The utility of Medicare claims data for measuring cancer stage

Secondary databases in gynecologic cancer research

Observational and cohort studies using large databases have made important contributions to gynecologic oncology. Knowledge of the advantages and potential limitations of commonly used databases benefits both readers and reviewers. In this review, researchers familiar with National Cancer Database (NCDB), Surveillance, Epidemiology, and End Results Program (SEER), SEER-Medicare, MarketScan, Healthcare Cost and Utilization Project (HCUP), National Surgical Quality Improvement Program (NSQIP), and Premier, describe each database, its included data, access, management, storage, highlights, and limitations. A better understanding of these commonly used datasets can help readers, reviewers, and researchers to more effectively interpret and apply study results, evaluate new research studies, and develop compelling and practice-changing research.

Use of cancer-directed therapy at the end of life among adolescents and young adults

BACKGROUND

Adolescents and young adults frequently receive chemotherapy near death. We know less about the use of targeted agents and immunotherapy or trends over time.

METHODS

We conducted a retrospective cohort study of 1836 adolescents and young adults with cancer who died between 2009 and 2019 after receiving care at 1 of 3 sites (Dana-Farber Cancer Institute, Kaiser Permanente Northern California, and Kaiser Permanente Southern California). We reviewed electronic health data and medical records to examine use of cancer-directed therapy in the last 90 days of life, including chemotherapy, targeted therapy, immunotherapy, and investigational drugs.

RESULTS

Over the study period, 35% of adolescents and young adults received chemotherapy in the last 90 days of life; 24% received targeted therapy, 7% immunotherapy, and 5% investigational drugs. Additionally, 56% received at least 1 form of systemic cancer-directed therapy in the last 90 days of life. After adjustment for patient sex, race, ethnicity, age, site of care, diagnosis, and years from diagnosis to death, the proportion of adolescents and young adults receiving targeted therapy (odds ratio [OR] = 1.05 per year of death, 95% confidence interval [CI] = 1.02 to 1.10; P = .006), immunotherapy (OR = 1.27, 95% CI = 1.18 to 1.38; P < .0001), and any cancer-directed therapy (OR = 1.04, 95% CI = 1.01 to 1.08; P = .01) in the last 90 days of life increased over time.

CONCLUSIONS

More than half of adolescents and young adults receive cancer therapy in the last 90 days of life, and use of novel agents such as targeted therapy and immunotherapy is increasing over time. Although some adolescents and young adults may wish to continue cancer therapy while living with advanced disease, efforts are needed to ensure that use of cancer-directed therapy meets preferences of adolescents and young adults approaching death.

Quality of End-of-Life Care Among Adolescents and Young Adults With Cancer

Adolescents, young adults with cancer receive limited psychosocial and spiritual support near death.

Discussions About Goals of Care and Advance Care Planning Among Adolescents and Young Adults With Cancer Approaching the End of Life

PURPOSE

Adolescents and young adults (AYAs) with cancer receive high rates of medically intensive measures at the end of life. This study aimed to characterize the prevalence and timing of conversations about goals of care and advance care planning among AYAs at the end of life as one potential influence on care received.

METHODS

This was a review of electronic health data and medical records for 1,929 AYAs age 12-39 years who died after receiving care at one of three sites between 2003 and 2019, including documented conversations about goals of care and advance care planning, and care received.

RESULTS

A majority of AYAs were female (54%) and White (61%); 12% were Asian, 8% Black, and 27% Hispanic. Most patients had documented discussions about prognosis (86%), goals of care (83%), palliative care (79%), hospice (79%), and preferred location of death (64%). When last documented goals of care were evaluated, 69% of patients wanted care focused on palliation; however, 29% of those with palliative goals spent time in the intensive care unit (ICU) in the last month of life, and 32% had multiple emergency room (ER) visits. When goals-of-care discussions happened earlier, >30 days before death, AYAs were less likely to receive chemotherapy in the last 14 days of life (P = .001), ICU care (P < .001), ER visits (P < .001), and hospitalizations in the last month (P < .001).

CONCLUSION

High rates of medically intensive measures among AYAs near the end of life do not appear to be the result of a lack of discussions about goals of care and advance care planning. Although some interventions may be used to support palliative goals, earlier discussions have potential to reduce late-life intensive measures.

Development of a Machine Learning Model to Identify Colorectal Cancer Stage in Medicare Claims

PURPOSE

Staging information is essential for colorectal cancer research. Medicare claims are an important source of population-level data but currently lack oncologic stage. We aimed to develop a claims-based model to identify stage at diagnosis in patients with colorectal cancer.

METHODS

We included patients age 66 years or older with colorectal cancer in the SEER-Medicare registry. Using patients diagnosed from 2014 to 2016, we developed models (multinomial logistic regression, elastic net regression, and random forest) to classify patients into stage I-II, III, or IV on the basis of demographics, diagnoses, and treatment utilization identified in Medicare claims. Models developed in a training cohort (2014-2016) were applied to a testing cohort (2017), and performance was evaluated using cancer stage listed in the SEER registry as the reference standard.

RESULTS

The cohort of patients with 30,543 colorectal cancer included 14,935 (48.9%) patients with stage I-II, 9,203 (30.1%) with stage III, and 6,405 (21%) with stage IV disease. A claims-based model using elastic net regression had a scaled Brier score (SBS) of 0.45 (95% CI, 0.43 to 0.46). Performance was strongest for classifying stage IV (SBS, 0.62; 95% CI, 0.59 to 0.64; sensitivity, 93%; 95% CI, 91 to 94) followed by stage I-II (SBS, 0.45; 95% CI, 0.44 to 0.47; sensitivity, 86%; 95% CI, 85 to 76) and stage III (SBS, 0.32; 95% CI, 0.30 to 0.33; sensitivity, 62%; 95% CI, 61 to 64).

CONCLUSION

Machine learning models effectively classified colorectal cancer stage using Medicare claims. These models extend the ability of claims-based research to risk-adjust and stratify by stage.

Use of endoscopic ultrasound for pancreatic cancer from 2000 to 2016

Background and study aims  Pancreatic cancer (PC) is the fourth most common cause of cancer death in the United States. Previous studies have suggested a survival benefit for endoscopic ultrasound (EUS), an important tool for diagnosis and staging of PC. This study aims to describe EUS use over time and identify factors associated with EUS use and its impact on survival. Patients and methods  This was a retrospective review of the Surveillance, Epidemiology and End Results (SEER) database linked with Medicare claims. EUS use, clinical and demographic characteristics were evaluated. Chi-squared analysis, Cochran-Armitage test for trend, and logistic regression were used to identify associations between sociodemographic and clinical factors and EUS. Kaplan-Meier and Cox proportional hazard ratios were used for survival analysis. Results  EUS use rose during the time period, from 7.4 % of patients in 2000 to 32.4 % in 2015. Patient diversity increased, with a rising share of older, non-White patients with higher Charlson comorbidity scores. Both clinical (receipt of other therapies, PC stage) and nonclinical factors (region of country, year of diagnosis) were associated with receipt of EUS. While EUS was associated with a survival improvement early in the study period, this effect did not persist for PC patients diagnosed in 2012 to 2015 (median survival 3 month ± standard deviation [SD] 9.8 months without vs. 4 months ± SD 8 months with EUS). Conclusions  Our data support previous studies, which suggest a survival benefit for EUS when it was infrequently used, but finds that benefit was attenuated as EUS became more widely available.

Development and validation of coding algorithms to identify patients with incident lung cancer in United States healthcare claims data

PURPOSE

Our aim was to develop and validate a practical US healthcare claims algorithm for identifying incident lung cancer that improves on positive predictive value (PPV) and sensitivity observed in past studies.

METHODS

Patients newly diagnosed with lung cancer in Surveillance, Epidemiology, and End Results (SEER) (gold standard) were linked with Medicare claims. A 5% Medicare "other cancer" sample and noncancer sample served as controls. A split-sample validation approach was used. Rules-based, regression, and machine learning models for developing algorithms were explored. Algorithms were developed in the model building subset. Rules-based algorithms and those with the highest F scores were evaluated in the validation subset. F scores were compared for 1000 bootstrap samples. Misclassification was evaluated by calculating the odds of selection by the algorithm among true positives and true negatives.

RESULTS

A practical single-score algorithm derived from a logistic regression model had sensitivity = 78.22% and PPV = 78.50% (F score: 78.36). The algorithm was most likely to misclassify older patients (ages ≥80 years) or with missing data in the SEER registry, shorter follow-up time in Medicare (<3 months), insurance through Veterans Affairs, >1 cancer in SEER, or certain Charlson comorbidities (dementia, chronic pulmonary disease, liver disease, or myocardial infarction).

CONCLUSION

In this dataset, a practical point-based algorithm for identifying incident lung cancer demonstrated significant and substantial improvement (7.9% and 23.9% absolute improvement in sensitivity and PPV, respectively) compared with a current standard.

Variations in hospitalization and emergency department/observation stays using the oncology care model methodology in Medicare data

OBJECTIVE

To assess variations in hospitalizations, emergency department/observational (ED/OB) stays not resulting in hospitalization, reasons for hospitalization, and hospitalization discharge destinations after chemotherapy, information not provided as part of Oncology Care Model (OCM) baseline data.

METHODS

OCM methodology was applied to the Medicare 20% sample data to identify 6-month patient episodes triggered by chemotherapy in 2012-2015. Proportions of episodes with hospitalization or ED/OB stays, reasons for hospitalization, and discharge destinations were summarized.

RESULTS

Of 485,186 6-month episodes for 255,229 patients in 13,823 practices, 25% of episodes led to ≥1 hospitalization (from 14% in breast cancer to 56% in acute leukemia), and 23% to ED/OB stays (from 18% in breast cancer to 36% in liver cancer). In 2995 practices with ≥20 total episodes, practice-level proportions of episodes with hospitalization ranged from 14% to 31% (20th-80th percentile) and with ED/OB stays from 17% to 29%. For all cancers combined, the most frequent reasons for hospitalization were infection (13%), anemia (7%), dehydration (5%), and congestive heart failure (3%); the most common discharge destinations were home (71%) followed by a skilled nursing facility (13%), death (6%), and hospice (5%). Reasons for hospitalization and discharge destinations varied by cancer type; acute leukemia episodes led to the highest rates of infection and anemia, and central nervous system tumor episodes to the highest proportions of death or hospice discharge.

CONCLUSION

The variations in frequency of and reasons for hospitalization, ED/OB stays, and hospitalization discharge destinations across cancer types should be considered when evaluating OCM practice performance.

Development of predictive models to identify advanced-stage cancer patients in a US healthcare claims database

BACKGROUND

Although healthcare databases are a valuable source for real-world oncology data, cancer stage is often lacking. We developed predictive models using claims data to identify metastatic/advanced-stage patients with ovarian cancer, urothelial carcinoma, gastric adenocarcinoma, Merkel cell carcinoma (MCC), and non-small cell lung cancer (NSCLC).

METHODS

Patients with ≥1 diagnosis of a cancer of interest were identified in the HealthCore Integrated Research Database (HIRD), a United States (US) healthcare database (2010-2016). Data were linked to three US state cancer registries and the HealthCore Integrated Research Environment Oncology database to identify cancer stage. Predictive models were constructed to estimate the probability of metastatic/advanced stage. Predictors available in the HIRD were identified and coefficients estimated by Least Absolute Shrinkage and Selection Operator (LASSO) regression with cross-validation to control overfitting. Classification error rates and receiver operating characteristic curves were used to select probability thresholds for classifying patients as cases of metastatic/advanced cancer.

RESULTS

We used 2723 ovarian cancer, 6522 urothelial carcinoma, 1441 gastric adenocarcinoma, 109 MCC, and 12,373 NSCLC cases of early and metastatic/advanced cancer to develop predictive models. All models had high discrimination (C > 0.85). At thresholds selected for each model, PPVs were all >0.75: ovarian cancer = 0.95 (95% confidence interval [95% CI]: 0.94-0.96), urothelial carcinoma = 0.78 (95% CI: 0.70-0.86), gastric adenocarcinoma = 0.86 (95% CI: 0.83-0.88), MCC = 0.77 (95% CI 0.68-0.89), and NSCLC = 0.91 (95% CI 0.90 - 0.92).

CONCLUSION

Predictive modeling was used to identify five types of metastatic/advanced cancer in a healthcare claims database with greater accuracy than previous methods.

Classifying Stage IV Lung Cancer From Health Care Claims: A Comparison of Multiple Analytic Approaches

PURPOSE

Cancer stage is a key determinant of outcomes; however, stage is not available in claims-based data sources used for real-world evaluations. We compare multiple methods for classifying lung cancer stage from claims data.

METHODS

Our study used the linked SEER-Medicare data. The patient samples included fee-for-service Medicare beneficiaries diagnosed with lung cancer from 2010 to 2011 (development cohort) and 2012 to 2013 (validation cohort) who received chemotherapy. Classification algorithms considered Medicare Part A and B claims for care in the 3 months before and after chemotherapy initiation. We developed a clinical algorithm to predict stage IV (v I to III) cancer on the basis of treatment patterns (surgery, radiotherapy, chemotherapy). We also considered an ensemble of claims-based machine learning algorithms. Classification methods were trained in the development cohort, and performance was measured in both cohorts. The SEER data were the gold standard for cancer stage.

RESULTS

Development and validation cohorts included 14,760 and 14,620 patients with lung cancer, respectively. Validation analyses assessed clinical, random forest, and simple logistic regression algorithms. The best performing classifier within the development cohort was the random forests, but this performance was not replicated in validation analysis. Logistic regression had stable performance across cohorts. Compared with the clinical algorithm, the 14-variable logistic regression algorithm demonstrated higher accuracy in both the development (77% v 71%) and validation cohorts (77% v 73%), with improved specificity for stage IV disease.

CONCLUSION

Machine learning algorithms have potential to improve lung cancer stage classification but may be prone to overfitting. Use of ensembles, cross-validation, and external validation can aid generalizability. Degradation of accuracy between development and validation cohorts suggests the need for caution in implementing machine learning in research or care delivery.

Lifetime Occurrence of Brain Metastases Arising from Lung, Breast, and Skin Cancers in the Elderly: A SEER-Medicare Study

BACKGROUND

The Surveillance, Epidemiology, and End Results (SEER) Program recently released data on brain metastases (BM) diagnosed during primary cancer staging workup ("synchronous" BM, or SBM); this study examines the incidence of SBM compared with that of lifetime BM (LBM) identified using Medicare claims for patients diagnosed with lung cancer, breast cancer, or melanoma.

METHODS

Incidence proportions (IP) and age-adjusted rates for each of SEER SBM and Medicare LBM are presented along with measures of concordance between the two sources of data, where Medicare LBM were defined by several combinations of diagnosis and putative diagnostic imaging procedure codes.

RESULTS

The SBM IP in lung, breast, and melanoma cancers were 9.6%, 0.3%, and 1.1%, respectively; the corresponding LBM IP were 13.5%, 1.8%, and 3.6%. The greatest SBM IP among patients with lung cancer was 13.4% for non-small cell lung cancer, and among patients with breast cancer was 0.7% for triple-negative breast cancer. The greatest LBM IP among lung cancers was 23.1% in small-cell lung cancer, and among breast cancers was 4.2% for cases of the triple negative subtype.

CONCLUSIONS

Using a large dataset that is representative of the elderly population in the United States, these analyses estimate synchronous and lifetime incidence of BM in lung cancers, breast cancers, and melanomas.

IMPACT

These and other population-based estimates may be used to guide development of BM screening policy and evaluation of real-world data sources.

Age, socioeconomic features, and clinical factors predict receipt of endoscopic retrograde cholangiopancreatography in pancreatic cancer

BACKGROUND

Endoscopic retrograde cholangiopancreatography (ERCP) is the recommended technique for biliary decompression in pancreatic cancer. Previous studies have suggested racial, socioeconomic and geographic differences in diagnosis, treatment and outcomes of pancreatic cancer patients.

AIM

To examine geographic, racial, socioeconomic and clinical factors associated with utilization of ERCP.

METHODS

Surveillance, Epidemiology and End Results and linked Medicare claims data were used to identify pancreatic cancer patients between 2000-2011. Claims data were used to identify patients who had ERCP and other treatments. The primary outcome was receipt of ERCP. Chi-squared analyses were used to compare demographic information. Trends in use of ERCP over time were assessed using Cochran Armitage test. Adjusted odds ratios (aORs) and 95% confidence intervals (CIs) for receipt ERCP were calculated using logistic regression, controlling for other characteristics.

RESULTS

Among 32510 pancreatic cancer patients, 14704 (45.2%) underwent ERCP. Patients who had cancer located in the head of the pancreas (aOR 3.27, 95%CI: 2.99-3.57), had jaundice (aOR 7.59, 95%CI: 7.06-8.17), cholangitis (aOR 4.22, 95%CI: 3.71-4.81) or pruritus (aOR 1.42, 95%CI: 1.22-1.66) and lived in lower education zip codes (aOR 1.14, 95%CI: 1.04-1.24) were more likely to receive ERCP. In contrast, patients who were older (aOR 0.88, 95%CI: 0.83, 0.94), not married (aOR 0.92, 95%CI: 0.86, 0.98), and lived in a non-metropolitan area (aOR 0.89, 95%CI: 0.82, 0.98) were less likely to receive ERCP. Compared to white patients, non-white/non-black patients (aOR 0.83, 95%CI: 0.70-0.97) were less likely to receive ERCP. Patients diagnosed later in the study period were less likely to receive ERCP (aOR 2004-2007 0.85, 95%CI: 0.78-0.92; aOR 2008-2011 0.76, 95%CI: 0.70-0.83). After stratifying by indications for ERCP including jaundice, racial differences persisted (aOR black patients 0.80, 95%CI: 0.67-0.95, nonwhite/nonblack patients 0.73, 95%CI: 0.58-0.91). Among patients with jaundice, those who underwent surgery were less likely to undergo ERCP (aOR 0.60, 95%CI: 0.52, 0.69).

CONCLUSION

ERCP utilization in pancreatic cancer varies based on patient age, marital status, and factors related to where the patient lives. Further studies are needed to guide appropriate biliary intervention for these patients.

Validity of Administrative Data in Identifying Cancer-related Events in Adolescents and Young Adults: A Population-based Study Using the IMPACT Cohort

BACKGROUND

Despite the importance of estimating population level cancer outcomes, most registries do not collect critical events such as relapse. Attempts to use health administrative data to identify these events have focused on older adults and have been mostly unsuccessful. We developed and tested administrative data-based algorithms in a population-based cohort of adolescents and young adults with cancer.

METHODS

We identified all Ontario adolescents and young adults 15-21 years old diagnosed with leukemia, lymphoma, sarcoma, or testicular cancer between 1992-2012. Chart abstraction determined the end of initial treatment (EOIT) date and subsequent cancer-related events (progression, relapse, second cancer). Linkage to population-based administrative databases identified fee and procedure codes indicating cancer treatment or palliative care. Algorithms determining EOIT based on a time interval free of treatment-associated codes, and new cancer-related events based on billing codes, were compared with chart-abstracted data.

RESULTS

The cohort comprised 1404 patients. Time periods free of treatment-associated codes did not validly identify EOIT dates; using subsequent codes to identify new cancer events was thus associated with low sensitivity (56.2%). However, using administrative data codes that occurred after the EOIT date based on chart abstraction, the first cancer-related event was identified with excellent validity (sensitivity, 87.0%; specificity, 93.3%; positive predictive value, 81.5%; negative predictive value, 95.5%).

CONCLUSIONS

Although administrative data alone did not validly identify cancer-related events, administrative data in combination with chart collected EOIT dates was associated with excellent validity. The collection of EOIT dates by cancer registries would significantly expand the potential of administrative data linkage to assess cancer outcomes.

Identification of cancer patients using claims data from health insurance systems: A real-world comparative study

Objective

To evaluate the accuracy of identifying cancer patients by use of medical claims data in a health insurance system in China, and provide the basis for establishing the claims-based cancer surveillance system in China.

Methods

We chose Hua County, Henan Province as the study site, and randomly selected 300 and 1,200 qualified inpatient electronic medical records (EMRs) as well as the New Rural Cooperative Medical Scheme (NCMS) claims records for cancer patients in Hua County People’s Hospital (HCPH) and Anyang Cancer Hospital (ACH) in 2017. Diagnostic information for NCMS claims was evaluated on an individual level, and sensitivity and positive predictive value (PPV) were calculated taking the EMRs as the gold standard.

Results

The sensitivity of NCMS was 95.2% (93.8%−96.3%) and 92.0% (88.3%−94.8%) in ACH and HCPH, respectively. The PPV of the NCMS was 97.8% (96.7%−98.5%) in ACH and 89.0% (84.9%−92.3%) in HCPH. Overall, the weighted and combined sensitivity and PPV of NCMS in Hua County was 93.1% and 92.1%, respectively. Significantly higher sensitivity and PPV in identifying patients with common cancers than non-common cancers were detected in HCPH and ACH separately (P<0.01).

Conclusions

Identification of cancer patients by use of the NCMS is accurate on individual level, and it is therefore feasible to conduct claims-based cancer surveillance in areas not covered by cancer registries in China.

Predictive model algorithms identifying early and advanced stage ER+/HER2- breast cancer in claims data

PURPOSE

Claims databases offer large populations for research, but lack clinical details. We aimed to develop predictive models to identify estrogen receptor positive (ER+) and human epidermal growth factor negative (HER2-) early breast cancer (ESBC) and advanced stage breast cancer (ASBC) in a claims database.

METHODS

Female breast cancer cases in Anthem's Cancer Care Quality Program served as the gold standard validation sample. Predictive models were developed from clinical knowledge and empirically from claims data using logistic and lasso regression. Model performance was assessed by classification rates and c-statistics. Models were applied to the HealthCore Integrated Research Database (claims) to identify cohorts of women with ER+/HER2- ESBC and ASBC.

RESULTS

The validation sample included 3184 women with ER+/HER2- ESBC and 1436 with ER+/HER2- ASBC. Predictive models for ER+/HER2- ESBC and ASBC included 25 and 20 factors, respectively. Models had robust discrimination in identifying cases (c-stat = 0.92 for ESBC and 0.95 for ASBC). Compared with a traditional a priori algorithm developed with clinical insight alone, the ER+/HER2- ASBC-predictive model had better positive predictive value (PPV) (0.91, 95% CI, 0.90-0.93, vs 0.69, 95% CI, 0.66-0.73) and sensitivity (0.54 vs 0.35). Models were applied to the claims database to identify cohorts of 33 001 and 3198 women with ER+/HER2- ESBC and ASBC.

CONCLUSION

We conducted a validation study and developed predictive models to identify in a claims database cohorts of women with ER+/HER2- ESBC and ASBC. The models identified large cohorts in the claims data that can be used to characterize indications in the evaluation of targeted therapies.

Accuracy of administrative databases in detecting primary breast cancer diagnoses: a systematic review

OBJECTIVE

To define the accuracy of administrative datasets to identify primary diagnoses of breast cancer based on the International Classification of Diseases (ICD) 9th or 10th revision codes.

DESIGN

Systematic review.

DATA SOURCES

MEDLINE, EMBASE, Web of Science and the Cochrane Library (April 2017).

ELIGIBILITY CRITERIA

The inclusion criteria were: (a) the presence of a reference standard; (b) the presence of at least one accuracy test measure (eg, sensitivity) and (c) the use of an administrative database.

DATA EXTRACTION

Eligible studies were selected and data extracted independently by two reviewers; quality was assessed using the Standards for Reporting of Diagnostic accuracy criteria.

DATA ANALYSIS

Extracted data were synthesised using a narrative approach.

RESULTS

From 2929 records screened 21 studies were included (data collection period between 1977 and 2011). Eighteen studies evaluated ICD-9 codes (11 of which assessed both invasive breast cancer (code 174.x) and carcinoma in situ (ICD-9 233.0)); three studies evaluated invasive breast cancer-related ICD-10 codes. All studies except one considered incident cases.The initial algorithm results were: sensitivity ≥80% in 11 of 17 studies (range 57%-99%); positive predictive value was ≥83% in 14 of 19 studies (range 15%-98%) and specificity ≥98% in 8 studies. The combination of the breast cancer diagnosis with surgical procedures, chemoradiation or radiation therapy, outpatient data or physician claim may enhance the accuracy of the algorithms in some but not all circumstances. Accuracy for breast cancer based on outpatient or physician's data only or breast cancer diagnosis in secondary position diagnosis resulted low.

CONCLUSION

Based on the retrieved evidence, administrative databases can be employed to identify primary breast cancer. The best algorithm suggested is ICD-9 or ICD-10 codes located in primary position.

TRIAL REGISTRATION NUMBER

CRD42015026881.

Development and Application of a Machine Learning Approach to Assess Short-term Mortality Risk Among Patients With Cancer Starting Chemotherapy

IMPORTANCE

Patients with cancer who die soon after starting chemotherapy incur costs of treatment without the benefits. Accurately predicting mortality risk before administering chemotherapy is important, but few patient data-driven tools exist.

OBJECTIVE

To create and validate a machine learning model that predicts mortality in a general oncology cohort starting new chemotherapy, using only data available before the first day of treatment.

DESIGN, SETTING, AND PARTICIPANTS

This retrospective cohort study of patients at a large academic cancer center from January 1, 2004, through December 31, 2014, determined date of death by linkage to Social Security data. The model was derived using data from 2004 through 2011, and performance was measured on nonoverlapping data from 2012 through 2014. The analysis was conducted from June 1 through August 1, 2017. Participants included 26 946 patients starting 51 774 new chemotherapy regimens.

MAIN OUTCOMES AND MEASURES

Thirty-day mortality from the first day of a new chemotherapy regimen. Secondary outcomes included model discrimination by predicted mortality risk decile among patients receiving palliative chemotherapy, and 180-day mortality from the first day of a new chemotherapy regimen.

RESULTS

Among the 26 946 patients included in the analysis, mean age was 58.7 years (95% CI, 58.5-58.9 years); 61.1% were female (95% CI, 60.4%-61.9%); and 86.9% were white (95% CI, 86.4%-87.4%). Thirty-day mortality from chemotherapy start was 2.1% (95% CI, 1.9%-2.4%). Among the 9114 patients in the validation set, the most common primary cancers were breast (21.1%; 95% CI, 20.2%-21.9%), colorectal (19.3%; 95% CI, 18.5%-20.2%), and lung (18.0%; 95% CI, 17.2%-18.8%). Model predictions were accurate for all patients (area under the curve [AUC], 0.940; 95% CI, 0.930-0.951). Predictions for patients starting palliative chemotherapy (46.6% of regimens; 95% CI, 45.8%-47.3%), for whom prognosis is particularly important, remained highly accurate (AUC, 0.924; 95% CI, 0.910-0.939). To illustrate model discrimination, patients were ranked initiating palliative chemotherapy by model-predicted mortality risk, and observed mortality was calculated by risk decile. Thirty-day mortality in the highest-risk decile was 22.6% (95% CI, 19.6%-25.6%); in the lowest-risk decile, no patients died. Predictions remained accurate across all primary cancers, stages, and chemotherapies, even for clinical trial regimens that first appeared in years after the model was trained (AUC, 0.942; 95% CI, 0.882-1.000). The same model also performed well for prediction of 180-day mortality (AUC for all patients, 0.870 [95% CI, 0.862-0.877]; highest- vs lowest-risk decile mortality, 74.8% [95% CI, 72.7%-77.0%] vs 0.2% [95% CI, 0.01%-0.4%]). Predictions were more accurate than estimates from randomized clinical trials of individual chemotherapies or the Surveillance, Epidemiology, and End Results data set.

CONCLUSIONS AND RELEVANCE

A machine learning algorithm using electronic health record data accurately predicted short-term mortality among patients starting chemotherapy. Further research is necessary to determine the generalizability and feasibility of applying this algorithm in clinical settings.

Development of Query Strategies to Identify a Histologic Lymphoma Subtype in a Large Linked Database System

Background

Large linked databases (LLDB) represent a novel resource for cancer outcomes research. However, accurate means of identifying a patient population of interest within these LLDBs can be challenging. Our research group developed a fully integrated platform that provides a means of combining independent legacy databases into a single cancer-focused LLDB system. We compared the sensitivity and specificity of several SQL-based query strategies for identifying a histologic lymphoma subtype in this LLDB to determine the most accurate legacy data source for identifying a specific cancer patient population.

Methods

Query strategies were developed to identify patients with follicular lymphoma from a LLDB of cancer registry data, electronic medical records (EMR), laboratory, administrative, pharmacy, and other clinical data. Queries were performed using common diagnostic codes (ICD-9), cancer registry histology codes (ICD-O), and text searches of EMRs. We reviewed medical records and pathology reports to confirm each diagnosis and calculated the sensitivity and specificity for each query strategy.

Results

Together the queries identified 1538 potential cases of follicular lymphoma. Review of pathology and other medical reports confirmed 415 cases of follicular lymphoma, 300 pathology-verified and 115 verified from other medical reports. The query using ICD-O codes was highly specific (96%). Queries using text strings varied in sensitivity (range 7–92%) and specificity (range 86–99%). Queries using ICD-9 codes were both less sensitive (34–44%) and specific (35–87%).

Conclusions

Queries of linked-cancer databases that include cancer registry data should utilize ICD-O codes or employ structured free-text searches to identify patient populations with a precise histologic diagnosis.

The Effect of False-positive Mammograms on Antidepressant and Anxiolytic Initiation

BACKGROUND

Despite reported increases in anxiety following a false-positive mammogram, there is little evidence the effect rises to the clinical level of initiating medication.

OBJECTIVE

To analyze the effect of a false-positive mammogram on antidepressant or anxiolytic initiation and identify subpopulations most at risk.

SUBJECTS

MarketScan commercial and Medicaid claims databases used to identify women ages 40-64 undergoing screening mammography with no prior antidepressant or anxiolytic claims.

RESEARCH DESIGN

Using a retrospective cohort design, we estimated the effects of a false-positive relative to a negative mammogram on the likelihood of initiating antidepressants or anxiolytics using multivariate logistic models estimated separately by insurance type.

RESULTS

At 3 months after a false-positive mammogram, the relative risk (RR) for antidepressant or anxiolytic initiation was 1.19 [95% confidence interval (CI), 1.06-1.31] for the commercially insured and 1.13 (95% CI, 0.96-1.29) in the Medicaid population. In addition, 4 subgroups were at particularly elevated risk: commercially insured women ages 40-49 (RR=1.33; 95% CI, 1.13-1.54) or whose false-positive required multiple tests to resolve (RR=1.37; 95% CI, 1.17-1.57), included a biopsy (RR=1.68; 95% CI, 1.18-2.17), or whose resolution took >1 week (RR=1.21; 95% CI, 1.07-1.34).

CONCLUSIONS

False-positive mammograms were associated with significant increases in antidepressant or anxiolytic imitation among the commercially insured. Follow-up resources may be particularly beneficial for cases taking longer to resolve and involving biopsies or multiple tests. The results highlight the need to resolve false-positives quickly and effectively and to monitor depressive symptoms following a positive result.

Sensitivity of Medicare Claims to Identify Cancer Recurrence in Elderly Colorectal and Breast Cancer Patients

BACKGROUND

Researchers are increasingly interested in using observational data to evaluate cancer outcomes following treatment, including cancer recurrence and disease-free survival. Because population-based cancer registries do not collect recurrence data, recurrence is often imputed from health claims, primarily by identifying later cancer treatments after initial treatment. The validity of this approach has not been established.

RESEARCH DESIGN

We used the linked Surveillance, Epidemiology, and End Results-Medicare data to assess the sensitivity of Medicare claims for cancer recurrence in patients very likely to have had a recurrence. We selected newly diagnosed stage II/III colorectal (n=6910) and female breast cancer (n=3826) patients during 1994-2003 who received initial cancer surgery, had a treatment break, and then died from cancer in 1994-2008. We reviewed all claims from the treatment break until death for indicators of recurrence. We focused on additional cancer treatment (surgery, chemotherapy, radiation therapy) as the primary indicator, and used multivariate logistic regression analysis to evaluate patient factors associated with additional treatment. We also assessed metastasis diagnoses and end-of-life care as recurrence indicators.

RESULTS

Additional treatment was the first indicator of recurrence for 38.8% of colorectal patients and 35.2% of breast cancer patients. Patients aged 70 and older were less likely to have additional treatment (P < 0.05), in adjusted analyses. Over 20% of patients either had no recurrence indicator before death or had end-of-life care as their first indicator.

CONCLUSIONS

Identifying recurrence through additional cancer treatment in Medicare claims will miss a large percentage of patients with recurrences; particularly those who are older.

Early death after discharge from emergency departments: analysis of national US insurance claims data

OBJECTIVE

 To measure incidence of early death after discharge from emergency departments, and explore potential sources of variation in risk by measurable aspects of hospitals and patients.

DESIGN

 Retrospective cohort study.

SETTING

 Claims data from the US Medicare program, covering visits to an emergency department, 2007-12.

PARTICIPANTS

 Nationally representative 20% sample of Medicare fee for service beneficiaries. As the focus was on generally healthy people living in the community, patients in nursing facilities, aged ≥90, receiving palliative or hospice care, or with a diagnosis of a life limiting illnesses, either during emergency department visits (for example, myocardial infarction) or in the year before (for example, malignancy) were excluded.

MAIN OUTCOME MEASURE

 Death within seven days after discharge from the emergency department, excluding patients transferred or admitted as inpatients.

RESULTS

 Among discharged patients, 0.12% (12 375/10 093 678, in the 20% sample over 2007-12) died within seven days, or 10 093 per year nationally. Mean age at death was 69. Leading causes of death on death certificates were atherosclerotic heart disease (13.6%), myocardial infarction (10.3%), and chronic obstructive pulmonary disease (9.6%). Some 2.3% died of narcotic overdose, largely after visits for musculoskeletal problems. Hospitals in the lowest fifth of rates of inpatient admission from the emergency department had the highest rates of early death (0.27%)-3.4 times higher than hospitals in the highest fifth (0.08%)-despite the fact that hospitals with low admission rates served healthier populations, as measured by overall seven day mortality among all comers to the emergency department. Small increases in admission rate were linked to large decreases in risk. In multivariate analysis, emergency departments that saw higher volumes of patients (odds ratio 0.84, 95% confidence interval 0.81 to 0.86) and those with higher charges for visits (0.75, 0.74 to 0.77) had significantly fewer deaths. Certain diagnoses were more common among early deaths compared with other emergency department visits: altered mental status (risk ratio 4.4, 95% confidence interval 3.8 to 5.1), dyspnea (3.1, 2.9 to 3.4), and malaise/fatigue (3.0, 2.9 to 3.7).

CONCLUSIONS

 Every year, a substantial number of Medicare beneficiaries die soon after discharge from emergency departments, despite no diagnosis of a life limiting illnesses recorded in their claims. Further research is needed to explore whether these deaths were preventable.

Validating a proxy for disease progression in metastatic cancer patients using prescribing and dispensing data

BACKGROUND

Routine data collections are used increasingly to examine outcomes of real-world cancer drug use. These datasets lack clinical details about important endpoints such as disease progression.

AIM

To validate a proxy for disease progression in metastatic cancer patients using prescribing and dispensing claims.

METHODS

We used data from a cohort study of patients undergoing chemotherapy who provided informed consent to the collection of cancer-treatment data from medical records and linkage to pharmaceutical claims. We derived proxy decision rules based on changes to drug treatment in prescription histories (n = 36 patients) and validated the proxy in prescribing data (n = 62 patients). We adapted the decision rules and validated the proxy in dispensing data (n = 109). Our gold standard was disease progression ascertained in patient medical records. Individual progression episodes were the unit of analysis for sensitivity and Positive Predictive Value (PPV) calculations and specificity and Negative Predictive Value (NPV) were calculated at the patient level.

RESULTS

The sensitivity of our proxy in prescribing data was 74.3% (95% Confidence Interval (CI), 55.6-86.6%) and PPV 61.2% (95% CI, 45.0-75.3%); specificity and NPV were 87.8% (95% CI, 73.8-95.9%) and 100% (95% CI, 90.3-100%), respectively. In dispensing data, the sensitivity of our proxy was 64% (95% CI, 55.0-77.0%) and PPV 56.0% (95% CI, 43.0-69.0%); specificity and NPV were 81% (95% CI, 70.05-89.0%) and 91.0% (95% CI, 82.0-97.0%), respectively.

CONCLUSION

Our proxy overestimated episodes of disease progression. The proxy's performance is likely to improve if the date of prescribing is used instead of date of dispensing in claims data and by incorporating medical service claims (such as imaging prior to drug changes) in the algorithm. Our proxy is not sufficiently robust for use in real world comparative effectiveness research for cancer medicines.

Care in the Final Month of Life among Adolescent and Young Adult Cancer Patients in Kaiser Permanente Southern California

BACKGROUND

Little is known about the care that adolescent and young adult (AYA) patients with cancer receive at the end of life (EOL).

OBJECTIVE

To examine care in the last month of life among AYA patients with cancer.

DESIGN

Medical record review of the last 30 days of life.

SETTING/SUBJECTS

One hundred eleven AYA patients aged 15-39 years at death with either stage I-III cancer and evidence of cancer recurrence or stage IV cancer at diagnosis. Patients received care in Kaiser Permanente Southern California, an integrated healthcare delivery system, and died from 2007 to 2010.

MEASUREMENTS

Use of intensive measures, including chemotherapy in the last 14 days of life and emergency room visits, hospitalizations, and intensive care unit admissions in the last 30 days; documented care preferences; symptom prevalence and treatment; advance care planning; hospice use; and location of death.

RESULTS

One hundred seven patients (96%) had documented care preferences in the last month of life. At first documentation, 72% of patients wished for life-prolonging care, 20% wished for care focused on comfort, and 8% were undecided. Forty-seven percent of patients had documented changes in preferences in the last month, with 40% wishing for life-prolonging care when preferences were last noted before death, 56% preferring comfort, and 4% undecided. Seventy-eight percent of patients received at least one form of intensive EOL care, including 75% of those who preferred comfort measures at last documentation.

CONCLUSIONS

Many AYA patients enter the last month of life wishing for life-prolonging care. While most ultimately wish for comfort, intensive care is prevalent even among such patients.

Variation in geographic access to chemotherapy by definitions of providers and service locations: a population-based observational study

Background

An aging population, with its associated rise in cancer incidence and strain on the oncology workforce, will continue to motivate patients, healthcare providers and policy makers to better understand the existing and growing challenges of access to chemotherapy. Administrative data, and SEER-Medicare data in particular, have been used to assess patterns of healthcare utilization because of its rich information regarding patients, their treatments, and their providers. To create measures of geographic access to chemotherapy, patients and oncologists must first be identified. Others have noted that identifying chemotherapy providers from Medicare claims is not always straightforward, as providers may report multiple or incorrect specialties and/or practice in multiple locations. Although previous studies have found that specialty codes alone fail to identify all oncologists, none have assessed whether various methods of identifying chemotherapy providers and their locations affect estimates of geographic access to care.

Methods

SEER-Medicare data was used to identify patients, physicians, and chemotherapy use in this population-based observational study. We compared two measures of geographic access to chemotherapy, local area density and distance to nearest provider, across two definitions of chemotherapy provider (identified by specialty codes or billing codes) and two definitions of chemotherapy service location (where chemotherapy services were proven to be or possibly available) using descriptive statistics. Access measures were mapped for three representative registries.

Results

In our sample, 57.2 % of physicians who submitted chemotherapy claims reported a specialty of hematology/oncology or medical oncology. These physicians were associated with 91.0 % of the chemotherapy claims. When providers were identified through billing codes instead of specialty codes, an additional 50.0 % of beneficiaries (from 23.8 % to 35.7 %) resided in the same ZIP code as a chemotherapy provider. Beneficiaries were also 1.3 times closer to a provider, in terms of driving time. Our access measures did not differ significantly across definitions of service location.

Conclusions

Measures of geographic access to care were sensitive to definitions of chemotherapy providers; far more providers were identified through billing codes than specialty codes. They were not sensitive to definitions of service locations, as providers, regardless of how they are identified, generally provided chemotherapy at each of their practice locations.

Using linked electronic data to validate algorithms for health outcomes in administrative databases

The validity of algorithms used to identify health outcomes in claims-based and administrative data is critical to the reliability of findings from observational studies. The traditional approach to algorithm validation, using medical charts, is expensive and time-consuming. An alternative method is to link the claims data to an external, electronic data source that contains information allowing confirmation of the event of interest. In this paper, we describe this external linkage validation method and delineate important considerations to assess the feasibility and appropriateness of validating health outcomes using this approach. This framework can help investigators decide whether to pursue an external linkage validation method for identifying health outcomes in administrative/claims data.

Improving treatment and survival: a population-based study of current outcomes after a hepatic resection in patients with metastatic colorectal cancer

BACKGROUND

Population-based studies historically report underutilization of a resection in patients with colorectal metastases to the liver. Recent data suggest limitations of the methods in the historical analysis. The present study examines trends in a hepatic resection and survival among Medicare recipients with hepatic metastases.

METHODS

Medicare recipients with incident colorectal cancer diagnosed between 1991 and 2009 were identified in the SEER(Surveillance, Epidemiology and End Results)-Medicare dataset. Patients were stratified into historical (1991-2001) and current (2002-2009) cohorts. Analyses compared treatment, peri-operative outcomes and survival.

RESULTS

Of 31.574 patients with metastatic colorectal cancer to the liver, 14,859 were in the current cohort treated after 2002 and 16,715 comprised the historical control group. The overall proportion treated with a hepatic resection increased significantly during the study period (P < 0.001) with pre/post change from 6.5% pre-2002 to 7.5% currently (P < 0.001). Over time, haemorrhagic and infectious complications declined (both P ≤ 0.047), but 30-day mortality was similar (3.5% versus 3.9%, P = 0.660). After adjusting for predictors of survival, the use of a hepatic resection [hazard ratio (HR) = 0.40, 95% confidence interval (CI): 0.38-0.42, P < 0.001] and treatment after 2002 (HR = 0.88, 95% CI: 0.86-0.90, P < 0.001) were associated with a reduced risk of death.

CONCLUSIONS

Case identification using International Classification of Diseases, 9th Revision (ICD-9) codes is imperfect; however, comparison of trends over time suggests an improvement in multimodality therapy and survival in patients with colorectal metastases to the liver.

Longitudinal Trends in Costs of Palliative Radiation for Metastatic Prostate Cancer

BACKGROUND

In recent years, palliative treatment of prostate cancer metastases has been characterized by the use of more complex radiation treatment, despite a lack of evidence demonstrating a clinical benefit of these technologies in the palliative setting. The impact of adoption of these technologies on the costs of palliative radiation treatment in patients with metastatic prostate cancer remains poorly understood in the general patient population.

METHODS

The study was a retrospective analysis of Surveillance, Epidemiology and End Results (SEER) Medicare data of men aged 66 and older who died from metastatic prostate cancer between 2000 and 2007 and received radiation therapy for bony metastases in the last year of life. Direct costs were obtained from Medicare carrier and outpatient facility payments for all radiation treatment claims and adjusted to 2008 dollars.

RESULTS

A total of 1705 men met study inclusion criteria. Total Medicare payments for radiation therapy for bony metastases in the last year of life increased by 44.4% from an average of $2,763 in 2000 to $3,989 in 2007, with the proportion of all payments accrued within hospital-based settings increasing from 48% to 57%. Complexity of radiation therapy techniques over the same period was characterized by use of less simple (30.1% to 23.3%) and more complex (59.9% versus 66.7%) radiation therapy. From 2000-2003 to 2004-2007, the use of shorter treatment courses (≤5 fractions) decreased from 22% to 14%, and the use of single fraction treatment courses decreased by half (6.3% to 2.9%; P≤.001).

CONCLUSIONS

Between 2000 and 2007, palliative radiation therapy for bony prostate cancer metastases was characterized by the use of more advanced treatment technologies and prolonged radiation treatment courses. Further research investigating barriers to cost-effective palliation is warranted.

End-of-Life Care Intensity Among Adolescent and Young Adult Patients With Cancer in Kaiser Permanente Southern California

IMPORTANCE

Cancer is the leading disease-related cause of death among adolescents and young adults (AYAs), but little is known about the care that AYA patients with cancer receive at the end of life (EOL).

OBJECTIVE

To evaluate the intensity of EOL care among AYA patients with cancer.

DESIGN, SETTING, AND PARTICIPANTS

Cross-sectional study of Kaiser Permanente Southern California (KSPC) cancer registry data and electronic health records for 663 AYA patients with either stage I to III cancer and evidence of cancer recurrence or stage IV cancer at diagnosis. All patients were treated within KSPC, an integrated health care delivery system, and died between 2001 and 2010 before age 40 years (age range at time of death, 15-39 years).

MAIN OUTCOMES AND MEASURES

(1) Chemotherapy use in the last 14 days of life; (2) intensive care unit (ICU) care in the last 30 days of life; (3) more than 1 emergency department (ED) visit in the last 30 days of life; (4) hospitalization in the last 30 days of life; and (5) a composite measure of medically intensive EOL care including any of the aforementioned measures.

RESULTS

Eleven percent of patients (72 of 663) received chemotherapy within 14 days of death. In the last 30 days of life, 22% of patients (144 of 663) were admitted to the ICU; 22% (147 of 663) had more than 1 ED visit; and 62% (413 of 663) were hospitalized. Overall, 68% of patients (449 of 663) received at least 1 medically intensive EOL care measure.

CONCLUSIONS AND RELEVANCE

Most AYA patients received at least 1 form of medically intensive EOL care. These findings suggest the need to better understand EOL care preferences and decision making in this young population.

Challenges and opportunities in measuring cancer recurrence in the United States

Cancer recurrence and disease-free survival are key outcomes for measuring the burden of illness, assessing the quality of cancer care, and informing decisions about increasingly costly cancer therapies. Yet information about recurrence is not collected in cancer registries or other population-based data sources. To address the lack of population-based recurrence information, researchers are increasingly using algorithms applied to health claims to infer recurrence. However, the validity of these approaches has not been comprehensively evaluated. In this commentary, we review existing studies and discuss options for improving the availability of recurrence data. We found that the validity of claims-based approaches appears promising in small, single institution studies, but larger population-based studies have identified substantial limitations with using claims to identify recurrence. With the increasing availability of health data, there are potential options that can be implemented to enhance information about recurrence. These options include design of software for the electronic medical record that enables rapid and standardized reporting of recurrence, use of electronic pathology reports to facilitate streamlined collection of recurrence by cancer registries, and mandates by insurers to require reporting of recurrence on health claims submitted by physicians. All of these options will require that governmental agencies, health insurers, professional societies, and other groups recognize the importance of population-based recurrence data and determine that this information is a priority for assessing cancer outcomes and costs.

The Improving Rural Cancer Outcomes (IRCO) Trial: a factorial cluster-randomised controlled trial of a complex intervention to reduce time to diagnosis in rural patients with cancer in Western Australia: a study protocol

INTRODUCTION

While overall survival for most common cancers in Australia is improving, the rural-urban differential has been widening, with significant excess deaths due to lung, colorectal, breast and prostate cancer in regional Australia. Internationally a major focus on understanding variations in cancer outcomes has been later presentation to healthcare and later diagnosis. Approaches to reducing time to diagnosis of symptomatic cancer include public symptom awareness campaigns and interventions in primary care to improve early cancer detection. This paper reports the protocol of a factorial cluster-randomised trial of community and general practice (GP) level interventions to reduce the time to diagnosis of cancer in rural Western Australia (WA).

METHODS AND ANALYSIS

The community intervention is a symptom awareness campaign tailored for rural Australians delivered through a community engagement model. The GP intervention includes a resource card with symptom risk assessment charts and local referral pathways implemented through multiple academic detailing visits and case studies. Participants are eligible if recently diagnosed with breast, colorectal, lung or prostate cancer who reside in specific regions of rural WA with a planned sample size of 1350. The primary outcome is the Total Diagnostic Interval, defined as the duration from first symptom (or date of cancer screening test) to cancer diagnosis. Secondary outcomes include cancer stage, healthcare utilisation, disease-free status, survival at 2 and 5 years and cost-effectiveness.

ETHICS AND DISSEMINATION

Ethics approval has been granted by the University of Western Australia and from all relevant hospital recruitment sites in WA.

RESULTS

Results of this trial will be reported in peer-reviewed publications and in conference presentations.

TRIAL REGISTRATION NUMBER

Australian New Zealand Clinical Trials Registry (ANZCTR). ACTRN12610000872033.

Limited validity of diagnosis codes in Medicare claims for identifying cancer metastases and inferring stage

PURPOSE

Researchers are using diagnosis codes from health claims to identify metastatic disease in cancer patients. The validity of this approach has not been established.

METHODS

We used the linked 2005-2007 Surveillance, Epidemiology and End Results (SEER)-Medicare data to assess the validity of metastasis codes at diagnosis from claims compared with stage reported by SEER cancer registries. The cohort included 80,052 incident breast, lung, and colorectal cancer patients aged 65 years and older. Using gold-standard SEER data, we evaluated sensitivity, specificity, positive predictive value, and negative predictive value of claims-based stage, survival by stage classification, and patient factors associated with stage misclassification using multivariable regression.

RESULTS

For patients with a registry report of distant metastatic cancer, the sensitivity, specificity, and positive predictive value of claims never simultaneously exceeded 80% for any cancer: lung (42.7%, 94.8%, and 88.1%), breast (51.0%, 98.3%, and 65.8%), and colorectal (72.8%, 93.8%, and 68.5%). Misclassification of stage from Medicare claims was significantly associated with inaccurate estimates of stage-specific survival (P < .001). In adjusted analysis, patients who were older, black, or living in low-income areas were more likely to have their stage misclassified in claims.

CONCLUSIONS

Diagnosis codes in Medicare claims have limited validity for inferring cancer stage and metastatic disease.

Incidence and predictors of bowel obstruction in elderly patients with stage IV colon cancer: a population-based cohort study

IMPORTANCE

Research has been limited on the incidence, mechanisms, etiology, and treatment of symptoms that require palliation in patients with terminal cancer. Bowel obstruction (BO) is a common complication of advanced abdominal cancer, including colon cancer, for which small, single-institution studies have suggested an incidence rate of 15% to 29%. Large population-based studies examining the incidence or risk factors associated with BO in cancer are lacking.

OBJECTIVE

To investigate the incidence and risk factors associated with BO in patients with stage IV colon cancer.

DESIGN AND SETTING

Retrospective cohort, population-based study of patients in the Surveillance, Epidemiology, and End Results and Medicare claims linked databases who were diagnosed as having stage IV colon cancer from January 1, 1991, through December 31, 2005.

PATIENTS

Patients 65 years or older with stage IV colon cancer (n = 12 553).

MAIN OUTCOMES AND MEASURES

Time to BO, defined by inpatient hospitalization for BO. We used Cox proportional hazards regression models to determine associations between BO and patient, prior treatment, and tumor features.

RESULTS

We identified 1004 patients with stage IV colon cancer subsequently hospitalized with BO (8.0%). In multivariable analysis, proximal tumor site (hazard ratio, 1.22 [95% CI, 1.07-1.40]), high tumor grade (1.34 [1.16-1.55]), mucinous histological type (1.27 [1.08-1.50]), and nodal stage N2 (1.52 [1.26-1.84]) were associated with increased risk of BO, as was the presence of obstruction at cancer diagnosis (1.75 [1.47-2.04]). A more recent diagnosis was associated with decreased risk of subsequent obstruction (hazard ratio, 0.84 [95% CI, 0.72-0.98]).

CONCLUSIONS AND RELEVANCE

In this large population of patients with stage IV colon cancer, BO after diagnosis was less common (8.0%) than previously reported. Risk was associated with site and histological type of the primary tumor. Future studies will explore management and outcomes in this serious, common complication.

Out-of-pocket health care expenditure burden for Medicare beneficiaries with cancer

BACKGROUND

There is increasing concern regarding the financial burden of care on cancer patients and their families. Medicare beneficiaries often have extensive comorbidities and limited financial resources, and may face substantial cost sharing even with supplemental coverage. In the current study, the authors examined out-of-pocket (OOP) spending and burden relative to income for Medicare beneficiaries with cancer.

METHODS

This retrospective, observational study pooled data for 1997 through 2007 from the Medicare Current Beneficiary Survey linked to Medicare claims. Medicare beneficiaries with newly diagnosed cancer were selected using claims-based diagnoses. Generalized linear models were used to estimate OOP spending. Logistic regression models identified factors associated with a high OOP burden, defined as spending > 20% of one's income during the cancer diagnosis and subsequent year.

RESULTS

The cohort included 1868 beneficiaries with and 10,047 without cancer. Compared with the noncancer cohort, cancer patients were older, had more comorbidities, and were more likely to lack supplemental coverage. The mean OOP spending for cancer patients was $4727. Cancer patients faced an adjusted $976 (P < .01) incremental OOP spending. Greater than one-quarter (28%) of beneficiaries with cancer experienced a high OOP burden compared with 16% of beneficiaries without cancer (P < .001). Supplemental insurance and higher income were found to be protective against a high OOP burden, whereas assets, comorbidity, and receipt of cancer-directed radiation and antineoplastic therapy were associated with a higher OOP burden.

CONCLUSIONS

Medicare beneficiaries with cancer face a higher OOP burden than their counterparts without cancer; some of the higher burden was explained by the higher comorbidity burden and lack of supplemental insurance noted among these patients. Financial pressures may discourage some elderly patients from pursuing treatment.

The economic burden of metastatic breast cancer: a U.S. managed care perspective

This study was performed to quantify the economic burden and identify drivers of direct costs of mBC. In a retrospective study of a de-identified administrative claims database of privately insured patients, women between 18 and 64 years of age were included if they had at least one claim with a diagnosis of breast cancer and subsequently one or more claims with a diagnosis of secondary malignancy between January 1, 2003 and December 31, 2009. The study sample was further classified into the following subgroups: (1) Endocrine therapy, (2) HER-2 targeted therapy, (3) Concomitant HER-2 targeted and endocrine therapy, (4) Cytotoxic chemotherapy, and (5) No-systemic therapy. Costs for medical resource utilization were calculated on a per patient per month (PPPM) basis. A total of 7,698 mBC patients were identified from 2003 to 2009 with an average age at index of ~52 years, and average follow up of 2.2 years. The average total direct medical costs for 7,698 mBC patients were $9,788 PPPM. Outpatient costs accounted for the majority of overall PPPM costs. Examining the five different mBC therapeutic subgroups revealed that patients who received no-systemic therapy had the highest costs at $13,926 PPPM, while patients who received systemic endocrine therapy had the lowest costs at $5,303 PPPM. This study demonstrated that mBC is associated with substantial healthcare costs in a non-Medicare patient population. Assuming average PPPM costs of $9,788 and an average life expectancy of 2.2 years, the total average expenditure to treat mBC would be ~$250,000 per patient.

Hospital morbidity data for determining spread of disease at diagnosis for colorectal cancer: a validation study

AIMS

There is currently no routine collection of cancer stages in population-based data in Australia. This study evaluates the accuracy of International classification of diseases (ICD) codes for secondary neoplasms recorded in hospital morbidity data to assign spread of disease at diagnosis for colorectal cancer.

METHODS

The reference (gold) standard was the Australian clinicopathological stage (ACPS) documented by a treating colorectal surgeon and derived from histopathology and clinical findings. To allow comparison with stages derived from the hospital morbidity data (HMD), ACPS was mapped to the spread of disease (local, regional and distant). Sensitivity, specificity and positive-predictive values were calculated to compare the accuracy of stage derived from HMD.

RESULTS

Data from both the reference standard and HMD were available for 499 patients. HMD slightly overestimated patients with local disease (62.3 vs 56.9%). There was a corresponding underestimation of regional and distant spread of disease. While sensitivity for regional and distant disease was moderate (66.4 and 71.4%, respectively), specificity was high (92.7 and 96.6%, respectively).

CONCLUSION

ICD codes for secondary neoplasms in HMD are limited in their utility for determining the spread of disease for colorectal cancer. Clinicians need to ensure that clinical coders are provided with enough information to accurately code for spread of disease. We recommend reporting histopathology in a synoptic format which includes background information on the presence or absence of distant metastasis and the tumor node metastasis stage.

Pricing of surgeries for colon cancer: patient severity and market factors

BACKGROUND

This study examined effects of health maintenance organization (HMO) penetration, hospital competition, and patient severity on the uptake of laparoscopic colectomy and its price relative to open surgery for colon cancer.

METHODS

The MarketScan Database (data from 2002-2007) was used to identify admissions for privately insured colorectal cancer patients undergoing laparoscopic or open partial colectomy (n = 1035 and n = 6389, respectively). Patient and health plan characteristics were retrieved from these data; HMO market penetration rates and an index of hospital market concentration, the Herfindahl-Hirschman index (HHI), were derived from national databases. Logistic and logarithmic regressions were used to examine the odds of having laparoscopic colectomy, effect of covariates on colectomy prices, and the differential price of laparoscopy.

RESULTS

Adoption of laparoscopy was highly sensitive to market forces, with a 10% increase in HMO penetration leading to a 10.9% increase in the likelihood of undergoing laparoscopic colectomy (adjusted odds ratio = 1.109; 95% confidence interval [CI] = 1.062, 1.158) and a 10% increase in HHI resulting in 6.6% lower likelihood (adjusted odds ratio = 0.936; 95% CI = 0.880, 0.996). Price models indicated that the price of laparoscopy was 7.6% lower than that of open surgery (transformed coefficient = 0.927; 95% CI = 0.895, 0.960). A 10% increase in HMO penetration was associated with 1.6% lower price (transformed coefficient = 0.985; 95% CI = 0.977, 0.992), whereas a 10% increase in HHI was associated with 1.6% higher price (transformed coefficient = 1.016; 95% CI = 1.006, 1.027; P < .001 for all comparisons).

CONCLUSIONS

Laparoscopy was significantly associated with lower hospital prices. Moreover, laparoscopic surgery may result in cost savings, while market pressures contribute to its adoption.

Surgery wait times and specialty services for insured and uninsured breast cancer patients: does hospital safety net status matter?

OBJECTIVE

To determine whether safety net and non-safety net hospitals influence inpatient breast cancer care in insured and uninsured women and in white and African American women.

DATA SOURCES

Six years of Virginia Cancer Registry and Virginia Health Information discharge data were linked and supplemented with American Hospital Association data.

STUDY DESIGN

Hierarchical generalized linear models and linear probability regression models were used to estimate the relationship between hospital safety net status, the explanatory variables, and the days from diagnosis to mastectomy and the likelihood of breast reconstruction.

PRINCIPAL FINDINGS

The time between diagnosis and surgery was longer in safety net hospitals for all patients, regardless of insurance source. Medicaid insured and uninsured women were approximately 20 percent less likely to receive reconstruction than privately insured women. African American women were less likely to receive reconstruction than white women.

CONCLUSIONS

Following the implementation of health reform, disparities may potentially worsen if safety net hospitals' burden of care increases without commensurate increases in reimbursement and staffing levels. This study also suggests that Medicaid expansions may not improve outcomes in inpatient breast cancer care within the safety net system.

Insurance and inpatient care: differences in length of stay and costs between surgically treated cancer patients

BACKGROUND

Early research demonstrated that patients' length-of-stay and inpatient costs varied according to their health insurance status. The authors of the current report studied a population-based sample of privately insured, Medicaid-insured, and uninsured inpatients ages 21 to 64 years who underwent surgical resection for either nonsmall cell lung cancer (NSCLC) (n = 781) or colorectal cancer (CRC) (n = 8190) or who underwent mastectomy (n = 6201) to compare length of stay and inpatient costs by insurance status.

METHODS

Data for this study were derived from all civilian, acute-care hospitals in Virginia from 1999 to 2005. Hierarchical generalized linear models were used to estimate the relation between the explanatory variables and lengths of stay and costs. All analyses controlled for patient characteristics and hospital random effects.

RESULTS

Medicaid-insured patients with NSCLC had longer lengths of stay (39% or 2.64 days longer) and higher inpatient costs (20% or $2479 higher costs). Uninsured and Medicaid-insured patients with CRC had longer lengths of stay and higher inpatient costs. In contrast, uninsured patients with breast cancer had 11% shorter lengths of stay and 12% lower inpatient costs than privately insured patients. Medicaid-insured patients had 10% lower inpatient costs than privately insured patients. Differences were no longer statistically significant when reconstruction was added to the models.

CONCLUSIONS

Health insurance affected the need for health care and the amount of health care received. Uninsured and Medicaid-insured patients with lung cancer and colon cancer who underwent resection had longer lengths of stay and higher inpatient costs than privately insured patients, but they had shorter lengths of stay when reconstruction was not provided. Among the patients with breast cancer, patients and/or providers economized on discretionary procedures.

Hospital-activity data inaccurate for determining spread-of-disease at diagnosis for non-small cell lung cancer

OBJECTIVE

Accurate information on spread-of-disease at diagnosis would increase the usefulness of hospital-activity data for cancer research. This study evaluates the accuracy of codes recorded in hospital-activity data to assign spread-of-disease at diagnosis for non-small cell lung cancer (NSCLC).

METHODS

The reference (gold) standard was TNM stage as assigned at a multi-disciplinary meeting. To allow comparison with hospital-activity data, TNM stage was mapped to spread-of-disease (local, regional, distant). Sensitivity, specificity and positive-predictive values were stratified by whether the patient had surgery.

RESULTS

Data from the reference standard and hospital-activity database were available for 2,184 patients. According to the reference standard, local disease was present for 57.0% of surgical patients and 12.6% of non-surgical patients at diagnosis. Hospital-activity data over-estimated patients with local disease (surgical: 71.9%, non-surgical: 48.5%). There was a corresponding underestimation of distant spread-of-disease: surgical (reference standard: 4.0%, hospital-activity data: 2.7%); non-surgical (reference standard: 45.9%, hospital-activity data: 36.8%). This meant that hospital-activity data had good sensitivity but poor specificity for local disease; and poor sensitivity, but good specificity for metastatic disease.

CONCLUSION

Secondary diagnosis codes in hospital activity data do not accurately capture spread-of-disease at diagnosis for patients with non-small cell lung cancer; even when the clinical notes contain TNM clinical stage as documented at a multidisciplinary meeting.

IMPLICATIONS

Changes are needed to coding rules, and the ICD codes themselves, to allow for coding of regional and distant spread without specification of the precise site.

Variation of health-care resource utilization according to GERD-associated complications

Complications associated with gastroesophageal reflux disease (GERD) can include esophageal stricture, Barrett's esophagus, gastrointestinal hemorrhage, and extraesophageal symptoms. The impact of GERD-associated complications on health-care utilization deserves further evaluation. We identified commercial enrollees 18-75 years old with claims for GERD (International Classification of Diseases, Ninth Revision, Clinical Modification Codes: 530.81 or 530.11) and subsequent usage of proton pump inhibitors from 01/01/05 to 06/30/09. The initial GERD diagnosis date was designated as the index date, and patients were studied for 6 months preindex and postindex. Eligible patients were subsequently stratified based on medical claims for GERD-associated complications as follows: stage A (GERD diagnosis, no other symptoms), stage B (GERD + extraesophageal symptoms), stage C (GERD + Barrett's esophagus), stage D (GERD + esophageal stricture), and stage E (GERD + iron-deficiency anemia or acute upper gastrointestinal hemorrhage). Patient characteristics, health-care utilization, and costs were compared between stage A and each stage with complicated GERD (B-D). Of the 174,597 patients who were eligible for analysis, 74% were classified as stage A, 20% stage B, 1% stage C, 2% stage D, and 3% stage E. Relative to stage A, patients in stages C, D, and E were significantly more likely to visit a gastroenterologist (13% vs. 68%, 71%, and 38%, respectively) and had higher rates of esophageal ulcers (0.3% vs. 8%, 5%, and 3%, respectively) and Nissen fundoplication (0.05% vs. 0.6%, 0.3%, and 0.2%, respectively). Six-month GERD-related costs ranged from $615/patient (stage A) to $1714/patient (stage D); all-cause costs ranged from $4195/patient (stage A) to $11,340/patient (stage E). Compared with stage A, all other cohorts had significantly higher all-cause and GERD-related costs (P < 0.0001 for all comparisons). While patients with more severe GERD represented a relatively small portion of the GERD cohort, they demonstrated significantly greater health-care costs and overall utilization than patients with uncomplicated GERD.

Medicaid enrollment at early stage of disease: the Breast and Cervical Cancer Prevention and Treatment Act in Georgia

This study linked data from the Georgia Comprehensive Cancer Registry to Medicaid enrollment and claims to test whether the Breast and Cervical Cancer Prevention and Treatment Act (BCCPTA), which provided a new and quicker pathway to Medicaid eligibility for low-income breast cancer patients, led to more patients enrolling at an early stage of disease. Results based on difference-in-differences analysis indicated that Georgia's BCCPTA increased by 11 percentage points the probability of breast cancer patients enrolling in Medicaid at an early stage (p = .024). This effect could mean more treatment options and higher survival rates for these patients.

Distance to the closest radiotherapy facility and survival after a diagnosis of rectal cancer in Queensland

OBJECTIVE

To determine whether an association exists between distance from radiotherapy facilities and survival outcomes of people diagnosed with rectal cancer.

DESIGN AND SETTING

Descriptive population-based study using data from the Queensland Cancer Registry.

PATIENTS

All patients aged 20-79 years (n = 6848) diagnosed with invasive rectal cancer between 1 January 1996 and 31 December 2006.

MAIN OUTCOME MEASURE

Cause-specific survival.

RESULTS

The 5-year cause-specific survival was 62% (95% CI, 61%-64%); it was strongly influenced by stage at diagnosis (American Joint Committee on Cancer, Stages I-IV), ranging from 86% (Stage I) to 9% (Stage IV). After adjusting for age, sex, and stage at diagnosis, patients who lived 100-199 km, 200-399 km and 400 km or more from a radiotherapy facility were 16%, 30%, and 25%, respectively, more likely to die from rectal cancer than patients living within 50 km of such a facility. On average, there was a 6% increase in mortality risk (95% CI, 3%-8%; P < 0.001) for each 100 km increment in distance from the nearest radiotherapy facility. Shared frailty models showed that this association persisted after adjusting for the correlation between individual cancer patients living in the same remoteness or area-level socioeconomic status categories.

CONCLUSIONS

While centralisation of cancer treatment services has merit, our study provides evidence of a shorter survival for people with rectal cancer who live relatively far from radiotherapy facilities. It remains a priority to develop and implement policy, cultural and clinical measures to reduce the burden faced by rural and remote patients with rectal cancer.

Healthcare costs in patients with metastatic lung cancer receiving chemotherapy

Background

To characterize healthcare resource utilization and costs in patients with metastatic lung cancer receiving chemotherapy in the US.

Methods

Using data from a large private multi-payer health insurance claims database (2000-2006), we identified all patients beginning chemotherapy for metastatic lung cancer. Healthcare resource use (inpatient, outpatient, medications) and costs were tallied over time from date of therapy initiation ("index date") to date of disenrollment from the health plan (in most instances, presumably due to death) or the end of the study period, whichever occurred first. Healthcare utilization and costs were characterized using Kaplan-Meier sample average methods.

Results

The study population consisted of 4068 patients; mean (SD) age was 65 (11) years. Over a median follow-up of 334 days, study subjects averaged 1.5 hospital admissions, 8.9 total inpatient days, and 69 physician office and hospital outpatient visits. Mean (95% CI) cumulative total healthcare costs were $125,849 ($120,228, $131,231). Costs of outpatient medical services and inpatient care constituted 34% and 20% of total healthcare costs, respectively; corresponding estimates for outpatient chemotherapy and other medication were 22% and 24%.

Conclusion

Our study sheds additional light on the burden of metastatic lung cancer among patients receiving chemotherapy, in terms of total cost thru end of life as well as component costs by setting and type of service, and may be useful in informing medical resource allocation in this patient population.

Infused therapy and survival in older patients diagnosed with metastatic breast cancer who received trastuzumab

We used Surveillance, Epidemiology, and End Results-Medicare data (2000-2006) to describe treatment and survival in women diagnosed with metastatic breast cancer (MBC) who received trastuzumab. There were 610 patients with a mean age of 74 years. Overall, 32% received trastuzumab alone and 47% received trastuzumab plus a taxane. In multivariate analysis, trastuzumab plus chemotherapy was associated with a lower adjusted cancer mortality rate (Hazard Ratio [HR] 0.54; 95% Confidence Interval [CI] 0.39-0.74; p < .001) than trastuzumab alone among patients who received trastuzumab as part of first-line therapy. Adding chemotherapy to first-line trastuzumab for metastatic breast cancer is associated with improved cancer survival.

Healthcare costs in women with metastatic breast cancer receiving chemotherapy as their principal treatment modality

Background

The economic costs of treating patients with metastatic breast cancer have been examined in several studies, but available estimates of economic burden are at least a decade old. In this study, we characterize healthcare utilization and costs in the US among women with metastatic breast cancer receiving chemotherapy as their principal treatment modality.

Methods

Using a large private health insurance claims database (2000-2006), we identified all women initiating chemotherapy for metastatic breast cancer with no evidence of receipt of concomitant or subsequent hormonal therapy, or receipt of trastuzumab at anytime. Healthcare utilization and costs (inpatient, outpatient, medication) were estimated on a cumulative basis from date of chemotherapy initiation ("index date") to date of disenrollment from the health plan or the end of the study period, whichever occurred first. Study measures were cumulated over time using the Kaplan-Meier Sample Average (KMSA) method; 95% CIs were generated using nonparametric bootstrapping. Findings also were examined among the subgroup of patients with uncensored data.

Results

The study population consisted of 1444 women; mean (SD) age was 59.1 (12.1) years. Over a mean follow-up of 532 days (range: 3 to 2412), study subjects averaged 1.7 hospital admissions, 10.7 inpatient days, and 83.6 physician office and hospital outpatient visits. Mean (95% CI) cumulative total healthcare costs were $128,556 ($118,409, $137,644) per patient. Outpatient services accounted for 29% of total costs, followed by medication other than chemotherapy (26%), chemotherapy (25%), and inpatient care (20%).

Conclusions

Healthcare costs-especially in the outpatient setting--are substantial among women with metastatic breast cancer for whom treatment options other than chemotherapy are limited.