Comparison of Hospitals Affiliated With PPS-Exempt Cancer Centers, Other Hospitals Affiliated With NCI-Designated Cancer Centers, and Other Hospitals That Provide Cancer Care

Key Points

Question  Are there differences between hospitals affiliated with Medicare Prospective Payment System (PPS)-exempt cancer centers, hospitals affiliated with National Cancer Institute–designated cancer centers (NCI-CCs), and hospitals that provide cancer care in the United States?

Findings  Hospitals affiliated with PPS-exempt cancer centers and NCI-CCs, but not other hospitals that provide cancer care, had generally similar hospital characteristics, basic cancer-related services, patient comorbidity burden, and cancer surgery outcomes.

Meaning  Additional transparency is required regarding how PPS-exempt cancer centers are selected and maintained; moreover, rather than limiting public reporting of cancer quality metrics to PPS-exempt hospitals, public reporting of cancer quality metrics for all hospitals would be beneficial to the medical community and the public.

Importance  Congress has exempted 11 specialized cancer centers in the United States from the Prospective Payment System (PPS). These centers are also exempt from reporting many of the process-of-care and outcome measures to the Centers for Medicare & Medicaid Services that are required for hospitals in the PPS. It is not known how hospitals affiliated with PPS-exempt cancer centers differ from other hospitals affiliated with National Cancer Institute cancer centers (NCI-CCs) or other US hospitals that provide cancer care.

Objective  To examine differences between hospitals affiliated with PPS-exempt cancer centers, other hospitals affiliated with NCI-CCs, and other hospitals that provide cancer care on metrics that could be used in public reporting.

Design, Setting, and Participants  This retrospective cohort study compared hospital characteristics and cancer-related services using data from the American Hospital Association Annual Survey and US News Best Hospitals rankings. With a 100% sample of Medicare beneficiaries who underwent 1 of 9 cancer operations (brain tumor resection, colorectal resection, cystectomy, esophagectomy, gastrectomy, liver resection, lung resection, pancreatic resection, prostatectomy) from January 1, 2011, to May 31, 2015, we used hierarchical logistic regression methods to compare differences in 18 postoperative outcomes. Data analysis was conducted from February 2018 to August 2018.

Main Outcomes and Measures  This study evaluated hospital characteristics, including cancer-specific services, patient comorbidity burden, and cancer surgery postoperative outcomes, from PPS-exempt cancer centers, NCI-affiliated cancer centers, and other US hospitals that provide cancer care.

Results  Hospitals affiliated with PPS-exempt cancer centers (n = 15) and NCI-CCs (n = 54) were similar in hospital characteristics, basic cancer-related services, and patient comorbidity burden. Compared with NCI-CCs, PPS-exempt cancer centers had significantly higher US News reputation scores (mean [SD], 17.5 [24.0] vs 2.6 [4.8]; P < .001) but no differences in oncology patient volume, patient safety ratings, comorbidity burden, nurse staffing, US News total cancer scores, or US News survival scores. Hospitals affiliated with PPS-exempt cancer centers and NCI-CCs had similar adjusted postoperative outcomes for 15 of 18 measures, including mortality, readmission, and surgical site infections. Compared with hospitals affiliated with PPS-exempt cancer centers, patients treated at NCI-CCs were more likely to have postoperative sepsis (3.1% vs 1.7%; P = .002), acute renal failure (6.2% vs 3.9%; P = .01), and urinary tract infection (6.4% vs 4.0%; P = .002). Compared with the other hospitals that provide cancer care (n = 3578), PPS-exempt cancer center status was associated with improved outcomes for 7 of 18 measures, including mortality, sepsis, acute renal failure, pulmonary failure, and failure to rescue.

Conclusions and Relevance  Hospitals affiliated with PPS-exempt cancer centers and NCI-CCs had generally similar hospital characteristics, patient comorbidity burden, and cancer surgery outcomes. These findings raise questions about why some cancer centers are designated as PPS-exempt and why most hospitals are not required to publicly report cancer-specific quality metrics.

The Social Security amendments of 1983 led to the exemption of 11 specialized cancer centers in the United States from the Medicare Prospective Payment System (PPS).^1-9 The exemptions became effective between 1984 and 1999. Instead of being paid according to the diagnosis-related group (DRG) methodology, these centers are reimbursed on a reasonable cost basis.^3,7,9,10 The difference in payment mechanism notwithstanding, these centers are exempt from reporting all of the process-of-care, outcome, and patient experience measures to the Centers for Medicare & Medicaid Services (CMS) through Hospital Compare or any of the other CMS pay-for-performance programs. In contrast, hospitals in the United States that are part of the PPS report these measures. The Box lists the PPS-exempt cancer centers and affiliated hospitals.

Box Section Ref ID

In 2015, the US Government Accountability Office (GAO) assessed 9 of the 11 PPS-exempt cancer centers (2 were excluded owing to lack of available data), and found that CMS paid those centers an average 42% more for inpatient services (range, 7% to 91%) and an average 37% more (range, 13% to 44%) for outpatient services than they would have been paid with PPS reimbursement.⁶ The GAO estimated that Medicare paid these 9 centers in excess of 0.5 billion dollars per year more than they would have been reimbursed under the PPS system, with “no incentive for cost containment or incentives for cost efficiency.”⁶

In 2010, the Affordable Care Act established the PPS-Exempt Cancer Hospital Quality Reporting (PCHQR) program.^3,5,10-12 The PCHQR program requires PPS-exempt cancer centers to publicly report quality metrics, including several oncology-specific measures, as well as results of the Hospital Consumer Assessment of Healthcare Providers and Systems Survey (HCAHPS). Additional quality measures, including health care–associated infections, are collected but not publicly reported. These reporting requirements, however, only apply to the PPS-exempt cancer centers.

Because there are very limited available data to compare PPS-exempt cancer centers with PPS hospitals, it is unknown how their performance compares with that of other cancer centers designated by the National Cancer Institute (NCI-CCs),¹³ or other hospitals that provide cancer care. In this study, we used the available public data to compare hospital characteristics, cancer services, patient comorbidities, and postoperative outcomes at hospitals affiliated with PPS-exempt cancer centers, other hospitals affiliated with NCI-CCs, and other hospitals that provide cancer care in the United States.

We obtained patient-level data from Medicare inpatient and outpatient claims data. We identified fee-for-service Medicare beneficiaries 65 years and older who had 1 of 9 types of cancer operations (brain tumor resection, colorectal resection, cystectomy, esophagectomy, gastrectomy, liver resection, lung resection, pancreatic resection, prostatectomy) between January 1, 2011, and May 31, 2015 (as of February, 2018, this was the most recent Medicare data available), using International Classification of Diseases, 9th Revision (ICD-9) codes (eTable 1 in the Supplement). These 9 operations are frequently performed, associated with substantial morbidity (eg, at least one-third of patients after pancreatic surgery will have a complication), or are commonly examined in analyses of surgical cancer policy. We obtained hospital characteristics from the 2014 annual survey of the American Hospital Association and from data reported by hospitals to US News & World Report for their annual Best Hospitals rankings for cancer hospitals.¹⁴ All analysis took place between February 2018 and August 2018. The Northwestern University institutional review board deemed the study exempt from human subjects review owing to its retrospective evaluation of deidentified patient data.

Hospital affiliation classification was mutually exclusive: the 11 PPS-exempt cancer centers and affiliated hospitals (n = 15) (Box), hospitals affiliated with NCI-CCs (n = 54, eTable 2 in the Supplement), and other hospitals that provide cancer care for any of the 9 procedures (n = 3578). For analyses with Medicare data, only those patients who underwent cancer surgery were included. The NCI designates cancer centers using criteria such as scientific leadership, resources, depth of basic, clinical, and population-based transdisciplinary research, and patient care efforts; every 5 years it validates and renews the designations through site visits.¹³ If a hospital was affiliated with both a PPS-exempt cancer center and an NCI-CC (13 of 15 PPS-exempt affiliated hospitals), the center was classified as a PPS-exempt center, as the PPS-exempt status takes precedence for reporting and financial purposes. We also performed sensitivity analyses to assess the robustness of this classification by including an additional hospital that could also be considered affiliated with a PPS-exempt cancer center.

Using data from the American Hospital Association annual hospital survey, we compared differences in hospital characteristics between the 3 hospital groups. We used data from the US News annual rankings for cancer hospitals to compare hospitals affiliated with PPS-exempt centers and NCI-CCs only, as these data are not uniformly available for all other hospitals with respect to cancer care. We examined oncology patient volume, reputation score, patient safety rating, nursing staff, total cancer volume, and survival score. US News creates reputation scores from a national survey of physicians identified using the American Medical Association Physician Masterfile and the Doximity Physician Database, asking where physicians would send their most difficult cases regardless of location and cost.¹⁵ Survival scores, a measure of 30-day all-cause mortality after admission among both medical and surgical patients, are based on the percentile distribution of the mortality ratio and range from 1 to 10, with a higher value reflecting a lower mortality ratio.^11,15

Using Medicare data, we compared patient demographics, comorbidities, and procedure type. We obtained area-based income and education measures using the patients’ Social Security Administration beneficiary state and county codes linked to Area Health Resources Files (AHRF) from 2014 to 2015. We determined preexisting comorbidities using the Elixhauser comorbidity index.¹⁶

We examined 18 postoperative outcomes from the date of the operation to 30 days after surgery (irrespective of whether the patient was still an inpatient, at home, or readmitted to another facility). These outcomes were mortality (30-, 60-, 90-day), readmission (30-, 60-, 90-day), sepsis, central venous catheter–related blood stream infection, acute renal failure, surgical site infection, acute myocardial infarction, pneumonia, postoperative hemorrhage, venous thromboembolism, pulmonary failure, urinary tract infection, length of stay, and failure to rescue. We calculated readmission to an acute care hospital as days from the date of discharge. We defined length of stay as 9 days or longer (≥75th percentile of all patient stays), and failure to rescue as mortality within 30 days of surgery following a major complication as previously defined using Medicare data.¹⁷ Outcomes were based on previously developed and validated definitions, including those from the Agency for Healthcare Research and Quality Patient Safety Indicators, which are used in CMS public reporting and pay-for-performance programs.^17,18 We assessed complications on the basis of ICD-9 diagnosis codes from inpatient and outpatient institutional files and condition present-on-admission indicators. If a complication was present at the time of the index admission, patients were precluded from being coded as experiencing the complication. Specifications of complications are detailed in the eMethods in the Supplement.

We compared differences in categorical variables with the χ² test, Fisher exact test, or Kruskal-Wallis test, and differences in continuous variables with the median test or t test. We developed hierarchical logistic regression models with hospital random intercepts to assess differences in postoperative outcomes by hospital type (PPS-exempt cancer center, NCI-CC, other hospitals that provide cancer care) with adjustment for age, sex, race, Elixhauser comorbidities, presence of disseminated cancer, emergency/urgent admission status, discharge year, and primary surgical procedure. We report P values without correction but use asterisks to denote P values <.05 after Bonferroni adjustment, which ranged from .002 to .006. Because correction for multiple comparisons could result in a bias in favor of finding no difference between hospital types (ie, the study hypothesis), results are shown with and without adjustments for multiple comparisons. All tests were 2-sided, and the significance level was set at <.05. All analyses were performed using STATA/MP 15.

Hospitals affiliated with PPS-exempt cancer centers or NCI-CCs had similar structural characteristics, except bed size; NCI-CCs were more frequently larger (≥500 beds: PPS-exempt, 26.7%; NCI-CCs, 81.1%) (Table 1). Hospitals affiliated with PPS-exempt cancer centers or NCI-CCs performed more operations and offered a number of additional cancer-related services compared with the other hospitals that provide care (Table 1). Mean US News reputation scores were higher at hospitals affiliated with PPS-exempt cancer centers compared with hospitals affiliated with NCI-CCs (mean [SD], 17.5 [(24.0)] vs 2.6 ([4.8)]; P < .001); however, there was no significant difference in the overall US News survival score reported (PPS-exempt, 9.3 [1.1] vs NCI-CC, 8.5 [1.2]; P = .07).

Of the 414 126 patients undergoing 1 of the 9 operations examined, 16 940 patients (4.1%) underwent surgery at a hospital affiliated with a PPS-exempt cancer center, 48 196 patients (11.6%) at an NCI-CC-affiliated hospital, and 348 990 patients (84.3%) at another hospital that provide cancer care (Table 2). Compared with patients at PPS-exempt cancer centers, NCI-CC-affiliated hospitals had a lower percentage of patients with distant metastases (14.5% vs 25.0%; P = .004). However, when evaluating individual Elixhauser comorbidities by hospital type (eTable 3 in the Supplement), a higher percentage of patients treated at NCI-CCs had congestive heart failure, diabetes, renal failure, weight loss, and fluid/electrolyte disorders. Compared with PPS-exempt cancer centers, NCI-CC-affiliated hospitals performed a relatively higher proportion of colorectal (17.9% vs 14.1%), pancreas (10.6% vs 6.3%), and brain cancer (10.1% vs 7.9%) operations but fewer lung cancer operations (24.4% vs 29.5%) (global P = .03) (Table 2).

Compared with PPS-exempt cancer center–affiliated hospitals, other hospitals that provide cancer care treated a higher proportion of patients who were older (≥85 years; 10.6% vs 4.1%; P = .01), with more Elixhauser comorbidities (≥6 comorbidities; 7.4% vs 3.7%; P = .01), and a lower proportion of patients with distant metastases (16.4% vs 25.0%; P < .001) (Table 2). In addition, compared with PPS-exempt cancer center affiliated hospitals, the other hospitals that provide cancer care performed a relatively higher proportion of colorectal cancer procedures (49.8% vs 14.1%) (Table 2).

Compared with hospitals affiliated with PPS-exempt cancer centers, patients treated at NCI-CCs were more likely to have postoperative sepsis (3.1% vs 1.7%; P = .002; odds ratio [OR], 1.60; 95% CI, 1.19-2.16), acute renal failure (6.2% vs 3.9%; P = .01; OR, 1.64; 95% CI, 1.26-2.12), and urinary tract infection (6.4% vs 4.0%; P = .002; OR; 1.58; 95% CI; 1.26-2.00) but none of the other 15 complications after Bonferroni correction (Table 3, Table 4). Without correction for multiple comparisons, patients treated at NCI-CC–affiliated hospitals were more likely to experience 1 additional complication, which was pulmonary failure (3.8% vs 2.5%; P = .04; OR, 1.43; 95% CI, 1.07-1.91). The NCI-CC–affiliated hospitals did not perform significantly better for any complication. When compared with PPS-exempt cancer centers, the other hospitals that provided cancer care had significantly worse outcomes for mortality (30-, 60-, and 90-day), postoperative sepsis, acute renal failure, and pulmonary failure after Bonferroni correction. Without correction for multiple comparisons, the other hospitals had a higher likelihood of failure to rescue but a lower likelihood of surgical site infection and venous thromboembolism. Sensitivity analyses, with the inclusion of an additional hospital as affiliated with a PPS-exempt cancer center, showed qualitatively similar results (eTable 4 and eTable 5 in the Supplement).

We found that hospitals affiliated with PPS-exempt cancer centers generally had similar hospital characteristics, patient comorbidities, and outcomes compared with hospitals affiliated with NCI-CCs. These findings raise questions about why some cancer centers and not others are designated as PPS-exempt cancer centers and why there are different cancer quality reporting requirements for the various hospitals in the United States that provide cancer care.

With respect to hospital characteristics, process measures, and patient comorbidities, these findings suggest that hospitals affiliated with PPS-exempt cancer centers and NCI-CCs are similar. Importantly, there were no differences in measurable cancer-related services between the 2 types of centers based on publicly available information. Although many of the hospital characteristics are relatively nonspecific structural measures, the majority of these cancer centers are large, urban institutions with multiple accreditations. The PPS-exempt hospitals had significantly higher US News & World Report reputation scores than NCI-CC affiliated hospitals. Reputation is a subjective assessment but may be a proxy for specific characteristics such as cancer-related services, specialists, trials, and technologies. Other more objective measures from US News include oncology patient volume, patient safety rating, total cancer and survival scores. None of these objective measures were found to significantly differ between hospitals affiliated with PPS-exempt cancer centers and NCI-CCs.

When examining outcomes, there were few significant differences between hospitals affiliated with PPS-exempt cancer centers and NCI-CCs after adjusting for differences in patient demographics, comorbidities, and case mix. The PPS-exempt status was associated with improved outcomes for acute renal failure, postoperative sepsis, and urinary tract infection (after correcting for multiple comparisons), but not with the other 15 outcomes including mortality, the US News survival score (which includes medical and surgical patients), or readmissions. Further analysis would be required to better understand the findings for acute renal failure, postoperative sepsis, and urinary tract infection; possible explanations include a real difference, flaws in the measures, unmeasured confounding, or selection bias. For example, there could be differential coding of complications; PPS hospitals are motivated to capture billing codes for sepsis, acute renal failure, and urinary tract infection under the DRG payment model because these codes affect reimbursement. Hospitals affiliated with PPS-exempt cancer centers, however, may have less motivation to code for complications because their reimbursement would not be affected.

Although the use of Medicare administrative data to analyze patient outcomes has limitations,^19-21 mortality and readmission are 2 of the most reliable measures. The influence of measurement error and reporting bias is limited. Notably, we found mortality and readmissions to be similar between hospitals affiliated with PPS-exempt cancer centers and NCI-CCs, but lower than for other hospitals that perform cancer surgery.

At present, the PPS is using no cancer-specific quality measures; only PPS-exempt cancer centers are required to report these metrics.¹² Thus, comparative quality data on cancer care are only available for 11 cancer centers and their affiliated hospitals, and not for most hospitals that provide cancer care. Our results suggest that the same cancer-specific quality measures should be collected for all cancer centers and hospitals in the United States, and that these quality measures should be publicly reported for individual institutions. Among other advantages, such public reporting may allow identification of more specific differences, if any, between different types of cancer centers, and an assessment of whether PPS-exempt status is achieving the intended goals.

This study has limitations. First, our primary study sample was limited to Medicare beneficiaries over 65 years of age. However, the majority of patients with the malignant diseases included in the study are over 65 years of age, so this sample is likely to be representative.²² As the PPS exemption for cancer hospitals and the PCHQR program are CMS policies, Medicare beneficiaries are a reasonable population to examine.

Second, our primary analyses using Medicare data were limited to 9 common operations that are performed at most cancer centers, examined in many cancer policy evaluations, and that result in considerable morbidity and mortality. Our results may not generalize to other operations or to nonsurgical cancer care. Our other analyses focused on hospital characteristics, available cancer services, and information from US News reports. For example, we reported data on the US News survival score, which include both medical and surgical patients; these data did not demonstrate a significant difference between cancer center hospital types.

Third, when adjusting for differences between hospital types, we were limited to the patient characteristics available in the Medicare data; thus, we may be unable to account for other possible comorbidities and cancer-related factors. For example, we were unable to adjust for differences in stage of disease, other than if the patient had disseminated cancer. Our prior work, however, has demonstrated that inclusion of cancer stage variables does not appreciably affect hospital quality comparisons for short-term postoperative outcomes.^23,24 We were also unable to adjust for differences in intraoperative case complexity between hospital types; our prior work suggests consideration of oncologic case complexity also has little influence on hospital quality comparisons.²⁵ Moreover, Medicare data are widely used by CMS in all of their public reporting and pay-for-performance programs.

Finally, our study addressed only short-term postoperative outcomes, and not other measures of quality, such as long-term outcomes (eg, survival and recurrence). There are few data sources available that allow comparison of PPS-exempt cancer center affiliated hospitals and other NCI-CCs with respect to longer-term outcomes, survival, or adherence to cancer-related process measures. For example, SEER-Medicare data could not be used because the selected SEER registry regions include fewer than half of the PPS-exempt cancer centers and NCI-CCs.²²

In exploring differences in hospital characteristics, patient comorbidities, and postoperative outcomes at hospitals affiliated with PPS-exempt cancer centers, NCI-CCs, and other hospitals that provide cancer care in the United States, we found that hospitals affiliated with PPS-exempt cancer centers were largely similar to hospitals affiliated with NCI-CCs. Although PPS-exempt cancer centers may serve an important purpose by advancing the science and quality of cancer care, limited information is available on the selection of PPS-exempt cancer centers, their reimbursement, and how these centers compare with other hospitals with regard to quality of care. Our findings suggest the need for additional transparency, periodic reviews of the program by CMS, and consideration of whether the classification of PPS exemption should continue. Moreover, the requirement to publicly report cancer care–quality metrics should be uniformly applied across all types of hospitals, not just PPS-exempt cancer centers.

Accepted for Publication: February 27, 2019.

Corresponding Author: Karl Y. Bilimoria, MD, MS, Surgical Outcomes and Quality Improvement Center (SOQIC), Department of Surgery and Center for Healthcare Studies, Feinberg School of Medicine, Northwestern University, 633 N St Clair St, 20th Floor, Chicago, IL 60611 (k-bilimoria@northwestern.edu).

Published Online: June 17, 2019. doi:10.1001/jamainternmed.2019.0914

Author Contributions: Drs Chung and Merkow had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Merkow, Yang, Bentrem, Bilimoria.

Acquisition, analysis, or interpretation of data: Merkow, Pavey, Sohn, Chung, Bilimoria.

Drafting of the manuscript: Merkow, Yang, Bilimoria.

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

Statistical analysis: Merkow, Pavey, Chung, Bilimoria.

Obtained funding: Merkow, Bentrem, Bilimoria.

Administrative, technical, or material support: Yang, Bentrem.

Study supervision: Merkow, Yang, Bilimoria.

Conflict of Interest Disclosures: Dr Merkow reported support from the Agency for Healthcare Research and Quality and the American Cancer Society. Dr Bilimoria reported support from the National Institutes of Health, Agency for Healthcare Research and Quality, American Board of Surgery, American College of Surgeons, Accreditation Council for Graduate Medical Education, Health Care. No other disclosures reported.

Funding/Support: This research was supported by the Northwestern Institute for Comparative Effectiveness Research in Oncology (NICER-Onc) of the Robert H. Lurie Comprehensive Cancer Center of Northwestern University.

Role of the Funder/Sponsor: The funding sponsors played no part in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. The sponsors had no access to the data and did not perform any of the study analysis.