Data-driven approach to predicting the risk of pressure injury: A retrospective analysis based on changes in patient conditions.

AIMS: To determine the risk of pressure injury development in the intensive care unit based on changes in patient conditions. DESIGN: This retrospective study was based on secondary data analysis. METHODS: Patient data from electronic health records were retrospectively obtained and we included 438 and 1752 patients with and without pressure injury, respectively, among those admitted to the medical and surgical intensive care units (ICUs) from January 2017-February 2020. Changes in patient conditions were analysed based on the first and last objective data values from the day of ICU admission to the day before the onset of pressure injury and categorised as follows: improved, maintained normal, exacerbated and unchanged. Logistic regression was performed to identify the significant predictors of pressure injury development based on 11 variables. RESULTS: The 11 selected variables were age, body mass index, activity, acute physiology and chronic health evaluation II score, nursing severity level, pulse and albumin, haematocrit, C-reactive protein, total bilirubin and blood urea nitrogen levels. The risk for a pressure injury was high with exacerbation of or persistently abnormal levels of nursing severity, albumin, haematocrit, C-reactive protein, blood urea nitrogen and pulse >100 beat/min. CONCLUSION: Periodic monitoring of haematological variables is important for preventing pressure injury in the intensive care unit. REPORTING METHOD: The study followed STROBE guidelines. PATIENT OR PUBLIC CONTRIBUTION: This study contributes to the utilisation of patient data from electronic health records. RELEVANCE TO CLINICAL PRACTICE: In addition to other pressure injury risk assessment tools, ICU nurses can help prevent pressure injuries by assessing patients' blood test results, thereby promoting patient safety and enhancing the efficacy of nursing practice.

Patient Factors That Influence Reliable Delirium Assessments in a Medical Intensive Care Unit During the Routine Clinical Practice.

This study aimed to determine patient and therapeutic characteristics of patients in the medical intensive care unit (MICU) that contribute to inconsistent results of delirium assessments performed during routine clinical practice. Therefore, electronic health records were reviewed and compared with secondary data collected from the same medical ICU patients who were assessed using the Confusion Assessment Method in the ICU (CAM-ICU). Of 5,241 cases involving 762 patients, 827 (15.78%) cases showed disagreement between assessments. Continuous renal replacement therapy, physical restraint use, and altered mental status were factors that increased the likelihood of inconsistencies between assessments. A significant positive correlation was found between the CAM-ICU disagreement rate and the total number of assessments per month. To maximize the reliability of delirium assessments, individual-targeted approaches considering the patient's level of consciousness and type of treatment implemented are required, along with ensuring a stable, and regulated working environment and customized educational programs.

Development and Validation of an Automated Sepsis Risk Assessment System.

Aggressive resuscitation can decrease sepsis mortality, but its success depends on early detection of sepsis. The purpose of this study was to develop and verify an Automated Sepsis Risk Assessment System (Auto-SepRAS), which would automatically assess the sepsis risk of inpatients by applying data mining techniques to electronic health records (EHR) data and provide daily updates. The seven predictors included in the Auto-SepRAS after initial analysis were admission via the emergency department, which had the highest odds ratio; diastolic blood pressure; length of stay; respiratory rate; heart rate; and age. Auto-SepRAS classifies inpatients into three risk levels (high, moderate, and low) based on the predictive values from the sepsis risk-scoring algorithm. The sepsis risk for each patient is presented on the nursing screen of the EHR. The AutoSepRAS was implemented retrospectively in several stages using EHR data and its cut-off scores adjusted. Overall discrimination power was moderate (AUC>.80). The Auto-SepRAS should be verified or updated continuously or intermittently to maintain high predictive performance, but it does not require invasive tests or data input by nurses that would require additional time. Nurses are able to provide patients with nursing care appropriate to their risk levels by using the sepsis risk information provided by the Auto-SepRAS. In particular, with early detection of changes related to sepsis, nurses should be able to help in providing rapid initial resuscitation of high-risk patients. © 2016 Wiley Periodicals, Inc.