Development of Computerized Clinical Decision Support to Assist in Detecting and Preventing Delirium in the Hospital Setting

Delirium is an acute confusional state commonly affecting hospitalized older patients. As a part of the Improving Care of Vulnerable Elders (ICOVE) project, interventions were designed to improve delirium screening and reduce use of strong anticholinergic medications while avoiding contributing to alert fatigue. Methods: Baseline compliance was assessed for each step in the workflow. The intervention was designed in cooperation with the clinicians who are responsible for those steps and built using the integrated decision support facilities of the electronic patient record system. Compliance after the intervention is assessed using the same computerized measures used to determine the baseline and using statistical process control charts. Results: Baseline compliance for delirium screening was 62.3%. Baseline compliance for avoiding strong anticholinergics was 84.3%, but varied from 17.2% to 97.3% for different therapeutic classes. Clinicians asked for a column added to a patient summary table to indicate whether delirium screening had been performed, and a non-interruptive onscreen alert for strong anticholinergic medications. However, the prescribing alert could only be implemented as a modal dialog alert or a passive alert accessed through clicking a tab in the patient record. We implemented the modal dialog alert for orders and the passive alert for active medications. Conclusions: We were able to perform computerized assessment of delirium screening and avoidance of strong anticholinergic medications, although both rules contained ambiguous terms which needed to be defined for the intervention. Assessment of multiple outcomes proved useful both for assessing the quality of care and for identifying where additional support was needed. Assessing the specificity of inferences about data not recorded in structured fields also revealed areas for improvement that would otherwise be missed. Additional functionality for flexibility in decision support in electronic pat- ent record systems may be needed to combat alert fatigue, although this must be weighed against the need to maintain a consistent user experience.