperative and operative groups, followed by Hispanic patients (Fig. 1 left). The percentage of black patients who did not undergo craniotomy was higher than the percentage of black patients who did undergo craniotomy (12.8% vs 9.6%). The male/female ratio for nonoperative and operative groups was not significantly different, demonstrating a slight female preponderance (54% female vs 46% male). Patient age differed between the noncraniotomy and craniotomy groups, with the nonoperative group averaging 3 years younger than the operative patients (9.1 ± 0.1 years of age vs 12.0 ± 0.1 years, p = 0.0001). Review of financial data revealed that patients who underwent craniotomy were more likely to be private payers, with a higher percentage of Medicaid in noncraniotomy admissions (35.8% vs 28.4%, p = 0.0001). Median family income did not differ between groups (Table 1).

TABLE 1.

Demographic, hospital, and outcome data among admissions for CVMs

All CVM Patients (n = 9655)CVM w/o Craniotomy (n = 7827)CVM w/ Craniotomy (n = 1828)p Value
Mean age, yrs9.6 ± 0.19.1 ± 0.112.0 ± 0.10.0001
Male sex, n (%)4405.9 (46.2)3580.0 (46.2)825.9 (46.0)0.9
Race, n (%)0.0001
 White4362.0 (55.4)3497.4 (55.1)864.6 (56.4)
 Black961.2 (12.2)813.8 (12.8)147.4 (9.6)
 Hispanic1713.8 (21.8)1352.3 (21.3)361.5 (23.6)
 Asian/Pacific Islander255.0 (3.2)194.0 (3.1)61.0 (4.0)
 Native American39.5 (0.5)27.5 (0.4)11.9 (0.8)
 Other546.3 (6.9)460.8 (7.3)85.5 (5.6)
Median family income, n (%)0.1
 $1–24,9992120 (22.8)1730.7 (23.0)389.7 (22.0)
 $25,000–34,9992213 (23.8)1818.3 (24.2)394.8 (22.2)
 $35,000–44,9992264 (24.4)1803.2 (24.0)460.5 (25.9)
 >$45,0002688 (28.9)2157.7 (28.7)529.9 (29.9)
Primary payer, n (%)0.0001
 Medicare28.4 (0.3)28.4 (0.4)0 (0.0)
 Medicaid3317.4 (34.4)2798.8 (35.8)518.7 (28.4)
 Private5281.6 (54.8)4177.8 (53.5)1103.8 (60.5)
 Self-pay350.5 (3.6)298.9 (3.8)51.6 (2.8)
 No charge25.5 (0.3)20.7 (0.3)4.8 (0.3)
 Other631.7 (6.6)485.6 (6.2)146.2 (8.0)
Mean no. of diagnoses, %4.9 ± 0.034.9 ± 0.044.5 ± 0.10.0001
Mean no. of procedures, %2.6 ± 0.042.2 ± 0.034.1 ± 0.10.0001
Major operative procedures, n (%)1963.4 (52.7)1197.9 (40.4)765.5 (100.0)0.0001
ICH, n (%)1019 (10.6)675 (8.6)344 (18.8)0.0001
Admission source, n (%)0.0001
 ED1516 (24.5)1313.3 (26.0)203.0 (17.7)
 Another hospital855 (13.8)744.8 (14.7)109.7 (9.5)
 Another facility163 (2.6)140.6 (2.8)22.3 (1.9)
 Court/law2 (0.03)0 (0.0)1.6 (0.1)
 Routine/birth/other3666 (59.1)2853.7 (56.5)812.6 (70.7)
Admission type, n (%)0.0001
 Emergency1996 (31.5)1762.6 (33.7)233.8 (21.0)
 Urgent1186 (18.7)1004.8 (19.2)181.7 (16.3)
 Elective2842 (44.8)2151.2 (41.1)691.1 (62.1)
 Newborn291 (4.6)291.2 (5.6)0 (0.0)
 Other25 (0.4)18.9 (0.4)6.1 (0.6)
Disposition, n (%)0.0001
 Routine8058 (83.5)6533.9 (83.5)1524.3 (83.4)
 Short-term hospital516 (5.3)475.0 (6.1)41.0 (2.2)
 Other facility515 (5.3)337.6 (4.3)177.9 (9.7)
 Home health281 (2.9)235.8 (3.0)45.7 (2.5)
 AMA11 (0.1)9.6 (0.1)1.3 (0.1)
 Died273 (2.8)234.6 (3.0)38.1 (2.1)
Patient transfer in, n (%)533 (14.4)464.9 (15.8)68.3 (8.9)0.0001
Patient transfer out, n (%)82 (4.4)72.5 (5.0)9.2 (2.3)0.0001
Hospital size, n (%)0.05
 Small999 (10.9)783.3 (10.5)215.3 (12.5)
 Medium2165 (23.6)1769.7 (23.8)395.2 (23.0)
 Large6000 (65.5)4894.7 (65.7)1105.5 (64.4)
Hospital type, n (%)0.0001
 Rural141 (1.5)136.3 (1.8)4.6 (0.3)
 Urban nonteaching1045 (11.4)944.3 (12.7)100.8 (5.9)
 Urban teaching7978 (87.1)6367.0 (85.5)1610.6 (93.9)
Hospital region, n (%)0.0001
 Northeast2000 (21.3)1700.8 (22.4)299.0 (16.9)
 Midwest1765 (18.8)1460.1 (19.2)304.9 (17.3)
 South2837 (30.3)2313.3 (30.4)523.7 (29.7)
 West2772 (39.6)2134.5 (28.1)637.1 (36.1)
Mean length of stay, days6.9 ± 0.16.4 ± 0.19.2 ± 0.30.0001
Mean total charges, $66,864.00 ± 1211.0054,501.00 ± 1251.50120,173.00 ± 3209.000.0001
Mean complication rate0.07 ± 0.0060.1 ± 0.0070.0001

AMA = discharge against medical advice; ED = emergency department.

Patient numbers are weighted. Mean values are presented as the mean ± SEM.

FIG. 1.
FIG. 1.

Pediatric CVM in the US. Left: Race distribution. Right: Concentration of CVM cases over time. Figure is available in color online only.

Admission Characteristics

Approximately 50% of all admissions were emergency or urgent, whereas approximately 45% were elective. At admission, approximately 11% of patients had a concomitant diagnosis of ICH (Table 1). There were significant differences in the admission characteristics between patients who did not undergo a craniotomy and those who did undergo a craniotomy. Patients with CVM who underwent craniotomy were more likely than those who did not have a craniotomy to have a routine admission (70.7% vs 56.5%, p = 0.0001) and had a smaller mean number of diagnoses (4.5 ± 0.1 vs 4.9 ± 0.04, p = 0.0001) (Table 1). The majority of craniotomies were elective admissions (62.1% vs 41.1%, p = 0.0001). Patients who underwent a craniotomy were significantly more likely to have a concomitant diagnosis of ICH than patients who did not undergo a craniotomy (18.8% vs 8.6%, p = 0.0001).

In contrast, patients who had CVM without craniotomy were more likely than patients who had a craniotomy to have emergency admissions (33.7% vs 21.0%, p = 0.0001) and were more likely to be admitted directly from the emergency department (26.0% vs 17.7%, p = 0.0001) or from another hospital (14.7% vs 9.5%, p = 0.0001).

Hospital characteristics and charges differed by admission as well. Patients with craniotomy were more likely to be admitted to an urban teaching hospital (93.9% vs 85.5%, p = 0.0001) than patients who did not have a craniotomy. Patient charges were higher in those undergoing surgical treatment than in those in the nonoperative group ($120,173.00 ± $3209.00 vs $54,501.00 ± $1251.50, p = 0.0001) (Table 1).

Hospital Volume

Hospitals were divided into low-, medium-, and high-volume centers. Admissions for CVM were concentrated at a small number of high-volume centers (Fig. 1 right). Whereas one-third of CVM admissions were dispersed over 955 low-volume hospitals (< 10 cases/year), a similar number of CVM admissions were seen by just 38 high-volume centers (> 40 cases annually) (Table 2). High-volume centers were exclusively urban hospitals (both teaching and nonteaching). Patients with CVM who were admitted to high-volume centers were more likely to undergo a craniotomy than those admitted to medium- and low-volume hospitals (22.2% vs 20.6% vs 13.7%, respectively, p = 0.0001). From 2009 through 2012, there has been an increase in admissions for CVM to low- and medium-volume centers and a decrease in admissions for CVM to high-volume centers, indicating a recent decentralization of care (Fig. 1).

TABLE 2.

Hospital characteristics by hospital CVM volume for 1997–2012

Hospital Vol, Cases/Yr
<10 (n = 3219)10–40 (n = 3357)>40 (n = 3079)p Value
No. of hospitals95516838
Mean age, yrs11.2 ± 0.19.4 ± 0.18.4 ± 0.10.0001
Male sex, n (%)1462.4 (47.7)1496.0 (44.9)1447.6 (46.0)0.07
Race, n (%)0.0001
 White1368.5 (55.3)1491.4 (54.0)1502.2 (56.9)
 Black352.4 (14.2)372.2 (13.5)236.6 (9.0)
 Hispanic514.9 (20.8)633.4 (22.9)565.5 (21.4)
 Asian/Pacific Islander81.3 (3.3)95.6 (3.5)78.1 (3.0)
 Native American5.4 (0.2)16.4 (0.6)17.7 (0.7)
 Other154.3 (6.2)153.6 (5.6)238.4 (9.0)
Mean length of stay, days7.0 ± 0.27.1 ± 0.26.7 ± 0.20.4
Major operative procedures, n (%)615.7 (43.6)741.7 (53.7)606.1 (64.8)0.0001
ICH, n (%)399.0 (13.0)348.6 (10.4)271.2 (8.4)0.0001
Craniotomy, n (%)422.6 (13.7)690.2 (20.6)715.6 (22.2)0.0001
Disposition, n (%)0.0001
 Routine2391.5 (77.7)2807.9 (83.6)2858.8 (88.8)
 Short-term hospital311.5 (10.1)141.8 (4.2)62.8 (2.0)
 Other facility194.0 (6.3)198.6 (5.9)122.9 (3.8)
 Home health94.0 (3.1)117.5 (3.5)70.0 (2.2)
 AMA7.6 (0.2)3.3 (0.1)
 Died80.3 (2.6)88.1 (2.6)104.3 (3.2)
Hospital size, n (%)0.0001
 Small294.8 (10.0)344.5 (11.1)359.4 (11.6)
 Medium657.4 (22.3)570.1 (18.3)937.5 (30.2)
 Large1994.5 (67.7)2202.4 (70.7)1803.4 (58.2)
Hospital type, n (%)0.0001
 Rural96.8 (3.3)44.2 (1.4)0 (0.0)
 Urban nonteaching787.5 (26.7)208.8 (6.7)48.8 (1.6)
 Urban teaching2062.3 (70.0)2863.9 (91.9)3051.4 (98.4)
Hospital region, n (%)0.0001
 Northeast553.5 (18.5)713.8 (22.2)732.6 (23.1)
 Midwest735.0 (24.6)546.0 (17.0)483.9 (15.3)
 South932.7 (31.2)1167.3 (36.3)737.1 (23.2)
 West768.5 (25.7)790.5 (24.6)1212.6 (38.3)
Mean total charges, $59,446.60 ± 1976.7070,061.40 ± 2378.0070,449.70 ± 1867.600.0001
Mean complications0.07 ± 0.0060.07 ± 0.0060.1 ± 0.0070.0001

Patient numbers are weighted. Mean values are presented as the mean ± SEM.

Hospital Volume Effect on Patient Outcomes

Hospital volume had a direct correlation with patient outcomes, including length of stay, charges, complications, and discharge disposition/death (Table 3). The results, including the difference between paired groups (i.e., effect size), are displayed in Table 3. Length of stay was longer in patients with CVM who carried a diagnosis of ICH (p = 0.05) (Table 3, Fig. 2A and B). For patients with ICH, the LOS was shorter for low-volume centers than for medium- or high-volume centers (Fig. 2C and D).

TABLE 3.

Evaluation of outcomes based on annual hospital volume

Hospital Vol, Cases/Yr*
Outcome Clinical Variable<10 (n = 3219)10–40 (n = 3357)>40 (n = 3079)p Value
Mean length of stay, days
 ICH10.8 ± 0.613.1 ± 0.612.4 ± 0.80.05
 No ICH6.4 ± 0.26.4 ± 0.26.2 ± 0.2
 Craniotomy9.3 ± 0.59.3 ± 0.49.1 ± 0.60.7
 No craniotomy6.6 ± 0.26.5 ± 0.26.0 ± 0.2
Mean total charges, $
 ICH93,137 ± 8222110,894 ± 8177124,780 ± 10,2000.2
 No ICH54,571 ± 190665,426 ± 246665,628 ± 1783
 Craniotomy134,978 ± 8281126,215 ± 5270106,282 ± 42370.0001
 No craniotomy47,705 ± 177855,680 ± 259060,234 ± 2028
Mean complications, %
 ICH0.19 ± 0.030.18 ± 0.030.13 ± 0.030.006
 No ICH0.05 ± 0.010.05 ± 0.0050.09 ± 0.01
 Craniotomy0.16 ± 0.030.11 ± 0.020.09 ± 0.010.0001
 No craniotomy0.05 ± 0.010.05 ± 0.010.1 ± 0.01
Nonroutine disposition, n (%)
 ICH168 (5.2)115 (3.4)73 (2.4)0.1
 No ICH519 (16.1)434 (12.9)287 (9.3)
 Craniotomy168 (5.2)115 (3.4)73 (2.4)0.2
 No craniotomy519 (16.1)434 (12.9)287 (9.3)
Death, n (%)
 ICH21 (0.7)18 (0.5)19 (0.6)0.99
 No ICH59 (1.8)70 (2.1)85 (2.8)
 Craniotomy14 (0.4)12 (0.4)11 (0.4)0.02
 No craniotomy66 (2.1)76 (2.3)93 (3.0)

Mean values are presented as the mean ± SEM.

A generalized linear model was used to evaluate the association of hospital volume and clinical variables on outcomes.

FIG. 2.
FIG. 2.

Hospital outcomes by CVM volume. A and B: Comparison of length of stay with or without craniotomy (A) and with or without ICH (B). C and D: Comparison of total charges with or without craniotomy (C) and with or without ICH (D). Dots are means and error bars the 95% CI. Values on the x-axis are number of cases/year. Figure is available in color online only.

The trend for charges correlating with hospital volume depended on whether or not a patient had a craniotomy. If no craniotomy was performed, then charges increased with higher-volume centers ($60,234 high-volume vs $55,680 medium-volume vs $47,705 low-volume, p = 0.0001). In contrast, if a craniotomy was performed, then higher-volume centers cost significantly less ($106,282 high-volume vs $126,215 medium-volume vs $134,978 low-volume, p = 0.0001) (Table 3, Fig. 2C). Overall charges were higher for CVM admissions at high-volume centers with or without ICH.

There was a strong association between higher volume and lower rates of patient complications depending on whether a patient underwent a craniotomy (p = 0.0001) or if the patient also had a diagnosis of ICH (p = 0.006). Craniotomy at high-volume centers had the lowest risk of complications (Table 3, Fig. 3A and B). Similarly, the complication profile for admissions with ICH was lowest for high-volume centers (0.13% ± 0.03% high-volume vs 0.18% ± 0.03% medium-volume vs 0.19% ± 0.03% low-volume, p = 0.006) (Table 3, Fig. 3C).

FIG. 3.
FIG. 3.

Hospital outcomes by CVM volume. Comparison of percentage of complications with or without craniotomy (A), mean number of complications with or without craniotomy (B), and percentage of complications with or without ICH (C). Figure is available in color online only.

Regarding discharge, higher volume was significantly associated with better rates of routine discharge (88.8% high-volume, 83.6% medium-volume, 77.7% low-volume, p = 0.0001, Table 2). Although it was not statistically significant (p = 0.1), there was a trend toward lower rates of nonroutine discharge with higher volume in both craniotomy and noncraniotomy patients (Table 3). Mortality rates were generally low and not significantly different among centers with different volumes, with the sole exception of a slightly higher mortality rate for noncraniotomy patients in high-volume centers (3.0% high-volume vs 2.3% medium-volume vs 2.1% low-volume, p = 0.02).

Discussion

The KID is the largest pediatric inpatient database and is comparable to the National Inpatient Sample (NIS), containing nearly 40% of all inpatient records for the years acquired. A major advantage of the KID database is that it allows for examination of national trends in utilization, charges, quality, and outcomes. This advantage is particularly pronounced when studying rare conditions such as CVM, providing large cohorts of patients while eliminating any population or practice bias inherent to single-center studies. Previous studies using KID data have looked at specific operative procedures (such as moyamoya surgery or admissions for pediatric cerebrovascular operations) and have demonstrated clear associations between higher volume and better procedural outcomes.21 This study differs from—and is complementary to—previous work, in that it examines volume relationships in both operative and nonoperative patient groups sharing a single ICD-9-CM code while adding demographic and socioeconomic data. We sought to use the KID to eliminate the knowledge gaps in describing the epidemiological landscape in children with CVMs and to better define the correlation between hospital volume and cost, treatment, and clinical outcomes. To our knowledge, this is the first study of its kind to examine pediatric patients with CVMs by employing these types of analyses using a national database.

Demographics

In this investigation, we identified baseline demographics for pediatric patients with CVM. Our data revealed that males and females are affected equally, that CVM affects children of all economic backgrounds without difference, and that the average age at presentation is about 9 years old. For the entire cohort, the data suggest that CVM affects all races equally, with presentation by race closely reflecting the existing percentages within the general US population.

However, an important discrepancy was revealed when the cohort of children who had a craniotomy and the cohort of those who did not have a craniotomy were analyzed by race and financial/insurer data. The percentage of black patients who did not undergo craniotomy was higher than the percentage of black patients who did undergo craniotomy (12.8% vs 9.6%, p < 0.05). There have been no previous studies on racial bias for CVM treatment—only regional variations based on population density. Analysis of insurance data revealed that primary payers for admissions with craniotomy were private, followed by Medicaid (60.5% vs 28.4%). Lastly, there is a nonsignificant trend toward higher relative rates of craniotomy with increasing family income. Taken together, these data suggest that further investigation into racial and financial barriers to operative treatment may be indicated, which could reveal factors such as lower socioeconomic status or bias/fear of surgical intervention.

Volume-Outcome Associations

There is significant literature analyzing hospital case volume and patient outcome in studies of adult and pediatric neurosurgical care. Recently, Titsworth et al.21 performed an analysis of pediatric moyamoya admissions and demonstrated that high-volume centers provide better care with lower mortality in moyamoya patients—a difference most pronounced in children requiring surgical revascularization. In the neurosurgical literature, several studies that have utilized the NIS have shown lower morbidity and mortality, shorter LOS, and lower hospital charges in the treatment of adults with cerebral aneurysms,3,12 meningiomas,9 intracranial metastasis,2 and transsphenoidal surgery for pituitary tumors.4 Studies in pediatric neurosurgery have mirrored these results. An analysis of unruptured cerebral aneurysms in children using KID demonstrated lower mortality rates in teaching hospitals (OR 0.13, 95% CI 0.03–0.46) and hospitals with more beds (OR 0.35, 95% CI 0.06–1.92).1 Additionally, an investigation of 4712 admissions for pediatric brain tumor craniotomies demonstrated a lower mortality rate in higher-volume hospitals (2.3% vs 1.4%).19 To date, however, this sort of analysis is lacking for pediatric CVMs.

We discovered that from 2009 through 2012 there has been an increase in admissions for CVM to low- and medium-volume centers and a decrease in admissions to high-volume centers, indicating a recent decentralization of care (Fig. 1 right). This could be due to increased pressure on low-volume centers to adopt or expand complex neurosurgical practices, thus increasing the threshold to transfer patients. Similar trends have been demonstrated in adult spinal cord injury, resulting in worsened outcomes for those patients.13 The data presented in this investigation indicate that most metrics relevant to cost and quality of care are better with higher-volume centers and that these differences are particularly significant and favorable for patients undergoing craniotomy or with ICH. When patients had ICH or needed a craniotomy, the rate of complications was almost twofold lower at high-volume centers. Similarly, high-volume centers had the highest rates of routine discharge after treatment. In addition to clinical metrics, financial outcomes were also affected by patient volume. When craniotomy was performed, hospital costs were 27% lower at high-volume centers compared with low-volume centers, suggesting that better clinical outcomes can be provided at lower cost when treated at a high-volume center.

Importantly, not all data favored high-volume centers. In particular, for patients with less complex cases (without surgery and/or without ICH), low-volume centers had lower rates of complications (without surgery: 0.05% low-volume vs 0.05% medium-volume vs 0.1% high-volume, p < 0.05) and death (without surgery: 2.1% low-volume vs 2.3% medium-volume vs 3.0% high-volume, p = 0.02), along with lower costs (without surgery: $47,705 low-volume vs $55,680 medium-volume vs $60,234 high-volume, p = 0.001). This is likely because patients with less complex disease (i.e., those without ICH or a need for craniotomy) treated at low-volume centers have low rates of complications and may indeed be appropriately treated at lower-volume centers. Once the patient has been diagnosed, treated, and discharged and potentially referred for craniotomy, the data may be reflected under the category of high-volume centers.

Limitations and Future Directions

There are several limitations to the current investigation. Inherently, the KID is a retrospective tool for collecting data. As with any administrative database, there can be variability in data collection and reporting. Previous studies have estimated an approximately 80% accuracy of ICD-9-CM codes in national administrative data sources;8 however, there is a growing literature base describing the pitfalls of research based on administrative databases.22 To address these issues, we have used craniotomy and ICH as coding points to help increase the confidence of the findings in the current analysis. Dasenbrock et al.10 described a similar algorithm at a single institution to identify decompressive craniectomy for stroke based on diagnosis codes and found a 97.8% sensitivity and 99.9% specificity. Despite this limitation, we believe the data presented are representative of a cross-sectional analysis of the trends in CVM care in the US and provide insights regarding trends in care.

Another significant limitation of the study is the heterogeneity of the ICD-9-CM designation 747.81. It encapsulates anomaly of the cerebral arteriovenous system, AVM (cerebral/congenital), cerebral AVM, congenital cerebral AVM, congenital cerebral AVM (at birth), and nonruptured congenital cerebral aneurysm. Importantly, the code excludes ruptured cerebral aneurysm. As mentioned previously, the most common clinically relevant lesions for this code are AVMs and cavernous malformations. Berman et al.5 found that 5 lesion types accounted for 99% of the diagnoses, and the 2 most common were AVM (66%) and cavernous malformations (13%). They also found that the sensitivity of the coding was 94% accurate for identifying CVMs.5 Ultimately, although this code is only a surrogate for AVM and its use means that our study will have some non-AVM patients, it remains the best means possible for capturing this population. We have made efforts to use the term “CVM” in our work to acknowledge this imprecision, but the data cited from other studies suggest that the vast majority of our patients will be our target population of pediatric AVMs. Even with this limitation, we believe that our results help provide vital information in the demographics of CVMs and the treatment patterns and outcomes in children.

An additional potential for bias when comparing low-, medium-, and high-volume centers can be seen when assessing elective versus emergency admissions. Elective admissions generally carry lower cost and complication rates, and subsequently improved disposition. This should be taken into consideration, specifically for AVMs that initially present with hemorrhage and undergo resection in a delayed fashion once recovery has taken place. Although specific data on whether elective admissions were more likely at high-volume centers were unavailable from the KID database, we performed additional univariate and multivariate analyses (not reported) that demonstrated that admission type—elective vs emergency—was not significantly associated with length of stay, total charges, complications, or nonroutine disposition.

The socioeconomic disparity in this study is concerning, although it may represent coding errors (e.g., patients with sickle cell disease coded as CVM). Thus, further detailed analysis, which is outside the scope of the current paper, is necessary.

An additional limitation lies in the inherent data and the accuracy of the coding. With respect to complications, as noted in Results, we discovered a decreased incidence of complications with craniotomy at high-volume centers but also a rise in the mean number of complications at high-volume centers without craniotomy. Internal validation of the accuracy of the data could not be performed based on the analysis parameters we set. Administrative databases propose certain inferences be made from the data; we encourage all practitioners to use the data reported here in this light, acknowledging the potential shortcomings of the data set. With large data sets, especially administrative databases, the sheer volume of data can lead to all comparisons having statistical significance; however, this may not equate to clinical significance.

The KID tracks individual admissions but does not have unique patient identifiers. This prevents long-term tracking of single patients outside of a given admission. Even with these limitations, the KID is a useful way to investigate CVM admissions in pediatric patients on a national platform. Nevertheless, further multicenter, potentially multinational studies are necessary to validate the findings of this investigation.

Conclusions

This is the first study to utilize a national database to analyze American pediatric CVM demographic epidemiological information and evaluate the effect of hospital volume on patient outcome. This investigation reveals novel socioeconomic characteristics in the population of US pediatric patients with CVM and suggests that further study may be needed into racial and economic barriers to surgical treatment of these lesions. We demonstrate that surgical treatment of pediatric CVM at high-volume centers is associated with significantly fewer complications, better dispositions, and lower costs, although for patients with less complex disease (not needing surgery and/or without ICH), low-volume centers had lower rates of complications and death and lower costs. The aforementioned findings may support the consideration of appropriate referral of CVM patients with ICH toward high-volume, specialized centers.

Acknowledgments

We thank Kristin Kraus, MSc, for editorial assistance in preparing this paper.

The study was supported by the American Association of Neurological Surgeons/Congress of Neurological Surgeons Joint Pediatric Section Resident Traveling Fellowship (2017). We also acknowledge The Lucas Warner AVM Research Fund, The Fellows Fund, and the Credit Unions Kids at Heart organization for their support.

Disclosures

The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

Author Contributions

Conception and design: Ravindra, Smith. Analysis and interpretation of data: Ravindra, Karsy. Drafting the article: Ravindra. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Ravindra. Administrative/technical/material support: Smith.

References

  • 1

    Alawi AEdgell RCElbabaa SKCallison RCKhalili YAAllam H: Treatment of cerebral aneurysms in children: analysis of the Kids’ Inpatient Database. J Neurosurg Pediatr 14:23302014

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 2

    Barker FG II: Craniotomy for the resection of metastatic brain tumors in the U.S., 1988–2000: decreasing mortality and the effect of provider caseload. Cancer 100:99910072004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3

    Barker FG IIAmin-Hanjani SButler WEOgilvy CSCarter BS: In-hospital mortality and morbidity after surgical treatment of unruptured intracranial aneurysms in the United States, 1996–2000: the effect of hospital and surgeon volume. Neurosurgery 52:99510092003

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Barker FG IIKlibanski ASwearingen B: Transsphenoidal surgery for pituitary tumors in the United States, 1996–2000: mortality, morbidity, and the effects of hospital and surgeon volume. J Clin Endocrinol Metab 88:470947192003

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5

    Berman MFStapf CSciacca RRYoung WL: Use of ICD-9 coding for estimating the occurrence of cerebrovascular malformations. AJNR Am J Neuroradiol 23:7007052002

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Beslow LALicht DJSmith SEStorm PBHeuer GGZimmerman RA: Predictors of outcome in childhood intracerebral hemorrhage: a prospective consecutive cohort study. Stroke 41:3133182010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7

    Brown RD JrWiebers DOTorner JCO’Fallon WM: Incidence and prevalence of intracranial vascular malformations in Olmsted County, Minnesota, 1965 to 1992. Neurology 46:9499521996

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Burns EMRigby EMamidanna RBottle AAylin PZiprin P: Systematic review of discharge coding accuracy. J Public Health (Oxf) 34:1381482012

  • 9

    Curry WTMcDermott MWCarter BSBarker FG II: Craniotomy for meningioma in the United States between 1988 and 2000: decreasing rate of mortality and the effect of provider caseload. J Neurosurg 102:9779862005

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 10

    Dasenbrock HHCote DJPompeu YVasudeva VSSmith TRGormley WB: Validation of an International Classification of Disease, Ninth Revision coding algorithm to identify decompressive craniectomy for stroke. BMC Neurol 17:1212017

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Gerosa MACappellotto PLicata CIraci GPardatscher KFiore DL: Cerebral arteriovenous malformations in children (56 cases). Childs Brain 8:3563711981

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12

    Hoh BLRabinov JDPryor JCCarter BSBarker FG II: In-hospital morbidity and mortality after endovascular treatment of unruptured intracranial aneurysms in the United States, 1996–2000: effect of hospital and physician volume. AJNR Am J Neuroradiol 24:140914202003

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13

    Holland CMMazur MDBisson EFSchmidt MHDailey AT: Trends in patient care for traumatic spinal injuries in the United States: a National Inpatient Sample study of the correlations with patient outcomes from 2001 to 2012. Spine (Phila Pa 1976) 42:192319292017

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14

    Jellinger K: Vascular malformations of the central nervous system: a morphological overview. Neurosurg Rev 9:1772161986

  • 15

    LaPar DJKron ILJones DRStukenborg GJKozower BD: Hospital procedure volume should not be used as a measure of surgical quality. Ann Surg 256:6066152012

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    McCormick W: Pathology of vascular malformations of the brain in Wilson CStein B (eds): Intracranial Arteriovenous Malformations. Baltimore: Williams & Wilkins1984 pp 4463

    • Search Google Scholar
    • Export Citation
  • 17

    Michelsen WJ: Natural history and pathophysiology of arteriovenous malformations. Clin Neurosurg 26:3073131979

  • 18

    Riordan CPOrbach DBSmith ERScott RM: Acute fatal hemorrhage from previously undiagnosed cerebral arteriovenous malformations in children: a single-center experience. J Neurosurg Pediatr 22:2442502018

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 19

    Smith ERButler WEBarker FG II: Craniotomy for resection of pediatric brain tumors in the United States, 1988 to 2000: effects of provider caseloads and progressive centralization and specialization of care. Neurosurgery 54:5535652004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 20

    Stone MLLapar DJKane BJRasmussen SKMcGahren EDRodgers BM: The effect of race and gender on pediatric surgical outcomes within the United States. J Pediatr Surg 48:165016562013

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Titsworth WLScott RMSmith ER: National analysis of 2454 pediatric moyamoya admissions and the effect of hospital volume on outcomes. Stroke 47:130313112016

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 22

    van Walraven CAustin P: Administrative database research has unique characteristics that can risk biased results. J Clin Epidemiol 65:1261312012

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 23

    Wang HHTejwani RZhang HWiener JSRouth JC: Hospital surgical volume and associated postoperative complications of pediatric urological surgery in the United States. J Urol 194:5065112015

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 24

    Wilkins RH: Natural history of intracranial vascular malformations: a review. Neurosurgery 16:4214301985

If the inline PDF is not rendering correctly, you can download the PDF file here.

Article Information

Correspondence Vijay M. Ravindra: Clinical Neurosciences Center, University of Utah, Salt Lake City, UT. neuropub@hsc.utah.edu.

INCLUDE WHEN CITING Published online August 2, 2019; DOI: 10.3171/2019.5.PEDS19155.

Disclosures The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

© AANS, except where prohibited by US copyright law.

Headings

Figures

  • View in gallery

    Pediatric CVM in the US. Left: Race distribution. Right: Concentration of CVM cases over time. Figure is available in color online only.

  • View in gallery

    Hospital outcomes by CVM volume. A and B: Comparison of length of stay with or without craniotomy (A) and with or without ICH (B). C and D: Comparison of total charges with or without craniotomy (C) and with or without ICH (D). Dots are means and error bars the 95% CI. Values on the x-axis are number of cases/year. Figure is available in color online only.

  • View in gallery

    Hospital outcomes by CVM volume. Comparison of percentage of complications with or without craniotomy (A), mean number of complications with or without craniotomy (B), and percentage of complications with or without ICH (C). Figure is available in color online only.

References

  • 1

    Alawi AEdgell RCElbabaa SKCallison RCKhalili YAAllam H: Treatment of cerebral aneurysms in children: analysis of the Kids’ Inpatient Database. J Neurosurg Pediatr 14:23302014

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 2

    Barker FG II: Craniotomy for the resection of metastatic brain tumors in the U.S., 1988–2000: decreasing mortality and the effect of provider caseload. Cancer 100:99910072004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3

    Barker FG IIAmin-Hanjani SButler WEOgilvy CSCarter BS: In-hospital mortality and morbidity after surgical treatment of unruptured intracranial aneurysms in the United States, 1996–2000: the effect of hospital and surgeon volume. Neurosurgery 52:99510092003

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Barker FG IIKlibanski ASwearingen B: Transsphenoidal surgery for pituitary tumors in the United States, 1996–2000: mortality, morbidity, and the effects of hospital and surgeon volume. J Clin Endocrinol Metab 88:470947192003

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5

    Berman MFStapf CSciacca RRYoung WL: Use of ICD-9 coding for estimating the occurrence of cerebrovascular malformations. AJNR Am J Neuroradiol 23:7007052002

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Beslow LALicht DJSmith SEStorm PBHeuer GGZimmerman RA: Predictors of outcome in childhood intracerebral hemorrhage: a prospective consecutive cohort study. Stroke 41:3133182010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7

    Brown RD JrWiebers DOTorner JCO’Fallon WM: Incidence and prevalence of intracranial vascular malformations in Olmsted County, Minnesota, 1965 to 1992. Neurology 46:9499521996

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Burns EMRigby EMamidanna RBottle AAylin PZiprin P: Systematic review of discharge coding accuracy. J Public Health (Oxf) 34:1381482012

  • 9

    Curry WTMcDermott MWCarter BSBarker FG II: Craniotomy for meningioma in the United States between 1988 and 2000: decreasing rate of mortality and the effect of provider caseload. J Neurosurg 102:9779862005

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 10

    Dasenbrock HHCote DJPompeu YVasudeva VSSmith TRGormley WB: Validation of an International Classification of Disease, Ninth Revision coding algorithm to identify decompressive craniectomy for stroke. BMC Neurol 17:1212017

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Gerosa MACappellotto PLicata CIraci GPardatscher KFiore DL: Cerebral arteriovenous malformations in children (56 cases). Childs Brain 8:3563711981

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12

    Hoh BLRabinov JDPryor JCCarter BSBarker FG II: In-hospital morbidity and mortality after endovascular treatment of unruptured intracranial aneurysms in the United States, 1996–2000: effect of hospital and physician volume. AJNR Am J Neuroradiol 24:140914202003

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13

    Holland CMMazur MDBisson EFSchmidt MHDailey AT: Trends in patient care for traumatic spinal injuries in the United States: a National Inpatient Sample study of the correlations with patient outcomes from 2001 to 2012. Spine (Phila Pa 1976) 42:192319292017

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14

    Jellinger K: Vascular malformations of the central nervous system: a morphological overview. Neurosurg Rev 9:1772161986

  • 15

    LaPar DJKron ILJones DRStukenborg GJKozower BD: Hospital procedure volume should not be used as a measure of surgical quality. Ann Surg 256:6066152012

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    McCormick W: Pathology of vascular malformations of the brain in Wilson CStein B (eds): Intracranial Arteriovenous Malformations. Baltimore: Williams & Wilkins1984 pp 4463

    • Search Google Scholar
    • Export Citation
  • 17

    Michelsen WJ: Natural history and pathophysiology of arteriovenous malformations. Clin Neurosurg 26:3073131979

  • 18

    Riordan CPOrbach DBSmith ERScott RM: Acute fatal hemorrhage from previously undiagnosed cerebral arteriovenous malformations in children: a single-center experience. J Neurosurg Pediatr 22:2442502018

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 19

    Smith ERButler WEBarker FG II: Craniotomy for resection of pediatric brain tumors in the United States, 1988 to 2000: effects of provider caseloads and progressive centralization and specialization of care. Neurosurgery 54:5535652004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 20

    Stone MLLapar DJKane BJRasmussen SKMcGahren EDRodgers BM: The effect of race and gender on pediatric surgical outcomes within the United States. J Pediatr Surg 48:165016562013

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Titsworth WLScott RMSmith ER: National analysis of 2454 pediatric moyamoya admissions and the effect of hospital volume on outcomes. Stroke 47:130313112016

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 22

    van Walraven CAustin P: Administrative database research has unique characteristics that can risk biased results. J Clin Epidemiol 65:1261312012

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 23

    Wang HHTejwani RZhang HWiener JSRouth JC: Hospital surgical volume and associated postoperative complications of pediatric urological surgery in the United States. J Urol 194:5065112015

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 24

    Wilkins RH: Natural history of intracranial vascular malformations: a review. Neurosurgery 16:4214301985

TrendMD

Metrics

Metrics

All Time Past Year Past 30 Days
Abstract Views 155 155 155
Full Text Views 20 20 20
PDF Downloads 10 10 10
EPUB Downloads 0 0 0

PubMed

Google Scholar