Read a longer, more technical methodology and its appendices.

For its analysis of surgical complication rates, ProPublica acquired Medicare billing records for in-patient hospital stays from 2009 through 2013. We focused on eight common elective surgeries – knee replacements, hip replacements, three types of spinal fusions, one in the neck and two in the lower back, gall bladder removals, prostate removals, and prostate resections. We chose these surgeries because they are typically performed on healthy patients and are considered relatively low risk.

To be fair to surgeons, we first excluded from our analysis trauma and other high-risk cases that are more likely to result in complications that are beyond a surgeon’s control. We also excluded surgeries on patients who were admitted via the emergency department or from another healthcare facility. We used accepted statistical methods to adjust for age, the health of each patient, luck, and the overall performance of each hospital.

To compile a list of complications that were specifically related to the surgery, we enlisted the help of two dozen physicians. For each of the eight procedures, a panel of at least five doctors, including specialists who perform the procedure, reviewed the principal diagnosis code on cases in which patients were readmitted within 30 days. The doctors were asked to assess whether the code was likely to signal a complication related to the surgery. In cases where a majority of experts did not agree that a particular principle diagnosis code represented a complication, we did not include it in our analysis. We gave additional weight to the opinion of the surgeon specialists on each panel. The result was a list of hundreds of complications related to each of these surgeries – problems like infections, blood clots, uncontrolled bleeding and misaligned orthopedic devices.

If a patient was readmitted to any hospital (not just the hospital where the surgery was performed) within 30 days of a surgery for one of the conditions we identified, we counted the case as a complication for the surgeon who performed the initial procedure. If a patient was readmitted more than once within 30 days, we counted it as a single complication. We also tallied any patient deaths during the initial surgical stay as a complication for the operating surgeon.

To calculate a surgeon’s raw rate of complications for a given procedure, we divided the number of the surgeon’s patients who suffered a complication by the total number of surgeries he or she performed.

We screened each patient’s record for signs of other health problems, like obesity and diabetes and then assigned them a health score. To calculate the adjusted complication rate, we used a statistical method called a mixed effects model, which takes into account fixed factors like age and health score, and variable effects, like the performance of surgeons and hospitals.

A high adjusted complication rate indicates that a surgeon’s patients suffered harm more often than his or her peers. A low rate indicates that a surgeon’s patients fared better overall. The model assumes that those surgeons with very high raw rates experienced some level of bad luck and adjusts their numbers downwards. It similarly adjusts the rates of surgeons who had no complications, which is why no doctor in the data is below 1.1 percent.

Our searchable database, Surgeon Scorecard, reports adjusted complication rates for 16,827 surgeons operating at 3,575 hospitals. No rate is reported if a surgeon performed an operation fewer than 20 times. For the eight surgeries combined, the analysis identified 63,173 patients who were readmitted with a complication and 3,405 who died. Medicare paid hospitals $645.3 million for these readmissions alone.

Where we can, our database provides the precise number of complications. However, our agreement with the Centers for Medicare and Medicaid Services bars us from reporting any number of complications from 1 to 10, or from reporting any data that can be used to calculate a number that is within that range. Officials said this was necessary to protect patient privacy.

As with all health care research, ProPublica’s analysis has some limitations. While Medicare data is frequently used by health researchers, it does not include hospital stays paid for by private insurance or other government programs, such as Medicaid. As a result, many cases that could move a surgeon’s complication rates up or down are not counted in the analysis.

It is not always possible to determine the exact cause of a surgical complication and whether a surgeon is to blame. However, the American College of Surgeons says in its statement of principles that surgeons are responsible for a patient’s entire course of care, from diagnosis to surgery to rehabilitation, a view shared by many surgeons and experts we interviewed.