A Case for Delirium Risk Prediction Models to Aid in Triaging Resources to those Most at Risk an Integrative Literature Review

Abstract Delirium is a complex syndrome resulting from compounding effects of acute illness, comorbidities, and the environment. It results in adverse outcomes: elevated mortality rates, length of stay, readmissions, institutionalization, long-term cognitive changes, and diminished quality of life. The rate of iatrogenic delirium is astounding, ranging from 10%-89%. There are no curative treatments; thus, primary prevention is the key. The purpose of this literature review is to identify and critique the research for the accuracy of risk stratification and feasibility in practice. Support for interventions that prevent delirium is mounting; however, interventions are resource-intensive and often not implemented. Researchers have responded to this problem by developing risk stratification tools to triage interventions toward those of the highest risk. There is evidence that some of the models\u27 implementation is successful; however, they are not yet widely operationalized. A compilation of seven published models of risk prediction was critiqued and compared using the Stetler Model of Evidence-Based Practice as a guiding model. The Newcastle-Ottawa Scale and the Critical Appraisal and the Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS checklist) are employed to aid in the critical appraisal, evaluation of the study\u27s quality, and aid in data abstraction. The models show the ability to stratify risk. Still, their effectiveness in practice cannot be studied without directed interventions because they risk prediction models are created to aid healthcare staff in making clinical decisions. Therefore, a complete clinical pathway with evidence-based interventions should be employed with a delirium risk prediction model to triage the interventions to patients at the highest risk. Recommendations are to implement an automated electronic model (automatic calculation using the EMR or a machine learning model) into clinical practice along with a delirium prevention care pathway. Electronic versions of risk scores allow for an opportunity to achieve clinical efficiency and show statistical superiority to the other models. Published evidence on the impact of the models is diminutive. Their ability to triage patients and aid in clinical decision-making should be published in an impact study. Keywords: Delirium, risk assessment, risk prediction, risk model, risk score, patient safety, patient-centered outcomes researc