2026 Southeastern Residency Conference

Background: Concomitant administration of carbapenems and valproic acid (VPA) is a well-documented interaction that leads to rapid reductions in serum VPA concentrations with effects observed within 24 hours and persisting up to 7-10 days after discontinuation. This interaction increases the risk of breakthrough seizures, mood destabilization, and is associated with adverse clinical outcomes, including prolonged hospital stays and increased mortality, particularly in critically ill patients. Despite awareness, concomitant prescribing persists in large medical centers. At Grady Health System (GHS), prescribers and pharmacists receive a drug interaction warning when VPA and carbapenems are ordered together. Prior studies report only ~37% compliance with similar alerts, with over 50% of patients developing subtherapeutic VPA levels. Although drug interaction warnings are designed to mitigate this risk, it remains unclear whether these alerts effectively influence healthcare providers' behavior in hospitalized patients. This study aims to evaluate the clinical effectiveness of drug interaction warning targeting the co-administration of carbapenem and VPA and subsequent clinical sequelae at GHS.
Methods: This single-center, retrospective chart review included adults (≥18 years) admitted to GHS between January 1, 2020, and April 1, 2025, who received at least one overlapping dose of VPA and a carbapenem with a triggered drug interaction alert. Patients were excluded if discharged within 48 hours, receiving VPA and carbapenems prior to admission, or if clinical sequelae were documented prior to concomitant exposure. Data collected included demographics, indication for therapy, alert response, and resulting clinical actions (dose adjustment, discontinuation, or substitution). The primary outcome was the proportion of alerts with action versus no action at order entry or verification. Secondary outcomes included the type and frequency of resulting actions, incidence of clinical sequelae, hospital length of stay, duration of antimicrobial therapy, and documented indications for both VPA and carbapenem therapy Descriptive statistics were used, and chi-square and Fisher’s exact analysis evaluated associations between responder type and alert response.
Results: A total of 299 drug interaction alerts across 56 unique patients were analyzed. Clinical action was taken for 65 alerts (22%), while 234 alerts (78%) resulted in no action at order entry or pharmacist verification. Pharmacists were more likely than other providers to override alerts (84.0% vs 71.5%; χ² = 6.73, p = 0.009). Internal Medicine most frequently accounted for the provider responsible for the must-action (45.9%), followed by Infectious Diseases (12.6%). Given that carbapenems are restricted antimicrobials, Infectious Diseases consultation was commonly involved. At the patient level, clinical sequelae occurred in 40% of patients in the action group and 48% in the no-action group (Fisher’s exact p= 0.59). Median hospital length of stay was 22 versus 17 days, and median antibiotic duration was 3 versus 5 days, respectively. Agitation and behavioral changes were the most common sequelae in both groups, while breakthrough seizures and ICU transfer occurred less frequently and at similar rates. When action was taken, the most common interventions were switching the carbapenem, discontinuing valproic acid, or increasing the valproic acid dose.
Conclusions: Among 299 alerts, concomitant VPA and carbapenem use remained common, with alerts most often resulting in no action with no statically significant difference in clinical sequelae. Notably, nearly half of the patients in the no action group experienced adverse clinical outcomes, emphasizing clinical concern. Future research should assess pharmacist education, provider feedback, and alert optimization to improve compliance.