I’ve been talking to my class this week (and it’s going to be a theme next week) about the difficulty of analyzing epidemiological data on cancer — that there are so many steps to cancer progression and so many environmental and genetic inputs to the disease that sorting them all out is extremely complex. What I haven’t mentioned yet, but definitely will now, is the factor big money plays in encouraging statistical fraud.
The University of Texas MD Anderson Cancer Center in Houston has been the top-ranked cancer center on US News & World Report’s best-of list for the past 7 years. But that top ranking was aided by a massive error in data used to evaluate its care.
The error in M.D. Anderson’s favor was made by–M.D. Anderson! Avery Comarow, who assembles the rankings at U.S. News, told The Cancer Letter that this was a huge "screw-up." The hospital systematically misclassified emergency patients, which led to the exclusion of nearly 40 percent of admissions, Paul Goldberg, The Cancer Letter’s editor, reported. He said the error was discovered in 2009, but no way could be found to correct it. "Since U.S. News averages data over three years, the results of the M.D. Anderson top rating by the magazine released July 16 are still partially based on tainted data," he wrote.
Is “error” actually the right word to use?