Delirium in Survivors of Cardiac Arrest Treated With Mild Therapeutic Hypothermia

We conducted a retrospective cohort study of patients admitted to the 27-bed cardiovascular ICU, a tertiary level critical care unit with an annual volume of 2800 patients, at Vanderbilt University Medical Center, Nashville, Tennessee. The study period began May 15, 2007, and ended January 1, 2014. Eligible patients included all patients admitted after cardiac arrest. We included the names of all patients treated with therapeutic hypothermia in a prospective registry (International Cardiac Arrest Registry). We then collected data retrospectively and input the information into a REDCap database (REDCap Consortium). We excluded patients if they were persistently comatose or died before awakening from coma.

Treating physicians determined the suitability of therapeutic hypothermia. Patients were considered for inclusion in the study if they had experienced a cardiac arrest with a primary cardiac etiology; therapeutic hypothermia could be started within 12 hours of return of spontaneous circulation (ROSC); and the patients were 18 years or older, remained unresponsive after experiencing ROSC, and had a time to ROSC less than 60 minutes.¹⁹  Physicians cooled patients eligible for therapeutic hypothermia by using an active surface cooling device to maintain a core body temperature of 32°C to 34°C for a total of 24 hours after ROSC and then rewarmed the patients at a rate of 0.25°C/h. Physicians also administered a neuromuscular blocker to all patients; cisatracurium and a sedative (midazolam or propofol) were used during active cooling. The appropriate institutional review board reviewed and approved the study.

In addition to nonmodifiable baseline patient characteristics, we investigated potentially modifiable factors that we a priori hypothesized would have the potential to affect the prevalence and duration of delirium in patients after cardiac arrest. We included factors related to severity of illness, initial resuscitation, and medication dose before a patient reached a temperature of 36°C. Patients’ characteristics included age and comorbid illnesses. We obtained all baseline characteristics via retrospective chart review. We characterized severity of illness using the Acute Physiology and Chronic Health Evaluation II.²⁰  Initial resuscitation factors included occurrence of bystander cardiopulmonary resuscitation (CPR), initial rhythm (ventricular fibrillation or tachycardia, pulseless electrical activity or asystole), time in minutes from arrest to CPR by medical personnel, and time in minutes from CPR by medical personnel to ROSC (determined by review of emergency medical service records). Doses of medication during the therapeutic hypothermia from ICU admission to patient’s arousal (score on the Richmond Agitation-Sedation Scale [RASS] > −4) included total micrograms of fentanyl, milligrams of midazolam, milligrams of propofol, and micrograms of cisatracurium.

We defined duration of delirium as the total number of days of delirium after initial arousal from coma. We assessed delirium status for each 24-hour interval, from the time the patient first reached an internal core temperature of 36°C until discharge from the ICU or death.

We assessed for coma and delirium daily by using the RASS and the Confusion Assessment Method for the ICU (CAM-ICU), respectively.^9,21–23  RASS and CAM-ICU are both previously validated tools with excellent interrater reliability and are suitable for determining the presence and type of delirium in broad populations of ICU patients.⁹  Bedside nurses obtained RASS and CAM-ICU scores a minimum of twice daily for more than 90% of patient days. For the 52 of 772 days (6.7%) for which we had no usable assessments, we used the patient’s status on the day before and the day after the missing day to compute the status for the missing day. Bedside nurses can reliably use the RASS and the CAM-ICU to detect delirious patients, and the RASS and CAM-ICU have been previously validated in the ICUs at Vanderbilt University Medical Center, including the cardiovascular ICU.²⁴ 

Our team considered patients with RASS scores of −5 (unresponsive to physical and verbal stimulus) or −4 (responsive only to physical stimulus) for each 24 hours as comatose and therefore not eligible for evaluation of delirium. Each such day was classified as coma. During a 24-hour period, if a patient’s RASS score reached or exceeded −3, we classified that day as delirious (CAM-ICU score positive at any time during a 24-hour period) or normal (CAM-ICU score negative during all assessments during a 24-hour period). We further categorized delirium, as assessed by the CAM-ICU, into hyperactive (RASS score +1 to +4) or hypoactive (RASS score 0 to −3). We classified a day as hypoactive if all CAM-ICU scores were positive during a 24-hour period with RASS scores from 0 to −3. We classified a day as hyperactive if all CAM-ICU scores were positive during a 24-hour period and the RASS scores were +1 to +4. We classified any day with all CAM-ICU scores positive during the 24-hour period with both hyperactive and hypoactive RASS scores as mixed.

We used means and standard deviations and medians and interquartile ranges (IQRs) as appropriate to describe the patient characteristics of the overall population. The primary objective was to characterize the prevalence of delirium among cardiac arrest patients who were treated with therapeutic hypothermia and successfully warmed and aroused from coma. We report the proportion of patients with at least 1 day with positive CAM-ICU for the total included cohort.

The second aim of our study was to evaluate the potentially modifiable risk factors for delirium associated with resuscitation or therapeutic hypothermia protocols. For this step, we constructed a proportional odds logistic regression model to assess baseline risk factors for duration of delirium. Because of the highly skewed distribution, we used a proportional logistic model to analyze duration. The measure of association between predictors and outcomes in this model type is an odds ratio (OR). Because of the sample size, we could not include all factors collected; doing so would yield model overfitting. In addition to nonmodifiable baseline patient characteristics, we decided a priori to investigate potentially modifiable factors that we hypothesized could affect the prevalence and duration of delirium in patients after cardiac arrest. We included factors related to severity of illness, initial resuscitation, and medication dose before a patient reached a temperature of 36°C. Specific model variables included age, severity of illness, ventricular fibrillation or tachycardia vs pulseless electrical activity or asystole, time to CPR, time from CPR to ROSC, and total doses of fentanyl, midazolam, and propofol during therapeutic hypothermia. We used R statistical software (R Foundation for Statistical Computing) for all statistical analyses.

To address whether or not delirium after warming was due to sedation alone, we divided the population across days between patients who received sedation (propofol, midazolam, or fentanyl) and those who did not. In addition, we analyzed patients given or not given a sedative, excluding fentanyl. Finally, we assessed for the prevalence of and subtypes for each of these subgroups.

A total of 251 consecutive patients were treated with therapeutic hypothermia after cardiac arrest. The analysis includes 107 patients (43%) from the cohort who survived and awoke from coma and excludes the other 144 patients who died before ICU discharge or never had a RASS score greater than −3 that allowed assessment of delirium (Figure 1). Table 1 displays the demographic and baseline characteristics of the study cohort. The median age of survivors to first awakening was 57 (IQR, 47–64) years. The median score on the Acute Physiology and Chronic Health Evaluation II was 21.0 (IQR, 18.0–23.5), reflecting a high severity of illness. The majority of arrests had shockable rhythms (79%) and occurred out of the hospital (77%). Most patients received fentanyl for pain during therapeutic hypothermia (67%, n = 72) and midazolam for sedation (84%, n = 90).

Delirium occurred in 100% of survivors from the start of rewarming until the end of the ICU stay, for a median number of 4.0 days of delirium (IQR, 2.0–7.5). Figure 2 shows the distribution of cognitive status of patients each day after warming (coma, hypoactive delirium, mixed delirium, hyperactive delirium, and normal). Because the number of patients decreases over time, the relative percentage of patients is given for each patient day. Most of the episodes of delirium were hypoactive: 90% of patients had at least 1 hypoactive delirium day, 21% had at least 1 or more hyperactive delirium days, and 64% had at least 1 mixed delirium day. Normal mental status and coma were less common. In sensitivity analyses, the presence of sedation after warming did not affect the prevalence of delirium (Figure 3).

Figure 4 shows the effects of modifiable risk factors for duration of delirium associated with resuscitation or therapeutic hypothermia protocols. In evaluation of baseline risk factors and number of delirium days, multivariable analysis indicated that age (OR, 1.72; 95% CI, 1.01–2.95; P = .05), and time from initiation of CPR to ROSC (OR, 1.52; 95% CI, 1.11–2.07; P = .01) correlated with more days of delirium. In contrast, use of propofol (OR, 0.02; 95% CI, 0.00–0.48; P = .02) correlated with shorter duration of delirium.

Our results primarily indicate that all survivors of cardiac arrest treated with therapeutic hypothermia who awoke from coma experienced at least 1 day of delirium between rewarming and subsequent discharge from the ICU. Although the high prevalence of delirium may seem surprising, our cohort differs from typical ICU patients. The degree of injury that cardiac arrest induced, the tremendous swings in metabolism from therapeutic hypothermia, and the large doses of psychoactive medications all most likely contributed to the higher prevalence. Independent risk factors for longer duration of delirium included increased age and longer times from CPR to ROSC. Total propofol dose during therapeutic hypothermia and before warming correlated with fewer days of delirium, whereas total doses of fentanyl and midazolam did not correlate with the duration of delirium. The high prevalence of delirium in this population calls for a need to understand the relationship between delirium and long-term outcomes in these patients and then, if warranted, interventions that will decrease the prevalence of delirium in the time after resuscitation.

Delirium in ICUs varies depending on the specific population of patients. Reported prevalences vary from 25% in our mixed-population cardiovascular ICU²⁵  to 80% in medical ICUs.⁹  Previous reports²⁶  have indicated that survivors of cardiac arrest who experienced a myocardial infarction have a higher prevalence of delirium than do ICU patients who did not experience a cardiac arrest. Compared with other ICU patients, survivors of cardiac arrest treated with therapeutic hypothermia have unique characteristics: they are older, have multiple comorbid conditions, have a high prevalence of coronary artery disease,²⁷  and experience an acute anoxic injury that typically results in multiorgan failure and a high in-hospital mortality. Although the percentage of subtypes of delirium in our cohort of patients was similar to that reported in previous studies,^8,9,13,21,25  the overall prevalence of delirium of 100% in our cohort (higher than that reported in other specific ICU populations) may reflect the unique nature of our population. Consistent with the findings of previous studies²⁵  in a general cardiovascular ICU population, the majority of our patients experienced hypoactive delirium. Physicians who care for patients after resuscitation from cardiac arrest should keep these data in mind, because the awareness may help focus care on decreasing the duration of delirium by developing and implementing evidence-based interventions.²⁸ 

We examined risk factors both before and after resuscitation that we hypothesized could influence the duration of delirium. Among prehospital risk factors, age and longer times from initiation of CPR to ROSC correlated with increased duration of delirium. Age has previously been included as a known risk factor for delirium.¹⁶  Longer times from initiation of CPR to ROSC might reflect the detrimental effect that hypoxia has on brain function after therapeutic hypothermia. This finding supports the need for continued training of medical personnel in high-quality CPR that includes timely direct cardioversion to reestablish a perfusing rhythm.

Our protocol for therapeutic hypothermia, similar to the protocol of many other institutions, calls for high doses of sedatives before warming because of the need for paralytics to prevent shivering. We hypothesized that the use of sedation after resuscitation but before rewarming (ie, sedation used during therapeutic hypothermia and before rewarming) could represent a modifiable risk factor that might increase the prevalence of delirium in our patients. Surprisingly, in our investigation of modifiable risk factors after resuscitation, higher doses of propofol during therapeutic hypothermia protected patients and correlated with decreases in the duration of delirium, and total doses of midazolam during therapeutic hypothermia did not correlate with increases in the duration of delirium. To our knowledge, no previous study has indicated an association between total propofol dose and a reduced number of days of delirium. Although previous reports^8,16  indicated an association between benzodiazepines and delirium in several different populations of ICU patients, the patients in our study clearly are a unique group who responded differently to sedatives than did other critically ill patients in previous reports. Type I error may explain these findings related to propofol. Alternatively, our data on benzodiazepines may be underpowered for determining truth.

A consideration of some additional hypotheses of why total propofol dose before warming and not midazolam correlated with duration of delirium is worthwhile. First, we evaluated the total dose of sedatives used during therapeutic hypothermia and did not analyze the effect of sedative use after rewarming. Therefore, we limited our ability to assess the effect of sedatives after warming. Second, we provided propofol more frequently to patients closer to the end of the study period, and therefore the results may reflect changes in practice that may have occurred in later years. Alternatively, the use of propofol may represent confounding by indication, because we preferentially provided propofol during therapeutic hypothermia to patients who were at lower risk of delirium.

To determine whether or not sedation after warming was the cause of the high prevalence of delirium in the study, we split the sample population into patients who were or were not given sedatives. Figure 3 shows that by day 4, the prevalence of delirium in the patients receiving sedatives was approximately equivalent to the prevalence in the patients not receiving sedatives. In addition, in the smaller group who received no sedatives in the first few days after warming, the prevalence of delirium remained high. This finding indicates that the high prevalence of delirium, especially at later time points, is not all due to sedation. Of note, to our knowledge, no data exist to suggest differences between “sedative delirium” and “nonsedative delirium” in patients treated with mild therapeutic hypothermia after cardiac arrest. Further study is needed to understand whether or not outcomes differ if delirium occurs with or without sedation.

Our study has several limitations. First, in 2013, Nielsen et al⁷  published results indicating that mild therapeutic hypothermia is not superior to targeted temperature management for care after resuscitation from cardiac arrest. The design used in that study differed from our study design, and therefore we cannot directly compare the findings of Nielsen et al with the findings of the landmark trials^3,4  from 2002 that established mild therapeutic hypothermia as the standard of care. Ongoing trials of the effect of different doses of therapeutic hypothermia may yield better guidelines in the future. Despite the results from the targeted temperature management trial,⁷  the results of our study still apply to current practice, because mild therapeutic hypothermia remains guideline-based therapy for care after resuscitation from cardiac arrest.

Second, we wished to examine factors exclusive to patients treated with therapeutic hypothermia, and we have no data on patients not treated with therapeutic hypothermia. Third, our institution limits us to a specific protocol for therapeutic hypothermia, which may limit the generalizability of our data. Fourth, our model included variables that were determined a priori. In order to prevent overfitting of our models, the list of confounders did not include all options, and unmeasured variables may exist that could alter our results.

Fifth, the prevalence and duration of delirium in our results may exceed those in previous reports because we do not have differential measurements of CAM-ICU scores before and after a spontaneous awakening trial, and our results may include patients with rapidly reversible delirium due to sedation.²⁹  This effect most likely is small because of the high prevalence of delirium among patients without exposure to sedatives in sensitivity analyses. Finally, we did not structure this study to assess the role of delirium in adverse outcomes, such as mortality or long-term cognitive impairment; this role remains unknown in patients treated with mild hypothermia after resuscitation from cardiac arrest.

The results of this retrospective observational study indicate a remarkably high prevalence of delirium during the ICU stay in patients treated with therapeutic hypothermia after cardiac arrest. Older ages and longer times from initiation of CPR to ROSC were associated with increased duration of delirium, and higher total propofol dose used during therapeutic hypothermia. Because of the marked effect of delirium on outcomes indicated in numerous other investigations, further prospective research is needed to understand the relationship between sedative and paralytic choice during therapeutic hypothermia and prevalence and duration of delirium, as well as the relationship between delirium and long-term outcomes in patients treated with therapeutic hypothermia after cardiac arrest.

This research was performed at Vanderbilt University Medical Center. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the Department of Veterans Affairs.