Crossroads of Antimicrobial and Diagnostic Stewardship: Assessing Risks to Develop Clinical Decision Support to Combat Multidrug-Resistant Pseudomonas

Abstract

Abstract Background Early detection of multidrug-resistant Pseudomonas aeruginosa (MDRP) remains challenging. Existing risk prediction tools are difficult to translate to bedside application. The goal of this study was to develop a simple electronic medical record (EMR)–integrated tool for prediction of MDRP infection. Methods This was a mixed-methods study. We conducted a split-sample cohort study of adult critical care patients with P aeruginosa infections. Two previously published tools were validated using c-statistic. A subset of variables based on strength of association and ease of EMR extraction was selected for further evaluation. A simplified tool was developed using multivariable logistic regression. Both c-statistic and theoretical trade-off of over- versus underprescribing of broad-spectrum MDRP therapy were assessed in the validation cohort. A qualitative survey of frontline clinicians assessed understanding of risks for MDRP and potential usability of an EMR-integrated tool to predict MDRP. Results The 2 previous risk prediction tools demonstrated similar accuracy in the derivation cohort (c-statistic of 0.76 [95% confidence interval {CI}, .69–.83] and 0.73 [95% CI, .66–.8]). A simplified tool based on 4 variables demonstrated reasonable accuracy (c-statistic of 0.71 [95% CI, .57–.85]) without significant overprescribing in the validation cohort. The risk factors were prior MDRP infection, ≥4 antibiotics prior to culture, infection >3 days after admission, and dialysis. Fourteen clinicians completed the survey. An alert providing context regarding individual patient risk factors for MDRP was preferred. Conclusions These results can be used to develop a local EMR-integrated tool to improve timeliness of effective therapy in those at risk of MDRP infections.