outcomes in head and neck oncologic surgery at academic medical centers in the united states

The LaryngoscopeVC 2013 The American Laryngological,Rhinological and Otological Society, Inc.

TRIOLOGICAL SOCIETYCANDIDATE THESIS

Outcomes in Head and Neck Oncologic Surgery at Academic MedicalCenters in the United States

Scharukh Jalisi, MD, MA, FACS; Shethal Bearelly, MD; Abdirahman Abdillahi, BS; Minh Tam Truong, MD

Objectives/Hypothesis: To evaluate the impact of case volume and other variables on outcomes after head and neckoncologic surgery was performed at academic medical centers in the United States.

Study Design: Cross Sectional Ecological Study.Methods: The University HealthSystems Consortium (UHC) database was analyzed for discharge data on all patients

who underwent surgery for head and neck cancers (excluding thyroid and skin cancer) at full- member academic medicalcenters between quarter 4 of 2006 and quarter 4 of 2009. Multivariate and linear regression analyses and chi-square testswere applied to evaluate significant associations between hospital surgical volume and other independent variables, and toevaluate the risk of mortality, mortality index, complications, length of stay (LOS), LOS index, cost, and cost index.

Results: Of 22,357 surgical cases, 11,573 met our inclusion criteria. The only outcome that was statistically significantbased on volume was a lower complication rate in high volume hospitals (P ¼ 0.0486) as compared to low volume hospitals.All Payer Refined–Diagnosis Related Group defined major severity of illness was the only independent variable significantlyassociated with higher complication rates, observed LOS, and observed cost (P <0.0001, P ¼ 0.0139, and P ¼ 0.0092, respec-tively). Management of male patients and black patients resulted in a lower cost index (P ¼ 0.0472) and a higher complica-tion rate (P ¼ 0.0297), respectively. Patients with private insurance had lower complication rates, observed LOS, andobserved cost (P ¼ 0.0401, P ¼ 0.0001, and P ¼ 0.0187, respectively).

Conclusions: After controlling for other factors, academic medical centers with a higher cumulative case volume havelower rates of complications.

Key Words: Outcome, head and neck surgery, cost, volume, complications.Level of Evidence: 2b.

Laryngoscope, 123:689–698, 2013

INTRODUCTIONThe surgical and postoperative management of

head and neck cancer is complex. The management ofpatients with head and neck cancer requires a coordi-nated effort, often involving a team comprised ofsurgeons, physician extenders, medical and radiationoncologists, nurses, and others. Although the immediatepostoperative period typically is fraught with thedangers of complications (e.g., pneumonia, pulmonaryembolus, myocardial infarction, and stroke) efficient,well-coordinated multidisciplinary care and managementcan reduce morbidity and mortality.

Currently, there are no widely accepted nationalbenchmark standards for the management of head andneck surgical oncology patients in terms of cost, lengthof stay (LOS), mortality, comorbidity, and complications.Moreover, the Agency for Healthcare Research andQuality (AHRQ) Quality Indicators (QIs) measure healthcare quality by using readily available hospital inpatientadministrative data. The Patient Safety Indicators(PSIs) are a tool to help health system leaders identifypotential adverse events occurring during hospitaliza-tion. There are currently 27 PSI measures.1 Fordischarges occurring on or after October 1, 2008, theCenter for Medicare and Medicaid Services declared thatInpatient Prospective Payment System (IPPS) hospitalswill not receive additional payment for cases when oneof the 12 selected conditions is acquired during hospitali-zation (i.e., was not present on admission).2

The notion is not new that case volume can beinversely related to rate of complications for a specifictype of surgery. Review of relevant reports published inthe medical literature shows that bariatric surgery doneat hospitals with higher case volume has lower morbid-ity and mortality, especially for those cases with ahigher complexity and multiple comorbidities.3–6 There-fore, realizing that there might be a relationshipbetween case volume and outcomes in management ofpatients with head and neck cancer, we designed a study

From the Division of Head and Neck Surgical Oncology andSkullbase Surgery (S.J., A.A.), the Department of Otolaryngology–Headand Neck Surgery (S.J., A.A.), the Department of Neurological Surgery (S.J.),and the Department of Radiation Oncology (M.T.T.), Boston University,Boston, Massachusetts; and the Department of Otolaryngology–Head andNeck Surgery (S.B.), University of California at San Francisco, SanFrancisco, California, U.S.A.

Editor’s Note: This Manuscript was accepted for publicationSeptember 26, 2012.

The authors have no funding, financial relationships, or conflictsof interest to disclose.

Send correspondence to Scharukh Jalisi, MD, FACS, MA, Director,Division of Head and Neck Surgical Oncology and Skullbase Surgery,Department of Otolaryngology–Head and Neck Surgery, 820 HarrisonAvenue, FGH4, Boston, MA 02118. E-mail: [email protected]

DOI: 10.1002/lary.23835

Laryngoscope 123: March 2013 Jalisi et al.: Outcomes in Head and Neck Oncologic Surgery

689

to compare the costs, morbidity, and mortality ofpatients undergoing head and neck cancer surgeryamong academic hospitals included in an accessible andsearchable database. We used the University HealthSys-tems Consortium (UHC) database. The UHC was formedin 1984, and is an alliance of 112 academic medical cen-ters and 256 of their affiliated hospitals, representingapproximately 90% of the nation’s nonprofit academicmedical centers.7

MATERIALS AND METHODSPrior to initiation of any research, Institutional Review

Board approval was obtained (BUMC Protocol H-29787).Approximately 90% of the academic medical centers in theUnited States contribute to the UHC database, making this ageneralizable database for academic medical centers. Membershave access to a Clinical Data Base with discharge data thatallows hospitals to compare clinical performance. The UHCdatabase is a collection of patient-level (Uniform Bill) UB-04billing data8 from all member academic health centers and theiraffiliate community hospitals. The UB-04 form is a standardizedform approved by Center for Medicare and Medicaid services tocollect inpatient discharge diagnoses and billing data. It is usedby facilities to submit bills to third party payers. The databaseincludes inpatient discharge data such as LOS, mortality, read-mission rates, cost, complication rates, and dischargedisposition. Demographic data is also gathered, including age,race, sex, and comorbidities. The major benefit of the UHC Clin-ical Data Base is that it provides risk-adjusted expected valuesfor LOS, mortality, and cost so that differences in case-mix canbe accounted. The UHC Data base performs this risk adjust-ment with four steps:9 1) assignment of a severity of illness; 2)identification of a Patient Population for Model Generation; 3)use of multiple regression techniques to predict LOS, cost, andprobability of mortality based on a normative patient popula-tion; and 4) assignment of an expected LOS, cost, and mortalityto every patient in the database. To further clarify the designa-tion of severity of illness, UHC uses the All Payer Refined–Diagnosis Related Group (APR–DRG). It allows the severity ofillness (SOI) and risk of mortality (ROM) to be disease-specific

when the significance of comorbid conditions is dependent onthe underlying primary condition. Thus, with these models,each patient is assigned a severity of illness and a risk of mor-tality level (minor, moderate, major, or extreme). APR-DRGsare clinical models that have been extensively tested by 3MVR

with historical data. The historical data used in the develop-ment of version 20.0 of the APR-DRGs was a nationwidedatabase of 8.5 million discharges, which included all payer dis-charges from 1,000 general hospitals from 10 states and allpayer discharges from 47 children’s hospitals in the UnitedStates.10

In-hospital mortality is defined as the percentage ofpatients who died during the admission period. Deaths thatoccur after discharge are not counted, even if they occur within30 days. LOS is defined by computing the difference betweenthe time of discharge and the index procedure. Using an algo-rithm, the UHC calculates an estimated value for mortality,cost, and LOS. Mortality index, cost index, and LOS index arecalculated by taking the ratio of the observed to expected valuesof each variable.

UHC estimates costs of patient care using a ratio of costto charges (RCC) methodology.11 The UHC obtains service linecosts and revenues from Centers for Medicare and Medicaidservices, from which it uses to calculate a RCC for each service.UHC then collects detailed patient charges for each service andmultiplies it by the RCC to determine an estimated cost.

Readmission rates are defined by readmissions for any rea-son occurring within a given time after discharge (30 days, 14days, or 7 days). However, readmissions to hospitals other thanwhere the index procedure was performed are not counted.

Comorbid conditions recorded in the UHC database aredefined by the Agency for Healthcare Research and Quality(AHRQ) based on research by Elixhauser, et al.12

We analyzed the UHC database for discharge data on allpatients aged 18 to 100 years who underwent surgery for headand neck cancers at UHC full-member academic medicalcenters between quarter 4 of 2006 to quarter 4 of 2009. Weincluded younger patients, due to the rising trend of humanpapilloma virus-induced cancer in younger patients. All hospi-talizations were identified on the basis of principal diagnosis, asspecified by the International Classification of Diseases, 9thEdition (ICD9), and principal procedure codes, as specified bythe International Classification of Diseases, 9th Edition, Clini-cal Modification (ICD-9-CM). The principal ICD-9 diagnosiscodes that were used are specified in Table I, and the ICD9-CMprocedure codes used are listed in Table II. All major cancers ofthe head and neck were included and yielded 22,357 cases. Wethen excluded surgery for thyroid cancer and skin cancer. Thesecancers were excluded because, from our experience, these

TABLE I.ICD 9 Codes.

ICD 9 Codes Description

1400–1409 Lip

1410–1419 Tongue

1420–1422; 1428–1429 Salivary gland

1430–1431; 1438–1439 Gum

1440–1441; 1448–1449 Floor of the mouth

1450–1459 Mouth

1460–1469 Oropharynx

1470–1473; 1478–1479 Nasopharynx

1480–1483; 1488–1489 Hypopharynx

1490–1491; 1498–1499 Ill-defined sites

1600–1605; 1608–1609 Nasal cavity, sinuses

1610–1613; 1618–1619 Larynx

1700 Skull, face

1701 Mandible

1930 Thyroid

ICD 9-International Classification of Diseases, 9th Edition.

TABLE II.ICD9-CM Codes.

ICD 9-CM Code Description

300–304 Larynx

213, 214 Nose, mouth, pharynx

251–254 Tongue

2629 Salivary

2731–2732; 2742–2743; 2749; 2772 Mouth, face

280–289 Tonsil

2933–2939 Pharynx

403–405 Neck dissection

ICD 9-CM-International Classification of Diseases, 9th Edition, Clini-cal Modification.


690

operations are very different in that the surgeries are shorter,have fewer complications, and require fewer days in the hospi-tal. Extraction of cases with at least one of the diagnosis codesand at least one of the procedure codes resulted in 11,573 casesof head and neck cancer during the 3-year time periodevaluated.

Hospital volume was the primary independent variable inthis study. Secondary independent variables available from theUHC database were age, sex, race, APR-DRG–defined severityof illness, and method of payment. Cancer staging, cancerpathology, and histologic subtypes were not available from theUHC database.

Prior to analysis and after the dataset was obtained, hos-pitals were divided into tertiles and labeled high, moderate, andlow volume hospitals. There currently exist no guidelines as towhat defines a high volume versus a low volume hospital forhead and neck cancer. The analysis showed that high volumecenters performed greater than 50 cases per year, moderatevolume centers performed 22 to 49 cases per year, and lowvolume centers performed 21 cases or less per year during thetime period evaluated.

AnalysisThe primary clinical outcomes for this study were mortal-

ity, mortality index, LOS, LOS index, complication rate, cost,and cost index. Secondary outcome measures included IntensiveCare Unit (ICU) stay and readmission rates at 7, 14, and30 days after procedure. Additional data on comorbidities werealso recorded.

Mean and standard deviation for continuous variables,frequencies, and proportions were calculated to describe thedifference between three groups of hospitals for each of the out-come measures and patients characteristics. The statisticalsoftware SAS Version 9.2 (SAS Institute Inc., Cary, NC) wasused. ANOVA was performed to test the difference between con-tinuous variables, and chi-square test was performed to testdifferences in rates or proportions between the three groups.Regression model analysis was conducted to examine the associ-ation between independent variables available from the UHCdatabase of volume, age, sex, race, APR-DRG defined severityof illness, and method of payment and clinical outcomes. Fiveoutcomes were examined using linear regression analysis:mortality rate index, LOS observed, cost observed, cost index,and LOS index. Logistic regression models were used formortality rate and complication rate. In the regression model,proportions of each patient characteristic for each hospital wereused. P value of less than 0.05 was considered significant.

RESULTSNinety-three hospitals were involved in this study

after application of our criteria based on diagnosis andprocedure codes. These hospitals were divided by tertilesinto three groups. There were 30 high-volume hospitals(n ¼ 7,506 cases), 30 moderate volume hospitals (n ¼2,963 cases), and 33 low volume hospitals (n ¼ 1,104cases). No individual data from patients or surgeons wasaggregated or available at the hospital level.

Patient CharacteristicsTable III shows the distribution of patient charac-

teristics by hospital volume. The majority of head andneck cancer patients were older than 51 years of age(80.77%), males (70.12%), and white (78.67%) when

aggregated across the three volume groups. Furtheranalysis and comparison of the high and moderate vol-ume hospitals to the low volume hospitals demonstratedthat low volume hospitals care for a proportionatelyhigher number of patients age 31 to 50 and 51 to 64(P ¼ 0.0043). On the other hand, high and moderatevolume hospitals care for a larger proportion of patientsgreater than 65 years old. Low volume hospitals had ahigher proportion of male patients (71.29%) compared tohigh volume and moderate volume hospitals, but it wasnot statistically significant. From a racial perspective,high volume hospitals cared for a statistically significantlower proportion of black and Hispanic patients andhigher proportion of white patients compared to moder-ate and low volume hospitals (P <0.0001). There was ahigher proportion of major severity of illness in the mod-erate volume group and minor severity of illness in thelow volume group (P <0.0001). The payer mix was alsoskewed among the three volumes of hospitals. The highvolume hospitals had a higher proportion (43.65%) oftheir patients with private insurance and a lower pro-portion of Medicaid and Medicare patients compared tothe lower volume groups (P <0.0001). There was ahigher proportion of Medicare recipients (43.81%) in themoderate volume hospitals compared to high volumeand low volume hospitals.

Patient OutcomesTable IV shows the difference in outcome measures

between the three volumes of hospitals. High volumehospitals had a lower mortality rate (0.61%) versus lowvolume hospitals (0.72%), but this was not statisticallysignificant. Mortality Index (observed mortality dividedby expected mortality) was lower in the higher volumehospitals, but this was not statistically significant. Theoverall complication rate was lower in low volume hospi-tals (19.42%) versus high volume hospitals (20.30%), butthis was not statistically significant. Length of stay, LOSindex (LOS observed divided by LOS expected), andIntensive Care Unit (ICU) stay rate were significantlylower in high volume hospitals than in lower volumehospitals (P ¼ 0.0401, P ¼ 0.0465, and P <0.0001,respectively).

Mean cost of admission was lower in the highvolume group versus the lower volume groups, with anaverage cost of $22,716 in high volume hospitals and$24,059 in low volume hospitals (P ¼ 0.6576). Similarly,the cost index (cost observed divided by cost expected)was higher in low volume hospitals (1.3 vs. 1.2 and 1.1)compared to high and moderate volume hospitals (P ¼0.4644). Mean readmission rates were 2.51%, 4.22%, and6.48% for 7 day, 14 day, and 30 day readmissions,though this was not significantly different across thethree groups.

In order to further evaluate the differences betweenthe groups, we performed regression analysis. Table Vshows the results from the regression analysis, withcoefficients for each variable and P values in brackets.Hospital volume was not significant for any outcome pa-rameters except complication rate. A lower complication


691

TABLE IV.Outcome of Patients by Hospital Volume.

PATIENT OUTCOMES High Volume Moderate Volume Low Volume P value

Number of patients 7,506 2,963 1,104

Mortality, N (%) 46 (0.61) 25 (0.84) 8 (0.72) 0.4269

Mortality Index 0.9 (13.30) 1.1 (13.18) 1.4 (24.63) 0.5698

LOS (days), Mean(STD) 6.8 (23.66) 7.7 (19.29) 7.9 (17.00) 0.0401

LOS Index (STD) 1.2 (3.13) 1.3 (3.18) 1.4 (2.27) 0.0465

ICU Stay, N (%) 2,678 (35.68) 1,228 (41.45) 455 (41.22) <0.0001

ICU Days, Mean (STD) 3.8 (24.43) 4.4 (17.34) 4.4 (18.42) 0.2443

30 Day Readmit, N (%) 504 (6.76) 190 (6.47) 56 (5.11) 0.1183

14 Day Readmit, N (%) 328 (4.40) 129 (4.39) 32 (2.92) 0.0707

7-Day Readmit, N (%) 196 (2.63) 78 (2.65) 17 (1.55) 0.0940

Complications, N (%) 1,521 (20.30) 623 (21.03) 214 (19.42) 0.4903

Comorbidities, N (%) 6,248 (83.24) 2,475 (83.53) 899 (81.43) 0.2624

Cost ($), mean (STD) 22,716 (84,751.9) 23,222 (55,952.5) 24,059 (64,244.4) 0.6576

Cost Index 1.2 (3.62) 1.1 (2.45) 1.3 (2.89) 0.4644

*Means were compared using analysis of variance (ANOVA).Variables with frequency were tested using chi-square test.For indexes, tests were performed using the indexes for each hospital, while the listed indexes were calculated from overall observed and expected.ICU ¼ Intensive Care Unit; LOS ¼ Length of Stay; STD ¼ Standard Deviation.

TABLE III.Characteristics of Patients by Hospital Volume.

High Volume N ¼ 30 Moderate Volume N ¼ 30 Low Volume N ¼ 33 P value

Number of patients: 7,506 2,963 1,104

Age, n (%) 0.0043

18–30 92 (1.23) 41 (1.38) 9 (0.82)

31–50 1,328 (17.69) 522 (17.62) 233 (21.11)

51–64 3,145 (41.90) 1,184 (39.96) 476 (43.12)

� 65 2,941 (39.18) 1,216 (41.04) 386 (34.96)

Sex, n (%) 0.3338

Male 5,229 (69.66) 2,099 (70.84) 787 (71.29)

Female 2,277 (30.34) 864 (29.16) 317 (28.71)

Races, n (%) <0.0001

White 6,168 (82.17) 2,201 (74.28) 736 (66.67)

Black 627 (8.35) 386 (13.03) 145 (13.13)

Asian 113 (1.51) 54 (1.82) 24 (2.17)

Hispanic 119 (1.59) 100 (3.37) 97 (8.79)

Other 479 (6.38) 222 (7.49) 102 (9.24)

SOI, n (%) <0.0001

Minor 2,221 (29.59) 763 (25.75) 379 (34.33)

Moderate 3,598 (47.93) 1,480 (49.95) 504 (45.65)

Major 1,314 (17.51) 559 (18.87) 174 (15.76)

Extreme 373 (4.97) 161 (5.43) 47 (4.26)

Payer, n (%) <0.0001

Private 3,276 (43.65) 894 (30.17) 349 (31.61)

Medicaid 649 (8.65) 513 (17.31) 194 (17.57)

Medicare 3,117 (41.53) 1,298 (43.81) 406 (36.78)

Other Govt. Plans 216 (2.88) 71 (2.40) 80 (7.25)

Other 248 (3.30) 187 (6.31) 75 (6.79)

Comorbidities, N (%) 6,248 (83.24) 2,475 (83.53) 899 (81.43) 0.2624

*P values from chi-square tests. SOI ¼ Severity of Illness.


692

rate was noted in higher volume hospitals (P ¼ 0.0486)when compared to low volume hospitals. We determinedthat overall mortality rate was higher in the 51 to 64year age group compared to other age groups (P ¼0.0123).

Interestingly, there was no difference in mortalityfor patients greater than 65 years of age compared toother age groups. Care of male patients was associatedwith a statistically significant lower cost index (P ¼0.00472). The complication rate was higher in the blackpopulation compared to other races (P ¼ 0.0297); other-wise, black race was not significant for mortality rate,LOS, cost, mortality index, cost index, or LOS index. Asexpected, the complication rate, LOS observed, and costobserved were significantly higher in the major severitygroup (P <0.001, P ¼ 0.0139, and P ¼ 0.0092, respec-tively). Private payer status resulted in a statisticallysignificant lower complication rate, LOS observed, andcost observed (P ¼ 0.0401, P ¼ 0.0001, and P ¼ 0.0187,respectively). The LOS was reduced by 7.31 days in theprivate payer group (P ¼ 0.0001).

ComorbiditiesThe major comorbidities in head and neck cancer

surgical patients are listed in Table VI. The most com-mon comorbidities seen across the three hospital volumegroups were hypertension, metastatic cancer, chronicpulmonary disease, diabetes, hypothyroidism, and fluidelectrolyte disorders. The only statistically significantcomorbidities were metastatic cancer and chronic pulmo-nary disease. A higher proportion of patients withmetastatic cancer (P ¼ <0.0001) were managed by highvolume hospitals. Moderate volume hospitals had ahigher proportion of chronic pulmonary disease patientscompared to other hospitals (P <0.0001). Other comor-bidities that were cared for more by moderate volumehospitals included depression, anemia, alcohol abuse,weight loss, and drug abuse (P ¼ 0.0049, P ¼ 0.0116,P ¼ 0.0042, P <0.0001 and P ¼ 0.0009, respectively).There were more patients with psychoses managed bylow volume hospitals (P ¼ 0.0152).

ComplicationsThe complications from head and neck cancer

surgery are delineated in Table VII across the three hos-pital volume groups. Miscellaneous complicationsincluded air embolism, blood incompatibility reactions, aforeign body accidentally left in patient during proce-dure, reaction to foreign body left during procedure,disruption of wound, and emphysema resulting from theprocedure. The rate of miscellaneous complications wassignificantly lower in high volume hospitals (P <0.0044).There was higher postoperative pulmonary compromisein high volume hospitals (P ¼ 0.0315). Additionally, theaverage rates of common complications after head andneck surgery are shown in this table. [Postoperativehemorrhage rates are 4.67%; postoperative pulmonaryrates are 3.76%; wound infection rates are 2.27%; aspira-tion pneumonia rates are 1.64%; postoperative deep

TABLEV.

Resultsfrom

RegressionModelAnalysis:Estim

ate

(Pvalue).

Model1:

MortalityRate

Model2:

Complication

Model3:

LOSobs

Model4:

Costobs

Model5:

MortalityIndex

Model6:

LOSindex

Model7:

CostIndex

Volume

HighversusLow

�0.29(0.3544)*

�0.13(0.0486)

�0.25(0.7009)

�204.1

(0.9368)

�0.40(0.5218)

�0.12(0.2690)

�0.09(0.4832)

Moderate

versusLow

�0.17(0.7077)

�0.03(0.7338)

�0.22(0.7492)

�1941(0.4844)

�0.01(0.9820)

�0.07(0.5137)

�0.12(0.3596)

Sexproportion

male

�1.61(0.5113)

�0.92(0.0781)

�3.10(0.3896)

�20169(0.1514)

3.18(0.3609)

�1.00(0.0965)

�1.37(0.0472)

Ageproportion

51-64

9.42(0.0123)

�0.13(0.8535)

4.46(0.3399)

19333(0.2907)

6.36(0.1595)

0.79(0.3061)

0.93(0.2985)

�65

5.42(0.0641)

0.50(0.3639)

�0.89(0.8169)

2424.1

(0.8754)

1.40(0.7059)

�0.18(0.7822)

0.06(0.9360)

Racesproportion

Black

�0.64(0.6229)

0.65(0.0297)

�0.41(0.8457)

�2615(0.7573)

�2.34(0.2556)

�0.14(0.6958)

�0.11(0.7977)

Seve

rity

proportion

Major

�1.13(0.5287)

3.88(<

0.0001)

6.31(0.0139)

26829(0.0092)

�1.79(0.4621)

�0.48(0.2495)

0.10(0.8417)

Paye

rproportion

Private

�1.42(0.2344)

�0.53(0.0401)

�7.31(0.0001)

�16970(0.0187)

�0.64(0.7045)

�0.50(0.0910)

�0.27(0.4363)

Numbers

ineachcellis:coefficient(P

value).

LOS¼

Length

ofStay;Obs¼

Observed.


693

venous thrombosis and pulmonary embolism rates are1.30%; postoperative infection (not pneumonia or woundinfection) rates are 0.92%; postoperative acute myocar-dial infection rates are 0.73%; postoperative rates are0.28%; and postoperative urinary tract complicationsrates are 0.23%.]

DISCUSSIONAs reforms unfold in the system of providing health

care in the United States, there is an increasing empha-sis being placed on the quality, cost, and measurableoutcomes of care provided. Approximately $3.2 billion isspent in the United States each year on treatment ofhead and neck cancers.13 Accordingly, with the advent ofpay for performance in many states, reimbursement isdirectly tied to outcomes with lower complication rates.Surgery for head and neck cancers is labor intensive,requiring many resources, and the coordinated efforts ofmany well-trained health care providers who must havethe education, training, and experience requisite to pro-vide an optimal level of care. While patients with headand neck cancer understandably might, for the sake of

convenience, have a natural tendency to seek care asclose to home as possible—perhaps in a local communityhospital—when considering all factors, the desired out-come for some patients might be more achievable if careis provided at an academic medical center with highcase volume.14 This concept prompts consideration ofwhat has been called ‘‘regionalization’’ of care.

Regionalization is defined as the delivery of care ata limited number of select provider sites.15 Regionaliza-tion is an important concept to consider with proposedhealth care reforms and insurance companies movingtoward Accountable Care Organizations, which not onlyare going to be responsible for their costs and budgets,but also for the quality of care. The relationship betweenvolume of surgery and outcomes has been establishedfor complex abdominal procedures such as esophagec-tomy, pancreatectomy, and even bariatric surgery.16–21

To our knowledge, there has been no nationalbenchmarking data against which academic head andneck surgical oncology units can compare and measuretheir own performance based on volume and outcomes.One article by Bhattacharya et al.22 reviewed medicalcomplications in patients undergoing head and neck

TABLE VI.Comorbidity Comparison.

COMORBIDITIES High Volume (%) Moderate Volume (%) Low Volume (%) P value

Hypertension 3,503 (46.67) 1,424 (48.06) 501 (45.38) 0.2487

Metastatic cancer 2,635 (35.11) 887 (29.94) 337 (30.53) <0.0001

Chronic pulmonary disease 1,353 (18.03) 666 (22.48) 189 (17.12) <0.0001

Diabetes without complication 955 (12.72) 395 (13.33) 152 (13.77) 0.5044

Hypothyroidism 800 (10.66) 293 (9.89) 110 (9.96) 0.4507

Fluid-electrolyte isorder 728 (9.70) 326 (11.00) 115 (10.42) 0.1282

Depression 560 (7.46) 222 (7.49) 53 (4.80) 0.0049

Solid tumor without metastasis 493 (6.57) 186 (6.28) 73 (6.61) 0.8514

Deficiency anemia 480 (6.39) 238 (8.03) 76 (6.88) 0.0116

Alcohol abuse 483 (6.43) 242 (8.17) 86 (7.79) 0.0042

Weight loss 393 (5.24) 250 (8.44) 37 (3.35) <0.0001

Peripheral vascular disease 321 (4.28) 127 (4.29) 37 (3.35) 0.3427

Obesity 274 (3.65) 134 (4.52) 44 (3.99) 0.1150

Congestive heart failure 261 (3.48) 97 (3.27) 39 (3.53) 0.8590

Valvular disease 250 (3.33) 87 (2.94) 29 (2.63) 0.3291

Other neurological disorders 227 (3.02) 90 (3.04) 42 (3.80) 0.3672

Renal failure 216 (2.88) 98 (3.31) 29 (2.63) 0.3973

Liver disease 170 (2.26) 66 (2.23) 22 (1.99) 0.8492

Rheumatoid arthritis–collagen vascular disease 119 (1.59) 52 (1.75) 15 (1.36) 0.6496

Psychoses 117 (1.56) 65 (2.19) 28 (2.54) 0.0152

Drug abuse 98 (1.31) 68 (2.29) 22 (1.99) 0.0009

Diabetes with complication 86 (1.15) 36 (1.21) 9 (0.82) 0.5530

Coagulopathy 90 (1.20) 36 (1.21) 15 (1.36) 0.9029

Pulmonary circulation disease 63 (0.84) 35 (1.18) 8 (0.72) 0.1990

Paralysis 34 (0.45) 15 (0.51) 5 (0.45) 0.9348

Lymphoma 34 (0.45) 9 (0.30) 2 (0.18) 0.2752

Acquired immune deficiency syndrome 21 (0.28) 10 (0.34) 1 (0.09) 0.4093

Chronic blood loss anemia 34 (0.45) 19 (0.64) 6 (0.54) 0.4696

Peptic ulcer disease 0 (0) 2 (0.07) 1 (0.09) 0.0577


694

surgery. Unfortunately, 40.8% of the cases in this reviewwere thyroid surgery patients, which are consideredlow-acuity procedures by most head and neck surgicalpractices.14 Hence, in designing this study we purposelyexcluded thyroid and skin cancer operations, since theyare relatively straightforward and do not present thesame postoperative challenges as posed by a mandibulec-tomy with a free tissue transplant reconstruction.Gourin et al.23 showed that high volume hospital care inthe state of Maryland (greater than 18 cases per year)was associated with a shorter length of hospitalizationand lower cost of care for laryngeal surgery alone.Recently, it has been shown that high volume hospitalcare is associated with a shorter length of hospitalizationand lower hospital-related cost of care for oropharyngealcancer surgery.24

Our study results demonstrated that after control-ling for all other factors, high volume hospitals have alower complication rate. Moreover, high and moderatevolume hospitals also deal with a higher level of comor-bidities. In studies comparing volume of surgery tooutcome, the main weakness has been the limited riskadjustment in the databases used.25 The advantage ofusing the UHC database is that it uses extensive riskadjustment methodology to assign severity of illness,risk of mortality, observed cost and LOS. Hence,expected mortality, expected cost and expected LOS canbe assigned to all hospital volume groups to control fordiffering patient comorbidities based on the APR-DRGgrouping system. Ratio of a variable (e.g. observed mor-tality divided by expected mortality) can be calculated as

an index to account for case mix differences. Hence, ofgreater interest is the association of severity of illness,mortality index, cost index and LOS index with out-comes. In our study, APR-DRG defined major severity ofillness was the only independent variable significantlyassociated with a higher complication rate, observedLOS and observed cost (P <0.0001, P ¼ 0.0139 and P ¼0.0092, respectively).

LOS, LOS observed, cost, cost index, mortality rate,and mortality index were not significantly differentacross the three volume groups based on our regressionanalysis.

Mean ICU stay was not significantly differentbetween higher and lower volume hospitals (P ¼0.2443), but there was lower utilization trend in highvolume hospitals versus low volume hospitals (35.68%versus 41.22%), and this was statistically significant(P <0.0001). When we couple this with the higher rateof postoperative pulmonary compromise in high volumehospitals, it suggests that there is better management ofpulmonary complications in high volume hospitals, withno increase in mortality rate, ICU stay or LOS. Thiswould suggest that there are better system processes atwork in high volume hospitals that help reduce overallutilization of resources and hence do not increase costs.

Other parameters of outcome including 30 day, 14day, and 7 day readmit rates were not statistically differ-ent across the 3 groups. However, this could also be alimitation of the UHC database since it only includesindex case inpatient data; no outpatient data is collected.Hence, if a patient is admitted to another institution

TABLE VII.Complication Comparison.

COMPLICATIONS High Volume (%) Moderate Volume (%) Low Volume (%) P value: Chi-square

Other Complication of Procedure 412 (5.50) 169 (5.71) 43 (3.90) 0.0630

Postop hemorrhage/hematoma 352 (4.70) 146 (4.93) 43 (3.90) 0.3843

Miscellaneous complication 349 (4.66) 179 (6.04) 68 (6.17) 0.0044

Postop pulmonary compromise 307 (4.10) 90 (3.04) 38 (3.45) 0.0315

Mechanical complication due to device or implant 245 (3.27) 117 (3.95) 41 (3.72) 0.2115

Wound infection 181 (2.42) 61 (2.06) 21 (1.91) 0.3741

Aspiration PNA 120 (1.60) 49 (1.65) 21 (1.91) 0.7594

DVT–PE 96 (1.28) 43 (1.45) 12 (1.09) 0.6297

Reopening of surgical site 70 (0.93) 25 (0.84) 6 (0.54) 0.4215

Postop infection (not PNA, wound) 73 (0.97) 22 (0.74) 12 (1.09) 0.4504

Postop AMI 55 (0.73) 24 (0.81) 6 (0.54) 0.6778

Procedure related hematoma/laceration 59 (0.79) 30 (1.01) 12 (1.09) 0.3884

Cellulitis–decubitus ulcer 40 (0.53) 19 (0.64) 4 (0.36) 0.5493

Postop stroke 21 (0.28) 9 (0.30) 3 (0.27) 0.9759

Postop urinary tract complication 19 (0.25) 6 (0.20) 3 (0.27) 0.8719

Postop physical–metabolic derangement 21 (0.28) 4 (0.14) 2 (0.18) 0.3562

Postop cardiac abnormality (not AMI) 6 (0.08) 0 (0.00) 1 (0.09) 0.2964

Shock, cardiorespiratory arrest 5 (0.07) 2 (0.07) 0 (0.00) 0.6912

Postop GI hemorrhage, ulceration 8 (0.11) 4 (0.14) 2 (0.18) 0.7763

Nervous system complication 6 (0.08) 2 (0.07) 2 (0.18) 0.5197

Postop coma, stupor 3 (0.04) 1 (0.03) 0 (0.00) 0.8001

AMI¼ acute myocardial infarction, DVT¼ deep venous thrombosis, GI¼ gastrointestinal, PE¼ pulmonary embolism, PNA¼ pneumonia.


695

after surgery, this would not be reflected in the primaryinstitution’s readmission rate.

The mean cost of head and neck cancer surgery was$1,343 less at high volume hospitals than the mean costat low volume hospitals (P ¼ 0.6576). The cost index(observed cost divided by expected cost) is another usefulparameter for comparison which takes into considerationpatient comorbidities in assigning expected costs. Therewas a lower CI for high volume and moderate volumehospitals compared to low volume hospitals (P ¼0.4644), though the difference was not statisticallysignificant.

According to our analysis, a higher proportion ofpatients with metastatic cancer (P ¼ <0.0001) were man-aged by high volume hospitals. Moderate volumehospitals had a higher proportion of chronic pulmonarydisease patients compared to other hospitals (P <0.0001).Other comorbidities that were significantly cared formore by the moderate volume hospitals included depres-sion, anemia, alcohol abuse, weight loss, and drug abuse.(P ¼ 0.0049, P ¼ 0.0116, P ¼ 0.0042, P <0.0001 and P ¼0.0009, respectively). All of these comorbidities were asso-ciated with a higher complication rate, which shouldtheoretically result in longer LOS, costs, and mortality.This was not shown in our study. Such patients requiremore utilization of resources and have a higher complica-tion rate but again points to better institutional processesin place that result in no difference in LOS, LOS index,cost, Cost Index, or mortality when comparing highervolume hospitals to low volume hospitals. This analysisstresses the importance of using the index methodologyrather than the observed costs since we can take intoaccount the expected costs based on a patient’s comorbid-ities, and the UHC database allows for incorporation ofthe case mix index, which has been a limitation in manystudies.5,6,14,15,23,24

On the other hand higher miscellaneous compli-cations were noted in lower volume hospitals.Miscellaneous complications included air embolism,blood incompatibility reactions, foreign body accidentallyleft during procedure, reaction to foreign body left dur-ing procedure, disruption of wound, and emphysemaresulting from the procedure (P ¼ 0.0044).

There have been numerous studies that have shownbetter patient outcomes for surgical procedures done athigh volume hospitals.18,23,26–33 This study confirms theassociation between higher hospital volume and lowercomplication, but this does not prove causation.Moreover, this study suggests that despite higher comor-bidities managed by higher volume (greater than 21cases per year) hospitals, they are still able to managepatients efficiently and without cost increases. On theother hand this study also shows that hospitals perform-ing less than 20 cases per year do not have astatistically significant higher LOS, LOS observed, cost,cost index, mortality rate, and mortality index as com-pared to higher volume hospitals.

Several studies have described that the reason forimproved outcomes and lower costs in high volumehospitals is the interplay of health care providers andinstitutional processes.3,14–16 The components at the

surgical team level include: special training and exper-tise of surgeons, nurses, and other providers who canhelp in early recognition and optimization of comorbid-ities, and hence in better selection of patients andperioperative clinical decision making, which in turn canhelp reduce medical errors. It is likely that an experi-enced surgical team has better ability to handleintraoperative complications that translate into betterpostoperative outcomes. Institutional processes that areimportant for the success of head and neck cancer sur-gery include: availability of intensive care staffing,operating rooms with adequate equipment, diagnosticand imaging modalities, rehabilitation facilities, andappropriately trained nursing staff. The presence of com-prehensive multidisciplinary care results in earlydetection and management of complications and errors.

We attempted to look for any inherent differencesin patient demographics. Our study shows that there isa statistically significant lower-cost index for managingmale patients (P ¼ 0.0472) compared to females. Thismay suggest a gender inequality in obtaining healthcare. Females may present with more advanced disease.The stage of cancer at diagnosis was not present in theUHC database, so this may be a source of futureresearch on gender differences in the management ofhead and neck cancer.

In this study, we showed that patients aged 51 to64 years had a statistically significant higher mortalityrate than other age groups. One would assume that thegeriatric patient (age 65 and higher) would be at higherrisk of mortality after head and neck surgery, but we didnot see this in our study. This may suggest that moreresources should be utilized in the care of the 51- to64-year-old patient to reduce mortality rates. This raisesan interesting question for future studies as to why thegeriatric population does not have higher complicationsor costs related to them.

Our study shows that blacks had a statistically sig-nificant higher rate of complication (P ¼ 0.0297)compared to other races. Since blacks may represent amore socially disadvantaged group with more comorbid-ities, a reason for this difference could be the differencein access to care. They may present with more advancedcancers resulting in a higher complication rate. Racialinequality differences need to be further addressed infuture research.

The payer mix in different hospitals also helps usunderstand the delivery of care in head and neck cancerpatients. The high volume hospitals have 43.65% of theirpatients with private insurance as compared to moder-ate and low volume hospitals (P < 0.0001). This mayrepresent insured patients seeking out the institutionsand health care providers that are known for surgicalexpertise. The private payer group also had a statisti-cally significant lower LOS by 7.31 days (P ¼ 0.0013),complication rate (P ¼ 0.0401), and observed cost (P ¼0.0187). This may be explained by the aggressive utiliza-tion reviews employed by many private insurers forinpatient hospitalization. Furthermore, eagerness andavailability of postoperative skilled nursing facilities toaccept privately insured patients reduces time spent in


696

an acute care hospital, and hence reduces hospital levelcosts.

This association requires an analysis of privateinsurance oversight of utilization and care of theirpatients that results in lower complications, LOS, andcosts. Further dissection of these methodologies mayhelp translate into lower health-care costs for the future.

The lower complication rate in the high volumegroup can be explained by many of the reasons alreadystated above. The high volume of cases allows an experi-ence-driven standardized approach to be used on thesehead and neck cancer patients. With experience, theymay be able to optimize comorbidities preoperatively,identify and treat complications early in the postopera-tive period, and hence improve outcomes. Moreover, withthe presence of multiple specialties in high volume hos-pitals, they are able to more efficiently manage highercomplexity patients.

The findings in this study may suggest that sur-geries for head and neck cancer should be regionalizedto high or moderate volume institutions. Third partypayers are constantly searching for ways to reduce costand improve outcomes. Regionalization has been sug-gested for other high-risk procedures such aspancreaticoduodenectomies. Gordon et al. found that84% of mortality, 15% of hospital days, and 17.2% of hos-pital charges would be avoided if Whipple procedureswere referred to high volume centers.15 Similarly, forhead and neck cancer cases, regionalization may resultin better patient outcomes for patients currently beingtreated at low volume hospitals.23,24 However, this maybe very difficult to implement. Patients would have totravel a further distance to find the hospital with largervolume, which may be costly and cumbersome. Also, thelarger volume hospitals may be operating at capacity,and they may have difficulty accommodating the extrainflux of patients. Furthermore, when looking at individ-ual hospital data, the majority of hospitals in the lowvolume group actually had a 0% mortality rate. As agroup, they fared worse than the high and moderatevolume hospitals, but in actuality, many hospitals in thegroup still had good outcomes.

Major head and neck oncologic resections also bene-fit from incorporation of clinical care pathways (CCP). ACCP is a structured patient health care plan that organ-izes daily interventions and goals for a specific diagnosisor procedure along a time line.34,35 Gendron et al.34

showed that the CCP reduced LOS from 13 days to 8days in the first year and from 10.5 to 6.4 days at 3years after implementation.

Median charges declined from $105,410 to $65,919.High volume hospitals may incorporate CCPs due to thelarge volume of patients that they encounter, and thisallows for a more seamless coordination of perioperativeservices including nursing, rehabilitation services,speech therapy, respiratory therapy, nutrition, and socialwork.

There are several limitations of this study. First,this is an ecological study design since individual patientdata was not available and we rely on hospital leveldata. We do not have data on previous therapy, stage of

cancer, or histology of the tumor. We are at risk for anecological fallacy, that is, we assume that individualmembers of a group have the average characteristics ofthe group at large. Second, the mortality and morbidityare calculated from in-hospital discharge data. There-fore, deaths or complications that occur after dischargeare not incorporated in the calculations. This artificiallylowers the mortality rate for all three groups. Third, anycosts incurred from readmissions are not included in theindex hospital stay cost, so this falsely lowers our costestimates for the hospitals. However, since readmissionswere similar among the three groups, the comparisonbetween the categories is still valid. Also, any consulta-tions, diagnostic tests, or imaging tests that were donepreoperatively were not included in the cost, whichagain falsely lowers our cost estimates for all threegroups.

Does this mean a patient should definitely choosethe hospital performing greater than 50 surgeries peryear? Clearly, the answer is no. A more comprehensivecomparison is required. This study does not serve as astrict guideline, but as further evidence to show theassociation of increased volume lowering hospital com-plications. The volume should only play a part in thedecision to choose a hospital, along with consideringcomfort with the nursing staff and ease of proper follow-up care. Moreover, we found that even hospitals per-forming less than 20 surgeries per year had nostatistical difference in their LOS, LOS observed, cost,cost index, mortality rate, and mortality index comparedto higher volume hospitals. Finally, this study only looksat index admission hospital level data. There is no longterm follow-up data available.

Despite these limitations, the large sample size ofthis study provides us with reliable data to serve as themost up-to-date and accurate benchmarking measuresthat are available for academic head and neck cancersurgeons. It incorporates comorbidity, severity of illness,and risk of mortality data for the first time in a studylooking exclusively at head and neck surgical oncologyoutcomes. More importantly, data is gathered on analyz-ing cost index, mortality index, and LOS index thatincorporates comorbidity and case mix data into themodel created by UHC. The results of this study are im-portant given the forthcoming global payments modelfor health care and accountable care organizations,whereby there will be a greater emphasis on quality andcosts of care. Moreover, there is discussion of makingquality data publicly available as in the case of cardiacsurgery in Massachusetts. All this makes for a betterargument for the regionalization of head and neck can-cer surgery. Head and neck surgical oncologydepartments may now compare their own performanceto that of other academic medical centers.

CONCLUSIONThis study used an externally managed national

database to examine discharge data from 2006 to 2009for patients who underwent surgery for head and neckcancer at 93 academic hospitals throughout the country.


697

We found that high volume hospitals (greater than 50cases per year) had a lower complication rate than thelow volume (less than 20 cases per year) hospitals, evenafter accounting for differing case mixes. On the otherhand, lower volume hospitals had no statistical differ-ence in their LOS, LOS observed, cost, cost index,mortality rate, and mortality index compared to highervolume hospitals. We conclude that care of head andneck surgical oncology patients at higher volume aca-demic medical centers is associated with a lowercomplication rate. More physicians, better staffed surgi-cal floors and ICUs, and better equipment may becontributing to these outcomes at high volume hospitals.Regionalization of complex cases to higher volume hospi-tals may result in better patient outcomes with similarcosts, but this may be difficult to implement sincepatient travel would be costly and follow up may beinconvenient.

BIBLIOGRAPHY

1. The Agency for Healthcare Research and Quality (AHRQ) QualityIndicators. Available at: http://www.qualityindicators.ahrq.gov/psi_overview.htm. Accessed January 22, 2011.

2. The Center for Medicare Services. Available at: https://www.cms.gov/HospitalAcqCond/Downloads/HACFactsheet.pdf. Accessed January 22,2011.

3. Nguyen NT, Silver M, Robinson M, et al. Result of a national audit of bari-atric surgery performed at academic medical centers. Arch Surg 2006;141:445–450.

4. Wouters MW, Wijnhoven BP, Karim-Kos HE, et al. High volume versuslow volume for esophageal resections for cancer: the essential role ofcase-mix adjustments based on clinical data. Ann Surg Oncol 2008;15:80–87.

5. Courcoulas A. Schuchert M. Gatti G, Luketich J. The relationship of sur-geon and hospital volume to outcome after gastric bypass surgery inPennsylvania: a 3-year summary. Surgery 2003;134:613–623.

6. Lopez J, Sung J, Anderson W, et al. Is bariatric surgery safe in academiccenters? Am Surg 2002;68:820–823.

7. University Healthsystem Consortium. Available at: https://www.uhc.edu/12443.htm. Accessed January 22, 2011.

8. Clinical Database/Resource Manager User Manual October 2010. Univer-sity Healthsystem Consortium Web site available at: https://cimprod.uhc.edu/CDPRM/web/reports/Welcome.aspx#. Accessed January 22, 2011.

9. Neikirk J. Risk Adjustment Methodology for the Clinical Database.University Healthsystem Consortium Web site available at: https://www.uhc.edu/docs/003716955_RiskAdjustmentMethodology.pdf. AccessedJanuary 22, 2011.

10. 3M APR DRG Classification System. Methodology Overview ver. 26.1.The Agency for Healthcare Research and Quality (AHRQ). AHRQ Website available at: http://www.ahrq.gov/qual/mortality/Hughessumm.pdf.Accessed August 8, 2012.

11. Keroack M. Methodology for estimating inpatient care costs. UniversityHealthsystem Consortium Web site available at: https://www.uhc.edu/docs/003716704_COSTMETHrevised_4.pdf. Accessed January 22, 2011.

12. Elixhauser, A, Steiner C, Harris RD, Coffey RM. Comorbidity measuresfor use with administrative data. Medical Care 1998;36:8–27.

13. Costs of Cancer Care. National Cancer Institute Web site available at:http://progressreport.cancer.gov/2007/doc_detail.asp?pid¼1&did¼2007&chid¼75&coid¼726&mid¼. Accessed January 22, 2011.

14. Weber RS, Lewis CM, Eastman SD, et al. Quality and performance indica-tors in an academic department of head and neck surgery. Arch Otolar-yngol Head Neck 2010;136:1212–1218.

15. Gordon TA, Burleyson GP, Tielsch JM, et al. The effects of regionalizationon cost and outcome for one general high-risk surgical procedure. AnnSurg 1995:221:43–49.

16. Nguyen NT, Paya M, Stevens M, Mavandadi S, Zainabadi K, Wilson SE.The relationship between hospital volume and outcome in bariatricsurgery at academic medical centers. Ann Surg 2004;240:586–93.

17. Patti MG, Corvera CU, Glasgow RE, Way LW. A hospital’s annual rate ofesophagectomy influences the operative mortality rate. J GastrointestSurg 1998;2:186–192.

18. Birkmeyer JD, Siewers AE, Finlayson EV, et al. Hospital volume and sur-gical mortality in the United States. N Engl J Med 2002;346:1128–1137.

19. Dimick JB, Cowan JA, Upchurch GR, Colletti LM. Hospital volume andsurgical outcomes for elderly patients with colorectal cancer in theUnited States. J Surg Res 2003;114:50–56.

20. Finlayson EV, Goodney PP, Birkmeyer JD. Hospital volume and operativemortality in cancer surgery: a national study. Arch Surg 2003;138:721–725.

21. Liu JH, Zingmond D, Etzioni DA, et al. Characterizing the performanceand outcomes of obesity surgery in California. Am Surg 2003;69:823–828.

22. Bhattacharyya N, Fried M P. Benchmarks for Mortality, Morbidity, andLength of Stay for Head and Neck Surgical Procedures. Arch Otolaryn-gology Head Neck Surg 2001;127:127–132.

23. Gourin CG, Forastiere AA, Sanguineti G, Koch WM, Marur S, BristowRE. Impact of surgeon and hospital volume on short term outcomes andcost of laryngeal cancer surgical care. Laryngoscope 2011;121:85–90.

24. Gourin CG, Forastiere AA, Sanguineti G, Marur S, Koch WM, BristowRE. Impact of surgeon and hospital volume on short term outcomes andcost of oropharyngeal cancer surgical care. Laryngoscope 2011;121:74

25. Berger DH, Ko CY, Spain DA. Society of university surgeons positionstatement on the volume-outcome relationship for surgical procedures.Surgery 2003;134:34–40.

26. Halm EA, Lee C, Chassin MR. Is Volume related to outcome in healthcare? A systematic review and methodologic critique of the literature.Ann Intern Med 2002;137:511–520.

27. Dardik A, Lin JW, Gordon TA, Williams GM, Perler BA. Results of electiveabdominal aortic aneurysm repair in the 1990s: a population-basedanalysis of 2335 cases. J Vasc Surg 1999;30:985–995.

28. Ellison LM, Heaney JA, Birkmeyer JD. The effect of hospital volume onmortality and resource use after radical prostatectomy. J Urol 2000;163:867–869.

29. Philips KA, Luft HS, Ritchie JL. The association of hospital volume of per-cutaneous transluminal coronary angioplasty with adverse outcomes,length of stay, and charges in California. Med Care 1995;33:502–514.

30. Choti MA, Bowman HM, Pitt HA, et al. Should Hepatic Resection be per-formed at high-volume referral centers? J Gastrointest Surg 1998;2:11–20.

31. Lieberman MD, Kilburn H, Lindsey M, Brennan MF. Relation of perioper-ative deaths to hospital volume among patients undergoing pancreaticresection for malignancy. Ann Surg 1995;222:638–645.

32. Schrag D, Panageas K, Riedel E, et al. Hospital and surgeon procedurevolume as predictors of outcome following rectal cancer resection. AnnSurg 2002;236:583–592.

33. Romano PS, Mark DH. Patient and hospital characteristics related toin-hospital mortality after lung cancer resection. Chest 1992;101:1332–1337.

34. Gendron KM, Lai SY, Weinstein GS, et al. Clinical care pathway for headand neck cancer. Arch Otolaryngol Head Neck 2002;128:258–262.

35. Husbands JM, Weber RS, Karpati RL, et al. Clinical care pathways:decreasing resource utilization in head and neck surgical patients.Otolaryngol Head Neck Surg 1999;121:755–759.


698

outcomes in head and neck oncologic surgery at academic medical centers in the united states

Documents