Relationship Between the Volume of Craniotomies for Cerebral Aneurysm Performed at New York State Hospitals and In-Hospital Mortality
Background and Purpose After a craniotomy for cerebral aneurysm, postoperative mortality can be significant. Previous studies have shown that hospitals performing frequent high-risk procedures (such as coronary artery bypass) have a lower mortality than hospitals where these procedures are performed infrequently.
Methods The Statewide Planning and Research Cooperative System of the New York State Department of Health reviewed all discharges in New York State from 1987 through 1993 for the diagnoses of subarachnoid hemorrhage and/or cerebral aneurysm and for patients with the procedure code for craniotomy for ruptured or unruptured cerebral aneurysm. In-hospital mortality and length of stay were examined in relation to the volume of craniotomies for aneurysm performed at each individual hospital.
Results A total of 15 376 discharges for subarachnoid hemorrhage and 5638 craniotomies for aneurysm were tabulated in 208 hospitals. For all patients who underwent craniotomy for ruptured cerebral aneurysm (n=4034), there was a 43% (95% confidence interval, 29% to 57%) reduction in mortality rate in hospitals performing more than 30 craniotomies per year for cerebral aneurysm compared with hospitals performing less surgery (8.8% versus 15.5%, P<.0001). For all patients who underwent craniotomy for unruptured cerebral aneurysm (n=1604), there was an identical 43% (95% confidence interval, 14% to 73%) reduction in mortality in hospitals performing more than 30 craniotomies per year for cerebral aneurysm (4.6% versus 8.1%, P=.0087).
Conclusions Hospitals that frequently perform aneurysm operations have lower mortality rates for patients undergoing craniotomy for cerebral aneurysm than hospitals that perform fewer operations.
The relationship between surgical volume and mortality rates has previously been well established for selected procedures such as CABG, hip replacement, and balloon angioplasty.1 2 3 4 5 For CABG procedures in New York State, publication of mortality-volume relationships has directly led to a change in state policy, hospital privileges, and referral patterns that subsequently translated into a decline in procedure-related mortality.2
Intracranial aneurysm surgery can be a high-risk neurosurgical operation that, like CABG surgery, carries some procedure-related mortality even in the best of hands. Published mortality figures in the past have generally originated from high-volume centers employing neurosurgeons with a subspecialty interest in cerebrovascular surgery.6 7 8 9 10 Factors such as aneurysm size, location, and neurological condition of the patient were considered the most important predictors of outcome.
The volume of intracranial aneurysm surgery performed has never been previously addressed as a predictor of surgical mortality. The hypothesis underlying the present study is that ongoing experience with craniotomy for cerebral aneurysm is required to achieve and maintain the lowest possible mortality rate.
Materials and Methods
SPARCS was implemented by the New York State Department of Health in 1979, with the cooperation and initial financial support of the US Department of Health and Human Services. SPARCS receives, processes, stores, and analyzes inpatient data from all general hospitals in New York State. It is a comprehensive, integrated information network for healthcare resource planning, financial analysis, and surveillance of hospital service in New York State.
SPARCS makes use of two existing data sources: the DDA and the UBF. The DDA data set collected by SPARCS is based on the Uniform Hospital Discharge Data Set, which was developed by the National Committee on Vital and Health Statistics. The UBF, which is used as the uniform claim form for all third-party payers in New York State, is also a reporting document for SPARCS. Both data sets are submitted to SPARCS electronically. Each hospital’s medical record department submits data to SPARCS DDA in a uniform, computer-readable format. These data are sent to the Department of Health either directly by the hospital or through one of a number of private information processing services. Every DDA received by SPARCS is edited to identify errors. Each data element must have a valid value before the record is accepted by the system. When a DDA needs correction, the hospital or processing service is notified. DDAs are also screened carefully for duplicate submission. When a UBF has been completed for an inpatient discharge, the hospital’s billing department submits data required by SPARCS in a uniform, computer-readable format. Errors are identified, and hospitals are notified of records needing correction.
The DDA and UBF have a number of common elements that enable matching of records concerning patient stays. Linking these records produces a merged file containing both financial and summary medical abstract information for each inpatient hospitalization.
To address the issue of mortality after craniotomy for aneurysm and the relationship to hospital volume and experience, a computer search of SPARCS data was undertaken for all patients with the discharge diagnosis of SAH or unruptured cerebral aneurysm (ICD-9-CM codes 430 and 437.3).11 From 1987 through 1993, the discharge data were then grouped according to the annual volume of craniotomies for aneurysm (ICD-9-CM code 3951) performed at each individual hospital in New York State. In-hospital operative mortality and length of stay were compared among hospitals based on the volume of aneurysm surgery performed.
Statistical analysis of the mortality data was performed with the use of the χ2 test, and LOS means were compared with the use of an unpaired two-tailed t test (P<.05 was considered significant).
During the 7-year period of 1987 through 1993, there were 15 376 discharges for SAH reported by 208 general hospitals in New York State and 5638 craniotomies for cerebral aneurysm. From this group, 98 hospitals reported no cerebral aneurysm surgery, and the following number of hospitals performed the following number of craniotomies for aneurysm per year, on average: 36 performed <1; 37 performed 1 to 5; 16 performed 6 to 10; 9 performed 11 to 20; 8 performed 21 to 30; and 4 performed >30 (1 of these hospitals performed >100 annual operations during this 7-year period).
In-hospital mortality rate for 4034 patients undergoing craniotomy for ruptured cerebral aneurysm (Fig 1⇓) was 16% in hospitals performing <6 annual craniotomies for cerebral aneurysm, 16% in hospitals performing 6 to 10 annual operations, 15% in hospitals performing 11 to 20 annual operations, 15% in hospitals performing 21 to 30 annual operations, 10% in hospitals performing 31 to 100 annual operations, and 7% in the hospital performing >100 annual operations. There was a 43% (95% confidence interval, 29% to 57%) reduction in operative mortality rate for hospitals performing >30 annual craniotomies for aneurysm compared with the rest of the state (8.8% versus 15.5%, P<.0001).
In-hospital mortality for 1604 patients undergoing craniotomy for unruptured cerebral aneurysm (Fig 2⇓) was 12% in hospitals performing <6 annual craniotomies for cerebral aneurysm, 11% in hospitals performing 6 to 10 annual operations, 7% in hospitals performing 11 to 20 annual operations, 5% in hospitals performing 21 to 30 annual operations, 6% in hospitals performing 31 to 100 annual operations, and 3% in the hospital performing >100 annual operations. There was a 43% (95% confidence interval, 14% to 73%) reduction in operative mortality rate for hospitals performing >30 annual craniotomies for aneurysm compared with the rest of the state (4.6% versus 8.1%, P=.0087). However, there was no significant difference with regard to unruptured aneurysm surgery between any hospital group performing >10 annual craniotomies for aneurysm. There was a 53% decrease in mortality rate for unruptured aneurysm surgery in the 21 hospitals performing >10 annual craniotomies for aneurysm compared with the 89 hospitals performing ≤10 annual aneurysm operations (5.3% versus 11.2%, P<.0001).
LOS after craniotomy for ruptured aneurysm did not show any group-related differences except in the single hospital that performed >100 annual craniotomies for aneurysm. This hospital had a mean LOS of 24 days compared with a statewide mean of 35 days (P<.001). LOS after craniotomy for unruptured cerebral aneurysm was also not different for the hospital groups except for the single hospital that performed >100 operations, in which the mean LOS was 17 days compared with the statewide mean of 22 days (P<.05). During the 7 years of this analysis, LOS for unruptured aneurysm surgery declined precipitously in New York State and averaged only 15.5 days by 1993. By comparison, the hospital with >100 aneurysm operations had a mean LOS in 1993 of 7.8 days. LOS for ruptured aneurysm surgery did not change significantly during the 7-year period.
This preliminary study suggests a well-defined relationship between surgical volume and operative mortality after craniotomy for cerebral aneurysm. Patients who underwent surgical aneurysm clipping at hospitals with the lowest volume of procedures had the highest mortality rates. This relation was observed for both ruptured and unruptured aneurysms.
Relation of the Volume of Craniotomies for Cerebral Aneurysm and Mortality in Patients With Ruptured Aneurysms
Although mortality after ruptured aneurysm surgery was not significantly different among hospitals performing <30 annual craniotomies for aneurysm, there was a major reduction in mortality rate (43%) in centers performing >30 craniotomies per year for cerebral aneurysm. This information suggests that ongoing experience with aneurysm surgery might be a positive factor predicting outcome in patients with ruptured cerebral aneurysms. There is a definite learning curve for surgeons operating on ruptured intracranial aneurysms. The average neurosurgeon may never see a volume of cases sufficient to hone the high level of technical skill required to successfully operate on this patient population. Intraoperative misadventures are extremely common in aneurysm surgery, and strategies to avoid intraoperative aneurysm rupture and large-vessel occlusion, the major direct causes of immediate postoperative complications, are maximized by experienced operative teams.
In addition, postoperative vasospasm and hydrocephalus related to the original SAH can account for a significant percentage of overall mortality and morbidity.12 13 14 Approximately 22% of postoperative mortality after aneurysm surgery is directly related to cerebral vasospasm and subsequent brain infarction.15 The treatment of vasospasm is a rapidly evolving field. Currently, the most effective approach to this problem appears to be the judicious use of prophylactic intravascular volume expansion supplemented by induced hypertension and hemodilution in patients who show signs of cerebral ischemia.16 17 18 19 20 To maximize good outcomes, patients require 24-hour neurological and hemodynamic monitoring, preferably in an ICU with a staff experienced in neurological critical care. This degree of invasive monitoring and intensive treatment may be required for up to 14 days after the initial SAH.
It is increasingly evident that neurosurgeons alone cannot provide state-of-the-art treatment for patients with aneurysmal SAH. An interventional neuroradiologist may be required in selected cases for aneurysm embolization or balloon angioplasty.21 22 A critical care neurologist can play an important role in the ICU management of the patient in vasospasm. Neurology and neurosurgery resident physicians can provide emergency therapy throughout the nights and weekends when attending physicians cannot be immediately at the bedside. Finally, an ICU that is dedicated to neurosurgery and neurological patients and staffed by nurses who are interested and experienced in the management of patients with acute intracranial events may offer benefit over general recovery rooms or general surgery ICUs where SAH is rarely seen.
These assumptions might explain why hospitals with large volumes of aneurysm patients can provide surgical treatment with reduced mortality. Surgeons are more experienced, and the ICU staff is better equipped to handle these critically ill high-risk patients. Furthermore, all of the hospitals in this study performing >30 annual aneurysm operations had neurosurgical residency training programs, a factor that might imply better round-the-clock coverage for SAH patients.
Relation of the Volume of Craniotomies for Cerebral Aneurysm and Mortality in Patients With Unruptured Aneurysms
The major transition in operative mortality for cases of unruptured cerebral aneurysm occurred between hospitals performing ≥10 annual craniotomies for aneurysm and those performing <10. Surgical clipping of an unruptured aneurysm is a less dangerous procedure than clipping a ruptured cerebral aneurysm. There is less threat of intraoperative rupture with previously unruptured aneurysms, less concern about the intraoperative use of temporary arterial occlusion to facilitate clipping, and a greatly simplified course of postoperative management.
Furthermore, hospitals that infrequently perform aneurysm surgery are likely to be highly selective in choosing patients for elective clipping of unruptured cerebral aneurysms. Small anterior circulation aneurysms are more likely to be selected for surgery at low-volume centers, and larger, more complex aneurysms, especially in the posterior circulation, are more likely to be referred to larger centers, treated conservatively, or referred for alternative treatments. These factors may explain the finding that hospitals with an intermediate volume of aneurysm surgery had mortality rates as good as high-volume centers.
The majority of New York State hospitals rarely perform aneurysm surgery (<10 craniotomies per year for cerebral aneurysm), yet these hospitals have more than twice the mortality rate for surgical treatment of unruptured aneurysms as seen in larger volume centers. There would be minimal economic impact or loss of prestige generated by a policy of transferring this small number of cases to centers with larger volume. The data might suggest that such a policy would significantly reduce mortality from cerebral aneurysm surgery.
Limitations and Future of This Analysis
This study can only be viewed as a preliminary analysis. The study is retrospective and dependent on coding done at each individual hospital, most likely by nonphysician staff. There are likely to be some errors in the data that are inescapable for this type of study. Nonetheless, the data provide an affirmation of the original hypothesis that surgical volume is related to outcome. The study no doubt needs to be enlarged, performed prospectively, and the data quality improved to adjust for risk factors for each individual patient.
The lack of variation in volume at the study hospitals limits this study to basically a comparison of a few high-volume hospitals compared with many low-volume hospitals. Therefore, the observed relationships may have been due to unique hospital characteristics as opposed to volume of the procedure. Numerous possible factors could be proposed to explain the observed differences in mortality rates, not the least of which is referral networks.23 Such information is not available in this type of study.
The methodology used in this study is severely limited. No data concerning the skill, training, and experience of the individual surgeons involved were obtained. Critical information that could not be tabulated with this type of retrospective study included the type of craniotomy performed; location, size, and type of aneurysm; preoperative neurological grade of the patient; associated medical illnesses; degree of vasospasm; and need for other procedures such as ventriculostomy. These issues limit the conclusions that can be drawn from this study, and only a prospective risk-adjusted outcome study can provide valid data on which to design future healthcare delivery systems.
This analysis can only be validated by future studies in which patient populations are stratified for differences in expected outcomes. No such stratification was possible in this preliminary study. Future analyses must include differences in patient age, sex, and ethnicity among other characteristics that may be associated with in-hospital mortality.
This type of analysis was previously performed in New York State for CABG with attention focused on individual surgeons and hospitals.1 2 Surgeon volume was found to be more important than hospital volume, as is also likely the case with cerebral aneurysms. Release of the cardiac surgery information into the public domain led to the development of a detailed clinical data bank in which risk-adjusted outcome figures could be obtained and compared with surgical volume for individual surgeons and hospitals. The educational efforts of the New York State Department of Health over the last several years to inform surgeons of the inferior outcomes experienced by low-volume surgeons led to a reduction in percentage of operations performed by surgeons and hospitals that were having poor results. During this period of time there was a 41% decline in overall statewide operative mortality rate for CABG surgery.
Future studies of patients undergoing cerebral aneurysm surgery might also be helpful in identifying criteria that make certain cases high risk. A prospective analysis could focus not only on surgeon volume but also on presenting clinical grade, aneurysm size, aneurysm location, and timing of surgery in relation to SAH. This study would likely identify types of cases that might be more appropriately referred to specialty centers.
An average of >14 000 annual CABG operations are performed in New York State in a total of 30 hospitals.2 Conversely, the current analysis of cerebral aneurysm surgery reveals that 739 annual craniotomies for intracranial aneurysms are performed in 110 hospitals in New York State. Future studies may indicate that regional oversight like that already in place for CABG surgery could effectively reduce mortality for cerebral aneurysm surgery.
There appears to be a threshold of experience with intracranial aneurysm surgery required to achieve the best results possible. The volume of cases needed to be proficient at treating ruptured aneurysms appears to be greater than with unruptured aneurysms, although case selection at middle-volume centers may explain the relative improvement of results with unruptured aneurysms.
We hope that publication of this preliminary review will prompt an ongoing analysis of risk-adjusted mortality for patients of surgeons and hospitals performing different volumes of aneurysm surgery. Previous studies with CABG surgery in New York State have shown that feedback of this type of data to hospitals has improved overall quality and reduced mortality.
Selected Abbreviations and Acronyms
|CABG||=||coronary artery bypass graft|
|DDA||=||Discharge Data Abstract|
|ICD-9-CM||=||International Classification of Diseases, 9th Revision, Clinical Modification|
|ICU||=||intensive care unit|
|LOS||=||length of stay|
|SPARCS||=||Statewide Planning and Research Cooperative System|
|UBF||=||Uniform Billing Form|
The authors acknowledge the contributions of Bennett M. Stein, MD, for manuscript review, Beverly Diamond, PhD, for statistical analysis, and George L. McVey for obtaining raw data.
- Received July 31, 1995.
- Revision received September 22, 1995.
- Accepted October 9, 1995.
- Copyright © 1996 by American Heart Association
Heros RC. Intracranial aneurysm: a review. Minn Med. 1990;73:27-32.
ICD-9-CM: International Classification of Diseases, 9th Revision, Clinical Modification. 3rd ed. Los Angeles, Calif: Practice Management Information Corp; 1991.
Broderick JP, Brott TG, Duldner JE, Tomsick T, Leach A. Initial and recurrent bleeding are the major causes of death following subarachnoid hemorrhage. Stroke. 1994;25:1342-1347.
Longstreth WT, Nelson LM, Koepsell TD, van Belle G. Clinical course of spontaneous subarachnoid hemorrhage: a population-based study in King County, Washington. Neurology. 1993;43:712-718.
Awad IA, Carter P, Spetzler RF, Medina M, Williams FW Jr. Clinical vasospasm after subarachnoid hemorrhage: response to hypervolemic hemodilution and arterial hypertension. Stroke. 1987;18:365-372.