What hospital star ratings measure

The overall hospital star rating from CMS rolls dozens of public measures into a single 1-to-5 score. Not every hospital gets one. A hospital has to report enough measures to be rated, and many critical access, psychiatric and VA hospitals end up unrated. That's not a sign they're bad.

What goes into it

Mortality. Death rates within 30 days for patients treated for conditions such as heart attack, heart failure, pneumonia and stroke, adjusted for how sick the patients were when they arrived.

Safety of care. Infections people picked up in the hospital, like bloodstream and urinary tract infections, plus serious complications.

Readmissions. How often patients came back within 30 days of going home.

Patient experience. A standard survey of recent patients covering how well nurses and doctors communicated, whether the room was clean and quiet, and how they'd rate the hospital overall.

Timely and effective care. Whether certain things happened when they should have. Sepsis treatment is the big one.

Reading the numbers on our hospital pages

Infection ratios compare actual infections with the number expected for a hospital of that size and type. A ratio of 1.0 means as many as expected, and lower is better.

"Better," "no different" and "worse than national" are CMS's own judgments and account for statistical uncertainty. Most hospitals land on "no different" for most measures, which is normal.

ER time is the typical number of minutes from walking into the emergency department to leaving, for patients who were sent home. It's not the wait to see a doctor.

What the stars won't tell you

A hospital-wide rating can hide big differences between departments. A hospital might be excellent at cardiac surgery and average at everything else. The data also covers past periods, often a year or more back. If you're planning a specific operation, ask the surgeon how many of that procedure the hospital does a year and how its patients do afterward. That tells you more than any overall score.