Study design and setting
This was a single-center retrospective observational study designed to evaluate the effect of VPA on agitated adult patients admitted to the medical, surgical and trauma ICU at the Montreal General Hospital (MGH). Patients hospitalized between January 1st 2015 and December 31st 2021 were eligible for inclusion. Adult patients admitted to the MGH ICU during this period were included if a psychiatry consultation for agitation was requested during the ICU admission by the treating team, if they remained in the ICU for a minimum of 2 days following consultation and, if VPA was prescribed, it was administered for a minimum of 2 days. Patients were excluded if they were already on VPA prior to hospitalization, had a documented allergy or intolerance to VPA, or a contraindication to VPA (pregnancy, breastfeeding, active hepatic disorder, urea cycle disorder, presence of a drug interaction with carbapenems if concomitant administration for more than 48 hours). Patients were censored at death or 2 days following VPA cessation. Deidentified data were collected from patients’ files hosted in the MUHC electronic archives and entered in the web-based REDCap database.
Objectives and Outcomes
The primary objective was to describe the evolution of agitation of ICU patients referred to a psychiatry consultation service due to agitation and treated with VPA as adjuvant therapy. The primary outcome was the proportion of patients with agitation-free days during follow-up period, from Day − 2 to Day 7 with Day 1 defined as day of VPA initiation (see electronic supplementary material 1). An agitation-free day was defined as a Richmond Agitation-Sedation Scale (RASS) score of less than 2 with no mention of agitation related terms in the nursing notes (see electronic supplementary material 2) or a transfer to a ward. Secondary outcomes were the proportion of patients with an agitation-free day on Day 3 and on Day 7 compared to Day 1, the daily proportion of patients with at least 1 agitation-related events (ARE) during follow-up period and the evolution of concomitant medication use, in terms of proportions and median dose received, during follow-up period. An ARE is defined as an unplanned extubation, a medical device removal (such as lines, catheters or drains), a fall from bed, an attempted escape from the ICU, removing physical restraints, self-harm or an assault on medical personnel.
The second objective was to identify demographic and clinical predictors of VPA response, namely an agitation-free day with RASS less than 2 on Day 7 following VPA initiation or a transfer to a ward during follow-up period. The outcome for this objective was the probability of being agitation-free on Day 7.
The third objective, exploratory in nature, was to evaluate the independent effect of VPA as adjuvant therapy for treatment of agitation in ICU patients. The outcome evaluated was agitation-free days for every day of follow-up period, in patients exposed to VPA versus those not exposed to VPA. The follow-up period for this control group was from Day − 2 to Day 7, with Day 1 defined here as the day of psychiatric consultation (see electronic supplementary material 1).
Data collection
Demographic and clinical variables, including cause of agitation (as documented in the psychiatry consultation), reason for ICU admission, psychiatric comorbidities, substance and tobacco use, Acute Physiology and Chronic Health Evaluation II score (APACHE II) and Injury severity score (ISS) for the traumatic subgroup, ICU and hospital length of stay and duration of mechanical ventilation were collected.
Total VPA daily dose, route of administration, loading dose, serum drug level, and response variables consisting of RASS score, use of co-medications and agitation-related events were collected daily starting 2 days prior to VPA exposure, up to 7 days after or until 2 days after VPA discontinuation or death.
Total daily doses of opioids, benzodiazepines, ketamine, dexmedetomidine, clonidine, propofol, gabapentinoids and antipsychotics including as needed doses, were documented. All opioids were expressed using parenteral hydromorphone equivalents while benzodiazepines were expressed using parenteral lorazepam equivalents.
Statistical analyses
For both primary and secondary objectives, patients exposed to VPA for at least 48 hours were included in the analyses. For the primary objective, descriptive statistics were presented as medians with interquartile ranges or frequencies with percentages, with error bars as 95% confidence interval as appropriate. Control of agitation was compared between Day 1 and Day 3, and between Day 1 and Day 7 using a McNemar’s test.
For the secondary objective, a logistic regression model was used to evaluate factors associated with a response to VPA on Day 7. Univariate logistic regression analyses were performed on all baseline characteristics collected and variables with p-value < 0.15 were included in the multivariate logistic regression model.
For the third objective, patients exposed and non-exposed to VPA were included in the analysis. All patients were followed from Day − 2 to Day 7, with Day 1 defined as the day of the psychiatry consultation. Generalized Estimating Equation (GEE) models were used to predict the odds of a non-agitation day in a delay of 2 days after exposure to VPA. Mann Whitney U tests, Student T-tests and Chi-square tests were performed on all baseline characteristics collected and variables with p-value < 0.15 were included in the multivariate model. The GEE models were successively adjusted for individual and combined co-medication exposures in the past 2 days (opioids, benzodiazepines, propofol, antipsychotics, alpha-2 receptor antagonists, all co-medications), baseline characteristics with p values < 0.1 (age, gender, APACHE II score, admission diagnosis, traumatic brain injury (TBI) for cause of agitation, delirium for cause of agitation, current or past alcohol consumption, drug withdrawal), an inverse weighted propensity score, and inverse weighted propensity score with co-medication exposures. We also calculated a non-adjusted raw GEE model. The propensity score evaluating the probability of being exposed to VPA was developed using all baseline variables with a multivariate stepwise forward logistic regression based on the likelihood ratio. If a patient was transferred to the ward during follow-up, a last-observation-carried-forward imputation was performed for all medications, besides propofol where a non-exposed imputation was performed, as propofol is only administered in the ICU at our institution. Statistics for the tertiary objective were performed by a registered statistician (M.C.). Statistical analyses were completed using SPSS version 27 and SAS version 9.4 softwares with statistical significance defined as p-value less than 0.05.