Background: Antibiotic prescribing for ICU patients is often done empirically, before lab tests results are available. The general prevalence of bacteria and antibiotic resistance in a hospital or the ICU are commonly used together with patients' characteristics to guide the choice of antibiotic treatment. However, prevalence of infection changes over time and location within a hospital. We investigated whether lab tests results and patient mobility data that are recorded routinely in hospital information systems could be used to calculate individual estimates of infection for patients in the ICU.

Purpose: To examine the viability of calculating individual bacteria and antibiotic resistance estimates for each patient in a ICU, thereby creating an accurate reflection of each patient's personal exposure.

Methods: The data included all 7893 microbiology lab tests results that were ordered for 960 patients in a 27 bed ICU of a university hospital from September 2004 to September 2005. Screening is performed routinely for MRSA (Methicillin-Resistant Staphylococcus Aureus), C.difficile and VRE (Vancomycin-Resistant Enterococcus). Tests results include a date, type, sampling site, bacterial isolate, and antibiogram. By aggregating over locations, times and results, we calculated a cumulative exposure index for each patient by pathogen.

Results: Twenty-three patients became infected or colonized with MRSA in the ICU, with 3 time/location clusters that reflected possible MRSA transmission. Other results varied by pathogen, but for many of the nearly 40 infectious organisms, changes in prevalence and possible transmission patterns were identified. Common pathogens like Klebsiella (80 patients) could be subdivided by specific strain information and antibiotic resistance profile to identify an increased risk of a resistant strain according to location and date. With rare pathogens identification of any case represented an increased risk to others. For example, 7 cases of Acinetobacter could be traced back to 2 or 3 infectious episodes.

Conclusion: A patient's individualized risk, based on past exposure, can be calculated from lab test results for other patients that are accessible in automated laboratory information systems. These individualized estimates of infection have the potential to improve prediction of the infectious pathogen and antibiotic resistance compared to the general prevalence. Utilizing this information may improve prescribing decisions.