Craniotomy for the resection of metastatic brain tumors in the U.S., 1988–2000

Decreasing mortality and the effect of provider caseload

Authors

  • Fred G. Barker II M.D.

    Corresponding author
    1. Brain Tumor Center, Neurosurgical Service, Massachusetts General Hospital, Boston, Massachusetts
    2. Division of Neurosurgery, Department of Surgery, Harvard Medical School, Boston, Massachusetts
    • Brain Tumor Center, Massachusetts General Hospital, Fruit Street, Boston, MA 02114
    Search for more papers by this author
    • Fax: (617) 724-8769


Abstract

BACKGROUND

To assist in selecting treatment for patients with brain metastases, the current study assessed the risk of adverse outcomes after contemporary resection of metastatic brain tumors in relation to patient, surgeon, and hospital characteristics, with particular attention to the volume of care and trends in outcomes.

METHODS

A retrospective cohort study of 13,685 admissions from the Nationwide Inpatient Sample between 1988–2000 was performed. Multivariate logistic, ordinal, and loglinear regression were used with endpoints of mortality, discharge disposition, length of stay, and total hospital charges.

RESULTS

The overall in-hospital mortality rate was 3.1% and an additional 16.7% of patients were not discharged directly home. In multivariate analyses, larger-volume centers were found to have lower mortality rates for intracranial metastasis resection (odds ratio [OR], 0.79; 95% confidence interval [95% CI], 0.59–1.03 [P = 0.09]). An adverse discharge disposition also was less likely at higher-volume hospitals (OR, 0.75; 95% CI, 0.65–0.86 [P < 0.001]). For surgeon caseload, mortality was lower with higher-caseload providers (OR, 0.49; 95% CI, 0.30–0.80 [P = 0.004]) and an adverse discharge disposition occurred significantly less frequently (OR, 0.51; 95% CI, 0.40–0.64 [P < 0.001]). The annual number of resections increased by 79% during the study period, from 3900 (1988) to 7000 (2000). In-hospital mortality rates decreased from 4.6% (1988–1990) to 2.3% (1997–2000), a 49% relative decrease. Length of stay was reported to be significantly shorter with higher-volume providers. Hospital charges were not found to be associated significantly with hospital caseload and were found to be significantly lower after surgery that was performed by higher-caseload surgeons.

CONCLUSIONS

The results of the current study found that higher-volume hospitals and surgeons provided superior short-term outcomes after resection of intracranial metastasis was performed, with shorter lengths of stay and a trend toward lower charges. Cancer 2004;100:999–1007. © 2004 American Cancer Society.

Evidence suggests that patient mortality and morbidity are lower in many instances when complex medical or surgical procedures are performed at high-volume centers or by high-volume physician providers. For example, in-hospital mortality is lower when cardiovascular surgeries,1 complex cancer surgeries,2, 3 and surgical treatment of intracranial aneurysms4 are performed at high-volume hospitals or by high-volume surgeons. Lower mortality rates and a shorter length of hospital stay after craniotomy for brain tumors, broadly defined, also have been shown to be characteristic of high-volume centers.5, 6 These investigations described the pooled results of craniotomies for many types of brain tumor, with to our knowledge little attempt made to address temporal trends in practice patterns or results.5, 6

The current analysis will describe the results of craniotomy for resection of metastatic brain tumors performed between 1988–2000 in a representative sample of nonfederal hospitals in the U.S. The mortality rate for the resection of brain metastases was examined for its relation with provider caseload and changes over time.

MATERIALS AND METHODS

The data source for the current study was the Nationwide Inpatient Sample (NIS) hospital discharge database for the years 1988–2000, obtained from the Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality (AHRQ), Rockville, Maryland.7 The NIS is a hospital discharge database that represents approximately 20% of all inpatient admissions to nonfederal hospitals in the U.S. For these years, the NIS contains discharge data concerning 100% of discharges from a stratified random sample of nonfederal hospitals in 8–28 states to approximate a representative 20% subsample of all U.S. nonfederal hospital discharges. Because the NIS contains data regarding all patients discharged from sampled hospitals during the year regardless of age or payer, it can be used to obtain the annual total volume of specified procedures performed at individual hospitals. For many states, the surgeon who performed the principal procedure during the hospital admission is identified using a unique masked code. An overview of the NIS is available at URL: http://www.ahcpr.gov/data/hcup/nisintro.htm [12/22/03].

Inclusion and Exclusion Criteria and Definition of Endpoints

An admission for craniotomy for the resection of an adult metastatic brain tumor was defined as follows: patient age ≥ 19 years, a primary International Classification of Diseases–Clinical Modification (9th revision) (ICD-9-CM) diagnosis code of 198.3 (secondary malignant neoplasm of brain or spinal cord), and a primary ICD-9-CM procedure code of 01.59 (excision or destruction of tissue or lesion of brain).

Two primary endpoints were examined: in-hospital mortality and discharge to institutions other than the patient's home. In-hospital mortality was coded directly in the NIS database and was analyzed using logistic regression. Discharge disposition was coded on a four-level scale as death, discharge to a long-term care facility, discharge to other facilities, or discharge home and was analyzed using ordinal logistic regression. Length of stay (LOS) and total hospital charges were coded in the NIS data and were analyzed only for patients who were discharged from the hospital alive. LOS and hospital charge data were highly positively skewed and were analyzed as logarithmic transforms. One patient with an LOS of 0 days was recoded as “LOS missing.”

Patient Characteristics

Patient age, gender, race, median household income in the postal code of the patient's residence, primary payer (Medicare, Medicaid, private insurance, self-pay, no charge, and other), type of admission (emergency, urgent, and elective), and admission source (emergency unit, transfer from another hospital, transfer from long-term care, and routine) were coded in NIS data. Nine patients (0.1%) with an admission type of “other” were recoded as routine admissions. Greater than 5% of the discharges had missing values for 3 variables used principally as stratification factors for other analyses: race (32% missing), admission type (12% missing), and whether the principal procedure was performed on the first hospital day (15% missing). When these variables were used as stratification factors, missing values for race and admission type were imputed as follows: missing race was set to white and missing admission type was set to “emergency” for admissions whose source was the emergency unit, “urgent” for admissions that were transfers from another hospital, and “routine” for admissions from other sources. Whether the principal procedure was performed on the first hospital day was not imputed, and when race or admission type was the focus of the analysis, imputed values were not used.

To assess the effect of general medical comorbidity, the set of 30 medical comorbidity markers described by Elixhauser et al.,8 excluding the 2 specific neurologic comorbidity variables (“paralysis” and “other neurological deficit”), and 3 comorbidity variables likely to represent postoperative conditions (“fluid and electrolyte disorders,” “blood loss anemia,” and “deficiency anemias”) were calculated using AHRQ software (available from URL: www.ahcpr.gov/data/hcup/comorbid.htm [12/22/03]) and summed to give a single comorbidity score ranging between 0–25.

Potential complications of metastatic brain tumor resection were identified using the following codes: postoperative neurologic complications, including those resulting from infarction or hemorrhage (997.00–997.09); hematoma complicating a procedure (998.1–998.13); hydrocephalus (331.3–331.4) or the performance of a ventriculostomy (02.2); mechanical ventilation (96.70–96.72); deep venous thrombosis, pulmonary embolism, or placement of an inferior vena cava filter (415, 415.11–19, 451.0–9, 453.0–9, and 38.7); and transfusion of packed red blood cells (99.04).

ICD-9-CM diagnosis codes were used to define possible primary sites of origin of cerebral metastases: lung carcinoma (ICD-9-CM code 162), gastrointestinal carcinoma (codes 150–159), melanoma (code 172), breast carcinoma (code 174), and renal cell carcinoma (code 189.0). The presence of other metastases was defined as ICD-9-CM codes 197–198, excluding 198.3 or 198.4. (This definition excludes metastases to lymph nodes.)

Provider and Hospital Characteristics

Hospital region (Northeast, Midwest, South, or West), location (rural or urban), teaching status, and bed size (small, medium, or large) were coded in NIS data. Hospital and surgeon volumes of the metastatic brain tumor resections performed were derived by counting the cases for each identified surgeon and hospital in the database. Because hospital and physician volumes were skewed positively, the logarithmic transforms were used when volume measures were entered into regression models.

Statistical Methods

Statistical methods included the Fishers exact and Wilcoxon rank sum tests; Spearman rank correlation; and loglinear least-squares, ordinary logistic, and proportional-odds ordinal logistic regression.9–11 To correct for possible clustering of similar outcomes within hospitals (which could cause falsely inflated estimates of the statistical significance of regression coefficients), a sandwich variance-covariance matrix was estimated from the data using methods developed by Huber and White, with adjustment for clustering by hospital.11 Length of stay and hospital charges were analyzed as logarithmic transforms using least-squares regression corrected for clustering as described earlier.

Extrapolations to the entire U.S. population were adjusted for the NIS stratified survey method using SAS PROC SURVEYMEANS.12 Linear regressions on extrapolated population values (i.e., to test trends in the annual number of admissions or mortality rates) were weighted by the inverse of the variance. Logistic and ordinal logistic regression analyses treated the sample as a simple random draw from an infinite possible population (i.e., without weighting or correction for a finite sampling fraction).

Calculations were performed using SAS (version 8.2; SAS Institute, Inc., Cary, NC) and S-plus (Version 3.3 for Windows; Insightful, Inc., Seattle, WA) software with the Hmisc and Design modeling function software libraries developed by Harrell11, 13 and the LOCFIT local-likelihood regression library developed by Loader.14, 15P values are two-tailed.

RESULTS

There were 13,685 admissions for the resection of metastatic brain tumors identified in the NIS database between 1988–2000. Clinical characteristics of the patients are shown in Table 1. Most patients were white, ages 50–70 years, and 49% were women. Half of the admissions were classified as routine, approximately 25% were classified as urgent, and the remaining 25% were classified as an emergency. Approximately two-thirds of the patients were admitted from home, with admissions through the emergency unit or as a transfer from another hospital comprising the remainder of patient admissions.

Table 1. Clinical Characteristics of 13,685 Patients Who Underwent Resection of Metastatic Brain Tumors
Age (yrs) 
 Mean60
 Median61
 Interquartile range51–69
 Range19–94
Male gender51%
Race (n = 9301) 
 White88.4%
 Black6.3%
 Hispanic3.1%
 Asian/Pacific Islands1.0%
 Native American0.05%
 Other1.1%
Median household income for postal code of residence (n = 13,070) 
 Quartile 1 (lowest)20%
 Quartile 223%
 Quartile 322%
 Quartile 4 (highest)36%
Primary payer for care (n = 13,599) 
 Medicare39%
 Medicaid7%
 Private insurance47%
 Self-pay3%
 No charge0.1%
 Other4%
Admission type (n = 12,108) 
 Emergency28%
 Urgent26%
 Routine45%
Admission source (n = 12,938) 
 Emergency unit23%
 Transfer from acute care hospital7%
 Transfer from long-term care facility2%
 Routine69%
Primary site 
 Lung carcinoma29%
 All other sites< 5%
Other metastases18%
Years of treatment 
 1988–199017%
 1991–199322%
 1994–199623%
 1997–200038%

In-hospital mortality for the entire cohort was 3.1% (95% confidence interval [95% CI], 2.8–3.4%). Approximately 5.5% of the patients were discharged to long-term care facilities, 11.2% were discharged to other facilities such as rehabilitation hospitals, and 80.2% were discharged directly home.

Patient Characteristics and Outcome

Age, gender, race, primary payer for care, median household income in the postal code of the patient's residence, admission type and source, timing of the surgical procedure, medical comorbidity score, and tumor histology were tested as predictors of mortality and discharge disposition (Table 2). Age was found to be an important predictor of death and discharge from the hospital other than to home (P < 0.001) (Fig. 1). Female patients had rates of death and discharge other than to home that were modestly, but significantly, lower than those of male patients. Black patients had higher mortality rates and a modestly worse discharge disposition. Private insurance was associated with significantly lower mortality and more favorable discharge disposition. Patients with higher medical comorbidity scores, emergency or urgent care admissions, admission through the emergency unit or as a transfer from another hospital, or those whose procedure was not performed on the first hospital day had a higher mortality rate and worse discharge disposition. All these variables were included in multivariate analyses, as well as stratification by hospital geographic region.

Table 2. Effect of Patient Characteristics on Mortality Rates and Outcome at the Time of Hospital Discharge
 Odds of deathOdds of worse outcome at hospital discharge (four-level scale)
  1. Univariate analyses are shown except for median income, which was stratified by year.

Age (yrs) (per decade)1.35 (1.21–1.50) P < 0.0011.77 (1.69–1.85) P < 0.001
Female gender0.66 (0.55–0.80) P < 0.0010.90 (0.83–0.99) P = 0.02
Race  
 White vs. other0.76 (0.53–1.1) P = 0.140.98 (0.84–1.15) P = 0.8
 Black vs. other1.63 (1.10–2.41) P = 0.011.18 (0.96–1.46) P = 0.12
Primary payer (private insurance vs. other)0.53 (0.44–0.65) P < 0.0010.41 (0.37–0.45) P < 0.001
Median income in postal code of residence (by quartile)0.90 (0.82–0.99) P = 0.040.96 (0.91–1.001) P = 0.06
Admission type (emergency vs. urgent vs. routine)1.49 (1.33–1.66) P < 0.0011.42 (1.34–1.50) P < 0.001
Admission source  
 Emergency unit vs. routine1.71 (1.40–2.11) P < 0.0011.86 (1.68–2.05) P < 0.001
 Transfer from acute care hospital vs. routine1.17 (0.79–1.73) P = 0.42.14 (1.87–2.45) P < 0.001
Surgery not performed on day of admission to hospital1.96 (1.52–2.51) P < 0.0011.88 (1.69–2.10) P < 0.001
Medical comorbidity score (per point)1.19 (1.06–1.33) P = 0.0041.35 (1.28–1.42) P < 0.001
Tumor histology (lung primary vs. other)0.97 (0.79–1.19) P = 0.81.04 (0.95–1.13) P = 0.4
Other metastases (present vs. absent)1.26 (1.002–1.59) P = 0.051.05 (0.93–1.17) P = 0.4
Figure 1.

Probability of death or discharge other than to home after craniotomy for metastasis resection, plotted against patient age. Dashed line: mortality; dotted line: death or discharge to long-term care facility (LTF); solid line: death or discharge to LTF or short-term care facility (STF).

Hospital and Surgeon Characteristics and Outcome

Patients were treated at 821 hospitals. Compared with hospitals in which resections of brain metastases were not performed, hospital characteristics found to be predictive of the performance of one or more craniotomies for brain metastases included urban location, positive teaching status, and larger bed size (P < 0.001 for all). For 50% of the admissions, 1606 treating surgeons were identified in the database.

Hospitals and surgeons varied widely with regard to the volume of resections for brain metastases performed. Analyzed on a per-patient basis, the median annual number of craniotomies performed for metastases was 7 per hospital (range, 1–96 admissions; 25th percentile: 4 admissions, 75th percentile: 14 admissions) or 3 per surgeon (range, 1–33 admissions; 25th percentile: 2 admissions, 75th percentile: 5 admissions). For 744 patients (5%), no other craniotomy for the resection of brain metastases was reported during that year at their hospital, and for 1599 patients (23%;), no other craniotomy for the resection of brain metastases was reported that year by their surgeon.

A larger hospital and larger surgeon caseload were associated with lower in-hospital mortality rates after craniotomy was performed for metastasis resection (Fig. 2). The analysis was adjusted for casemix using patient age, gender, race, primary payer for care, median household income in the postal code of the patient's residence, geographic region, admission type and source, timing of procedure (first hospital day vs. later), medical comorbidity score, tumor histology (lung carcinoma vs. other), presence of other metastases, and year of treatment. Odds ratios (ORs) for the importance of hospital and surgeon caseload were reported for a tenfold difference in caseload, because this approximated the difference between the 25th and 75th percentiles for caseload. In-hospital mortality was lower at higher-volume hospitals (OR, 0.79; 95% confidence interval [95% CI], 0.59–1.03 [P = 0.09]). The mortality rate at the lowest-volume-quintile hospitals (1 or 2 admissions per year) was 4.4%, compared with 1.8% at the highest-volume-quintile hospitals (≥ 18 admissions per year). Mortality rates were significantly lower when craniotomies were performed by higher-caseload surgeons (OR, 0.49; 95% CI, 0.30–0.80 [P = 0.004]). The mortality rate for surgeries performed by lowest-volume-quintile surgeons (1 admission per year) was 3.9%, compared with 1.4% for highest-volume-quintile surgeons (≥ 7 admissions per year). Adjustment for the presence of hemiparesis, hemiplegia, or aphasia did not appear to have any effect on the ORs or degrees of statistical significance in these analyses.

Figure 2.

In-hospital mortality rates after craniotomy for metastasis resection plotted against hospital and surgeon caseload (grouped by quintile). The relation between a larger caseload and lower mortality rates was not found to be significant on multivariate analysis for hospitals (P = 0.09) but was significant for surgeons (P = 0.004).

An adverse discharge disposition was also less likely after craniotomy for the resection of metastases at high-volume hospitals (OR, 0.75; 95% CI, 0.65–0.86 [P < 0.001]) (Fig. 3). Casemix adjustment was performed using multivariate regression, as described earlier. The rate of discharge not directly home was reported to be 24.9% at lowest-volume-quintile hospitals (1 or 2 admissions per year) compared with 16.7% at highest-volume-quintile hospitals (> 18 admissions per year). For surgeon caseload, higher volumes also were found to be predictive of a better outcome at the time of hospital discharge (OR, 0.51; 95% CI, 0.40–0.64 [P < 0.001]). The rate of discharge not directly home was 23.7% after surgeries performed by lowest-volume-quintile surgeons (1 admission per year) compared with 12.6% for highest-volume-quintile surgeons (≥ 7 admissions per year). Again, adjustment for the presence of hemiparesis, hemiplegia, or aphasia appeared to have no effect on the ORs or degrees of statistical significance in these analyses.

Figure 3.

Rates of discharge not directly home after craniotomy for metastasis resection plotted against hospital and surgeon caseload (grouped by quintile). The relation between a larger caseload and lower mortality rates was found to be significant on multivariate analysis for both hospitals (P < 0.001) and surgeons (P < 0.001).

Patient characteristics were tested as potential predictors of hospital or surgeon caseload. Older patients were found to have lower-caseload hospitals and surgeons (P < 0.001 for both). White patients appeared to have higher-caseload surgeons compared with black patients (P = 0.04), but there was no significant difference with regard to hospital caseloads between the two groups. The primary payer for care was a significant predictor of both hospital and surgeon caseload (P < 0.001 for both); patients with private insurance tended to have higher-caseload hospitals and surgeons than those with other primary payers. Patients from higher-income areas of residence had higher-volume hospitals and surgeons (P < 0.001 for both). Emergency or urgent admissions were more common to lower-volume hospitals and surgeons (P < 0.001 for both). Patients with more medical comorbidities tended to have lower-volume hospitals and surgeons (P < 0.001 for both). Patients with lung carcinoma metastases and patients with other known metastases appeared to have lower-volume hospitals and surgeons (P < 0.002 for all).

Trends over Time

Projected to the U.S. population, the total annual number of admissions increased from 3900 in 1988 to 7000 in 2000, a 79% relative increase, or 190 additional resections annually (P = 0.002) (Fig. 4). The median patient age decreased slightly during this period, from 62 years in 1988–1990 to 60 years in 1997–2000 (P < 0.001). Patients treated in later years were less likely to be white (8% nonwhite in 1988–1990 vs. 13% nonwhite in 1997–2000; P < 0.001). Elective admissions became more common (39% in 1988–1990 vs. 48% in 1997–2000; P < 0.001). The probability of lung carcinoma histology did not appear to change during the study period (P = 0.25), and the presence of other coded metastases became more common (P = 0.03).

Figure 4.

Trends over time for craniotomy for metastasis resection, projected to all nonfederal U.S. hospitals. The top panel shows the increasing total annual number of craniotomies performed for metastasis resection (P = 0.002) and the bottom panel shows the decrease in the in-hospital mortality rates (P < 0.001).

Projected to the U.S. population, there was a decrease noted in mortality rates during the study period, from 4.6% in 1988–1990 to 2.3% in 2000, a 49% relative decrease (Fig. 4). In a multivariate logistic analysis based on the study cohort that adjusted for age, gender, race, primary payer, median household income in the postal code of the patient's residence, admission type and source, procedure timing, comorbidity score, geographic region, tumor histology, and hospital caseload, the decrease in mortality rates was found to be statistically significant (P < 0.001). For each year, the relative decrease in mortality rates was 6% (95% CI, 3–9%). The relative decrease in mortality rates was found to be uniform across hospital caseload quintiles; mortality rates at lowest-volume-quintile hospitals were reported to decrease 47% during the study period (4.9% for 1988–1990 vs. 2.6% for 1997–2000) whereas they decreased 55% at highest-volume-quintile hospitals (3.2% for 1988–1990 vs. 1.4% for 1997–2000).

LOS and Hospital Charges

LOS was found to have decreased significantly during the study period (by 4.4% per year; P < 0.001); the median LOS was reported to be 11 days in 1988–1990 and was reported to have decreased to 6 days in 1997–2000. After multivariate adjustment for the variables described earlier and stratification by treatment year, the LOS was found to be significantly shorter at larger-volume hospitals (P < 0.001). In a similar multivariate model, a larger surgeon caseload also was found to be associated with a shorter LOS (P < 0.001).

Total hospital charges were found to have increased significantly during the study period, from a median of $16,200 in 1988 to $26,000 in 2000 (by 5.3% per year; P < 0.001). After multivariate adjustment for the variables described earlier and stratification by treatment year, there was no significant relation detected between hospital caseload and total hospital charges (charge differential for a 10-fold larger caseload +2.4%; 95% CI,-0.2% to +5.0% [P = 0.07]). Total hospital charges were found to be significantly lower for patients of higher-caseload surgeons (charge differential for a 10-fold larger caseload,-4.8%; 95% CI,-1.1% to-8.4% [P = 0.01]).

DISCUSSION

The current analysis included 13,685 admissions for the resection of intracranial metastatic tumors, performed at nonfederal U.S. hospitals between 1988–2000. The in-hospital mortality rate was 3.1% and an additional 17% of patients were not discharged directly home. After multivariate adjustment for differences in the casemix, care provided by higher-volume hospitals and surgeons was followed by lower mortality rates and better outcome at the time of hospital discharge, with shorter LOS and unchanged or lower total hospital charges. These benefits were found to be enjoyed more often by patients who were younger, healthier, and white; those who had private insurance; and those who resided in wealthier areas. The annual number of admissions for the resection of intracranial metastases were found to have increased 79% during the study period, and in-hospital mortality rates decreased by 49%.

The tendency toward superior outcomes after care provided by physicians or hospitals with large current caseloads is called the volume-outcome effect. The volume-outcome effect has been demonstrated previously for other intracranial procedures such as clipping of intracranial aneurysms,4, 16 as well as in two studies that addressed the surgical treatment of brain tumors.5, 6 Both these studies included intracranial tumors of many types, including primary brain tumors and meningiomas, in addition to metastatic lesions. Although patients with metastases formed a small subgroup in each of these studies, mortality rates were found to be correlated inversely with hospital caseload (defined as the volume of craniotomies for any tumor type) in each. Other measures of provider volume (such as hospital or surgeon caseload of resections of metastases) and other endpoints (such as LOS and hospital charges for metastases resections) were not examined in these studies.

The results of the current study confirm and extend these prior findings. As reported previously, in-hospital mortality was less than half as frequent at highest-volume-quintile hospitals when compared with lowest-volume-quintile hospitals. A larger surgeon caseload was found to have a similar association with lower mortality rates. In addition, both hospital and surgeon caseload were found to be significantly associated with less frequent adverse hospital discharge disposition, a shorter LOS, and (for surgeon caseload) with lower total hospital charges. These findings were found to be unaffected by multivariate adjustment for casemix using patient age, gender, race, primary payer for care, median household income in the postal code of the patient's residence, geographic region, admission type and source, timing of procedure (first hospital day vs. later), medical comorbidity score, tumor histology, presence of other coded metastases, and year of treatment. However, important tumor-related variables that affect surgical risk for metastasis resection, such as tumor size and location and preoperative performance status, were not included in our data source and hence could not be included in the current analysis. Biased distribution of these risk factors between patients treated by high-volume and those treated by low-volume providers could have affected the results of the current study. However, adjustment for coded variables that most likely reflected preoperative neurologic deficits (such as hemiparesis, hemiplegia, and aphasia) appeared to have no effect on the volume-outcome effect demonstrated in the current study.

Demonstrations of better outcomes after surgical procedures performed at higher-volume centers often lead to a call for the concentration of care into the hands of a subset of providers, such as the regionalization of care at specialized centers. The limitations of a study such as the current one should be carefully considered before its conclusions are accepted as supporting the regionalization of intracranial metastasis resection. Because a hospital discharge database in which patient identities are masked was used as the data source for the current study, long-term outcomes (such as functional status at 30 days or 6 months, or survival after discharge) were not available for study. Because some patients discharged to short-term rehabilitation centers return to normal functional levels when recovery is complete, using short-term outcomes most likely exaggerates the difference in long-term functional outcomes between patients treated at high-volume and those treated at low-volume hospitals. In addition, volume-outcome studies, for which randomization is difficult or impossible, are observational studies of competing treatments (i.e., treatment at high-volume or low-volume centers) and therefore are liable to selection bias. The current analysis demonstrated some evidence of a lower-risk population treated at high-caseload centers (such as younger patient age, less severe medical comorbidities, less frequent emergency or urgent admissions, and less frequently coded presence of other metastases). Although the analysis adjusted for these risk factors using multivariate regression, it is likely that the differences in baseline risk between patients treated at high-volume and those treated at low-volume centers were larger than what was estimated in the current study on the basis of observed variables.

One change in neurosurgical practice that could have affected the results of the current study was the increasing use during the study period of radiosurgery to treat cerebral metastases.17 If higher-risk patients were increasingly referred for radiosurgery rather than open resection, mortality rates for surgery would be expected to decrease with time. In addition, some of the difference reported with regard to outcome between high-volume and low-volume centers could have resulted from the selection of the highest-risk patients for radiosurgery treatment, rather than surgical resection, at large-volume centers. However, a comparison between patients in the current study who were treated with surgery between 1995–2000 and 1461 others contained in the NIS who were treated with radiosurgery during the same period suggested that patients treated with radiosurgery were a lower-risk population compared with those treated with surgery. Specifically, patients treated with radiosurgery were younger and had more frequent elective admissions, lower rates of hemiplegia (excluding patients with complications of surgery), fewer medical comorbidities, and similar rates of other coded metastatic disease (unpublished data). The general similarity in risk profile between patients with brain metastases who were treated using surgery and those treated with radiosurgery has been confirmed by several single-institution studies.18–22 This suggests that the increasing use of radiosurgery does not explain the decrease in surgical mortality noted over the study period. Similarly, unless a higher proportion of low-risk patients are referred for radiosurgery from low-volume institutions than from high-volume institutions (which are more likely to have radiosurgical treatment facilities), the increasing use of radiosurgery does not explain the volume-outcome effect observed in the current study.

The increase in the annual number of cases and the decrease in mortality rates observed in the current study remain unexplained. An annual caseload of 7000 resections in the U.S. represents a small subset of all patients with brain metastases, for which the U.S. annual incidence is reported to be as high as 170,000 cases.23 Less stringent selection of the most-favorable cases for surgery could explain the increasing total caseload but most likely is not consistent with the younger patient age and lower mortality rates observed in more recent years. More frequent surgeries performed in asymptomatic patients with small lesions discovered through screening examinations may be a better explanation for these findings.

Conclusions

The current study included a large representative sample of patients who underwent craniotomy for the resection of brain metastases in the U.S. between 1988–2000. After multivariate adjustment for other risk factors, adverse outcomes (mortality and adverse hospital discharge disposition) were found to be less frequent after surgery at high-volume centers or performed by high-volume surgeons. An increasing overall number of annual cases treated and a substantial decline in the mortality rates at centers of all sizes also were observed. There was a bias toward the more frequent treatment of low-risk patients at high-volume centers. The results of the current study require confirmation, preferentially using a database with more clinical information than available herein, before their use in formulating population-wide health policies. However, the current study findings may have significance for those patients with intracranial metastases who seek high-quality surgical care, and for those physicians responsible for advising them.

Ancillary