Overlooked pitfalls in multi-class machine learning classification in radiation oncology and how to avoid them.

In radiation oncology, Machine Learning classification publications are typically related to two outcome classes, e.g. the presence or absence of distant metastasis. However, multi-class classification problems also have great clinical relevance, e.g., predicting the grade of a treatment complication following lung irradiation. This work comprised two studies aimed at making work in this domain less prone to statistical blindsides. In multi-class classification, AUC is not defined, whereas correlation coefficients are. It may seem like solely quoting the correlation coefficient value (in lieu of the AUC value) is a suitable choice. In the first study, we illustrated using Monte Carlo (MC) models why this choice is misleading. We also considered the special case where the multiple classes are not ordinal, but nominal, and explained why Pearson or Spearman correlation coefficients are not only providing incomplete information but are actually meaningless. The second study concerned surrogate biomarkers for a clinical endpoint, which have purported benefits including potential for early assessment, being inexpensive, and being non-invasive. Using a MC experiment, we showed how conclusions derived from surrogate markers can be misleading. The simulated endpoint was radiation toxicity (scale of 0-5). The surrogate marker was the true toxicity grade plus a noise term. Five patient cohorts were simulated, including one control. Two of the cohorts were designed to have a statistically significant difference in toxicity. Under 1000 repeated experiments using the biomarker, these two cohorts were often found to be statistically indistinguishable, with the fraction of such occurrences rising with the level of noise.

Domestic Job Shortage or Job Maldistribution? A Geographic Analysis of the Current Radiation Oncology Job Market.

PURPOSE: To examine whether permanent radiation oncologist (RO) employment opportunities vary based on geography. METHODS AND MATERIALS: A database of full-time RO jobs was created by use of American Society for Radiation Oncology (ASTRO) Career Center website posts between March 28, 2016, and March 31, 2017. Jobs were first classified by region based on US Census Bureau data. Jobs were further categorized as academic or nonacademic depending on the employer. The prevalence of job openings per 10 million population was calculated to account for regional population differences. The χ2 test was implemented to compare position type across regions. The number and locations of graduating RO during our study period was calculated using National Resident Matching Program data. The χ2 goodness-of-fit test was then used to compare a set of observed proportions of jobs with a corresponding set of hypothesized proportions of jobs based on the proportions of graduates per region. RESULTS: A total of 211 unique jobs were recorded. The highest and lowest percentages of jobs were seen in the South (31.8%) and Northeast (18.5%), respectively. Of the total jobs, 82 (38.9%) were academic; the South had the highest percentage of overall academic jobs (35.4%), while the West had the lowest (14.6%). Regionally, the Northeast had the highest percentage of academic jobs (56.4%), while the West had the lowest (26.7%). A statistically significant difference was noted between regional academic and nonacademic job availability (P=.021). After we accounted for unit population, the Midwest had the highest number of total jobs per 10 million (9.0) while the South had the lowest (5.9). A significant difference was also observed in the proportion of RO graduates versus actual jobs per region (P=.003), with a surplus of trainees seen in the Northeast. CONCLUSIONS: This study presents a quantitative analysis of the RO job market. We found a disproportionately small number of opportunities compared with graduates trained in the Northeast, as well as a significant regional imbalance of academic versus nonacademic jobs. Long-term monitoring is required to confirm these results.

Del consentimiento como prueba / Informed consent as evidence

No disponible

The 2009 devaluation of radiosurgery and its impact on the neurosurgery-radiation oncology partnership.

Neurosurgeons, radiation oncologists, and, increasingly, other surgical specialists recognize that radiosurgery is an important tool for managing selected disorders throughout the body. The partnership between neurosurgeons and radiation oncologists has resulted in collaborative studies that have established the clinical benefits of radiosurgery. Today, however, a range of political and financial issues is straining this relationship and thereby undermining the practice of radiosurgery. Neurosurgeons and radiation oncologists recently restricted the definition of radiosurgery to include only cranial- and spine-focused radiation treatments. Meanwhile, organized radiation oncology decided unilaterally that radiosurgery administered to other parts of the body would be termed stereotactic body radiation therapy. Finally, neurosurgical and radiation oncology coding experts developed new Current Procedural Terminology codes for cranial vault and spine radiosurgery, which were approved for use by the Relative Value Scale Update Committee as of 2009. The authors suggest that the neurosurgery strategy-which included 1) reasserting that all of the tasks of a radiosurgery procedure remain bundled, and 2) agreeing to limit the definition of radiosurgery to cranial vault and spine-has failed neurosurgeons who perform radiosurgery, and it may jeopardize patient access to this procedure in the future. The authors propose that all of the involved medical specialties recognize that the application of image-guided, focused radiation therapy throughout the body requires a partnership between radiation and surgical disciplines. They also urge surgeons to reexamine their coding methods, and they maintain that Current Procedural Terminology codes should be consistent across all of the different specialties involved in these procedures. Finally, surgeons should consider appropriate training in medical physics and radiobiology to perform the tasks involved in these specific procedures; ultimately all parties should receive equivalent reimbursement for similar assigned tasks, whether performed individually or jointly.

Who enrolls onto clinical oncology trials? A radiation Patterns Of Care Study analysis.

PURPOSE: To identify factors significantly influencing accrual to clinical protocols by analyzing radiation Patterns of Care Study (PCS) surveys of 3,047 randomly selected radiotherapy (RT) patients. METHODS AND MATERIALS: Patterns of Care Study surveys from disease sites studied for the periods 1992-1994 and 1996-1999 (breast cancer, n = 1,080; prostate cancer, n = 1,149; esophageal cancer, n = 818) were analyzed. The PCS is a National Cancer Institute-funded national survey of randomly selected RT institutions in the United States. Patients with nonmetastatic disease who received RT as definitive or adjuvant therapy were randomly selected from eligible patients at each institution. To determine national estimates, individual patient records were weighted by the relative contribution of each institution and patients within each institution. Data regarding participation in clinical trials were recorded. The factors age, gender, race, type of insurance, and practice type of treating institution (academic or not) were studied by univariate and multivariate analyses. RESULTS: Overall, only 2.7% of all patients were accrued to clinical protocols. Of these, 57% were enrolled on institutional review board-approved institutional trials, and 43% on National Cancer Institute collaborative group studies. On multivariate analysis, patients treated at academic facilities (p = 0.0001) and white patients (vs. African Americans, p = 0.0002) were significantly more likely to participate in clinical oncology trials. Age, gender, type of cancer, and type of insurance were not predictive. CONCLUSIONS: Practice type and race significantly influence enrollment onto clinical oncology trials. This suggests that increased communication and education regarding protocols, particularly focusing on physicians in nonacademic settings and minority patients, will be essential to enhance accrual.

From concept to CPT code to compensation: how the payment system works.

All radiologists and radiation oncologists provide medical services to patients every day with the full anticipation that these services will be appropriately reimbursed. Yet most take this process for granted. Few have even a rudimentary idea how the system works by which a coding mechanism and reimbursement schedule are developed and maintained for the vast array of services they provide. Clearly, this is not good business. You need not stay in the dark any longer! This article describes (1) the fundamental structure of reimbursement for radiology and radiation oncology services; (2) the multiple steps required as a new procedure advances from a research concept to the assignment of a code in the American Medical Association's Current Procedural Terminology; (3) the process by which the new procedure and code are assigned a reimbursement value in the Medicare Fee Schedule, which acts as the base for over 75% of current medical reimbursement; and (4) the maintenance of this system for existing procedures.

To code, or not to code?

In summary, it is also important to remember the hidden rules: 1) Just because there is a code in the manual, it doesn't mean it can be billed to insurance, or that once billed, it will be reimbursed. 2) Just because a code was paid once, doesn't mean it will ever be paid again--or that you get to keep the money! 3) The healthcare provider is responsible for knowing all the rules, but then it is impossible to know all the rules! And not knowing all the rules can lead to fines, penalties or worse! New codes are added annually (quarterly for OPPS), definitions of existing codes are changed, and it is the responsibility of healthcare providers to keep abreast of all coding updates and changes. In addition, the federal regulations are constantly updated and changed, making compliant billing a moving target. All healthcare entities should focus on complete documentation, the adherence to authoritative coding guidance and the provision of detailed explanations and specialty education to the payor, as necessary.

The correct usage of the new radiation oncology codes. 77395--3-dimensional simulation; 77419--conformal weekly radiation therapy treatment management; 77432--stereotactic treatment management.

In summary, radiation oncology now has three new related codes describing the 3-dimensional simulation and treatment of relatively small tumor volumes. These codes, when properly used and completely documented, are reimbursable at rates higher than the conventional simulation and treatment delivery codes that they replace. The physician should be cautioned however, that the indiscriminate use of these codes without accurate documentation of the medical necessity could result in penalties and/or pay back in the event of an audit. As with all new codes, we may rest assured that the insurance carrier will be looking very carefully at the documentation of these new and expensive procedures.