Hospital infections kill more people than car crashes. Here’s how to cover them better

A database maintained by the CDC is an underused journalistic tool

Ebola has killed one person in the United States, but about 75,000 people die each year from infections they picked up in a hospital. As Dr. Peter Pronovost of Johns Hopkins—one of the champions of checklists in medicine—notes frequently, that is more deaths than from breast cancer or car crashes.

Despite the public health risk they pose, these infections don’t draw a lot of media coverage. But the national focus on Ebola gives journalists a news hook to write about the domestic infections that kill scores patients in their own backyards. And a database maintained by the federal government—one of the newest, most detailed, and underused health data sets around—makes it easier than ever.

Here’s how that data set works: More than 3,000 hospitals report infections to the Centers for Disease Control and Prevention. The CDC then weeds out the ones that patients may have had before they were admitted; the remainder are known as Healthcare-Associated Infections, or HAIs. The HAI database currently attributes more than 125,000 infection cases to specific hospitals.

It also lists the CDC’s evaluations of each hospital for six categories of infections:

  • Central-line associated bloodstream infections (CLABSI), which occur when germs travel through flexible tubes that are put into veins to deliver drugs or nutrients
  • Catheter-associated urinary tract infections (CAUTI), which come from tubes put into bladders to drain urine from patients
  • Methicillin-resistant Staphylococcus aureus (MRSA), a staph bacteria that many antibiotics don’t work on
  • Clostridium difficile (C. diff.), which often strikes in the intestines after antibiotics wipe out normal, healthy bacteria
  • Infections that occur at the site of a hysterectomy surgery
  • Infections that occur at the site of a colon surgery

For journalists, this data can pinpoint which hospitals in your region are having the most trouble averting infections, which hospitals are doing the best, and which infections are providing the biggest challenge. You can choose to look at all six infections, or burrow into just one—if you focus on MRSA or C. diff, you can expand your story to examine how disturbingly prevalent these germs likely are around your community. And this fall, Medicare will be penalizing hospitals with high rates of patient harm, including some of these infections, so that gives you another way to get into the story.

When you’re ready to dig in to the data, there are a couple possible routes to follow. If you’re only looking at a few hospitals, you can see how they perform on Medicare’s Hospital Compare website. After you locate the hospitals, go to the “Readmissions, complications & deaths” tab and then click on “Healthcare-associated infections”. This will show whether the hospital performed better, worse, or no different from national benchmarks the CDC set for each of the six measures. Don’t be surprised if a hospital is rated “no different” from the benchmark; most are, as the government uses conservative statistical techniques. It’s also not unusual for a hospital to do better than average for one type of infection and worse for another.

If you’re looking at a lot of hospitals, you can download a PDF or a CSV file of all ratings that we put together at Kaiser Health News, where I’m a senior correspondent. Or you can pull the data directly from the Healthcare-Associated Infections – Hospital file posted on It’s a huge file, with more than 165,000 records, but you can filter it down to your city, county, state, or other criteria, and then export the data you want.

For each of the six measures, the file shows how many infections a hospital had (Observed Cases), and how many infections the CDC predicted it should have had, given the national benchmark, and the kind of patients and services at each hospital (Predicted Cases). For statistical reasons, a hospital can have more infections than the CDC predicted and still be labelled no different than the benchmark, so be careful in throwing around specific numbers.

The database also tells you how many patients were potentially vulnerable to getting to a particular infection; this is expressed as Patient Days. (If you want to understand how the CDC determines a hospital’s infection rate, there’s a concise explanation here, and you can learn about confidence intervals and standardized infection ratios.)

To take a specific example: Riverview Regional Medical Center in Gadden, AL, had 1,593 central line days, which counts every day that a patient had a central line inserted in the intensive care unit. Seven patients with central lines inserted got infections over that year at Riverview. That may not seem like a lot—but many infection control experts believe there’s never a reason for a patient to suffer one of these infections, since hospitals have known for years how to prevent them. The number of infections was higher than the benchmark estimate, so the CDC rated Riverview as worse. That’s something for journalists in the community to follow up on. 

Hospitals do have qualms about the CDC data that are worth taking into consideration. One concern is whether hospitals that are very assertive in looking for infections are more likely to find them, leading the CDC to conclude they have high rates. Also, while some states double-check the accuracy of the numbers, others do not, and the federal government doesn’t do much oversight.

Also, though the CDC takes a lot of mitigating factors into account—including whether the hospital is affiliated with a medical school, the size of the hospital, and the length of surgery—many hospitals complain the risk adjustment doesn’t fully capture their mix of patients and services. And, since the current data stop at September 2013, hospitals are likely to say they’ve improved since then. In that case, you can always ask them to provide more recent numbers.

There may be elements of truth in these reservations, but the HAI measures are about as solid as you get in public health care data. These measures have been approved by industry groups, and of course the CDC stands behind them. With 1 in 25 hospital patients getting an infection, these often-preventable illnesses are valid concerns for anyone who steps into a hospital and for the journalists who cover these institutions. The HAI database is a good place to start.

Has America ever needed a media watchdog more than now? Help us by joining CJR today.

Jordan Rau is a senior correspondent for Kaiser Health News, a nonprofit news service that is an editorially independent program of the Henry J. Kaiser Family Foundation.