|
|
 |
|
How HAI Statistics Are Calculated |
What does the Department of Health do with the data to produce these reports?
We follow guidelines issued by the National Quality Forum and other organizations to produce meaningful, accurate, and dependable information. We use established methods of epidemiology to produce the type of rates suitable for comparison purposes. We use statistical methods to determine whether differences in rates are large enough or persistent enough to be significant. If we find significant differences, we work with individual hospitals to investigate their root causes. We also consider the state of their infection surveillance and reporting.
How do we know the numbers are accurate?
Reporting doesn’t have to be perfect to be useful, but it has to be accurate enough to be meaningful. Accuracy is checked in three ways. First, the Department of Health looks at all reported information every couple of weeks to see if it makes sense. We look for internal inconsistencies that suggest errors. This starts with inspecting a list of conferred rights to be sure hospitals have not changed our ability to see their information in the NHSN system. We confirm that they have entered data current to within 30-60 days of each month-end reporting period. Exploratory data analysis methods are then used to see whether numerator and denominator data are within reasonable ranges for the size and type of each hospital. Finally, we look for consistency and trends from month to month within the rates of infection.
In addition to this, the Department will begin to conduct spot checks in 2010 by sending one of its program consultants into selected hospitals each month where they will review randomly selected cases to see if those cases were coded and reported correctly. Our consultant will check three things:
- What percentage of actual infections was detected and reported (technically, this is the sensitivity of a surveillance program);
- The extent to which patients who do not fit surveillance definitions were incorrectly coded as infected (technically, the specificity of a surveillance program will be computed); and
- Reasons for any miscoding detected (e.g. mistakes in interpretation of NHSN infection definitions, lapses in communication between information sources within a hospital, not enough staff to do all the work, etc.).
This information will be used to develop corrective action plans where necessary.
Finally, hospitals rely on their own infection surveillance program findings to support decisions about policies, practices and special studies. Therefore, it is a hospital management responsibility to monitor and maintain sufficient accuracy in their surveillance program. The Department of Health provides instructions and resources to Washington’s hospitals so that they all have a uniform way of conducting their own monitoring. These instructions are based on a statistical sampling plan, much as is used in many other industries for quality assurance sampling. For even more details about the sampling plan, please contact the Department’s HAI Program staff.
What is a "confidence interval" or a "significant difference"?
A confidence interval is a statistical measure of the precision of a rate estimate. It is a plus-or-minus range around the infection rate reported. For example, a 95% confidence interval means that if we repeated the sampling of rates over more periods of time, 95 times out of 100 we would expect the value to fall within the range shown. In order to ensure that differences between infection rates are real and not random variation, we measure rate difference in terms of how wide the confidence intervals are. This allows us to see if a difference is statistically significant.
Since larger hospitals treat more patients, they can report their infection rates with more precision than smaller hospitals. The larger the sample size, in this case the number of patients, the narrower the confidence interval range. We can be more confident that a hospital’s performance falls within the range shown by its confidence interval, but less sure that its rate of performance is exactly the rate shown.
Why aren't all the infection rates shown as simple percentages?
We show incidence rates, which means the rate at which new infections occur during each time period shown. We use two kinds of incidence rates: an attack rate and an incidence density rate. An attack rate tells what percentage of patients is affected. We use this to measure surgical site infection rates because the risk is having surgery and the percentage of patients who become infected is meaningful for comparisons. An incidence density rate considers how long patients remain at risk. When we calculate rates for other sites of infections, like central line-associated bloodstream infections and ventilator-associated pneumonia, we consider the number of days that patients are exposed to those devices. An incidence density rate allows us to make meaningful comparisons between hospitals or between different patient care areas within a single hospital. Incidence rates do not include infections that began in an earlier period but are still present.
|
|
|
