Longitudinal varying coefficient single-index model with censored covariates

It is of interest to health policy research to estimate the population-averaged longitudinal medical cost trajectory from initial cancer diagnosis to death, and understand how the trajectory curve is affected by patient characteristics. This research question leads to a number of statistical challenges because the longitudinal cost data are often non-normally distributed with skewness, zero-inflation, and heteroscedasticity. The trajectory is nonlinear, and its length and shape depend on survival, which are subject to censoring. Modeling the association between multiple patient characteristics and nonlinear cost trajectory curves of varying lengths should take into consideration parsimony, flexibility, and interpretation. We propose a novel longitudinal varying coefficient single-index model. Multiple patient characteristics are summarized in a single-index, representing a patient's overall propensity for healthcare use. The effects of this index on various segments of the cost trajectory depend on both time and survival, which is flexibly modeled by a bivariate varying coefficient function. The model is estimated by generalized estimating equations with an extended marginal mean structure to accommodate censored survival time as a covariate. We established the pointwise confidence interval of the varying coefficient and a test for the covariate effect. The numerical performance was extensively studied in simulations. We applied the proposed methodology to medical cost data of prostate cancer patients from the Surveillance, Epidemiology, and End Results-Medicare-Linked Database.

The Influence of Hospital Characteristics on Patient Survival in Surgically Managed Metastatic Disease of Bone: An Analysis of the SEER-Medicare Linked Database

OBJECTIVES

We investigated whether patients receiving surgical treatment for metastatic disease of bone (MDB) at hospitals with higher volume, medical school affiliation, or Commission on Cancer accreditation have superior outcomes.

MATERIALS AND METHODS

Using the Surveillance, Epidemiology, and End Results-Medicare database, we identified 9413 patients surgically treated for extremity MDB between 1992 and 2014 at the age of 66 years or older. Cox proportional hazards models were used to calculate the hazards ratios (HR) for 90-day and 1-year mortality and 30-day readmission according to the characteristics of the hospital where bone surgery was performed.

RESULTS

We observed no notable differences in 90-day mortality, 1-year mortality, or 30-day readmission associated with hospital volume. Major medical school affiliation was associated with lower 90-day (HR: 0.88, 95% confidence interval [CI]: 0.80-0.96) and 1-year (HR: 0.92, 95% CI: 0.87-0.99) mortality after adjustments for demographic and tumor characteristics. Surgical treatment at Commission on Cancer accredited hospitals was associated with significantly higher risk of death at 90 days and 1 year after the surgery. This effect appeared to be driven by lung cancer patients (1-year HR: 1.17, 95% CI: 1.07-1.27).

CONCLUSIONS

Our findings suggest surgical management of MDB at lower-volume hospitals does not compromise survival or readmissions. There may be benefit to referral or consultation with an academic medical center in some tumor types or clinical scenarios.

Pancreatic cancer treatment costs, including patient liability, by phase of care and treatment modality, 2000-2013

OBJECTIVES

Our study provides phase-specific cost estimates for pancreatic cancer based on stage and treatment. We compare treatment costs between the different phases and within the stage and treatment modality subgroups.

METHODS

Our cohort included 20,917 pancreatic cancer patients from the Surveillance, Epidemiology, and End Results (SEER)-Medicare database diagnosed between 2000 and 2011. We allocated costs into four phases of care-staging (or surgery), initial, continuing, and terminal- and calculated the total, cancer-attributable, and patient-liability costs in 2018 US dollars. We fit linear regression models using log transformation to determine whether costs were predicted by age and calendar year.

RESULTS

Monthly cost estimates were high during the staging and surgery phases, decreased over the initial and continuing phases, and increased during the three-month terminal phase. Overall, the linear regression models showed that cancer-attributable costs either remained stable or increased by year, and either were unaffected by age or decreased with older age; continuing phase costs for stage II patients increased with age.

CONCLUSIONS

Our estimates demonstrate that pancreatic cancer costs can vary widely by stage and treatment received. These cost estimates can serve as an important baseline foundation to guide resource allocation for cancer care and research in the future.

Secondary Surgery Versus Chemotherapy for Recurrent Ovarian Cancer

OBJECTIVE

The best course of treatment for recurrent ovarian cancer is uncertain. We sought to determine whether secondary cytoreductive surgery for first recurrence of ovarian cancer improves overall survival compared with other treatments.

MATERIALS AND METHODS

We assessed survival using Surveillance, Epidemiology and End Results-Medicare data for advanced stage ovarian cancer cases diagnosed from January 1, 1997 to December 31, 2007 with survival data through 2010 using multinomial propensity weighted finite mixture survival regression models to distinguish true from misclassified recurrences. Of 35,995 women ages 66 years and older with ovarian cancer, 3439 underwent optimal primary debulking surgery with 6 cycles of chemotherapy; 2038 experienced a remission.

RESULTS

One thousand six hundred thirty-five of 2038 (80%) women received treatment for recurrence of whom 72% were treated with chemotherapy only, 16% with surgery and chemotherapy and 12% received hospice care. Median survival of women treated with chemotherapy alone, surgery and chemotherapy, or hospice care was 4.1, 5.4, and 2.2 years, respectively (P<0.001). Of those receiving no secondary treatments, 75% were likely true nonrecurrences with median survival of 15.9 years and 25% misclassified with 2.4 years survival. Survival among women with recurrence was greater for those treated with surgery and chemotherapy compared with chemotherapy alone (hazard ratio=1.67; 95% confidence interval, 1.13-2.47). Women who were older with more comorbidities and high-grade cancer had worse survival.

CONCLUSIONS

Secondary surgery with chemotherapy to treat recurrent ovarian cancer increases survival by 1.3 years compared with chemotherapy alone and pending ongoing randomized trial results, may be considered a standard of care.

Patient, Hospital, and Geographic Disparities in Laparoscopic Surgery Use Among Surveillance, Epidemiology, and End Results-Medicare Patients With Colon Cancer

BACKGROUND

Surgical resection is the primary treatment for colon cancer, but use of laparoscopic approaches varies widely despite demonstrated short- and long-term benefits.

OBJECTIVE

The purpose of this study was to identify characteristics associated with laparoscopic colon cancer resection and to quantify variation based on patient, hospital, and geographic characteristics.

DESIGN

Bayesian cross-classified, multilevel logistic models calculated adjusted ORs and CIs for patient, surgeon, hospital, and geographic characteristics and unexplained variability (predicted vs. observed values) using adjusted median odds ratios for hospitals and counties.

SETTINGS

The Surveillance, Epidemiology, and End Results-Medicare claims database (2008-2011) supplemented with county-level American Community Survey (2008-2012) demographic data was used.

PATIENTS

A total of 10,618 patients ≥66 years old who underwent colon cancer resection were included.

MAIN OUTCOME MEASURES

Nonurgent/nonemergent resections for colon cancer patients ≥66 years old were classified as laparoscopic or open procedures.

RESULTS

Patients resided in 579 counties and used 950 hospitals; 47% of patients underwent laparoscopic surgery. Medicare/Medicaid dual enrollment, age ≥85 years, and higher tumor stage and grade were negatively associated with laparoscopic surgery receipt; proximal tumors and increasing hospital size and surgeon caseload were positively associated. Significant unexplained variability at the hospital (adjusted median OR = 3.31; p < 0.001) and county levels (adjusted median OR = 1.28; p < 0.05) remained after adjustment.

LIMITATIONS

This was an observational study lacking generalizability to younger patients without Medicare or those with Health Maintenance Organization coverage and data set did not reflect national hospital studies or hospital volume. In addition, we were unable to account for specific types of comorbidities, such as obesity, and had broad categories for surgeon caseload.

CONCLUSIONS

Determining sources of hospital-level variation among poor insured patients may help increase laparoscopic resection to maximize health outcomes and reduce cost. See Video Abstract at http://links.lww.com/DCR/A363.

Metabolic syndrome increases risk of Barrett esophagus in the absence of gastroesophageal reflux: an analysis of SEER-Medicare Data

GOALS

To evaluate the association between metabolic syndrome (MetS) and risk of Barrett esophagus (BE) using the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database compared with 2 control groups--Medicare population controls and endoscopy controls.

BACKGROUND

BE principally arises as an adaptation to the proinflammatory state induced by gastroesophageal reflux disease (GERD). The relationship between obesity and BE is presumed to be mediated by GERD. However, evidence suggests central adiposity also increases risk of BE independent of GERD. Central adiposity is one risk factor defining MetS, which confers a systemic proinflammatory state--a potential GERD-independent mechanism by which obesity could increase the risk of BE.

STUDY

MetS was defined as diagnosis of at least 3 of the following conditions: obesity, elevated triglycerides, high blood pressure, and elevated fasting glucose. Multivariable logistic regression was used to estimate adjusted odds ratios and 95% confidence intervals.

RESULTS

In 2198 incident BE cases, prior MetS was significantly associated with BE (odds ratio, 1.20; 95% confidence interval: 1.07, 1.36) compared with population controls. However, GERD status modified the association; among those without prior GERD, MetS increased risk of BE by 34%; however, no association was observed among those with a prior GERD diagnosis (P-value for effect modification <0.001). MetS was not associated with risk of BE compared with endoscopy controls.

CONCLUSIONS

MetS increased the risk of BE compared with population controls, an association driven by and confined to the non-GERD stratum. MetS may mediate an association between central adiposity and BE for those without GERD.

Survival outcomes in patients with early stage, resected pancreatic cancer - a comparison of gemcitabine- and 5-fluorouracil-based chemotherapy and chemoradiation regimens

PURPOSE

We conducted a comparative survival analysis between patients with resected pancreatic cancer who received adjuvant treatment with either gemcitabine- or 5-fluorouracil-based chemotherapy and chemoradiation regimens.

PATIENTS AND METHODS

The Surveillance, Epidemiology and End Results (SEER)-Medicare database was used to identify patients with pancreatic cancer diagnosed from 1998 to 2005 who received curative surgery and adjuvant chemotherapy with either 5-fluorouracil or gemcitabine. These groups were subdivided by treatment with radiotherapy. Patients were followed until death, study end-point or a maximum of 5 years after diagnosis.

RESULTS

Three hundred and fifty-nine patients received 5-fluorouracil and 346 received gemcitabine. Compared with chemoradiation with 5-fluorouracil, outcomes for patients who received chemoradiation with gemcitabine did not differ. Patients who received gemcitabine without radiation had increased hazards (poorly differentiated tumours: HR = 1.50, p = 0.01; moderately differentiated tumours, HR = 1.28, p = 0.11). However, outcomes of patients who received 5-fluorouracil without radiation varied with tumour grade. In moderately differentiated tumours, patients had better outcomes with 5-fluorouracil when compared with chemoradiation with 5-fluorouracil (HR = 0.42, p = 0.02). In poorly differentiated tumours, the opposite was true (HR 2.10, p = 0.09).

CONCLUSION

Patients with low-grade resected pancreatic cancer may have better outcomes with 5-fluorouracil-based chemotherapy without radiation when compared with 5-fluorouracil with radiation.

Patient, hospital, and geographic disparities associated with comanagement during hospitalization for colorectal cancer surgery

BACKGROUND

Comanagement of surgical patients has increased, but information regarding detailed characteristics of patients receiving comanagement during hospitalization for colorectal cancer (CRC) surgery is lacking.

OBJECTIVE

To examine the use of and characteristics associated with comanagement of patients hospitalized for CRC surgery.

DESIGN

This study used a population-based cross-sectional design.

SETTING

We used the linked 2000 to 2005 Surveillance, Epidemiology, and End Results and Medicare claims data.

PATIENTS

We included 37,065 patients aged 66 years or older, hospitalized for definitive CRC surgery following stage I to III diagnosis.

MEASUREMENTS

The outcome of interest was comanagement during hospitalization for CRC surgery, and we examined the association between several patient and hospital characteristics. Comanagement was defined as having a relevant physician (ie, internal medicine hospitalist/generalist) submit a claim for evaluation and management services on 70% or more of the days of hospitalization of the patient.

RESULTS

During hospitalization for CRC surgery, 27.6% of patients were comanaged, but this percentage varied widely across hospitals (from 1.9% to 83.2%). Several patient and hospital characteristics were associated with the use of comanaged care, of which important characteristics included older age at diagnosis, presence of comorbidity, emergency surgery, and hospital volume.

CONCLUSIONS

Extensive variability existed in comanagement use across patients and hospitals, likely reflecting the lack of evidence for its clinical effectiveness.

Survival after diagnosis of localized T1a kidney cancer: current population-based practice of surgery and nonsurgical management

OBJECTIVE

To compare overall and cancer-specific survival (CSS) of patients who undergo nonsurgical management (NSM), partial nephrectomy (PN), and radical nephrectomy (RN). NSM is being increasingly used for older patients with early-stage kidney cancer and competing risks of death. However, survival is poorly characterized for this approach compared with surgery with PN or RN.

METHODS

The Surveillance, Epidemiology and End Results-Medicare database from 1995 to 2007 was used to identify patients aged 65 years or older diagnosed with localized T1a kidney cancer treated with PN, RN, or NSM. We used Cox proportional hazards regression, Fine and Gray competing risks regression, and propensity score matching to adjust for patient and tumor characteristics.

RESULTS

Of 7177 Medicare beneficiaries meeting the inclusion criteria, 754 (10.5%) underwent NSM, 1849 (25.8%) PN, and 4574 (63.7%) RN, with 436 (57.8%), 389 (21.0%), and 1598 (34.9%) patients dying from any cause, respectively, at a median follow-up of 56 months. Overall survival favored PN and RN compared with NSM (hazard ratio [95% CI]: 0.40 [0.34-0.46] and 0.50 [0.45-0.56], respectively) as did CSS (hazard ratio [95% CI]: 0.42 [0.27-0.64] and 0.62 [0.46-0.85], respectively). However, there was no difference in CSS between any 2 treatment groups for younger patients (<75 years), whereas there was an excess of kidney cancer deaths for NSM patients aged 75-79 years and an attenuated difference for patients aged 80 years or older.

CONCLUSION

NSM was associated with an increased risk of kidney cancer death among Medicare beneficiaries aged 75-79 years. Evolving active surveillance protocols will need to develop robust selection criteria to maximize cancer survival for older patients with kidney cancer.

Preoperative breast MRI and surgical outcomes in elderly women with invasive ductal and lobular carcinoma: a population-based study

Existing evidence suggests that preoperative breast magnetic resonance imaging (MRI) might not improve surgical outcomes in the general breast cancer population. To determine if patients differentially benefit from breast MRI, we examined surgical outcomes-initial mastectomy, reoperation, and final mastectomy rates-among patients grouped by histologic type. We identified women diagnosed with early-stage breast cancer from 2004 to 2007 in the SEER-Medicare dataset. We classified patients as having invasive ductal carcinoma (IDC), invasive lobular carcinoma (ILC), mixed ductal/lobular carcinoma (IDLC) or other histologic type. Medicare claims were used to identify breast MRI and definitive surgeries during the initial surgical treatment episode. We used propensity score methods to account for the differential likelihood of exposure to MRI. Of the 20,332 patients who met our inclusion criteria for this study, 12.2 % had a preoperative breast MRI. Patients with ILC as compared to other histologic groups were most likely to receive MRI [OR 2.32; 95 % CI (2.02-2.67)]. In the propensity score-adjusted analyses, breast MRI was associated with an increased likelihood of an initial mastectomy for all patients and among all histologic subgroups. Among patients with ILC, having a breast MRI was associated with lower odds of a reoperation [OR 0.59; 95 % CI (0.40-0.86)], and an equal likelihood of a final mastectomy compared to similar patients without a breast MRI. Overall and among patients with IDC and IDLC, breast MRI was not significantly associated with a likelihood of a reoperation but was associated with greater odds of a final mastectomy. Our study provides evidence in support of the targeted use of preoperative breast MRI among patients with ILC to improve surgical planning; it does not provide evidence for the routine use of breast MRI among all newly diagnosed breast cancer patients or among patients with IDC.

Differential receipt of sentinel lymph node biopsy within practice-based research networks

BACKGROUND

Practice-based research networks (PBRNs) are promising for accelerating not only research, but also dissemination of research-based evidence into broader community practice. Sentinel lymph node biopsy (SLNB) is an innovation in breast cancer care associated with equivalent survival and lower morbidity, as compared with standard axillary lymph node dissection. We examined the diffusion of SLNB into practice and whether affiliation with the Community Clinical Oncology Program (CCOP), a cancer-focused PBRN, was associated with more rapid uptake of SLNB.

RESEARCH DESIGN

Surveillance Epidemiology and End Results-Medicare data were used to study women diagnosed with stage I or II breast cancer in the years 2000-2005 and undergoing breast-conserving surgery with axillary staging (n=6226). The primary outcome was undergoing SLNB. CCOP affiliation of the surgical physician was ascertained from NCI records. Multivariable generalized linear modeling with generalized estimating equations was used to measure association between CCOP exposure and undergoing SLNB, controlling for potential confounders.

RESULTS

Women treated by a CCOP physician had significantly higher odds of receiving SLNB compared with women treated by a non-CCOP physician (OR 2.68; 95% CI, 1.35-5.34). The magnitude of this association was larger than that observed among patients treated by physicians operating in medical school-affiliated hospitals (OR 1.76; 95% CI, 1.30-2.39).

CONCLUSIONS

Women treated by CCOP-affiliated physicians were more likely to undergo SLNB irrespective of the hospital's medical school affiliation, suggesting that the CCOP PBRN may play a role in the rapid adoption of research-based innovation in community practice.

Utilization and determinants of adjuvant therapy among older patients who receive curative surgery for pancreatic cancer

OBJECTIVE

We conducted a population-based study to describe the utilization, determinants, and survival effects of adjuvant therapies after surgery among older patients with pancreatic cancer.

METHODS

Using Surveillance, Epidemiology, and End Results-Medicare data, we identified patients older than 65 years who received surgical resection for pancreatic cancer during 1992-2002. We constructed multiple logistic regression models to examine patient, clinical, and hospital factors associated with receiving adjuvant therapy. Cox proportional hazards models were used to examine the effect of therapy on survival.

RESULTS

Approximately 49% of patients received adjuvant therapy after surgery. Patient factors associated with increased receipt of adjuvant therapy included more recent diagnosis, younger age, stage II disease, higher income, and geographic location. Hospital factors associated with increased receipt of adjuvant therapy included cooperative group membership and larger size. Adjuvant treatments associated with a significant reduction in 2-year mortality (relative to surgery alone) were chemoradiation or radiation alone but not chemotherapy alone.

CONCLUSIONS

Our findings suggest that adjuvant chemoradiation and, to a lesser degree, radiation only are associated with a reduction in the risk of mortality among older patients who undergo surgery for pancreatic cancer. However, receipt of adjuvant therapy varied by period and geography as well as by certain patient and hospital factors.