IDSA guidelines recommend that antimicrobial stewardship programs measure their institution's antimicrobial use. Although the standard benchmark for measurement of antimicrobial use is medication administration record (MAR) data, administrative data are often more accessible. To determine whether administrative data was reliable for the evaluation of antimicrobial use, we examined performance characteristics of administrative data relative to MAR data.
We obtained 1) administrative pharmacy data from the Pediatric Health Information System (PHIS) and 2) MAR data for all systemic antibiotics associated with inpatient and observation visits for patients discharged from a large, tertiary care children's hospital between 1/1/2009 and 12/31/2013. Rates of antibiotic use, aggregated by month and quarter, were compared between these sources. We also calculated performance characteristics of the administrative data for several measures of antibiotic use, with the MAR as the reference.
We matched 143,325 admissions and 851,986 patient days. Prior to a new electronic medical record (EMR) that integrated billing and pharmacy processes, administrative data overestimated antibiotic use (Figure 1). With the new EMR, administrative data was 99.7% sensitive and 93.0% specific for any antibiotic administration, regardless of specific drug, during a given admission. For a given patient day, administrative data was 87.4% sensitive and 95.9% specific for antibiotic administration, also regardless of particular drug. At the level of patient day, for exact drug, specificity was> 98.5% for all antibiotics, while sensitivities were 1.7-96.1%; 25/35 antibiotics had sensitivities ≥ 85% (Figure 2, antibiotics with ≥ 250 total days of therapy included).
Integration of pharmacy and billing processes improved the accuracy of administrative data for measurement of antibiotic utilization. Although the relationship between pharmacy and billing systems likely varies across health systems, administrative data can be an effective surrogate for true antibiotic utilization with good sensitivities and high specificities. The reliability of these data, however, varies across specific antibiotics.
J. M. Beus,
T. A. Metjian, None
J. S. Gerber, None
See more of: Poster Abstract Session