Development and evaluation of the automated risk assessment system for multidrug-resistant organisms (autoRAS-MDRO)

Abstract

A high proportion of infections acquired in hospitals are caused by multidrug-resistant organisms (MDROs). The priority in MDRO prevention is to detect high-risk patients and implement preventive intervention as soon as possible. To develop an automated risk assessment system for MDROs (autoRAS-MDRO) to screen for patients at MDRO infection risk and evaluate the predictive validity of the autoRAS-MDRO. Data for 4200 variables were extracted from the electronic health records (EHRs) for constructing the MDRO risk-scoring algorithm, which was based on a logistic regression model. The autoRAS-MDRO was designed such that the MDRO risk classification (high, moderate, low risk) could be automatically displayed on the nursing Kardex screen in the EHRs system. For the development of the MDRO risk-scoring algorithm, 1000 patients with MDROs and 4000 patients without MDROs were selected; similarly, for the evaluation, 2173 and 8692 patients with and without MDROs, respectively, were selected. The predictive validity of the autoRAS-MDRO was as follows: (i) at the 6-month evaluation: sensitivity, 81%; specificity, 79%; positive predictive value (PPV), 49%; negative predictive value (NPV), 94%; and Youden index, 0.60; (ii) at the 12-month evaluation: sensitivity 79%, specificity 78%, PPV 47%, NPV 94%, and Youden index, 0.57. The autoRAS-MDRO had moderate predictive validity. It could be useful in redirecting nurses' time and efforts required for MDRO risk assessment and implementation of infection control measures, and in reducing the incidence of MDRO infection in hospitals, thereby contributing to patient safety.