NHSN recently published a final document, “Your guide to the standardized infection ratio”. (Centers for Disease Control, 2010). This document tries to explain what a new statistic, the Standardized Infection Ratio (SIR),for HAIs is and why it is useful. The problem is that they use statistical jargon that is hard to decipher, even for those of us who have taken statistics courses. This post is an attempt to put the explanation into terms that are easier to understand (and hopefully, easier to explain to others).
Background
Before explaining the new statistic, let’s review the way that HAIs were reported before in the annual HNSN statistical report. As shown in the 2009 report (AJIC), results were reported according to clinical service locations where HAIs were developed (e.g. Inpatient wards/adult step down units/post critical care) and sometimes by the type of organization the units were located in (e.g. medical/surgical/major teaching hospital). The problem was that the comparisons could be made only between locations. All IPs knew that there are more specific factors that contribute to the development of HAIs than only where the patient was located. NHSN agreed with this and developed a new statistic that takes many factors into account.
That is where the new statistic, SIR, comes in. The SIR is a number that takes into account many more factors than just the location where the HAI developed. Unfortunately, the road to the SIR is a little twisty. If you can bear with me, I hope that you will understand it by the end of this posting.
(The explanation given in the NHSN document (CDC, 2010) describes how SIRs were developed for SSIs. So even though the first SIRS reports will be for CLABSIs, I will the SSI example information to explain what they did.)
The Process
First, they collected SSI information submitted to NHSN between 2006 and 2008. They looked not only at location but at other factors that may have contributed to the development of the SSI including ASA score, age, length of surgery, etc. They fed all the information they had about risk factors into a computer for each kind of surgical procedure. They then asked the computer to list the factors that seemed to be the most important for each type of surgery. Every surgery came out with its own combination of important factors (we’ll call it a risk profile).
After they had identified the risk profile for each surgery, they went back to the 2006 – 2008 data and looked at different levels of risk factor for each surgery. For instance, they looked at the number of SSIs that had occurred in craniotomy patients who were 40 years old, had an ASA score of 1, and a surgery lasting longer than average. Then they looked at the number of infections in patients who were 50 years old with the same ASA scores, and same length of surgery, etc., etc. until all combinations had been reviewed. (Because of the many, many calculations, you can see why a computer would have to do this.)
Each computer analysis returned a percentage that showed how much of a chance a person with a given level of factors had of developing an SSI. For instance, the computer might report that patients like the 40 year old patient described above had a 5% rate of developing an SSI and that those like the 50 year old had a 10% chance.
After all of these calculations NHSN felt very confident in saying that these past calculations could be used to “predict” future events. They call their predictions EXPECTED values. And they compare these expected numbers to your new submissions (called the OBSERVED value) to see how well your facility is doing. (You might ask if it is “fair” to apply old information to new situations. After all, new procedures will be developed that may change the risks. The answer is yes. Scientists agree that carefully analyzed historical data is the best information we have to apply to the future. In addition, as new data come in they will be added to the pool of information and rates will be recalculated.)
To calculate the SIR, NHSN uses expected cases instead of percentages. (Not sure why they do this since they could compare percentages, but they apparently though the numbers of cases was easier to understand). The formula for the SIR is: observed cases/expected cases.
The SIR is usually a fraction because the number of expected cases is often not a whole number. And, you read the number “backwards”. That is any number above 1.0 means you are doing worse than expected. Any number less than 1.0 means you are performing up to or better than standards. Let’s use some examples.
If the observed number is 10 and the expected number is 4.5, your SIR will be 2.2. That means that you are having more than twice as many events as you should be having. If the observed number is 4.5 and the expected is 10, your SIR will be 0.45. That means that you are having less than half the number of events that would be expected.
I realize that this was a long explanation. But I hope that it helped you to understand how NHSN came up with the number and how you can use it to measure your performance. I also realize that what may seem perfectly clear to me, may be pretty muddy to someone else. So if you have any questions, please feel free to post them to the blog. (And remember the person who posts is usually the one who is willing to say what other people are thinking. No question is silly.)
