Association of Hospital Volume With Laryngectomy Outcomes in Patients With Larynx Cancer

Key Points

Question  What is the association between hospital volume and outcomes for laryngectomy surgery?

Findings  Among 45 156 patients at 5516 hospitals in this cross-sectional study, a minimum hospital volume threshold of more than 6 cases per year was associated with reduced odds of postoperative complications, with a greater reduction in the odds of complications with increasing volume. High-volume hospital care (>28 cases per year) was also associated with lower odds of in-hospital mortality, the mean incremental length of hospitalization, and costs.

Meaning  Laryngectomy outcomes appear to be associated with hospital volume, with reduced morbidity associated with a minimum hospital volume threshold and with reduced mortality, morbidity, length of hospitalization, and costs associated with higher hospital volume.

Importance  A volume-outcome association exists for larynx cancer surgery, but to date it has not been investigated for specific surgical procedures.

Objectives  To characterize the volume-outcome association specifically for laryngectomy surgery and to identify a minimum hospital volume threshold associated with improved outcomes.

Design, Setting, and Participants  In this cross-sectional study, the Nationwide Inpatient Sample was used to identify 45 156 patients who underwent laryngectomy procedures for a malignant laryngeal or hypopharyngeal neoplasm between January 2001 and December 2011. The analysis was performed in 2018. Hospital laryngectomy volume was modeled as a categorical variable.

Main Outcomes and Measures  Associations between hospital volume and in-hospital mortality, complications, length of hospitalization, and costs were examined using multivariate logistic regression analysis.

Results  Among 45 156 patients (mean age, 62.6 years; age range, 20-96 years; 80.2% male) at 5516 hospitals, higher-volume hospitals were more likely to be teaching hospitals in urban locations; were more likely to treat patients who had hypopharyngeal cancer, were of white race/ethnicity, were admitted electively, had no comorbidity, and had private insurance; and were more likely to perform flap reconstruction or concurrent neck dissection. After controlling for all other variables, hospitals treating more than 6 cases per year were associated with lower odds of surgical and medical complications, with a greater reduction in the odds of complications with increasing hospital volume. High-volume hospitals in the top-volume quintile (>28 cases per year) were associated with decreased odds of in-hospital mortality (odds ratio, 0.45; 95% CI, 0.23-0.88), postoperative surgical complications (odds ratio, 0.63; 95% CI, 0.50-0.79), and acute medical complications (odds ratio, 0.63; 95% CI, 0.48-0.81). A statistically meaningful negative association was observed between very high-volume hospital care and the mean incremental length of hospitalization (−3.7 days; 95% CI, −4.9 to −2.4 days) and hospital-related costs (−$4777; 95% CI, −$9463 to −$900).

Conclusions and Relevance  Laryngectomy outcomes appear to be associated with hospital volume, with reduced morbidity associated with a minimum hospital volume threshold and with reduced mortality, morbidity, length of hospitalization, and costs associated with higher hospital volume. These data support the concept of centralization of complex care at centers able to meet minimum volume thresholds to improve patient outcomes.

A favorable association exists between hospital volume and patient outcomes for complex surgical procedures. Hospitals performing high volumes of bariatric surgery, cardiovascular surgery, lung cancer surgery, and gastrointestinal cancer surgery have been shown to have better survival outcomes with fewer surgical complications, with similar observations made for high-volume surgeon care.^1-7 Building on this research around the volume-outcome association, the Leapfrog Group,⁸ a coalition of large Fortune 500 employers and other health care purchasers, adopted a minimum surgical volume standard for selected high-risk surgical procedures. In 2015, in response to a US News & World Report article highlighting the volume-outcome association for surgery,⁹ leaders at the Johns Hopkins Hospital and Health System, Dartmouth-Hitchcock Medical Center, and the University of Michigan Health System publicly pledged to limit certain complex elective procedures to surgeons and hospitals meeting volume criteria.¹⁰ The “volume pledge” requires surgeons and hospitals to meet modest minimum volume standards for 10 procedures that have the strongest link between hospital volume and patient mortality.¹¹ This approach will effectively result in regionalization of high-risk surgical care to high-volume centers within a health care system, and the Johns Hopkins Health System makes these data publicly available.¹²

A volume-outcome association has been demonstrated for head and neck cancer surgical care, with improved outcomes reported for high-volume hospitals.^13-27 These reports often include a variety of head and neck primary sites and are limited by low perioperative mortality rates and the inclusion of a range of both low-risk and high-risk surgical procedures, and minimum volume standards for specific head and neck cancer procedures have not been established to date. Procedures that are most sensitive to a volume-outcome association are high-risk, low-volume procedures with outcomes that show wide variation across hospitals and for which hospital volume can serve as a proxy measure for outcomes.²⁸ This association may be most pronounced for laryngectomy surgery, which has decreased in recent years primarily because of a decreased incidence of larynx cancer concomitant with an increased use of primary nonoperative treatment, with a growing proportion of laryngectomy procedures performed after prior radiotherapy associated with an increase in postoperative complications.^22,29-34 Flap reconstruction is increasingly used in this setting to improve postoperative outcomes, which further increases the complexity of surgery. Because of this trend, centralization of laryngectomy procedures appears to be occurring at high-volume and academic centers,^22,29,31 yet laryngectomy was removed as a key indicator procedure representative of resident training by the American Board of Otolaryngology–Head and Neck Surgery in 2011 because of decreasing numbers even at such institutions.³⁵

We hypothesized that laryngectomy—as a high-risk, increasingly low-volume surgery—would have a strong link between hospital volume and patient morbidity and mortality. We sought to characterize the volume-outcome association for laryngectomy surgery by examining associations between volume and in-hospital mortality, complications, length of hospitalization, and costs and aimed to identify a possible minimum hospital volume threshold for high-volume care.

A cross-sectional study of patients with a diagnosis of larynx cancer was performed using discharge data from the Nationwide Inpatient Sample (NIS),³⁶ the largest all-payer inpatient care database in the United States, containing data from approximately 8 million hospital stays each year from a stratified sample of 20% of nonfederal US hospitals from participating states. The NIS database provides information regarding the index hospital admission and includes patient demographic data, primary and secondary diagnoses, primary and secondary procedures, hospital characteristics, and inpatient and discharge mortality rates. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes were used to identify adult patients (≥18 years) who specifically underwent laryngectomy for a malignant laryngeal or hypopharyngeal neoplasm between January 2001 and December 2011 (eTable 1 in the Supplement). The analysis was performed in 2018. Additional procedures, such as neck dissection or pedicled or free flap reconstruction, were recorded. Prior radiotherapy was derived from codes for previous exposure to therapeutic or other ionizing radiation (ICD-9-CM code V15.3). This protocol was reviewed and approved by the Johns Hopkins Medical Institutions Institutional Review Board. Informed consent was inapplicable because the data are from a national deidentified database.

The primary clinical end points (dependent variables) were in-hospital mortality, complications, length of hospitalization, and costs. Codes for specific comorbid illnesses were used to create categories for surgical and acute medical complications. Acute medical complications were derived from codes for acute cardiac events, acute pulmonary edema or failure, acute renal failure, acute hepatic failure, acute cerebrovascular events, sepsis, pneumonia, gastrointestinal bleeding, deep vein thrombosis/pulmonary embolism, and urinary tract infection assigned at the time of hospital discharge (eTable 2 in the Supplement). Surgical complications were derived from codes for complications directly resulting from surgical procedures assigned at the time of hospital discharge. Hospital-related charges for each index admission were converted to the organizational cost of providing care using cost-to-charge ratios for individual hospitals. Cost-to-charge ratios were calculated using information from the detailed reports by hospitals to the Centers for Medicare & Medicaid Services, providing an estimate of the all-payer inpatient cost-to-charge ratio by hospital, and were multiplied by each patient’s charge to obtain the cost per admission.³⁷ All costs were adjusted for inflation based on US Bureau of Labor Statistics indexes,³⁸ with the results converted to 2018 US dollars. To obtain national cost estimates, all discharges were reweighted to account for cases in which cost estimates were missing.

Hospital volume was examined as an independent variable. The average annual number of laryngectomy cases performed per year of surgical activity was obtained by calculating the mean of the number of cases performed each year for each individual hospital for the years in which that hospital performed at least 1 laryngectomy. Hospital laryngectomy volume was modeled as a categorical variable. Annual volumes were divided into quintiles, with high volume defined as hospitals above the 80th percentile. Based on the quintile distribution of the annual number of cases per hospital, hospitals were categorized as very low volume (≤3 cases per year), low volume (4-6 cases per year), medium volume (7-15 cases per year), high volume (16-28 cases per year), or very high volume (>28 cases per year). Secondary independent variables included were the following: age, race/ethnicity, sex, nature of admission (elective or urgent/emergent), comorbidity, payer (commercial, private/health maintenance organization, Medicare, Medicaid, self-pay, or other), hospital teaching status, hospital ownership/control (not-for-profit or other), hospital location (rural or urban), geographic region, hospital bed size, prior radiotherapy, and procedure. The American Joint Commission on Cancer tumor stage, tumor grade, histological subtype, and outcome after discharge were not available from the NIS database. Comorbidity was graded using the Romano adaptation of the Charlson Comorbidity Index, excluding ICD-9-CM codes for the index cancer diagnosis from the solid tumor category, as previously described.²² Because cancer staging information is not available in the NIS, ICD-9-CM codes for metastases were excluded because these are not a reliable surrogate for disease stage.

Data were analyzed using statistical software (Stata, version 12; StataCorp LP). Associations between variables were analyzed using cross-tabulations, multivariate logistic regression analysis, and generalized linear regression modeling. Data were weighted, and modified hospital and discharge weights to correct for changes in sampling over time were applied. Variance estimation was performed using procedures for survey data analysis with replacement. Strata with 1 sampling unit were centered at the population mean. Variables with missing data for more than 10% of the population were coded with a dummy variable to represent the missing data in regression analysis. The primary clinical end points were evaluated using multiple logistic regression analysis. Generalized linear regression modeling with a log link was used to analyze length of hospitalization and hospital-related costs because these variables were not normally distributed.

There were 45 156 laryngectomy cases between 2001 and 2011 performed at 5516 hospitals (Table 1). The mean patient age was 62.6 years (age range, 20-96 years). The majority of patients were of white race/ethnicity and male, were admitted electively, had no associated comorbidity, and received care at teaching hospitals located in urban locations. A history of prior radiotherapy was documented in 7.6% of cases. Flap reconstruction was performed in 10.1% of patients, and concurrent neck dissection was performed in 15.1% of patients. Medical complications occurred in 23.6% of patients, and surgical complications occurred in 32.2% of patients. In-hospital mortality occurred in 1.0% of patients.

The distribution of the number of hospitals by hospital volume quintile is shown in Figure 1. Very high-volume hospitals comprised 1.7% of all hospitals and performed a mean of 42 cases per year, while very low-volume hospitals comprised 68.2% of all hospitals and performed a mean of 2 cases per year (Figure 1). Higher-volume hospitals were more likely to be not-for-profit teaching hospitals in urban locations; were more likely to treat patients who had hypopharyngeal cancer, were of white race/ethnicity, were admitted electively, had no comorbidity, and had private insurance; and were more likely to perform flap reconstruction or concurrent neck dissection. Rates of mortality and postoperative complications were lowest for hospitals in the very high-volume quintile (Figure 2).

Multiple logistic regression analysis of variables associated with in-hospital mortality, postoperative surgical complications, and acute medical complications, after controlling for all other variables, demonstrated that, compared with hospitals in the bottom quintile, hospitals in the top quintile (>28 cases per year) were associated with lower odds of in-hospital mortality (OR, 0.45; 95% CI, 0.23-0.88) and the lowest odds of postoperative surgical complications (OR, 0.63; 95% CI, 0.50-0.79) and acute medical complications (OR, 0.63; 95% CI, 0.48-0.81) (Table 2). Hospitals treating more than 6 cases per year (medium-volume and high-volume hospitals) also had lower odds of postoperative surgical and acute medical complications but were not associated with differences in in-hospital mortality.

Multivariate generalized linear regression analyses of independent variables associated with length of hospitalization and hospital-related costs are summarized in Table 3. After controlling for all other variables, a clinically meaningful negative association was observed between surgery at higher-volume hospitals (>15 cases per year) and length of hospitalization, with a decrease in the mean incremental length of hospitalization at hospitals in the top-volume quintile (−3.7 days; 95% CI, −4.9 to −2.4 days) that exceeded the increase in the mean incremental length of hospitalization associated with patient variables. Surgery at hospitals in the top-volume quintile only was associated with lower mean incremental hospital-related costs (−$4777; 95% CI, −$9463 to −$900).

These data demonstrate that there is a significant volume-outcome association for laryngectomy, with improved outcomes seen with increasing hospital volume. A threshold of greater than 6 cases per year was associated with a clinically meaningful lower odds of postoperative surgical complications, acute medical complications, and reduced length of hospitalization, with hospitals in the top-volume quintile that performed more than 28 cases per year also associated with decreased odds of in-hospital mortality and lower costs. These findings are important in an era of health care reform, when increasing attention is focused on health care value, defined as outcomes relative to costs. Because laryngectomy is increasingly becoming a low-volume procedure that is performed more often for salvage and with an increased risk of complications, the use of a minimum hospital volume threshold for laryngectomy may be an appropriate approach to improve value in head and neck cancer care.

To our knowledge, this is the first analysis of minimum hospital volume thresholds for a specific surgical procedure of interest in head and neck cancer. An association between hospital volume and patient outcomes for head and neck cancer surgery has been previously established but is not procedure specific. Studies^14,18,19 evaluating surgery for oral cavity, pharynx, larynx, hypopharynx, and salivary gland neoplasms have demonstrated an association between hospital volume and long-term survival, with high-volume hospital care defined as ranging from 60 to 75 cases per year and including a variety of procedures in volume calculations. Using similar cutoffs for volume, high-volume hospital care has been associated with lower in-hospital mortality and decreased failure-to-rescue rates after surgery for oral cavity, pharynx, hypopharynx, and larynx cancer.^26,27 Studies evaluating larynx cancer surgery specifically have shown improved long-term survival for high-volume hospital care, defined as 12 or more cases per year,^24,25 and lower length of hospitalization and decreased costs at a volume threshold of 18 cases per year,^20,22 but the studies combined a variety of surgical procedures exclusive of biopsy to define volumes. In Ontario, Canada, policy changes have been implemented based on such observations. Ontario has a universal single-payer health insurance system and, in 2014, implemented organizational standards for head and neck cancer treatment that include qualitative descriptors of a head and neck cancer center.³⁹ These standards link head and neck cancer center funding to minimum institutional volume cutoffs, effectively regionalizing head and neck cancer surgical care to high-volume hospitals to improve outcomes.

The use of hospital volume as a proxy for quality is controversial in the United States, with opponents arguing that it supports regionalization of care, limits access, and may increase travel burden and care fragmentation, particularly for patients in rural areas.²⁸ Furthermore, it is argued that volume does not accurately reflect the process measures associated with surgical outcomes, including patient selection, preoperative evaluation and preparation, adherence to evidence-based medical guidelines, and postoperative care and does not reflect individual hospital or surgeon performance.⁴⁰ However, sample size is the greatest limitation to direct process and outcome measurement for procedures that are performed infrequently at individual hospitals because of low case numbers in the denominator of process or outcome measures. For high-risk, low-volume procedures, a structural measure (eg, volume) is a proxy for outcomes if the association between structure and process or between structure and outcome is established and sample sizes are small.^28,41 For low-risk, high-volume head and neck cancer procedures, such as neck dissection, outcomes have been shown to be associated with process measures, such as lymph node yield, which neutralizes the volume-outcome association.^42,43 The use of volume-based standards requires specific volume-based thresholds for specific procedures to avoid generalizations to low-risk procedures that are more influenced by process or outcome measures than volume.

While process measures may underlie some of the observed association between laryngectomy volume and outcomes, volume may be a stronger predictor of outcomes for laryngectomy. Variables reflecting process measures of care, including patient selection, preoperative evaluation and preparation, adherence to stage-specific evidence-based treatment guidelines, and institutional culture, are not available in the NIS, preventing evaluation of what the effect of process is on outcomes and whether this effect modulates the observed volume-outcome association. Eskander et al⁴⁴ have reported that guideline adherence for head and neck cancer surgical care in Ontario was associated with hospital volume, but adherence rates were only moderate for high-volume hospitals, which suggests that other factors may underlie the observed favorable volume-outcome association they observed in their population.¹⁸ Our core group has previously reported a significant association between hospital volume and adherence to evidence-based guidelines for larynx cancer treatment in Surveillance, Epidemiology, and End Results Medicare data, which was associated with improved survival; however, after controlling for quality of care, a survival advantage remained for high-volume larynx cancer surgical care.²⁵ The findings of these studies suggest that, for larynx cancer surgery, hospital volume may have a greater influence on patient outcomes than evidence-based guideline process measures, which may reflect unmeasured perioperative processes that favorably alter outcomes.

Given the significant association between hospital volume and outcomes, a key policy issue for clinicians and policymakers to explore is regionalization of care in this era of health care reform. Regionalization may reduce operative risks, assure access for all patients with complex conditions to the highest-quality care, and provide an optimal educational environment for surgical trainees. These benefits have to be balanced against the desire of many patients to receive care in their home communities, as well as the cost and logistical challenges of having patients travel to regional centers for complex procedures, which is a particular challenge for patients in rural environments and for those with limited financial means. Federal and local leaders need to develop solutions that balance these different goals and meet the needs of their community. Given patient preferences for receiving care locally, when possible, and the strong economic and reputational incentives associated with the performance of surgery, we recognize that regionalizing care will not be easy and will require committed leadership. One possible starting point for regionalization of care might occur in metropolitan areas, where low-volume hospitals could seek to move their volume to geographically proximate medium- and high-volume hospitals, which may reduce preventable deaths and adverse outcomes while still maintaining patient access and satisfaction. This approach will require a cultural shift that may be most implementable within a health care organization in the current climate of increasing hospital consolidation.

There are several limitations to the use of hospital discharge data that may have influenced our findings. The NIS database provides no follow-up data beyond the index admission, is limited to a 30-day postoperative window, and contains no information on stage of disease, grade, subtype, or survival, precluding analysis of long-term outcomes. The NIS database does not contain information regarding readmission, previous surgical procedures, or prior chemotherapy, which could potentially have influenced the results with regard to the extent of surgery and outcomes. The ability to adequately control for case mix is limited if discharge diagnoses from administrative databases are used, which has been done in almost all studies investigating volume and outcome associations to date, and the process measures associated with preoperative care and surgical decision making, as well as long-term outcomes, cannot be assessed. Prior radiotherapy, which is known to influence outcomes, may have been undercoded. Postoperative complications may not have been apparent at the time of discharge; as a result, the incidence of complications may have been underreported. Cost analysis was based on hospital-related charges, adjusted for institutional expense-to-revenue ratios, and did not include physician-related costs because these data are not contained in the NIS database. However, the NIS is the largest database of inpatient surgical procedures that includes hospital-level and cost data and, unlike registry data, represents the spectrum of community, small, and large hospitals, providing a more accurate picture of hospital care across the United States.

Our findings demonstrate that there is a meaningful association between hospital volume and laryngectomy surgery outcomes, and we describe minimum hospital volume thresholds that are associated with in-hospital mortality, surgical and medical complications, length of hospitalization, and costs of care. Institutional leaders and policymakers may want to consider opportunities to shift laryngectomy procedures from lower-volume hospitals to higher-volume hospitals to improve outcomes and reduce costs.

Accepted for Publication: September 6, 2018.

Corresponding Author: Christine G. Gourin, MD, MPH, Department of Otolaryngology–Head and Neck Surgery, Johns Hopkins Medical Institutions, 601 N Caroline St, Ste 6260, Baltimore, MD 21287 (cgourin1@jhmi.edu).

Published Online: November 21, 2018. doi:10.1001/jamaoto.2018.2986

Author Contributions: Dr Gourin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Gourin, Stewart, Pitman, Austin.

Acquisition, analysis, or interpretation of data: Gourin, Frick, Fakhry, Eisele, Austin.

Drafting of the manuscript: Gourin.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Gourin.

Obtained funding: Gourin.

Administrative, technical, or material support: Gourin, Pitman, Eisele.

Supervision: Gourin, Fakhry, Austin.

Conflict of Interest Disclosures: None reported.