Discussion

Using the SAS program created here (Appendix), all US hospitals can set 75^th percentile HAI threshold counts, with freedom to explore alternative thresholds such as top decile or tertile SIR. Our analysis also showed that at least 2/3 of US hospitals must achieve zero HAI counts in order to perform in the top quartile SIR nationally for five of six HAIs. For example, 70% of US hospitals must have zero CAUTIs per year in order to achieve top-quartile performance on CAUTI SIR. It should be noted that this analysis is based solely on National Healthcare Safety Network (NHSN)-reported units, which does not necessarily include all HAIs occurring in a hospital (i.e. those occurring in non-NHS reportable units would not be included in threshold counts). This tool differs from existing tools such as the NHSN Targeted Assessment for Prevention Reports in that our program allows for a user-specified threshold target at any national percentile from 0 to 100 (eg: 25^th percentile, 50^th percentile, 66^th percentile, 95^th percentile, etc.) rather than having to select and interpret a meaningful SIR-based target, which can be difficult to contextualize in the absence of national percentile performance data. There was notable variability and zero-inflated distributions for mean HAI threshold counts across hospitals, which points to the need for the percentile-based threshold benchmarking offered by our SAS program versus an SIR-based approach which may not be as intuitive.

At our institution, quality leadership has found that using count-based thresholds is more practical for communicating performance goals to clinical department chairs and contextualizes that “zero harm” truly means zero infections if we wish to perform in the top quartile. The results of this analysis highlighted that top-quartile SIR performance can only be achieved by allowing zero HAIs per year for the majority of US hospitals. Also notable is the inter-hospital variation in count thresholds. Whereas 2 out of every 3 hospitals cannot allow a single CLABSI in order to perform in the top quartile, the highest volume US hospital can allow 46 CLABSIs per year and still achieve a top quartile SIR. Though this is a well-documented statistical artifact of the HAI methodologies, ^{Reference Masnick, Morgan, Sorkin, Macek, Brown, Rheingans and Harris2} we suspect that most hospital quality leaders have never contemplated the extent of variation through this lens. This is also a limitation of tools such as our, as they may be irrelevant for small or critical access hospitals with expected event counts of zero, however for these hospitals lower percentile thresholds such as the 26^th percentile (i.e. better than worst quartile) could be explored using our tool. Our analysis highlights that “zero-count” 75^th percentile goals may be implausible for both large and resource-limited hospitals, and a focus on processes for HAI prevention may be a more pragmatic focus. Alternatively, phased reduction goals such as 20^th percentile performance in the first year, followed by 30^th percentile the next year, may be a better aim. This analysis ultimately raises questions about the practicality of national HAI quality improvement efforts for small hospitals, and may support a national cumulative attributable difference strategy targeting larger hospitals. ^{Reference Soe, Gould, Pollock and Edwards3} Regardless, the program provided here is intended to be used for both practical and educational purposes at a hospital-specific level rather than nationally. With minimal edits to column and file names, our program could also be used to set percentile-based and count-based goals for any of the hospital-level data on CMS Care Compare, including patient safety indicators, patient experience, readmissions, and mortality. Using the SAS program shared here, all US hospitals can calculate hospital-specific HAI threshold counts for achieving a pre-specified benchmarked SIR performance percentile.