How we tracked MRSA cases

For this project, we confronted a big hurdle from the get-go: If the state and federal governments don't bother to count MRSA cases, how could we do it ourselves?

We turned first to a database — kept by the state Department of Health — called the Comprehensive Hospital Abstract Reporting System, or CHARS.

Without naming patients, this database provides a wealth of information about individuals discharged from Washington hospitals. Using three distinct coding groups, CHARS offers such information as the patient's ZIP code; whether a patient was insured; what the patient was diagnosed with; how the patient was treated; and how much it cost.

The state uses this data to identify health trends and to analyze costs.

But as powerful a tool as CHARS is, one drawback stood out. The database has 13,367 diagnosis codes — but not a single one for MRSA (methicillin-resistant Staphylococcus aureus). That, in itself, is telling: MRSA has been killing people in hospitals for decades, but the CHARS coding system, used in Washington and only about a dozen other states, fails to account for it.

So drawing upon what medical researchers have done elsewhere, we developed a methodology to sift these cases. Our first filter was a diagnostic code labeled V090. If that code pops up, the patient was diagnosed with some kind of resistant organism. The code doesn't say which organism — only that it was antibiotic resistant.

Our second step was to take those patients with the V090 code and look for any of nine other codes that indicate a staphylococcus infection. The pairing of the V090 code with any of these other nine codes signals MRSA.

But there's a possible exception. The V090 code could mean that some other germ was resistant — E. coli, for example. Or pseudomonas. Or klebsiella. We identified 12 other germs that the V090 code might apply to other than staph. If any of these other 12 germs showed up, we excluded the patient from our MRSA count — unless staph was specifically designated as the primary diagnosis.

By doing this sorting, we were able to identify Washington hospital patients — year by year — who had been treated for MRSA. In 2007, the number of admitted patients with MRSA turned out to be 4,723. Ten years before, the number had been 141.

Using CHARS data, we also were able to do narrower sorting — determining, for example, how many of these MRSA patients acquired the infection after an operation. How many wound up having an amputation. How many lost their sternum because of the infection.

In our reporting, we also relied upon a second database: death certificates in Washington state. This database, likewise, is kept by the state Department of Health.

Because we wanted to get a historical view of MRSA in Washington, we obtained death-certificate data going back to 1980.

We expected to get counts of MRSA deaths through the years. Instead, we failed to find so much as one. The problem? This database — like CHARS — included no code for MRSA. And if a cause of death doesn't have a code, it doesn't get counted.

Then we discovered that the Department of Health (DOH) has, since 2003, kept a separate death-certificate database — without ever disclosing it to the public.

This database included information that went beyond codes. It included any notes written by doctors or medical examiners on the death-certificate forms that expounded on the circumstances under which the person died. Within these notes, words like "methicillin-resistant" signaled that MRSA contributed to the person's death.

We demanded this database, and after a week, the DOH relented. We then were able to use it to count MRSA deaths in Washington since 2003. For the years before that, we relied on the CHARS database.

We reviewed our methodology and numbers with epidemiologists and the state Department of Health, who concurred with our method.

By analyzing the CHARS and death-certificate databases side by side, we found significant holes in each. Sometimes MRSA would show up in an individual patient's death certificate — but not in the same patient's diagnoses in CHARS. Sometimes, we found the reverse.

What these holes suggest is that the counts used in this series — for hospital patients treated for MRSA, and for deaths attributed to MRSA — are baseline figures. In both instances, the actual number is almost certainly higher.

Our reporting went beyond such counts, however. To research how Washington hospitals have fared with basic infection-control measures, we reviewed tens of thousands of pages of inspection reports and documents from individual complaints.

We also reviewed dozens of articles in medical journals to get a historical picture of MRSA's sweep in Washington and across the country.