Hospital volume is associated with survival but not multimodality therapy in Medicare patients with advanced head and neck cancer

Authors

  • Arun Sharma MD, MS,

    1. Department of Otolaryngology, Head and Neck Surgery, University of Washington, Seattle, Washington
    Search for more papers by this author
  • Stephen M. Schwartz MPH, PhD,

    1. Program in Epidemiology, Fred Hutchinson Cancer Research Center, Seattle, Washington
    2. Department of Epidemiology, University of Washington, Seattle, Washington
    Search for more papers by this author
  • Eduardo Méndez MD, MS

    Corresponding author
    1. Department of Otolaryngology, Head and Neck Surgery, University of Washington, Seattle, Washington
    2. Surgery and Perioperative Care Service, VA Puget Sound Health Care System, Seattle, Washington
    3. Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, Washington
    • Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue, North Mailstop D4-100, Seattle, WA 98109-1024; Fax: (206) 667-5948

    Search for more papers by this author

Abstract

BACKGROUND:

Given the complexity of management of advanced head and neck squamous cell carcinoma (HNSCC), this study hypothesized that high hospital volume would be associated with receiving National Comprehensive Cancer Network (NCCN) guideline therapy and improved survival in patients with advanced HNSCC.

METHODS:

The Surveillance, Epidemiology, and End Results (SEER)-Medicare database was used to identify patients with advanced HNSCC. Treatment modalities and survival were determined using Medicare data. Hospital volume was determined by the number of patients with HNSCC treated at each hospital.

RESULTS:

There were 1195 patients with advanced HNSCC who met inclusion criteria. In multivariable analyses, high hospital volume was not associated with receiving multimodality therapy per NCCN guidelines (odds ratio = 1.02, 95% confidence interval = 0.66-1.60), but showed a nearly significant inverse association with survival in a model adjusted for National Cancer Institute–designated cancer center status, age, sex, race, socioeconomic status, marital status, comorbidity, year of diagnosis, tumor site, and tumor stage (hazard ratio = 0.85, 95% confidence interval = 0.69-1.04).

CONCLUSIONS:

Medicare patients with advanced HNSCC treated at high-volume hospitals were not more likely to receive NCCN guideline therapy, but had nearly statistically significant better survival, when compared with patients treated at low-volume hospitals. These results suggest that features of high-volume hospitals other than delivery of NCCN guideline therapy influence survival. Cancer 2013. © 2013 American Cancer Society.

INTRODUCTION

It is estimated that 52,610 Americans will be newly diagnosed with head and neck squamous cell carcinoma (HNSCC) in 2012.1 Many patients with HNSCC have locally advanced disease with nodal metastases at the time of diagnosis, which carries a much poorer prognosis compared with early-stage disease.2 Patients with advanced disease require multidisciplinary management by a collaborative team comprising multiple physician specialties and disciplines.3 National Comprehensive Cancer Network (NCCN) guidelines, based on data from randomized controlled trials, recommend multimodality therapy (either surgery followed by adjuvant therapy or primary chemoradiation) for all advanced HNSCC (either T3, T4, N2, or N3), except T3N0 laryngeal cancers, which can be treated with unimodality therapy or multimodality therapy.4-7 Cisplatin-based chemotherapy is recommended for patients with advanced or high-risk HNSCC.8-10 Given the complex treatment and coordination required for patients with advanced HNSCC, suboptimal care and outcomes may be more likely to occur in these patients compared with those who require less complex care. In addition to their complexity, treatment modalities for advanced HNSCC have significant toxicities,11 which pose an additional barrier for receiving NCCN guideline therapy.

Prior studies in diseases other than HNSCC have shown that hospital volume12 and physician volume13 influence outcomes. Treatment at a high-volume facility is associated with improved long-term survival in patients with early- or late-stage laryngeal cancer.14, 15 However, no studies have examined whether hospital factors are associated with receiving the following key aspects of NCCN guideline therapy: 1) multimodality therapy for patients with advanced HNSCC; and 2) cisplatin-based chemotherapy when indicated.

The objective of this study was to determine whether hospital volume and other hospital characteristics influence treatment type and mortality in patients with advanced HNSCC. We hypothesized that high hospital volume would be associated with receiving NCCN guideline therapy and improved survival in patients with advanced HNSCC.

MATERIALS AND METHODS

Data Sources

This study was based on the Surveillance, Epidemiology, and End Results (SEER)-Medicare database, which was generated and is maintained by the National Cancer Institute's Health Services and Economic branch (http://healthservices.cancer.gov/seermedicare/).16 The SEER data consist of individual cancer diagnoses from multiple US population-based cancer registries.17 Each SEER cancer diagnosis is linked to Medicare claims data, yielding the SEER-Medicare database. This use of the SEER-Medicare database was approved by the University of Washington Institutional Review Board.

Patient Sample

Patients with head and neck cancer diagnosed in 2003 through 2007 were identified in the SEER database. Patients with prior diagnosis of cancer were excluded to ensure that the treatment being captured in Medicare claims data was targeted at the patient's advanced HNSCC, not a different malignancy. Codes from the third edition of the World Health Organization International Classification of Disease for Oncology (ICD-O-3) were used to limit the sample to patients with invasive HNSCC of the oral cavity, oropharynx, hypopharynx, or larynx (histology codes 8052, 8070, 8071, 8072, 8073, 8074, 8075, 8076, 8078, 8083, and 8084; site codes C01.9-C02.4, C02.8-C03.1, C03.9-C04.1, C04.8-C05.2, C05.8-C06.2, C06.8, C06.9, C09.0, C09.1, C09.8-C10.3, C10.8, C10.9, C12.9-C13.2, C13.8, C13.9, C32.0-32.3, C32.8, and C32.9). Patients younger than 66 years were excluded to ensure that patients in the study would have Medicare claims data. Only patients enrolled exclusively in Medicare Parts A and B for 1 year prior to diagnosis (to be able to determine comorbidities) and 4 months after diagnosis (to be able to determine treatment modalities) were included. Patients without staging data available were excluded. Patients with distant metastatic disease were also excluded, because the goal of the study was to study patients with advanced HNSCC who should be receiving multimodality therapy with curative intent. Only patients with advanced stage were included, with advanced stage defined as T3, T4, N2, or N3 (except T3N0 laryngeal squamous cell carcinoma) stage disease.

Patient and Tumor Characteristics

The Patient Entitlement and Diagnosis Summary File (PEDSF) of the SEER-Medicare database includes tumor site and stage data. From 2004 onward, the PEDSF contains American Joint Committee on Cancer staging data. For patients diagnosed prior to 2004, Extent of Disease variables contained in PEDSF were available and used to determine American Joint Committee on Cancer stage. Demographic characteristics, such as age, sex, and race, are reliably recorded within the PEDSF.18 In addition, education and income data for each patient's census tract and ZIP code from the 2000 census are incorporated into the PEDSF as indicators of socioeconomic status (SES).18 Medicare inpatient and physician claims were used to construct a Charlson-based comorbidity score by reviewing claims 12 calendar months prior to the SEER diagnosis date, as described.19-21

Determination of Cancer Treatment Modality

Treatment was determined using physician, inpatient, and outpatient claims. Among patients treated with chemotherapy, those who received cisplatin were also identified through use of agent-specific billing codes. We included only claims for treatment administered within 4 months of diagnosis to ensure we captured the initial treatment regimen as opposed to treatment for progressive disease. To determine whether treatment was administered within this interval, we used the hospitalization date from inpatient claims because it is more reliably recorded than the procedure date,12 and the service date from physician and outpatient claims.

Determination of Hospital Volume and Characteristics

The key objective of the study was to study the relationship of hospital volume to treatment modality and survival. We defined hospital volume on all invasive HNSCC patients aged 66 years or older diagnosed from 2003 through 2007 without prior cancers who exclusively had Medicare Parts A and B for 4 months after diagnosis (“Volume Cohort”). We required patients to have 4 months of Medicare coverage after diagnosis to ensure that we could determine the entire initial course of treatment they received and the treating hospital. When patients were treated at more than 1 hospital, the first hospital at which they received treatment was used. Hospitals located outside of SEER registry catchment areas were excluded to avoid underestimating their volume, because the SEER-Medicare database would not include all patients seen at these hospitals. Hospital volume was computed as the number of patients with HNSCC treated by each hospital and classified as a dichotomous variable (ie, low and high) using the median value.

Patient Survival

Date of death, obtained from the Social Security Administration, is included in Medicare data and was used to determine patient survival and time from diagnosis until death. The end of follow-up for analytic purposes was December 2009.

Data Analysis

Treatment modalities were compared between groups of patients. Because patients are “clustered” to the hospitals at which they receive treatment, it is important to correct for the effect of clustering to avoid overestimating the effect of hospital characteristics.22 For that reason, multivariable models in which more than 2 treatment modalities were considered as outcomes were fit through use of clustered multinomial logistic regression. Multivariable analyses were performed using generalized estimating equations when examining whether patients received any multimodality treatment or did not and whether patients received cisplatin or did not.23 Kaplan-Meier survival curves were estimated, and the log-rank test was used to compare survival between groups. The shared frailties method was used within the Cox proportional hazards model to estimate univariate and multivariable associations between hospital characteristics and survival.24 All confidence intervals were calculated using 2-sided α = .05 and the threshold for statistical significance was set at P < .05. All analyses were conducted using STATA version 11.1 Intercooled (StataCorp, College Station, Tex).

RESULTS

There were a total of 53,155 patients identified with head and neck cancer diagnosed from 2003-2007 in SEER, of which 1195 met all study criteria. There were 5740 patients who met criteria to be included in the Volume Cohort, which was used to determine hospital volume (Fig. 1). Characteristics of the 1195 patients and their tumors meeting criteria for the study are shown in Table 1.

Figure 1.

Algorithm shows definition of study cohort. Asterisk (*) indicates HNSCC was defined as squamous cell carcinoma of the oral cavity, oropharynx, hypopharynx, or larynx. Double asterisk (**) indicates advanced stage was defined as either T3, T4, N2, or N3 (except T3N0 laryngeal squamous cell carcinoma). Abbreviations: HMO, health maintenance organization; HNSCC, head and neck squamous cell carcinoma; SEER, Surveillance, Epidemiology, and End Results.

Table 1. Selected Characteristics of Patients With Advanced HNSCC by Hospital Volume, SEER-Medicare Database, 2003-2007
CharacteristicAll Advanced HNSCC (n = 1195)Treated at Low-Volume Hospitals (n = 315)Treated at High-Volume Hospitals (n = 490)
  • a

    Percent of the population living below the poverty level in the patient's census tract.

  • b

    The exact number of patients was suppressed in cells in which there were fewer than 11 patients to ensure patient confidentiality and comply with the SEER-Medicare Data Use Agreement.

  • Abbreviations: HNSCC, head and neck squamous cell carcinoma; IQR, interquartile range; SEER, Surveillance, Epidemiology, and End Results; SES, socioeconomic status.

Age, median (IQR)73 (69-79)74 (69-80)73 (69-78)
Sex: male816 (68%)219 (69%)334 (68%)
Race: white1080 (90%)287 (91%)436 (89%)
Ethnicity: non-Hispanic1147 (96%)299 (95%)476 (97%)
SES; % poverty, median (IQR)a7.2 (4.1-13.4)7.6 (4.5-14.2)6.6 (3.7-11.7)
Marital statusb   
 Single469 (39%)
 Married692 (58%)
 Unknown34 (3%)
Charlson comorbidity index   
 0803 (67%)203 (64%)321 (66%)
 1241 (20%)72 (23%)106 (22%)
 ≥2151 (13%)40 (13%)63 (13%)
Year of diagnosis   
 2003197 (16%)53 (17%)70 (14%)
 2004245 (21%)52 (17%)95 (19%)
 2005244 (20%)79 (25%)95 (19%)
 2006256 (21%)65 (21%)111 (23%)
 2007253 (21%)66 (21%)119 (24%)
Tumor Site   
 Oral cavity386 (32%)80 (25%)176 (36%)
 Oropharynx442 (37%)130 (41%)166 (34%)
 Larynx251 (21%)79 (25%)96 (20%)
 Hypopharynx116 (10%)26 (8%)52 (11%)
Tumor T stage   
 1145 (12%)45 (14%)58 (12%)
 2270 (23%)72 (23%)114 (23%)
 3112 (9%)26 (8%)48 (10%)
 4668 (56%)172 (55%)270 (55%)
Tumor N stage   
 0327 (27%)90 (29%)122 (25%)
 1120 (10%)34 (11%)43 (9%)
 2689 (58%)172 (55%)300 (61%)
 359 (5%)19 (6%)25 (5%)
Months of follow-up, median (IQR)26 (11-45)26 (12-46)28 (12-46)

Of the 1195 patients in the study, 1035 had a billing code corresponding to treatment, indicating that the other 160 patients did not receive treatment within 4 months of diagnosis. Of the patients receiving treatment, the treating hospital could be determined in 908 patients. Hospital volume could be ascertained for 805 patients; the treating hospitals of the remaining patients were not located within SEER catchment areas. Median hospital volume was 15 patients with invasive HNSCC who met criteria for the Volume Cohort; hospitals treating 15 or more patients from the Volume Cohort were classified as high-volume institutions. Patients treated at high-volume hospitals had higher SES, as measured by poverty level in the patient's census tract, were more likely to have oral cavity cancers and were less likely to have oropharyngeal cancers, but did not differ with respect to any of the other patient or tumor characteristics (Table 1).

Treatment modalities between patients treated at low- and high-volume hospitals were compared (Table 2). Patients treated at high-volume hospitals were more often treated with surgery (56% versus 46%), less often with cisplatin (18% versus 23%), and more often with surgery and radiation only (27% versus 20%). A single multivariable model with specific treatment modalities was generated, showing no statistically significant differences in treatment type by hospital volume (Table 3). A separate multivariable model showed that patients treated at high-volume hospital were less likely to receive cisplatin (odds ratio [OR] = 0.63, 95% confidence interval [CI] = 0.41-0.98). Patients were more likely to receive cisplatin if they were younger than 75 years (OR = 3.68, 95% CI = 2.33-5.82) and had oropharyngeal (OR = 2.24, 95% CI = 1.26-3.98) or hypopharyngeal (OR = 3.35, 95% CI = 1.63-6.90) tumors. More recent year of diagnosis was associated with receiving cisplatin (2003: baseline; 2004: OR = 2.15, 95% CI = 1.02-4.54; 2005: OR = 2.09, 95% CI = 1.02-4.28; 2006: OR = 2.44, 95% CI = 1.19-5.01; 2007: OR = 3.23, 95% CI = 1.59-6.56).

Table 2. Treatment Modalities of Patients With Advanced HNSCC by Hospital Volume, SEER-Medicare Database, 2003-2007
TreatmentAll Advanced HNSCC (n = 1195)Treated at Low-Volume Hospital (n = 315)Treated at High-Volume Hospital (n = 490)
  • a

    The exact number of patients was suppressed in cells in which there were fewer than 11 patients to ensure patient confidentiality and comply with the SEER-Medicare Data Use Agreement.

  • Abbreviations: HNSCC, head and neck squamous cell carcinoma; SEER, Surveillance, Epidemiology, and End Results.

Surgery511 (43%)146 (46%)277 (56%)
Surgery only110 (9%)28 (9%)58 (12%)
Radiation905 (75%)285 (90%)427 (87%)
Radiation only123 (10%)41 (13%)44 (9%)
Chemotherapy574 (48%)184 (58%)257 (52%)
Chemotherapy onlya15 (1%)
Cisplatin207 (17%)74 (23%)86 (18%)
Cisplatin onlya<11
Multimodality therapy787 (66%)245 (78%)385 (79%)
 Surgery, radiation, and chemotherapy168 (14%)55 (17%)86 (18%)
 Surgery, radiation, and cisplatin72 (6%)26 (8%)35 (7%)
 Surgery and radiation only228 (19%)62 (20%)131 (27%)
 Surgery and chemotherapy onlya<11
 Surgery and cisplatin onlya<11
 Radiation and chemotherapy only386 (32%)127 (40%)166 (34%)
 Radiation and cisplatin onlya131 (11%)
No treatment160 (13%)
Table 3. Multivariable Model of Hospital Volume and Other Characteristics Associated With Receipt of Different Treatments in Patients With Advanced HNSCC, SEER-Medicare Database, 2003-2007
 High Hospital Volume
TreatmentaOR (95% CI)P
  • a

    The association between hospital volume and each treatment was assessed in a single multivariable model, which included hospital volume, National Cancer Institute–designated cancer center status, age, sex, race, socioeconomic status, marital status, comorbidity as measured by Charlson comorbidity index, year of diagnosis, tumor site, tumor T stage, and tumor N stage as covariates.

  • Abbreviations: CI, confidence interval; HNSCC, head and neck squamous cell carcinoma; OR, odds ratio; SEER, Surveillance, Epidemiology, and End Results.

Surgery only (reference)1.00
Radiation only0.52 (0.24-1.09).082
Surgery and radiation only0.93 (0.57-1.98).797
Radiation and chemotherapy only0.59 (0.30-1.18).135
Surgery, radiation, and chemotherapy0.55 (0.25-1.17).122

A multivariable model of hospital, patient, and tumor characteristics in relation to use of multimodality therapy as the outcome was fit (Table 4). Hospital characteristics were not associated with receiving multimodality therapy. Patients who were younger than 75 years, had oropharyngeal or laryngeal tumors, and had high N stage were more likely to receive multimodality therapy. Year of diagnosis was also associated with multimodality therapy, with increasing ORs over time.

Table 4. Multivariable Model of Hospital, Patient, and Tumor Characteristics in Relationship to Receipt of Multimodality Therapy in Patients With Advanced HNSCC, SEER-Medicare Database, 2003-2007
CharacteristicOR (95% CI)P
  • a

    Percent of the population living below the poverty level in the patient's census tract.

  • Abbreviations: CI, confidence interval; HNSCC, head and neck squamous cell carcinoma; NCI, National Cancer Institute; OR, odds ratio; SEER, Surveillance, Epidemiology, and End Results; SES, socioeconomic status.

High hospital volume1.02 (0.66-1.60)0.919
NCI-designated cancer center1.23 (0.57-2.65)0.605
Age < 75 y2.53 (1.71-3.72)<.001
Sex: female1.20 (0.79-1.84)0.391
Race: nonwhite1.70 (0.79-3.63)0.172
SES; poverty ≥ 7.2%a1.34 (0.92-1.96)0.129
Marital status: married1.15 (0.77-1.70)0.49
Charlson comorbidity index  
 0 (reference)1
 11.23 (0.77-1.97)0.377
 ≥20.69 (0.41-1.17)0.17
Year of diagnosis  
 2003 (reference)1
 20041.58 (0.87-2.89)0.134
 20051.61 (0.91-2.84)0.103
 20061.98 (1.10-3.55)0.022
 20072.29 (1.26-4.18)0.007
Tumor site  
 Oral cavity (reference)1
 Oropharynx1.76 (1.08-2.88)0.024
 Larynx2.39 (1.37-4.18)0.002
 Hypopharynx1.34 (0.69-2.60)0.382
Tumor T stage  
 1 (reference)1
 21.51 (0.78-2.91)0.221
 31.66 (0.72-3.85)0.238
 41.66 (0.83-3.33)0.154
Tumor N stage  
 0 (reference)1
 10.65 (0.36-1.18)0.157
 21.93 (1.09-3.42)0.024
 32.10 (0.72-6.13).176

Patient survival was compared on the basis of hospital characteristics, with slightly better survival at higher volume hospitals and hospitals designated as National Cancer Institute (NCI) cancer centers (Fig. 2). The multivariable survival model in Table 5 shows that these characteristics were nearly statistically significant (high hospital volume: hazard ratio [HR] 0.85, 95% CI = 0.69-1.04; NCI-designated cancer center: HR = 0.88, 95% CI = 0.64-1.21). Patients who had higher Charlson comorbidity index, higher T stage tumors, and higher N stage tumors had worse survival; patients who were younger than 75 years and had oropharyngeal tumors had better survival. In order to explore potential factors that may account for the association between high hospital volume and survival, we fit a separate multivariable model that included treatment modalities. This model showed that patients whose treatment included surgery or cisplatin had better survival (surgery: HR = 0.75, 95% CI = 0.56-1.01; cisplatin: HR = 0.76, 95% CI = 0.57-1.02). There were no associations between radiation or chemotherapy and survival (radiation: HR = 1.01, 95% CI = 0.56-1.65; chemotherapy: HR = 1.06, 95% CI = 0.75-1.49). When surgery alone was compared with all other treatments in a multivariable model, it was not associated with survival (HR = 1.03, 95% CI = 0.74-1.43).

Figure 2.

Kaplan-Meier survival curve is shown for survival by hospital characteristics in patients with advanced head and neck squamous cell carcinoma, SEER (Surveillance, Epidemiology, and End Results)-Medicare Database, 2003-2007.

Table 5. Multivariable Survival Model of Hospital, Patient, and Tumor Characteristics in Patients With Advanced HNSCC, SEER-Medicare Database, 2003-2007
CharacteristicHR (95% CI)P
  • a

    Percent of the population living below the poverty level in the patient's census tract.

  • Abbreviations: CI, confidence interval; HNSCC, head and neck squamous cell carcinoma; HR, hazard ratio; NCI, National Cancer Institute; SEER, Surveillance, Epidemiology, and End Results; SES, socioeconomic status.

High hospital volume0.85 (0.69-1.04)0.123
NCI-designated cancer center0.88 (0.64-1.21)0.443
Age < 75 y0.65 (0.53-0.79)<.001
Sex: female1.05 (0.84-1.31)0.681
Race: nonwhite0.82 (0.58-1.14)0.241
SES; poverty ≥ 7.2%a0.94 (0.77-1.15)0.542
Marital status: married0.89 (0.72-1.09)0.243
Charlson comorbidity index  
 0 (reference)1
 11.44 (1.15-1.81)0.002
 ≥21.69 (1.27-2.25)<.001
Year of diagnosis  
 2003 (reference)1
 20040.90 (0.65-1.24)0.529
 20051.06 (0.78-1.44)0.699
 20060.99 (0.72-1.36)0.971
 20070.94 (0.68-1.30)0.723
Tumor site  
 Oral cavity (reference)1
 Oropharynx0.53 (0.41-0.70)<.001
 Larynx0.78 (0.60-1.02)0.073
 Hypopharynx0.84 (0.60-1.18)0.314
Tumor T stage  
 1 (reference)1
 21.41 (0.97-2.03)0.071
 32.24 (1.47-3.41)<.001
 41.86 (1.28-2.70)0.001
Tumor N stage  
 0 (reference)1
 11.24 (0.86-1.81)0.25
 21.87 (1.41-2.47)<.001
 32.48 (1.57-3.92)<.001

DISCUSSION

Survival among patients with HNSCC is influenced by patient and tumor characteristics, such as performance status, comorbidity, tumor size, and presence of regional lymph node metastasis.25, 26 The objective of this study was to examine the role of hospital characteristics in survival. We were particularly interested in patients with advanced HNSCC, given the complexity of management when compared with patients who had early HNSCC. Because multimodality therapy is recommended for patients with advanced HNSCC,4 we used it as an outcome of interest, in addition to survival. We found that patients treated at high-volume hospitals were not more likely to receive multimodality therapy, but were more likely to have improved survival, when compared with patients treated at low-volume hospitals.

Prior studies, many using SEER-Medicare data, have shown relationships between surgical volume and survival for many major cancer-directed surgical procedures.12, 27 However, studies examining the association of hospital volume and other hospital characteristics with treatment and survival in patients with HNSCC have been limited. Prior work by Chen et al using the National Cancer Data Base has shown that early- and late-stage laryngeal cancer patients treated at high-volume hospitals have improved survival.14, 15 In addition, Gourin et al used the Maryland Health Service Cost Review Commission database to show that an increasing proportion of patients with laryngeal cancer received surgical treatment at high-volume hospitals between 1990 and 2009. Patients treated at high-volume hospitals had shorter duration of hospitalization and lower hospital-related costs of care.28, 29 However, from these studies, the role of confounding by patient comorbidities, effect of hospital characteristics on delivery of NCCN guideline therapy, and greater generalizability to the advanced HNSCC patient population were still unknown. The current study examined the relationships between hospital characteristics and delivery of therapy as well as survival in patients with advanced HNSCC. Because multiple patient and tumor characteristics, such as patient comorbidities, have been shown to be associated with survival, these were included in our multivariable models. We accounted for clustered data in our analyses to avoid overestimating the effect of hospital volume.22 We found a similar relationship between hospital volume and survival as the studies by Chen et al, but in a sample of patients with advanced HNSCC.

Contrary to one of our hypotheses, we found that the volume of HNSCC treated at a hospital was not associated with receiving NCCN guideline therapy (ie, multimodality therapy and cisplatin-based chemotherapy). However, we found that the use of multimodality therapy and cisplatin-based chemotherapy increased from 2003 to 2007; this finding may be related to studies published in 2004 that showed improvements in locoregional control and disease-free survival with the addition of cisplatin in patients whose cancers had adverse features.8, 9 The specific modality of treatment patients received varied on the basis of hospital volume, although these differences were not statistically significant in a multivariable model. These differences may represent different practice patterns and treatment preferences at high-volume hospitals in comparison with low-volume hospitals. Alternatively, this finding could be an artifact of referral patterns, with patients who were more amenable to surgical therapy being referred to high-volume hospitals for surgery. The latter explanation is supported by our finding that high-volume hospitals were more likely to be treating patients with oral cavity cancers and less likely to be treating patients with oropharyngeal cancers than low-volume hospitals.

In addition to hospital volume, we examined the role of NCI-designated cancer center status. Patients treated at these centers were more likely to receive multimodality therapy compared to other patients, but this result did not achieve statistical significance. In addition, the multivariable survival analysis indicated slightly favorable survival at NCI-designated cancer centers.

Although the primary objective of this study was to examine the effect on hospital characteristics on receiving NCCN guideline therapy and survival, we found in multivariable survival analysis that patients whose treatment regimens included either surgery or cisplatin had better survival. The association between surgery and survival did not remain when surgery was the only treatment, indicating a benefit to the use of surgery with additional modalities. Another retrospective database study by Chen et al14 also found an association between surgery and survival in patients with advanced laryngeal cancer. However, database analyses cannot account for certain important confounding factors, such as whether tumors are surgically resectable and whether patients can tolerate surgery or cisplatin-based chemotherapy. For that reason, further studies could examine the choice of treatment modalities in patients with advanced HNSCC in a prospective setting.

In considering the interpretation of this study, it is important to remember the limitations inherent to studies using cancer registries and claims databases. Generalizability of results is a concern because study patients were limited to those individuals meeting criteria for the study. Therefore, our findings may only be valid in the elderly Medicare population rather than the larger population of Americans with advanced HNSCC. The data we used did not contain any information about patient preferences or choices; these could easily influence whether patients receive treatment and their survival. However, we included other patient characteristics (including SES status) in our multivariable models. Finally, we used the first hospital at which a patient was treated to determine whether its characteristics influence treatment and survival. Moreover, when calculating hospital volume, we only counted patients as being treated by a certain hospital if that was the first hospital at which they were treated. Although this strategy (which was chosen a priori) is valid when patients receive their entire treatment regimen at a single hospital, results could have differed when patients are treated at multiple hospitals, as was the case for some patients in the study. In sensitivity analysis, we found that our results did not differ when all hospitals at which patients were treated were used to determine volume and to perform the multivariable analyses (data not shown).

This study is the first to show that hospital volume influences outcomes in Medicare patients with advanced HNSCC. We found that patients treated at high-volume hospitals have improved survival; this difference was nearly statistically significant. Despite the improved survival at high-volume hospitals, the proportion of patients receiving multimodality therapy, which is recommended for all patients with advanced HNSCC and has been shown to influence survival, was similar between low- and high-volume hospitals. Although this study does not necessarily mean that all patients with advanced HNSCC should be treated at high-volume hospitals or at NCI-designated cancer centers, it does suggest that features of these hospitals, such as a multidisciplinary team approach or other institutional factors, play a critical role in influencing survival without influencing whether patients receive NCCN guideline therapy. Studying and implementing these institutional factors may be effective in narrowing the disparity in survival across hospitals that was found in this study.

FUNDING SOURCES

This work was funded by National Institute on Deafness and Other Communication Disorders Grant 2T32DC000018; National Cancer Institute Contract N01-PC-35142; and National Center for Research Resources Grant 5KL2RR025015.

CONFLICT OF INTEREST DISCLOSURES

Dr. Méndez has received speaking fees/honoraria from Intuitive Surgical, Inc. All other authors made no disclosures.

Ancillary