In my earlier post I commented on the liberation of health data and its use by entrepreneurs and health professionals to persuade policy makers, payers and providers to reward quality over quantity. In a recent study on lung cancer, this data was once again put to good use. An analysis of Medicare costs, enrollment, and demographic data from the Centers for Medicare & Medicaid Services (CMS) beneficiary files estimated the cost-effectiveness of screening the Medicare population at high risk for lung cancer.
The data found that lung cancer today claims the lives of more people annually than the next four most lethal cancers combined, which are colon, breast, pancreas, and prostate.1,2 In the United States, an estimated 224,210 people will be diagnosed with lung cancer. An estimated 159,260 people will die of the disease in 2014.3
The study concluded that not only is lung cancer screening a cost-effective strategy4, but from the standpoint of cost per life-year saved, lung cancer meets or exceeds the value of colon, breast, pancreas and prostate screenings. While CMS is still deliberating whether lung cancer screening will be reimbursed in 2015, mounting evidence fueled by this liberated data suggests the decision is imminent.
Computer Aided Detection
If lung cancer screening is reimbursed by CMS we likely will faced an even bigger challenge. Screening approximately 4.9M, and growing, high-risk Medicare beneficiaries using an already overburdened radiologist workforce. One solution some radiologists are hoping for is to draft CAD (Computer Aided Detection) into their team, since lung nodule detection is extremely heterogeneous and comparison is hard, time-consuming work. Multiple studies have found that radiologist nodule detection skills are inferior to that of CAD5. For example, in a recent study Dr. Geoffrey Rubin of Duke University and his colleagues digitally embedded 157 nodules 5 mm and larger into 41 CT datasets. Thirteen experienced radiologists read the data without CAD and only found 49% of the nodules. Yet despite CAD’s known ability to boost radiologist performance6 and a decade of CAD fine tuning, these systems haven’t found a place in the daily practice of most imaging facilities.
So if help is needed and it is, and CAD is the promise, as demonstrated in many studies, then why is CAD not widely adopted by most radiologists?
Could cloud be an answer?
As I’ve met with several radiologists in the past month, the answer to this question seems mixed. Detractors point to CAD itself highlighting too many false positives that wastes time and infers the algorithm itself needs improvement. Others say post processing takes too long and they can’t wait, while others suggest CAD workflow needs to be simultaneous (Rad with CAD) versus sequential (CAD then Rad). Supporters want to see CAD ergonomics and GUI improvements such as CAD results embedded directly into the radiologist report, while others would like to see CMS and other payers force the use of CAD by reimbursing only those studies that have gone through CAD.
It’s a fascinating dilemma. Is it possible that cloud technology can solve some of this? What if the algorithms and post processing were done in the cloud? Would that speed up post processing, and could we improve the algorithm with an easy feedback loop? What if CAD results vetted by radiologists were embedded in to reporting tools? Would love to hear what you think.
1 American Cancer Society. Lung cancer (non-small cell) overview. Revised April 30, 2014. www.cancer.org /Cancer /LungCancer-Non-SmallCell/OverviewGuide/lung-cancer-non-small-cell-overview-key-statistics. Accessed June 6, 2014. 2 American Lung Association. Lung cancer fact sheet. www.lung.org/lung-disease/lung-cancer/resources/facts-figures/lung-cancer-fact-sheet.html#Mortality. Accessed July 21, 2014. 3 American Cancer Society. What are the key statistics about lung cancer? Revised February 11, 2014. www.cancer.org/cancer/lungcancer-smallcell/detailedguide/small-cell-lung-cancer-key-statistics. Accessed June 17, 2014. 4 Offering Lung Cancer Screening to High-Risk Medicare Beneficiaries Saves Lives and Is Cost-Effective: An Actuarial Analysis: August 2014 Vol 7, No 5 – Regulatory 5 (Radiology, January 2005, Vol. 234:1, pp. 274-283). 6 Godoy and colleagues (American Journal of Roentgenology, January 2013, Vol. 200:1, pp. 74-83)