Customize your JAMA Network experience by selecting one or more topics from the list below.
In 2013, the Institute of Medicine (IOM) released a consensus report entitled “Delivering High-Quality Cancer Care: Charting a New Course for a System in Crisis,”1 which provides a blueprint for improving the quality of cancer care in the United States. In particular, its central focus is the need for delivering patient-centered cancer care and ensuring that patients have an opportunity to receive effective, high-value, and safe treatments that are consistent with their individual needs, values, and preferences.2 The IOM report’s goals and recommendations are hierarchical and move from those perceived to be most achievable—those directly associated with the patient-clinician encounter—to those that are more difficult to accomplish, such as the elimination of disparities in care and comprehensive delivery-system reform, including new payment models. The first 2 recommended goals of the IOM report focus on the importance of engaged patients. Goal 1 emphasizes the critical role of the cancer care team in providing patients and their families with understandable information on cancer prognosis, treatment benefits and harms, palliative care, psychosocial support, and estimates of the total and out-of-pocket costs of cancer care. Goal 2 stresses the importance, in the setting of advanced cancer, of providing patients with end-of-life care consistent with their needs, values, and preferences.1
In this issue of JAMA Oncology, Brooks et al3 describe the development of a clinical prediction model to assess the risk for chemotherapy-related hospitalization (CRH) in patients with advanced cancer initiating palliative chemotherapy. The model was developed using a retrospectively designed nested case-control study drawing on data from a single community hospital clinical registry. The authors make the case for the potential value of information regarding the risk for CRH in this patient population, noting that the ability of oncologists to predict CRH risk with standard indicators (eg, performance status) is poor. They propose that being able to identify the profile of patients most at risk for this clinically adverse event may “improve the chemotherapy informed-consent process, allow for modification of treatment regimens to reduce the risk of toxic effects, and identify patients who may benefit from aggressive supportive care around the time of chemotherapy initiation.”3 The authors acknowledge that there has been considerable research already focused on development of strategies to assess the risks associated with chemotherapy toxic effects; however, they posit that the risks and burdens associated with CRH make it a more meaningful end point than a laboratory value in the toxic range. That being said, whether a hospital admission was related to recent (within the past 30 days) administration of chemotherapy was determined by adjudication and consensus of a team of oncology care clinicians, including representatives from medical oncology, nursing, and pharmacy.
The clinical registry used in this study was developed for quality-improvement purposes, and the authors had to retrospectively collect demographic and medical data from the clinical records of case and control patients to develop their risk model. Important data that should have been provided in this report include the number of medical oncologists treating patients during the registry observation period and their distribution between cases and controls, the matching of the cases and controls in terms of year of treatment, and whether cases and controls received guideline-concordant support for white blood cell (WBC) growth factor during treatments.4 A large population-based study conducted during this same period showed both underuse and overuse of WBC growth-factor support in patients receiving chemotherapy for lung and colorectal cancer.5
Finally, because this is a single-institution study, and the annual number of patients included in the registry was relatively small (1579 patients during 9 years from 2003 to 2011), the findings may reflect the practice patterns of only a few physicians, and the generalizability of the findings to other practice settings is a concern.
In this study, the model development relied on data that were retrospectively abstracted, and it is not surprising that the univariate examination of demographic and medical variables found greater comorbidity, more frequent use of multiagent regimens, and greater likelihood of gastrointestinal and/or mucosal toxic effects and neutropenia among the case patients than among controls. Similarly, the pretreatment laboratory abnormalities that were noted more often in the case patients with CRH were leucopenia, anemia, thrombopenia, and liver, renal, albumin level, and calcium level abnormalities.
The study included a relatively small patient sample along with multiple variables potentially associated with CRH. Using these components, the authors built a more parsimonious prediction model, reducing the number of variables from 15 to 7 through a “supervised” backward selection process that involved clinical reasoning to combine or eliminate variables and arrive at those included in the final model. Thus, the model was not purely data driven and might have been influenced by potential clinical bias from the research team members, some of whom were medical oncologists practicing at the hospital.
The final prediction model is reasonably robust, but the sensitivity is low given the threshold the authors choose to present. The specific variables identified as significant in the model are not surprising, with the exception of younger age, but this might be accounted for by clinical decisions not to administer chemotherapy to older frail patients in this community setting.
The retrospective design and use of a clinical registry rather than a prospective inception cohort of all patients with advanced cancer limits a full understanding of the context of care, especially how treatment decisions were made. We are able to gain a much better perspective on the natural history of treatment decisions and shared decision making from the population-based CanCORS study,6 in which patients with advanced lung and colorectal cancer received initial chemotherapy,7-10 and the patient and treatment characteristics are more fully described.11
Finally, the authors do not describe the frequency of emergency department visits or the length of hospitalizations among the patients in this clinical registry, which would provide additional important information to be able to share with patients.
How should we use the findings from this study? Do they help inform the patient-clinician encounter goals recommended in the recent IOM report?1 There are a number of limitations to the use of this model for enhancing patient-centered care and shared decision making. First, the model is derived from a single community hospital in 1 region of the country and has not been validated in other settings. Second, it will be challenging to validate the model because the CRH outcome variable is not readily available in administrative data, and few clinical registries of this sort exist.
Finally, the data from this study were collected during a time when the concurrent delivery of cancer treatment with integrated palliative care was uncommon. Indeed, the American Society of Clinical Oncology Provisional Clinical Opinion12 regarding the integration of palliative care into standard oncology care was published in 2012 and reflected emerging high-level evidence of the benefits of integrated care. The inclusion of palliative care concurrent with all cancer care is also strongly recommended in the IOM report.1 As a result of these evolving recommendations, patients today treated with chemotherapy for advanced cancer are more likely than were patients during the period reviewed by Brooks et al3 to receive concurrent palliative and supportive care, which holds the promise of preventing CRH events.
While the risk-prediction model described by Brooks et al,3 if validated, could potentially help to identify higher-risk patients, it is clear that all patients initiating chemotherapy deserve the necessary supports from their cancer care team to avoid unnecessary hospitalizations. Chemotherapy-related hospitalizations may well be prevented through more careful discussions with patients about their prognoses and the risks and benefits of treatments. In addition, patients’ social support and living situations should be better evaluated before the chemotherapy is initiated, and after therapy, 24-hour access to clinical support services should be provided to preemptively manage chemotherapy toxic events that would otherwise lead to emergency department visits and hospitalizations. This comprehensive system of care is the goal of the recently announced Centers for Medicare and Medicaid Services (CMS) Oncology Care Model13 that will provide supplemental episode-of-care payments to clinicians whose patients are receiving chemotherapy. Coordination of care and patient engagement, as delineated in the IOM goals 1 and 21 are explicitly called out by CMS, with the aim of improving health outcomes by appropriately aligning financial incentives with delivery of high-quality cancer care. It is hoped that this model, and possibly others, will demonstrate how oncology care in the United States can be transformed to provide higher value and better quality care. We must wait and see. In the meantime, we must do our best to actively engage patients and their families in their advanced cancer care decisions and ensure that the necessary palliative and psychosocial supports are integrated into their treatment and end-of-life care.
Corresponding Author: Patricia A. Ganz, MD, Center for Cancer Prevention & Control Research, Jonsson Comprehensive Cancer Center, 650 Charles Young Drive South, Room A2-125 CHS, Los Angeles, CA 90095-6900 (email@example.com).
Published Online: April 30, 2015. doi:10.1001/jamaoncol.2015.0832.
Conflict of Interest Disclosures: None reported.
Ganz PA. Delivering Patient-Centered Care in the Setting of Advanced Cancer: What Does a Clinical Risk-Prediction Model Have to Do With It? JAMA Oncol. 2015;1(4):430–432. doi:10.1001/jamaoncol.2015.0832
Create a personal account or sign in to: