Page 80 - JSOM Summer 2019

prolonged ketamine infusion among patients admitted to a from SAS (version 13, Cary, NC). We compared study vari-
burn ICU. ables between patient groups by using a Student t-test for
continuous variables, Wilcoxon rank sum test for ordinal
2
Study Goal variables, and χ test for nominal variables. We reported con-
We aimed to compare inhospital mortality between intubated tinuous variables as means and standard deviations. We re-
burn ICU patients receiving prolonged ketamine infusion last- ported ordinal variables as medians and interquartile ranges.
ing ≥7 days or until death versus patients before the imple- We used logistic regression models to determine associations
mentation of the novel ketamine protocol who were receiving with in-hospital mortality by calculating odds ratios (ORs)
other agents. with 95% confidence intervals (CIs). We further stratified
analyses by burn severity categories based on the TBSA of full-
thickness burns: up to 10%, 11–30%, and >30%.
Methods
The Brooke Army Medical Center regulatory office reviewed Results
protocol C.2016.129n and determined the study met exemp-
tion criteria. We obtained only deidentified data. The initial search yielded 3464 subjects within the data-
base. After the removal of nonburn subjects (i.e., those with
Subjects and Setting Stevens-Johnson syndrome, etc.) there were 2394 subjects
We performed this study at the San Antonio Military Medical available for analysis. Of these, 475 were in the ketamine
Center (SAMMC) in conjunction with the US Army Institute group and 1919 were in the control group.
of Surgical Research (USAISR) burn ICU at Joint Base San
Antonio–Fort Sam Houston, Texas. Clinical providers entered The median age of subjects receiving ketamine was 42 (IQR
data into an Essentris-based electronic medical record (Health 28–56) years versus 40 (IQR 27–56) years in the control group
IT Outcomes, Erie, PA) from which investigators retrieved (p = .465). Subjects in the ketamine group had higher average
study variables using the CliniComp Database (CliniComp In- fluid requirements within the first 24 hours (11,344 ± 8305mL
ternational, San Diego, CA). versus 3553 ± 2963mL, p < .001). Subjects in the ketamine
group had higher rates of concomitant analgesic administra-
In 2012, the USAISR burn ICU implemented a protocol us- tion: 99.8% (n = 474) versus 96.9% (n = 1860, p < .001);
ing prolonged ketamine infusions for pain control for patients higher rates of benzodiazepine administration: 94.3% (n =
with severe burn injuries who were at risk for hemodynamic 448) versus 36.9% (n = 708, p < .001); and higher rates of
instability. We used a time-based cohort design—before and antipsychotic administrations: 40.6% (n = 193) versus 10.1%
after implementation of this protocol. The burn ICU protocol (n = 193, p < .001). Of the total, 1533 in the <10% group,
for ketamine infusion was as follows: For procedural sedation, 586 were in the 11–30% group, and 281 were in the >31%
12.5–25mg IV push was provided every 30 minutes as needed categories (table 1).
for a maximum of 100mg. This was followed by a 5μg/kg/min
infusion that was titrated to a maximum dosage of 30μg/kg/ The median number of ventilator-free days within the first 30
min for continuous sedation. Prior to initiation of ketamine, days did not vary significantly between the ketamine group
patients were premedicated with 0.5–1mg of lorazepam IV and the control group: 8.5 days (IQR 1–16) versus 8 days
push to mitigate against dissociative symptoms. Patients re- (IQR 3–13, p = .442). Subjects in the ketamine group did have
ceiving continuous ketamine infusion were provided 1mg of longer lengths of hospital stay: median 27 days (IQR 14–49)
lorazepam every 6 hours during the course of the infusion. versus 9 days (IQR 4–15, p < .001). Overall, subjects receiving
ketamine had higher mortality rates: 59.4% (n = 117) versus
We included subjects undergoing infusions between January 40.6% (n = 80), p < .001), with an odds ratio for death of 7.51
2012 and April 2016. We searched for all subjects who un- (95% CI 5.53–10.20, p < .001). We then performed a multi-
derwent intubation within the first 24 hours of admission and variable logistic regression analysis controlling for factors that
received a ketamine infusion (intervention cohort) for at least were significantly associated with death: TBSA category, ven-
7 days or from admission until death, whichever came first. tilator days and vasopressor administration. When controlling
We then searched for similar subjects from January 2008 to for these confounders, there was no longer an association be-
December 2011 who received any drug except for ketamine to tween ketamine and in-hospital mortality: OR 0.66 (95% CI
serve as the control cohort. We excluded subjects if they died 0.41–1.05, p = .08).
within 24 hours of admission or they underwent admission for
pathologies unrelated to thermal burns (e.g., Stevens-Johnson Discussion
syndrome). Our primary outcome was in-hospital mortality.
In this study, we report the outcomes of 475 subjects receiv-
CliniComp Database ing a prolonged ketamine infusion for analgesia versus 1919
CliniComp provides a secondary database for reporting pur- matched controls in a burn ICU. Overall, subjects receiving
poses called Global Data Repository (GDR). This database ketamine infusions were more critically ill with more extensive
comprises an Oracle 11g database that receives transactions burns, higher proportions of inhalational injuries, higher rates
from the main Essentris system in a near-real-time fashion. of concomitant vasopressor administration and higher mor-
Standard Structured Query Language (SQL) client applica- tality rates. However, when controlling for observable con-
tions extract data from this database. founders we noted no association between ketamine receipt
and in-hospital mortality in this critically ill group.
Data Analysis
We performed all statistical analysis by using Microsoft Ex- In contrast to many previous reports that focused on short-
cel (version 10, Redmond, WA) and JMP Statistical Discovery term markers such as vital signs or asthma metrics, 4–7,11 we

78 | JSOM Volume 19, Edition 2 / Summer 2019