Once officials at the Fred Hutchinson Cancer Research Center figured out that we were working on a story that might not be flattering, they became quite secretive.
We wanted to know more about financial conflicts of interest in experimental trials, yet they flatly refused to give us any information about the financial holdings of top researchers or even the center itself. The fact that some of these doctors had financial entanglements with the treatments they were researching - or that patients died from these treatments - didn't sway them at all.
The state open records act was useless because the Hutch is a private institution, even though nearly 70 percent of its funding comes from federal grants. And an intriguing amendment to the federal Freedom of Information Act - one intended to open records on projects funded with government grants - was rendered worthless as well when research lobbyists prevailed in getting new regulations that severely undercut the law.
We seemed to be stuck.
Unfortunately, there are no off-the-shelf databases on physicians' financial holdings. We knew that there was a lot of information to be gleaned from Securities and Exchange Commission records, patent records and the like, but we faced the nightmare scenario of reading through thousands - perhaps hundreds of thousands - of pages looking for names.
Then it hit us. With a little bit of computer programming, reading through those documents could be automated.
It wouldn't be as easy as joining the sex-offender database to the day-care worker database. But with the wealth of data easily accessible on the Internet and the sophisticated tools we have available today, it was clearly doable.
The results appear as part of our five-part series, "Uninformed Consent: What patients at the Hutch weren't told about the experiments in which they died." The series, written by Duff Wilson and myself and edited by David Boardman, documents two failed clinical trials in which doctors had financial interests and in which patients died from risky treatments without being properly informed of the risks.
To get information about financial conflicts, first we needed researchers' names. Then we would search for those names in a slew of financial documents from the SEC, an insider-trading database and the U.S. patents full-text database. We'd finally add our results to the few financial disclosure forms that we were able to obtain.
How we did it:
First we headed to the library to gather up every biennial scientific report the Hutch has ever published. These included the names of their entire faculty, including post-doctoral fellows. We sent photocopies to a vendor to be keyed into a database. That was the easy part.
Next we tackled SEC documents. Those are free through the Edgar database at www.sec.gov. But we had two problems: How do we download all documents for companies involved in the health-care industry? And how do we search those documents for names?
The answer to the download problem was to write a program in Visual Basic. If I ever doubted the value of learning a bit of programming to do CAR work, this experience resolved those doubts. VB includes functions that allow you to go out onto the Web and grab files. Other languages do too.
It took me two solid days to write the program in one of those work binges where you forget if it's almost dawn. But the payoff was running the thing and actually seeing it work. I annoyed my wife and plenty of friends by showing off the finished product.
The list of all health-care companies came from Edgar's own documentation pages. Using my program, I grabbed every annual report and proxy statement for all years and put them into one massive text file.
The next problem was what to do with this massive file. I could have written another Visual Basic program to search for each researcher's name in the file, but that would have been deadly slow. What I really needed was a text-indexing tool to make the search speedy. My first thought was to use Lotus Notes but I was lucky to have a tool called Texis, a document retrieval database that we use at the Seattle Times for our own electronic clips library.
I won't bore you with the details here. I'll just say that Texis is extremely powerful and fast. Best of all it's a relational SQL database that can do joins. So once I got the text file loaded, parsed by company name, document type, year and the full text of each document, I could run a fairly simple join with the researchers' names. The query would search each document's full text looking for name matches, like an automated clip search. We searched for 1,800 names.
I had a big assist on this from research editor Tom Boyer, who developed our electronic library. For more information on Texis, go to www.thunderstone.com.
The result of the query was a new file with a large chunk of text on either side of a name match. I loaded that dataset into SQL Server and then started the manual process of verifying each match. With a few exceptions, this was easier than it sounds because typically the Hutch would be mentioned somewhere in the text.
Patents
The next challenge was to get all patents by these researchers. The data are available for free at www.uspto.gov/patft/index.html Again, a modification of the Visual Basic program allowed me to grab off the Web all patents in which Hutch researchers were listed as inventors. I limited this search to inventors from the state of Washington, which helped to produce more clean matches.
Again, I had to weed out false hits manually. If the person invented a new type of typewriter ribbon or garden tool, I automatically eliminated that patent. However, a patent for radiation delivery to lymphoid and marrow tissues was a good candidate for a true hit. Finding true matches wasn't terribly difficult.
We searched an insider-trading database with the help of a company called First Call, www.firstcall.com One last piece came from investigative reporter Deborah Nelson, who unearthed these stories but left for The Washington Post before she could complete the project. She had requested all financial disclosure forms for medical faculty at the University of Washington who also had positions at the Hutch. That information - useful but incomplete - was then keyed into a database. All of these data were loaded into SQL Server. I developed an interface in Access to pull all the data we had on each researcher together on one page. We created an 'A' list of the researchers that we wanted to focus on. As we learned of other financial connections, we were able to key these in manually through the Access interface.
A quick note: to add researchers to the 'A' list later, I used an incredible tool in SQL Server called triggers. FoxPro has the same capability. The trigger would automatically launch a stored procedure to create a unique ID for any name we added to the 'A' list and then automatically search all datasets and add that unique ID to other tables to allow for joins. In short, you type in a name and suddenly all the data about that person magically appears.
Triggers may be routine for database developers, but I'd never heard of them being used by CAR folks. I was awestruck by them and encourage you to investigate.
Medline
In the course of our reporting, we also downloaded abstracts of medical journal articles written by Hutch researchers in the past decade. These data are available through a National Institutes of Health site called Medline, www.ncbi.nlm.nih.gov/pubmed/
Downloading them involved a total of about 18,000 searches. My original program ran as fast as possible, and I hadn't really thought about the demand that this would put on NIH servers until the annoyed database administrator blocked our access. I called to let him know what we were doing and he gave us access again providing we spaced our searches five seconds apart and ran the program after 9 p.m.
We pulled full copies of articles for researchers with financial interests the old-fashioned way, photocopying them at a medical library. This effort produced a mere paragraph in our series about how researchers don't disclose their interests in medical journals, even when the articles are about their patents.
We also made heavy use of our death certificate database in tracking down patients. Other data included all grants from the NIH and all clinical trials ever advertised by the Hutch on the National Cancer Institutes' Web site.
After doing the CAR work, I assumed a more traditional role of a shoe-leather reporter on the series. This project is a good example of the seamless blending of CAR work with investigative reporting. Even though the series was aided by the most sophisticated data work I've ever done, the focus was not on data but on the compelling stories of the patients, their families, a whistleblower, a financier, a Nobel Prize winner and the Hutch's doctors.
David Heath can be reached by e-mail at dheath@seattletimes.com