The estimated survival rate of 8% to 10% after out-of-hospital cardiac arrest (OHCA) remains dismal. Few studies have addressed predictors of functional neurologic outcome after successful resuscitation. The objective of the study was to identify variables associated with favorable neurologic outcomes, defined by a Glasgow Coma Scale of 14 or 15, after OHCA. We used a propensity analysis and classification and regression tree model of 184 OHCA patients surviving to hospital admission at a cardiac arrest receiving center in Los Angeles County from 2008 to 2013. Forty-three patients (23%) had a favorable outcome, median age was 65 years (interquartile range [IQR] 54 to 76), and 98 (53%) were men. Sixty-six patients (36%) presented with a shockable rhythm. The majority were witnessed, either by a civilian (n = 115, 63%) or a paramedic (n = 25, 14%). Bystander cardiopulmonary resuscitation was performed on 84 patients (46%); median dose of epinephrine was 2 mg (IQR 1 to 3). Median time to return of spontaneous circulation was 21 minutes (IQR 16 to 29); the median lactate level was 5.2 mmol/L (IQR 2.8 to 9.2). Lower epinephrine doses (<1.5 mg) and lactate levels <5 mmol/L were predictive of a normal Glasgow Coma Scale, with 90.7% sensitivity (95% confidence interval [CI] 76.9% to 96.9%), 47.5% specificity (95% CI 39.1% to 56.1%), a positive predictive value of 34.5% (95% CI 31.6% to 46.1%), a negative predictive value of 94.4% (95% CI 85.5% to 98.2%), and an area under the curve of 0.89. The propensity score–adjusted logistic regression model demonstrated that receiving <1.5 mg of epinephrine was associated with a favorable neurologic outcome (odds ratio 3.3, 95% CI 1.1 to 10, p = 0.04). In conclusion, for patients surviving to hospital admission, a good neurologic outcome was associated with having received <1.5 mg of epinephrine and a lactate level <5 mmol/L.
Highlights
- •
We evaluated patients surviving to hospital admission after out-of-hospital cardiac arrest.
- •
We performed a propensity score–adjusted analysis to evaluate epinephrine’s effect on outcomes.
- •
Administration of <1.5 mg of epinephrine was associated with better neurologic outcomes.
- •
Patients whose lactate level was <5 mmol/L were more likely to have good neurologic outcomes.
After the introduction and eventual acceptance of “mild” or “therapeutic” hypothermia (TH) in the early 2000s for cerebral preservation after resuscitation, studies have focused on the outcome of neurologic function after hospital discharge, rather than just survival. Such studies have largely assessed the predictive value of systemic markers of neuronal injury or the utility of imaging methods to predict outcome after hypothermia. TH is now commonly used in the care of patients resuscitated from out-of-hospital ventricular fibrillation because of its demonstrated benefit in this patient group. An increasing number of patients resuscitated from non–ventricular fibrillation arrest are also receiving this intervention because of observational studies suggesting improved neurologic outcome. TH is resource intensive as is post-rewarming care and serial or repeated neurologic assessment and management of those patients who have failed the intervention, that is, recurrent seizures, myoclonus, and so on. Using classification and regression trees (CARTs), we identify potential predictors of favorable neurologic outcome in adults resuscitated from out-of-hospital cardiac arrest (OHCA) using only information or data available within the first hour after restoration of spontaneous circulation (ROSC). Such predictors may identify which patients are most likely to benefit from hypothermia. We hypothesized that lower lactate levels and epinephrine requirements are associated with better periarrest perfusion and neurologic outcome.
Methods
This is an observational, retrospective, cohort study of all consecutive adult patients (≥18 years) with OHCA who presented to a single public hospital in Los Angeles County from January 1, 2008, to June 30, 2013. The institutional review board at the participating hospital provided ethics approval with waiver of informed consent.
Harbor-UCLA is a 553-bed general municipal teaching hospital located in southwestern Los Angeles County, which has a population of 11 million residents. With a catchment area of 27 square miles, the hospital has full capabilities for percutaneous coronary intervention and the implementation of hypothermia after cardiac arrest. In Los Angeles County, emergency medical services are provided by way of a 2-tiered response activated by a central dispatch 911 network with >3,500 licensed paramedics, certified in advanced cardiac life support. Paramedics are authorized to initiate protocols adhering to advanced cardiac life support guidelines, with online medical control.
Three abstractors (AMH, DO, and JTN), who were blinded to the hypothesis of the study, were trained to review and cull data from paramedic run sheets, nursing charts, and physician medical records. Cardiac arrest subjects were included only if they survived to hospital admission, as defined by ROSC that is achieved either in the field or in the emergency department (ED) and sustained such that they are assigned a hospital bed. Exclusion criteria included age <18 years, patients with traumatic arrest, and those with an arrest related to a definite respiratory cause, a drug overdose, strangulation, electrocution, or drowning. The abstractors used a standardized abstraction form to record data, which included age, gender, first documented rhythm on paramedic arrival, witnessed or nonwitnessed arrest, presence or absence of bystander cardiopulmonary resuscitation (CPR), administration of epinephrine by emergency medical service personnel, where ROSC was achieved (field vs in the ED), lactate level within 60 minutes of ED arrival, and the Glasgow Coma Score (GCS) on hospital discharge. A priori precise operational definitions for the variables were outlined. The primary end point was survival to hospital discharge with favorable neurologic outcome, defined as a GCS of 14 or 15. The data sheets were randomly checked by the senior abstractor, and disagreements were discussed until consensus was achieved.
The data were entered into a Microsoft Excel (Seattle, Washington) spreadsheet, using DBMS/Copy, version 8 (DataFlux Corporation, Cary, North Carolina) to convert the file into an SAS v.9.3 (Cary, North Carolina) database. Descriptive statistics and univariate comparisons were done to evaluate baseline known factors for survival to hospital discharge after resuscitation from cardiac arrest: gender, age, rhythm on paramedic arrival, witnessed or nonwitnessed arrest, and presence or absence of bystander CPR. We assessed the association of time to ROSC, epinephrine dose, and lactate level with neurologic outcomes, with a type I error threshold of p <0.05. Odds ratios (ORs) with respective 95% confidence intervals (CIs) were calculated. In general, continuous numerical variables are summarized using medians and interquartile ranges and were compared using the Wilcoxon rank sum test. Proportions were compared using Fisher’s exact test or ORs with 95% CIs, as appropriate, and no adjustment was made for multiple comparisons. The Salford systems ( http://www.salford-systems.com —Salford Predictive Mining Suite, 2011) were used for the CART recursive partitioning analysis to identify predictors and cut points for continuous variables for a favorable neurologic outcome. CART has multiple potential advantages over multivariate logistic regression. The classification tree is cross-validated 10 times using a bootstrapping technique in which the tree is developed using 90% of the patients and then tested using patients not included in the development of the tree. The following variables were selected to develop a prediction model: age, gender, witnessed arrest (yes or no), bystander CPR (yes or no), arrest rhythm (shockable or not), time to ROSC, epinephrine administration, and serum lactate level. The sensitivity and specificity of the model with respective 95% CIs were determined.
Because epinephrine administration was not randomly assigned in the study population, we also assigned a propensity score, which is a conditional probability from 0 to 1, that a subject will be treated based on an observed group of covariates. Because patients who received epinephrine may differ systematically from those who did not, the propensity score was used to minimize the differences and improve the comparability between the groups. Finally, we developed a logistic regression model, which included the propensity score, as well as the variables identified by CART, for the end point of favorable neurologic outcome. Model fit was assessed with the Hosmer-Lemeshow goodness-of-fit statistic.
A target sample size of 184 subjects was determined based on the number of factors predicting a good neurologic outcome (for each factor predicting a GCS of 14 or 15, we anticipated requiring at least 10 patients) and the anticipated number of patients with the outcome (approximately 40% or 20% among those who survive to hospital admission). Thus, 4 risk factors could be assessed in a robust fashion.
Results
A total of 184 patients met our inclusion criteria. Forty-three patients (23%) had a GCS of 14 or 15 on hospital discharge. See Table 1 for the details of our study patients. On univariate analysis, the predictors of a favorable neurologic status on discharge included the following ( Table 2 ): younger age, shockable rhythm, witnessed arrest, having bystander CPR, achieving ROSC in the field (vs the ED), lower dose of epinephrine administered, a shorter time to resuscitation, and a lower serum lactate level. The results of the adjusted ORs for a multivariate analysis including significant univariate predictors are described in Table 3 . The recursive partitioning yielded 2 notable predictor variables: epinephrine administration and serum lactate with cut points of 1.5 mg and 5 mmol/L, respectively. Thirty of the 43 subjects with a favorable neurologic discharge received <1.5 mg of epinephrine. Of the remaining 13 subjects with a normal discharge GCS, 9 subjects had a serum lactate level <5 mmol/L. The sensitivity of this 2-predictor model was 90.7% (95% CI 76.9% to 96.9%), a specificity of 47.5% (95% CI 39.1% to 56.1%), a positive predictive value of 34.5% (95% CI 31.6% to 46.1%), a negative predictive value of 94.4% (95% CI 85.5% to 98.2%).
| Characteristics | |
|---|---|
| Age (years) | Median = 65, IQR 54–76 |
| Men | 98 (54%) |
| Women | 86 (47%) |
| Ventricular fibrillation or ventricular tachycardia | 66 (36%) |
| Pulseless electrical activity | 66 (36%) |
| Asystole | 52 (28%) |
| Citizen witnessed | 116 (63%) |
| Paramedic witnessed | 25 (14%) |
| Unwitnessed | 43 (24%) |
| Bystander cardio-pulmonary resuscitation | 59 (32%) |
| Field return of spontaneous circulation | 145 (79%) |
| Emergency department return of spontaneous circulation | 39 (21%) |
| Time to resuscitation (minutes) | Median = 21, IQR 16–29 |
| Lactate (mmol/L) | Median = 5, IQR 2.8–9.2 |
| Survival to hospital discharge | 75 (41%) |
| Survival with GCS of 14 or 15 | 43 (23%) |
| Characteristic | GCS 14 or 15 (n = 43) | GCS <14 (n = 141) | p-Value |
|---|---|---|---|
| Age (years) | Median = 59, IQR 49–70 | Median = 65, IQR 56–78 | 0.03 |
| Men | 25 (58%) | 73 (52%) | 0.5 |
| Shockable rhythm | 30 (70%) | 36 (26%) | <0.0001 |
| Witnessed arrest | 38 (88%) | 103 (73%) | 0.04 |
| Bystander cardiopulmonary resuscitation | 26 (61%) | 59 (42%) | 0.03 |
| Field return of spontaneous circulation | 42 (98%) | 103 (73%) | <0.01 |
| Time to resuscitation (minutes) | Median = 17.5, IQR 11–20 | Median = 23, IQR 18–32 | <0.0001 |
| Epinephrine administered (mg) | Median = 1, 0–2 | Median = 2, IQR 2–3.3 | <0.0001 |
| Lactate (mmol/L) | Median = 3.3, IQR 2.3–4.8 | Median = 6.1, IQR 3.4–10.2 | <0.01 |
| Survival to hospital discharge | 43 (100%) | 32 (23%) | <0.0001 |
| Variable | Odds Ratio, 95% CI, p-Value |
|---|---|
| Age | 0.98 (0.94–1.01), p = 0.2 |
| Witnessed arrest | 1.1 (0.3–4.1), p = 0.9 |
| Shockable rhythm | 3.7 (1.4–10.0), p = 0.01 |
| Bystander cardiopulmonary resuscitation | 1.6 (0.5–4.6), p = 0.4 |
| Time to return of spontaneous circulation | 1.0 (0.9–1.0), p = 0.2 |
| Emergency department as site of resuscitation | 1.3 (0.1–17.3), p = 0.8 |
| Lactate | 0.9 (0.7–1.1), p = 0.2 |
| Epinephrine | 0.8 (0.5–1.2), p = 0.3 |
As noted in Table 4 , there were differences in the cohort that received epinephrine versus those that did not. Those receiving epinephrine had longer times to resuscitation, higher lactate levels, and were less likely to have the following characteristics: male gender, shockable arrest rhythm, bystander CPR, witnessed arrest, and survival to discharge. A propensity score–adjusted model (for receiving epinephrine), which included lactate level and epinephrine administration, demonstrated that receiving <1.5 mg of epinephrine was associated with a favorable neurologic outcome (OR 3.3, 95% CI 1.1 to 10, p = 0.04).
| Variable | Epinephrine (n = 160) | No Epinephrine (n = 24) | OR (95% CI) | p-Value | ||
|---|---|---|---|---|---|---|
| N (median) | % IQR | N (median) | % (IQR) | |||
| Age (years) | 65 | 55–77 | 60.5 | 52–68 | — | 0.3 |
| Male gender | 79 | 50 | 18 | 75 | 0.3 (0.1–0.9) | 0.03 |
| Shockable arrest rhythm | 48 | 30 | 18 | 75 | 0.1 (0.1–0.4) | <0.0001 |
| Witnessed arrest | 117 | 74 | 23 | 96 | 0.1 (0.1–0.9) | 0.02 |
| Bystander CPR | 67 | 42 | 17 | 71 | 0.3 (0.1–0.8) | <0.01 |
| Field ROSC | 121 | 76 | 24 | 100 | — | <0.01 |
| Time to resuscitation (minutes) | 22 | 18–32 | 13.5 | 9–18 | — | <0.0001 |
| Lactate (mmol/L) | 5.8 | 3.0–9.9 | 2.9 | 1.8–4 | — | <0.001 |
| Survival to hospital discharge | 56 | 35 | 19 | 79 | 0.1 (0.1–0.4) | <0.0001 |
| Survival with GCS of 14 or 15 | 27 | 17 | 16 | 67 | 0.1 (0.1–0.4) | <0.0001 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
