Antimicrobial stewardship programs (ASPs) require robust assessments of antimicrobial use (AU) to demonstrate impact.Reference Pollack and Srinivasan 1 The preferred metric for AU is days of therapy (DOT) because of several advantages described previously.Reference Polk, Fox, Mahoney, Letcavage and MacDougall 2 – Reference Moehring, Anderson, Cochran, Hicks, Srinivasan and Dodds Ashley 4 One DOT is counted when a single antimicrobial agent is administered on a calendar day regardless of the number of administrations, resulting in whole day counts even for partial days of exposure.Reference Polk, Fox, Mahoney, Letcavage and MacDougall 2 When DOTs are aggregated, an AU rate is calculated over a denominator of person time at risk.Reference Ibrahim and Polk 5 Days of stay on an inpatient unit are considered person time at risk for hospital antimicrobial exposures.
Traditionally, person time has been measured in patient days, a manual or electronic count of the number of patients in the location measured at the same time each day. 6 This metric may miss a partial day of patient exposure either at the beginning or end of a patient stay depending on the time of the daily count. Therefore, the National Healthcare Safety Network (NHSN) AU Option recently introduced a new metric termed “days present” as an alternate measure of person time to capture partial days in hospital locations. 7 Days present is the count of calendar days when a patient is present in the given location for any portion of the calendar day. Days present calculations are challenging because they require electronic capture of continuous admission–discharge–transfer (ADT) data and extensive data cleaning.Reference Moehring, Anderson, Cochran, Hicks, Srinivasan and Dodds Ashley 4 The impact of using the days present metric on hospital- and unit-level estimates of person time at risk has not been described previously. In this study, we aimed to compare patient days and days present to a “gold standard” of person time to quantify how choice of denominator may affect AU rates.
METHODS
We analyzed bed flow data from 5 community hospitals and 2 academic medical centers that participated in the Benefits of Terminal Room Disinfection Study from April 2012 to July 2014.Reference Anderson, Chen and Weber 8 Data from inpatient units were included in the analysis, and emergency department, observation, and procedural unit data were excluded. Bed flow data were prospectively validated using samples of manually documented patient movements. Bed flow data included date–time of room entry and exit measured to the minute. Duplicate room entries were excluded. Extremely short unit stays of <2 hours were excluded because many of these events represented administrative actions and not true patient movements upon validation of bed flow data. Unit type was defined by local infection preventionists using NHSN definitions. 9
Person time was calculated by subtracting date–time of room exit from date–time of room entry. Patient days were calculated using a midnight census count. Unit-level days present were counted if the patient was on an inpatient unit for any portion of a calendar day. When aggregated at the hospital level, an individual patient counted 1 day present on each calendar day; between-unit transfers did not result in double counting for hospital-level estimates, as specified in the NHSN AU option. 7 Percent relative differences (RDs) for patient days and days present were compared to person time among hospitals and units. The RDs were also calculated between days present and patient days.
RESULTS
More than 1.7 million patient days were evaluated during the 28-month period (Table 1). Median length of stay was 2.9 days per stay (interquartile range [IQR], 2.5–4.9) among the 7 hospitals and 3.5 days per stay (IQR, 2.8–4.6) among the 120 hospital units. For the hospital-level calculations, patient days were close underestimates of person time, whereas days present calculations overestimated person time (median RD, 33%; IQR, 24%–37%). Compared with community hospitals, the 2 academic centers had larger patient volumes, longer length of hospital stay, higher numbers of between-unit transfers, and lower RD comparing days present to person time. A hypothetical exercise applying these hospital-level denominators to a DOT numerator is provided in the Supplemental Material.
TABLE 1 Relative Differences comparing Days Present and Patient Days Among 7 Hospitals, 2012–2014Footnote a
NOTE. C, community hospital; A, academic hospital; IQR, interquartile range; ICU, intensive care unit; SD, standard deviation.
a Relative differences were calculated as follows: relative difference between days present and person time = (days present – person time)/person time. Average length of stay was calculated using person time.
In unit-level analyses, days present also overestimated person time. However, the magnitude of the RD differed by unit type. The highest RDs were seen in unit types with short stays and historically lower AU (eg, cardiology step-down units and labor and delivery units). The lowest RDs were seen in unit types with long stays (eg, bone marrow transplant units and burn units).
DISCUSSION
To our knowledge, this is the first comparative description of 2 denominators used to represent patient time at risk for antimicrobial use. Patient days, the traditional infection prevention denominator that counts at a single time each calendar day, may miss a partial day at risk on the day of admission or discharge, depending on the time of the daily census count. The newer days-present metric attempts to address this by counting all partial days. When aggregated, the additional time resulting from partial days increased AU rate denominator counts substantially. In our analysis, days present counts were approximately one-third higher than person time rounded to the nearest minute. Relative differences varied among hospitals and units and was highly correlated to length of stay.
Our findings have important implications for AU assessments. First, AU estimates using days present will be substantially lower than those using patient days. Thus, stewards need to carefully delineate days present versus patient days when interpreting and time-trending local data and when comparing local AU estimates to published literature or publically available AU estimates. Similarities in these terms and abbreviations may cause confusion. Second, the impact of short-stay patients has implications for hospital and unit comparisons. The presence of extra time in aggregated days present estimates will result in lower AU estimates in locations that care for patients with short stays. High-volume units with short stays (eg, labor and delivery wards and nurseries) have been considered lower-risk areas for antimicrobial exposure, but these had the highest RDs in our study. Stenejhem et alReference Stenehjem, Hersh and Sheng 10 described how the inclusion of these “miscellaneous” units inflated facility-wide denominators and affected the utility of facility-level comparisons; ultimately, they decided to exclude those units when benchmarking. Similarly, the NHSN AU SAARs (standardized antimicrobial administration ratios) exclude all units except general medical/surgical wards and medical/surgical intensive care units. 7 Eliminating units from analyses limits information gained about these patient populations and excludes them from assessment for improvement opportunities. Third, we observed that the RDs between patient days and days present varied among hospitals and that academic hospitals had lower RDs. We hypothesize that this observation is related to complex case mix and its association with longer length of stay. Risk adjustment methods for hospital benchmarking may help improve comparisons; however, the large effect from length of stay may be difficult to fully overcome with risk adjustment.
The advantages of the days present metric include the ability to participate in the NHSN AU option and access national data for comparisons. Alternatively, patient days are readily available, actively used by infection prevention, and do not require additional resources from information technology (Table 2). Individual hospitals should choose a single AU denominator metric based on their available resources and needs, then standardize terminology to most effectively interpret and externally share analyses of ASP impact.
TABLE 2 Antimicrobial Use Denominator Metric Summary
NOTE. NHSN, National Healthcare Safety Network; AU antimicrobial use.
This study has several limitations. The 7-hospital study sample may not be large enough to fully describe the comparisons. Included hospitals are in the southeastern United States and may differ from hospitals in other geographic locations and practice settings. Existing data within the NHSN could be used to validate these findings.
In summary, days present denominators increased days at risk estimates by approximately one-third when compared with patient days. This effect differed among hospitals and units and was highly influenced by short length of stay.
ACKNOWLEDGMENTS
Financial support: The Benefits of Terminal Room Disinfection study was funded by the Centers for Disease Control and Prevention Epicenters Program (grant no. 1U54C000164 to D.J.S.). R.W.M. was supported by the Agency for Healthcare Research and Quality (AHRQ grant no. K08 HS023866). D.J.A. was supported by the National Institutes of Health (NIH grant no. K23 AI095357).
Potential conflicts of interest: All authors report no conflicts of interest related to this article.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2018.54.
