Protecting critically ill patients from drug-related harm and adverse neuropsychiatric outcomes

Abstract

Patients admitted to the intensive care unit (ICU) are critically ill and depend on life-sustaining therapies, exposing them to multiple potent intravenous drugs. This places them at risk of drug-related harm and adverse outcomes. This thesis is framed by two developments in intensive care: an increased attention to patient safety and to neuropsychiatric outcomes. Its overarching objective was to improve outcomes of critically ill patients by protecting them from drug-related harm and adverse neuropsychiatric outcomes, investigating interventions across the medication use process in adult, pediatric and neonatal ICUs, and the association between drug exposure and neuropsychiatric outcomes. Patient safety. Adverse drug events (ADEs) may be non-preventable adverse drug reactions or preventable medication errors (MEs), which can occur at any stage: prescribing, preparation and dispensing, administration and monitoring. Prescribing. Prescription errors account for over half of MEs in hospitalized patients, and children are especially vulnerable. Clinical decision support systems (CDSSs) can reduce MEs but risk alert fatigue. In Chapter 2, a CDSS in the pediatric ICU reduced protocol deviations from 0.89% to 0.49% of prescriptions (p=0.02), with prescriptions needing adjustment falling 77% (p=0.03). Preparation and dispensing. Dispensing IV drugs as ready-to-administer products reduces errors, but aseptic preparation limits shelf life to 31 days, contributing to drug waste — a system-level harm given climate change. In Chapter 3, replacing aseptically prepared syringes with prefilled sterilized syringes reduced drug waste from 31% to 5% (p<0.0001). Administration. Between 32% and 56% of ICU MEs occur during administration. In the neonatal ICU, physicochemical incompatibilities are critical. In Chapter 4, alprostadil and vancomycin were physically compatible with neonatal parenteral nutrition, whereas insulin aspart and lidocaine formed subvisible particles exceeding thresholds; in-line filters are advised when co-administration is unavoidable. Monitoring. Therapeutic drug monitoring optimizes dosing but requires defined toxicity thresholds. In Chapter 5, a critically ill adult treated with benzylpenicillin experienced seizures after cessation of hemofiltration, with a serum concentration of 96 mg/ml — identifying the culprit drug and suggesting a plausible toxicity threshold. Neuropsychiatric outcomes. As more patients survive, focus has shifted to how they survive. Delirium. Delirium is associated with adverse short- and long-term outcomes; drug exposure is a modifiable risk factor. In Chapter 6, delirium risk was higher after deep sedation with midazolam (adj. HR 1.32) or midazolam plus propofol (adj. HR 1.29) versus propofol alone. Long-term impairments. Over half of survivors report physical, cognitive or mental health problems (post-intensive care syndrome). Inflammatory pathways may underlie these. In Chapter 7, no association was found between glucocorticoids and PTSD symptoms one year later (adj. OR 0.99). In Chapter 8, IL-6 inhibitors in COVID-19 patients were beneficially associated with PTSD/anxiety/depression symptoms (adj. OR 0.46). Conclusions. Chapter 9 clusters recommendations as Share, Standardize and a Specialized clinical pharmacist in the ICU, with future perspectives on CDSS contextualization, a greener ICU, transitions of care, and drug-associated delirium. While risks remain, continued research can meaningfully improve outcomes toward a safer, more sustainable ICU.