 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |  |
 |
|
|
|
||||||||||||||||
|
|
|
||||||||||||||||
 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
Canada Communicable Disease Report
- Supplement INFECTION CONTROL GUIDELINESPreventing Infections Associated with Indwelling Intravascular Access DevicesAPPENDIX IV Calculating Infection Rates Different ways to calculate rates of infections associated with intravascular access devices have been used in past years. However, in order to make meaningful comparisons, data must be presented in an epidemiologically valid format(19). This involves three objectives:
The first two objectives can be achieved by adopting validated criteria to define infections, allocating sufficient resources to monitor clinical and laboratory records for infection and device use, and periodically validating the surveillance system through independent review. The third objective can be achieved by expressing infection rates in terms of infections per 1,000 device-days, keeping separate statistics for distinctly different devices. A common practice of dividing number of infections by number of patients with devices (an "attack rate" [see Appendix III]) fails to adjust for the fact that some patients have intravascular access devices for longer than others. I. Calculating Infections per 1,000 Device-Days The following example for central venous line surveillance in an ICU has been provided to facilitate understanding of how to calculate infection rates. In example A, patient #1 had 3 days of exposure to risk, patient #2 had 4 days, patient #3 had 7 days, and so forth. The total number of days at risk for the 10 patients is 38 days. The number of infections (1) divided by the number of device-days (38) multiplied by a scale factor (1,000) yields an overall estimate for the incidence density rate of 26.32 infections per 1,000 device-days. II. Interpreting Results The best surveillance approach should strike a balance: use long-term periods to estimate baseline rates as precisely as possible for comparison with benchmark or literature based targets, but monitor for trends frequently. Confidence intervals and statistical process control (SPC) charts are useful tools for these two tasks.
Periodically, an institution's rate of infection should be compared with benchmark targets (best of class) obtained from organizations with similar data and with published rates in the literature (Table 5). This will allow each institution to determine how its rate of infection compares with the "best of class" or a "larger population" average rate. Since variations in infection rates will be different between institutions, even within patient groups, an upper and lower infection-rate range of acceptable rates should be determined appropriate to the individual institution or setting. A confidence interval (CI) is a statistical tool used to estimate the expected width of such a variation. The width of confidence intervals decreases as the sample size increases, (e.g., year's accumulation of device-days provides a more accurate estimate of the true rate than month's accumulation). For example, a target infection rate for central venous catheters on a medical-surgical ICU may be 1.3 infections/1,000 device-days with a 95% CI lower limit of 0 infections/1,000 device-days and a 95% CI upper limit of 8.7 infections/1,000 device-days (obtained from the literature). If the patient group in Example A was of medical-surgical patients with central venous catheters, the infection rate would be 1 infection/38 device-days or 26.3 infections/1,000 device-days (1/38 x 1,000 = 26.3) and since this does not fall within the lower and upper CI limits (between 0 and 8.7 infections/1,000 device-days) this patient group's infection rate can be said to be statistically significantly different from the target rate. When an institution's infection rate is higher than the upper confidence interval it is due to either a) a statistically significant higher infection rate than the target benchmark or b) the sample size or number of device-days (only 38 device-days in Example A) being too small and not representative of the "true" infection rate. In cases such as b), statistical adjustments can be made for small samples. If the small size of the sample (in this case the number of device days) may be influencing the results, then consult a statistician or epidemiologist for advice.
At more frequent intervals, each facility should compare its current rates of infection with its previous rates. This section describes how trends can be detected through the technique of statistical process control (SPC) charting(131). If each month's rate of infection (or just the number of infections, if the number of device-days remains relatively uniform) were marked sequentially on a chart, some variation would be expected from month to month with some values above average and some below. SPC charting plots just such a chart, but also adds objective interpretive criteria. There are different kinds of SPC charts, depending on which mathematical distribution best fits the data; the "c" and "u" SPC charts add control limits based on the Poisson distribution. By setting warning limits at 2 standard deviations above the average count, a monthly incidence above this limit can be identified as indicating potential problems and appropriate investigation can be initiated(132,133). This approach may be further refined by applying supplementary decision rules or more sophisticated charting techniques (e.g., "g" or "h" charts)(134).
[Previous]
[Table of Contents] [Next] |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|||
|
Last Updated: 2002-11-08 |  |
|
