What Is Nosocomial? Large Variation in Hospital Choice of Numerators and Denominators Affects Rates of Hospital-Onset Methicillin-Resistant Staphylococcus aureus

We calculated hospital-onset methicillin-resistant Staphylococcus aureus (HO-MRSA) rates for Orange County, California, hospitals using survey and state data. Numerators were variably defined as HO-MRSA occurring more than 48 hours (37%), more than 2 days (30%), and more than 3 days (33%) postadmission. Survey-reported denominators differed from state-reported patient-days. Numerator and denominator choices substantially impacted HO-MRSA rates.


Infect Control Hosp Epidemiol 2012;33(11):1166-1169
Widespread mandates for surveillance and reporting of hospital-onset methicillin-resistant Staphylococcus aureus (HO-MRSA) assume that reported rates can and should be compared. 1,2 However, rate comparisons between hospitals have many caveats. First, case definitions vary. The Centers for Disease Control and Prevention (CDC) definition of healthcare-associated events occurring more than 48 hours after admission 3 is often operationalized as more than 2 or 3 calendar-days after admission to define cases. 4,5 Thus, hospitals using more than 3 calendar-days will classify fewer cases as HO-MRSA than those using more than 2 calendar-days.
Beyond variations in numerator definitions, it is widely assumed that patient-day denominators are uniformly measured. This assumption has not been evaluated. We sought to assess this assumption by comparing data collected by our study survey to line-item length-of-stay data from mandatory state-reported discharge data.
Additionally, HO-MRSA rates, typically reported as cases per 10,000 total patient-days, should restrict denominators to when patients are at risk to become a case. 6,7 For example, if HO-MRSA cases are defined as newly detected MRSA events occurring more than 2 calendar-days after admission, then denominators should exclude the first 2 days of hospitalization because events occurring within 2 days of admission are ineligible to become cases. We also sought to assess how variations in numerator definitions could be mit-igated by using at-risk patient-days calculated from mandatory state-reported discharge data.

METHODS
We surveyed all 32 acute care hospitals in Orange County, California, for infection prevention program numerator and denominator definitions of HO-MRSA rates. Using hospitalspecific definitions, each program provided monthly counts of newly detected HO-MRSA cases (colonization or infection) occurring during a hospitalization or within 2 days of discharge for a consecutive 12-month period between January 1, 2007, and December 31, 2008. Screening cultures were excluded because active MRSA screening was variably implemented across hospitals at the time of the survey. Monthly hospital-wide patient-day denominators were requested to calculate HO-MRSA rates. Patient-day denominators were also obtained from the mandatory California discharge data set, 8 which includes line-item admission and demographic information from all California hospitals. The Institutional Review Boards of the University of California Regents and California State Committee for the Protection of Human Subjects approved this study.
We assessed the proportion of hospitals using HO-MRSA case definitions occurring more than 48 hours, more than 2 calendar-days, or more than 3 calendar-days after admission. We then compared survey-reported patient-day denominators to those calculated from the mandatory California discharge data set. For hospitals defining HO-MRSA as cases occurring after hospital day 2 or 3, we determined at-risk patient days by summing patient-days after removing the first 2 or 3 days of every admission, respectively, using mandatory state discharge data. For the subset of hospitals using calendar-day definitions, we also calculated annual HO-MRSA incidence density per 10,000 patient-days using survey-reported cases as the numerator and 3 separate denominators: survey-reported and state-reported total patient-days, and state-reported at-risk patient-days.
We calculated the percent difference between hospitalspecific survey and state-reported total patient-days and compared these denominators using 2-tailed f tests. For the subset of hospitals using calendar-day definitions, we assessed the percent difference between (a) survey-reported and statereported total patient-days and (b) state-reported total versus at-risk patient-day denominators and HO-MRSA rates using 2-tailed t tests.

RESULTS
Of the 32 total hospitals in Orange County, 5 did not respond. Three hospitals did not provide patient-day denominators by survey, and state denominator data were unavailable for 2 additional hospitals. Among the 27 responding hospitals, 10 (37%), 8 (30%), and 9 (33%) hospitals used case definitions of events occurring more than 48 hours, more than 2 calendar-days, and more than 3 calendar-days after admission, respectively. Significant denominator differences were also found between survey-reported patient-days (mean 39,648; median 38,150) and patient-days derived from mandatory California discharge data (mean 49,014; median 41,163) (n = 22, paired t test, P = .02).
After excluding 1 hospital with incomplete numerator data, 16 hospitals remained that used calendar-day definitions. For this subset, we assessed HO-MRSA rates using 3 denominator choices (survey-reported total, state-reported total, and statereported at-risk patient-days; Table 2). Mean HO-MRSA rates using at-risk patient-days were significantly higher than rates using state-reported total patient-days (mean, 6.5 vs 3.7, paired f test, P = .002). Compared to HO-MRSA rates using at-risk patient-days, there was a larger reduction in HO-MRSA rates when using total patient-days for hospitals using a more than 3 calendar-day versus more than 2 calendar-day definition (50% vs 30%, t test, P = .03).

D I S C U S S I O N
Despite national attempts at standardization, we found that many hospitals used varying methods of HO-MRSA case finding to comply with CDC recommendations for using a window greater than 48 hours. Across a large metropolitan county, approximately one-third of hospitals each used case definitions of 48 hours, more than 2 calendar-days, and more than 3 calendar-days. These differences in definition are not limited to HO-MRSA and will affect the interhospital comparability of all publicly reported healthcare-associated infections. Our data support the need for a standardized definition for comparisons.  " Total patient-days obtained from mandatory California discharge data set. 5 b Adjusted for hospital-specific HO-MRSA definitions. Denominators excluded the first 2 or 3 days of admission in hospitals using >2 or >3 calendar-day definitions, respectively.
Our work not only suggests variation in the collection of denominator-days for healthcare-associated infection rates, but also provides insight into the different ways in which hospital patient-days are calculated. Some hospitals included discharge days and all licensed acute care beds (including psychiatric and rehabilitation beds), whereas others did not. Because hospitals assess daily census at different times of day, this discrepancy will persist despite CDC recommendation to count patients at the same time each day. 9 Additionally, infection prevention programs were often unaware of how patient-day values were calculated. Surveyed programs reported obtaining denominators for hospital-associated infection reporting from numbers posted on internal websites or received from other departments.
The differences in numerator and denominator choices compared to a uniform strategy of calculating patient-days is germane to the public reporting of nearly all healthcareassociated infections. Although many are familiar with inconsistencies in numerator definitions, widespread discrepancies in patient-day denominators were unexpected. Differences in denominator choice caused an average discrepancy in HO-MRSA rates of 20%, with a maximum discrepancy of 90% in 1 hospital. Consistent with prior reports, 6,7 adjusting denominators for at-risk patient-days significantly increased HO-MRSA rates. Importantly, the impact of denominator choices on HO-MRSA rates may be magnified by continued reductions in hospital length-of-stay, with over half of hospitalizations of 3 days or more. 10 This study has important limitations. First, we assessed hospitals in only 1 county. Nevertheless, this study is larger than prior work that was limited to single-institution data. 6 Second, patient-days occurring after conversion to MRS A positivity were not excluded. Finally, it is likely that some case counts represented prevalent carriers. However, this limitation does not affect the study objective.
In summary, we show that use of nonstandardized numerators and denominators for reporting healthcare-associated metrics is widespread and significantly impacts healthcare-associated infection rates. Although internal consistency in numerator and denominator choices can provide hospitals with internal benchmarks for healthcare-associated infection rates, lack of standardization is increasingly problematic in an era of public reporting and interfacility comparisons. The economic and public opinion implications of hospital rates for healthcare-associated infection make the use and implementation of standard definitions critical for healthcare facilities and future comparative research.

ACKNOWLEDGMENTS
Financial support. This study was funded by the National Institutes of Health (MIDAS, grant 1U01 GM076672, to S.S.H.).
Potential conflicts of interest. All authors report no conflicts of interest relevant to this article. All authors submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest, and the conflicts that the editors consider relevant to this article are disclosed here.