You’d think drug and medical device makers would know how to spell the names of their own products.
But when companies submitted data to the federal government last year on their payments to doctors, some got the product names wrong. Forest Laboratories misspelled its depression drug, Fetzima, as "Fetziima" 953 times — in more than one-third of all the reports on the drug. Medical device company Amedica Corp. sometimes called its Preference screw system "Preferance."
Amid much anticipation and after a lengthy delay, the government in September unveiled its Open Payments database, saying it would bring transparency to relationships between physicians and the drug and medical device industries. But this openness has been clouded by numerous errors that detract from its usefulness.
The database mistakes surfaced as ProPublica developed an app — rolled out this month with The Upshot, a New York Times policy-analysis initiative — to identify the drugs and medical devices that were most heavily promoted to doctors in the last five months of 2013, the period covered by the data. The ProPublica review found that many of the drugs with the highest spending weren’t cures or even medical breakthroughs, but rather "me-too" drugs that were little different from other drugs on the market.
As we tried to get accurate tallies of all payments associated with each drug and device, we encountered widespread problems with companies’ submissions. We also learned that the Centers for Medicare and Medicaid Services, which administers the Open Payments database, does not double-check what companies submit.
"We are very committed to not altering data," said Shantanu Agrawal, director of the agency’s Center for Program Integrity. "Our role is not to spell-check for the industry. The act of transparency also will improve the data itself."
As companies’ errors are exposed, Agrawal explained, they will work to fix them.
While misspelling product names was an obvious, infrequent error, we encountered a variety of other problems. Companies, for instance, routinely recorded payments associated with a single drug under multiple names.
Take H.P. Acthar Gel, an expensive injectable drug used to treat multiple sclerosis, kidney disease, lupus and other conditions. The drug’s maker, Questcor Pharmaceuticals, logged payments related to the drug under eight names, including Acthar, Acthar-Pulm, Acthar-IS, Acthar-Rheum and Acthar-MS. The payments associated with each name didn’t stand out much. But when they were all added together, the drug ranked in the top 20 for spending on doctors. For our analysis, we grouped them together. (Questcor has since been acquired, and its new owner said it would report all future payments under one name.)
Sanofi and Genzyme US Companies reported more than 31,800 payments totaling $1.7 million associated with Lantus, its long-acting insulin, but listed the drug under a variety of names: Lantus Timely Insulin, Lantus-Vial, Lantus TML and Lantus Capture Insulin. We combined them as Lantus. (We tabulated Lantus SoloStar payments separately because the drug comes in a pen delivery system, not a vial.)
Another complication was that about 8.5 percent of the 4.3 million general payments companies reported to the government were not connected to specific products. More than 140 companies did not list products at all in their payment records. We were not able to say which products those payments related to in our analysis.
Other companies didn’t leave product fields blank but, in some cases, might as well have. Gilead Sciences listed the product associated with eight payments as "unrecognized" and connected two dozen more to "Gilead Corporate." Novo Nordisk and its subsidiaries attributed 90 payments totaling $133,000 to a product called "Disease State," a phrase that is physician talk for a medical condition like diabetes, not a particular drug. We created a catchall category for such payments, listing them under "Other."
We also learned that some companies listed products in the wrong category, or listed them as both drugs and devices in the same payment records. Luitpold Pharmaceuticals Inc., a drugmaker, did this in more than 8,000 payment records.
Companies were able to associate each payment with up to five drugs and five devices, so summing up spending by drug and device was tricky. When a payment was connected to more than one product, companies did not break out how much was attributable to each. Rather than guessing how to apportion such payments, ProPublica applied the full payment amount to each product. Since some companies routinely listed multiple drugs for individual payments, excluding such spending would have substantially distorted their totals.
Sometimes, when we dealt with individual payments attributed to multiple products, we found that companies had listed the same products more than once.
Janssen Pharmaceuticals, a division of Johnson & Johnson, did this hundreds of times with its arthritis drug Simponi, listing it as both drug No. 1 and as drug No. 2 in records for the same payments. A company spokeswoman said although Janssen submitted the same name twice, the second reference was to Simponi Aria, a distinct product. As a result, we had to write code to count these payments only once.
In a number of cases, multiple subsidiaries of a company reported payments related to the same drugs. Five subsidiaries of Johnson & Johnson reported payments associated with the diabetes drug Invokana: Janssen Research & Development LLC; Johnson & Johnson Health Care Systems Inc.; Animas Corp.; Janssen Pharmaceuticals Inc. and LifeScan Inc. (When we calculated payments for individual products, we included every company making payments on its behalf.)
There’s no indication that companies were being deliberately evasive — indeed, payments could have come from different subsidiaries. But the complexity shows how difficult it would be for the general public to get a true sense of payments from the publicly released data sets.
Even companies that experienced reporting problems said they were proud of their efforts. For example, several dozen of the 20,000 payments Purdue Pharma reported for its pain drug Butrans and its sleeping pill Intermezzo were attributed to "BUP" and "INT," but company officials were quick to point out that translates to an error rate of less than 1 percent.
"I’m really pleased that we were able to achieve such high accuracy with our first submission, but given the complexity of the reporting and tight deadlines the industry faced, it’s natural that there would be gaps in the overall data set," Maggie Feltz, who leads Purdue’s Open Payments initiative, said in a written statement. "I’m confident the industry will continue refining its methods and we’ll have an even better outcome next year."
The Pharmaceutical Research and Manufacturers of America, the industry trade group, said mistakes were understandable since this had been the first year companies reported this information publicly and the government repeatedly changed its guidance for companies. Officials hope the coming year goes more smoothly. Data covering the full year 2014 is expected to be released this June.
Given the large amount of information the Centers for Medicare and Medicaid Services asked companies to submit and the agency’s evolving rules and guidance, "it’s not surprising to me that there are some errors out there," said John Murphy, PhRMA’s assistant general counsel. As time goes on, he said, "I’d suspect this stuff will get much more streamlined and much better."