Outcome of Adult Cardiopulmonary Resuscitations at a Tertiary Referral Center Including Results of "Limited" Resuscitations

Background  The results of in-hospital resuscitations may depend on a variety of factors related to the patient, the environment, and the extent of resuscitation efforts. We studied these factors in a large tertiary referral hospital with a dedicated certified resuscitation team responding to all cardiac arrests.

Methods  Statistical analysis of 445 prospectively recorded resuscitation records of patients who experienced cardiac arrest and received advanced cardiac life support resuscitation. We also report the outcomes of an additional 37 patients who received limited resuscitation efforts because of advance directives prohibiting tracheal intubation, chest compressions, or both.

Main Outcome Measures  Survival immediately after resuscitation, at 24 hours, at 48 hours, and until hospital discharge.

Results  Overall, 104 (23%) of 445 patients who received full advanced cardiac life support survived to hospital discharge. Survival was highest for patients with primary cardiac disease (30%), followed by those with infectious diseases (15%), with only 8% of patients with end-stage diseases surviving to hospital discharge. Neither sex nor age affected survival. Longer resuscitations, increased epinephrine and atropine administration, multiple defibrillations, and multiple arrhythmias were all associated with poor survival. Patients who experienced arrests on a nursing unit or intensive care unit had better survival rates than those in other hospital locations. Survival for witnessed arrests (25%) was significantly better than for nonwitnessed arrests (7%) (P = .005). There was a disproportionately high incidence of nonwitnessed arrests during the night (12 AM to 6 AM) in unmonitored beds, resulting in uniformly poor survival to hospital discharge (0%). None of the patients whose advance directives limited resuscitation survived.

Conclusions  Very ill patients in unmonitored beds are at increased risk for a nonwitnessed cardiac arrest and poor resuscitation outcome during the night. Closer vigilance of these patients at night is warranted. The outcome of limited resuscitation efforts is very poor.

SEVERAL FACTORS affect survival after cardiac arrest. Witnessing of arrests, early initiation of resuscitation, return of cardiac function within 20 minutes, and patient youth are associated with higher survival rates.¹ We were interested in other factors that might affect survival after cardiac arrest within a single large teaching institution with a dedicated resuscitation team responding to arrests throughout the hospital. In addition to the details of the resuscitations themselves, we examined other factors: whether patients were monitored at the time of cardiac arrest (intensive care unit [ICU] vs nursing floor), the outcome of arrests in miscellaneous and remote locations (radiology suites, dialysis unit, cafeteria, etc), the correlation of outcome with the patient's primary disease process, and the time of day when the arrest occurred. We also examined the interactions of these and other factors on survival at various times after resuscitation.

Since the passage of the Patient Self-Determination Act in 1991, advance directives regarding the extent of resuscitation have become a reality in the care of patients.² Patients, with the guidance of physicians and/or social workers, can limit resuscitative measures in the event of a sudden arrest or critical worsening from a previously stable physical condition. This has resulted in a "limited code" phenomenon. In this study we report the outcome of the small cohort of patients who, on admission to the hospital, executed advance directives that limited resuscitative measures.

The cardiac resuscitation team at the Cleveland Clinic Foundation, Cleveland, Ohio, consists of a senior internal medicine resident, a senior anesthesia resident, a respiratory therapist, and registered nurses. One nurse is dedicated to recording all of the resuscitation efforts, procedures, and medications on a standardized cardiopulmonary resuscitation (CPR) data form during the event. This form contains patient demographic information; main diagnoses; and time, day, location, and immediate precipitating event of the cardiac arrest. It also lists beginning and finishing times of resuscitation, resuscitation outcome, and all the administered drugs and performed interventions. Immediately after resuscitation, the senior physician present and the nurse in charge review this information for accuracy. The form is printed in duplicate; the original remains a part of the patient's record and the copy is reviewed and kept by the hospital's CPR coordinator. All participants in the resuscitation are certified in advanced cardiac life support. Standardized resuscitation carts are located throughout the hospital, in all patient units and other care areas. All resuscitations are announced overhead and by activating "code" pagers distributed to appropriate personnel. Response time, even in remote locations, is less than 5 minutes.

We examined the medical records of all patients who underwent CPR from January 1, 1994, to July 1, 1995, at the Cleveland Clinic Foundation. Criteria for study inclusion were age 18 years or older, adherence to the advanced cardiac life support protocol, excepting intubation if it was not deemed necessary, chest compressions, and no limitation of resuscitation efforts because of do not resuscitate orders. We report an outcome analysis of 445 consecutive resuscitations that met these criteria. If a patient sustained more than 1 arrest, only the results of the first resuscitation were included.

We also searched the CPR database for limited codes, defined as a resuscitation with advance directives prohibiting 1 or more of the following: intubation, chest compressions, electrical defibrillation, or prolonged resuscitation efforts.

Data were analyzed with respect to the primary outcomes: survival immediately after resuscitation, at 24 hours, at 48 hours, and at hospital discharge. Independent variables assessed for association with survival include age, sex, and primary preexisting medical condition (cardiovascular disease, end-stage disease, infectious disease, and all others) (Table 1). Whether the patient was being ventilated at the time of arrest, whether the arrest was witnessed, and duration of resuscitation were all analyzed with respect to survival rates. Electrocardiographic (ECG) arrhythmias noted during resuscitation were analyzed in 2 ways: (1) by ECG event (asystole, ventricular fibrillation, pulseless electrical activity, or any combination of these events) and (2) by the number of ECG events recorded during a single resuscitation (patients with only one type of arrhythmia, those with any combination of 2 arrhythmias, or those with all 3 arrhythmias). Other data regarding the details of resuscitation included the number of defibrillations performed, use of a pacemaker (internal or external), use of open-chest cardiac massage, the number of ampoules of atropine and epinephrine used, and whether the patient required tracheal intubation. The relationship between the location of the arrest (nursing unit, ICU, emergency department, and miscellaneous sites) and the survival rate was also assessed by time of day (night [12 AM to 6 AM] or day and evening [6 AM to 12 AM]).

Associations with survival rates immediately after resuscitation, at 24 hours, at 48 hours, and at hospital discharge were assessed with univariate and multivariate logistic regression analyses. Results are reported as percentages surviving, with odds ratios and 95% confidence intervals.

Multivariate models for immediate and hospital discharge survival rates were fit using backward selection and considering factors significant at P<.15 univariately. Interactions among the considered variables were deemed significant at P<.10. Main effects were retained in the model if P<.05. Each continuous variable was assessed for linearity in the outcome using moving average plots, and variables were categorized if moderate or severe nonlinearity was observed.

A moving average plot was calculated to examine the relationship between time of arrest and survival rate. Because the outcome is dichotomous, this type of plot was used to smooth out the relationship and remove some of the arbitrariness of choosing a specific starting time. This curve gives a "smoothed" representation of the changes in survival rates throughout the day. Patients were ordered by the time the arrest occurred, and the proportion surviving was calculated for groups of 50 sequential patients. Each successive point on the curve represents a recalculation shifting 10 patients at a time.

Results are reported as mean ± SD, medians with 25th and 75th percentiles, or odds ratios and 95% confidence intervals. The probability (with correction for multiple comparisons when necessary) of a type I error being less than .05 is considered significant. Most analyses were performed using statistical software (SAS version 8.0; SAS Institute Inc, Cary, NC).

During the period of this study, the Cleveland Clinic Foundation had an average of 863 adult (aged >17 years) beds available with a 77% occupancy, for an average daily inpatient census of 665 patients. The ICUs accounted for 38 beds with an 89% occupancy, resulting in a mean ICU census of 34 patients. The age of adult patients was 61 ± 15 years, and 53% were men. Eighty-eight percent of patients were white and 10% were black. Medical services admitted 35% of the patients, and 21% of them were on the cardiology service. Surgical services admitted 51% of the patients, with 15% being cardiothoracic patients. The remaining 14% of the patients were admitted to miscellaneous services.

Resuscitation records for 445 patients met the criteria for inclusion in this analysis. Overall survival to hospital discharge was 23% (104/445). In our analysis, neither age nor sex affected survival rates at any time: immediate, 24 hours, 48 hours, or at hospital discharge (Table 2, Table 3, and Figure 1). Whether an arrest occurred while a patient was being ventilated in an ICU did not alter survival rates (Table 2). At all observation points, patients with a witnessed arrest were more likely to survive (Table 2). At hospital discharge, survival for witnessed arrests was 25% and for nonwitnessed arrests was 7%. The duration of the resuscitation was significantly related to survival rates at all times. Survival for patients whose resuscitations lasted for 15 minutes or less was 40% vs 18% for longer resuscitations (Table 2). From another viewpoint, the median duration of resuscitations for patients who survived to hospital discharge was 20 minutes vs 29 minutes for nonsurvivors (P<.001) (Table 3). Neither open-chest cardiac massage nor insertion (internal pacing) or application (external pacing) of a pacemaker seemed to affect survival. The 19 patients who had open-chest cardiac massage were all post–coronary artery surgery patients in the cardiac ICU who had not responded to several minutes of closed-chest massage. Only 33 of the 445 patients did not require endotracheal intubation as part of the resuscitation; however, 73% of them survived to hospital discharge compared with 19% of those who required tracheal intubation (P<.001) (Table 2). Mean code duration for patients who were not intubated and survived was 17 ± 12 minutes, significantly shorter than for those who died (P<.001) (Table 3). Furthermore, 23 of the 24 survivors who were not intubated had witnessed arrests.

At all points, the outcome of a resuscitation at night was worse than during the day: only 14% of patients who experienced an arrest between 12 AM and 6 AM survived to hospital discharge, whereas almost twice as many (26%) survived if the arrest occurred during the day or evening (P = .006) (Table 2).

The primary preexisting medical condition did not affect the immediate survival rate or the survival rate at 24 hours; however, it was associated with the 48-hour survival rate and the rate to hospital discharge (Table 4). Discharge survival for the cardiovascular disease group (30%) was significantly better than that for the infectious disease group (15%) and those with end-stage disease (8%) (P<.001). Survival for all other primary diagnoses was 20%.

The location within the hospital where arrest and resuscitation occurred significantly affected survival rates at 48 hours and hospital discharge. Twenty-five percent of patients who had an arrest on nursing units (floors and ICUs) survived to discharge. Survival after arrests in miscellaneous areas (16%) and especially in the emergency department (5%) was significantly lower (P<.007) (Table 4). Response time for the resuscitation team at all locations was less than 3 minutes 96% of the time and less than 5 minutes in all cases. Response time did not affect outcome in this analysis. The number of ECG events recorded on the CPR data form (ventricular fibrillation, asystole, and pulseless electrical activity) was associated with survival: of patients who had only 1 ECG event, 28% survived to hospital discharge, whereas survival in patients with 2 and 3 ECG events was 10% and 0%, respectively (Table 4). Patients who received multiple defibrillations had lower survival rates at all times (P<.001). Likewise, increasing doses of epinephrine and atropine were associated with lower survival rates at all times (Table 3).

A decreased immediate survival rate was statistically associated with the following predictors: nighttime arrest, increased number of defibrillations, increased dose of epinephrine, and use of atropine during resuscitation (P = .01) (Table 5). Patients whose only ECG event was ventricular fibrillation were an estimated 3.6 times more likely to survive than those who had asystole (P = .002). Patients who had pulseless electrical activity only or a combination of ECG events had odds of survival similar to patients with asystole (P = .58 and .20, respectively).

A decreased survival rate to hospital discharge (Table 6) was associated with the following predictors: longer resuscitation times (P = .04) and the necessity to perform endotracheal intubation (P = .001). Medical therapy with atropine and increasing doses of epinephrine decreased the odds of surviving to hospital discharge. Cardiac arrest due to primary cardiac disease was associated with a better chance of surviving to discharge than cardiac arrest in end-stage disease (P = .01) but was not different from patients with infectious disease or miscellaneous diseases (P = .80 and .28, respectively).

The hospital location of the cardiac arrest played a significant role in survival to hospital discharge: emergency department and miscellaneous locations seemed to have lower survival rates than regular nursing floors and ICUs.

The times when cardiac arrests occurred were distributed evenly throughout any 24-hour period. However, the immediate survival rate for patients who experienced arrest during the night was approximately half that for arrests occurring at other times of the day (P<.008). In fact, this was true at all 4 survival periods examined, with 14% of patients who had nighttime arrests and 26% of those with arrests occurring at other times surviving to hospital discharge (Table 2). There was a significant disparity between the percentage of arrests not witnessed during the day vs those not witnessed at night: 9% (31/328) vs 21% (25/117). Figure 2 shows immediate survival rates and survival rates to hospital discharge (moving averages) as a function of the time of day. During the day, patients whose cardiac arrest was not witnessed had approximately one third the chance of survival to hospital discharge as those whose arrest was witnessed (odds ratio, 0.38). At night, the chances of survival for a nonwitnessed arrest were less than one tenth of those that were witnessed (P = .02).

There were 37 patients (18 women and 19 men) who had advance directives restricting the extent of resuscitation efforts, referred to as limited code. The mean age of these patients was 61 ± 23 years. Twenty-four of these patients had advance directives forbidding the use of 1 or more of the following: intubation, chest compressions, or electrical defibrillation. The remaining 13 patients requested time-limited CPR (maximum of 10 minutes). Only 6 of these patients (16%) survived the initial event, only 2 (5%) continued to survive at 48 hours, and none survived to hospital discharge. Of these patients, 45% had cardiovascular disease, 5% had terminal cancer, 35% had end-stage disease (renal and liver), and 15% were in the miscellaneous group.

In this study we analyzed the results of 445 resuscitations, one of the largest recent series from a single institution using a standardized protocol and reporting procedure. Our analysis of the resuscitation data yielded results that are similar to those of other authors³ except in one aspect. We found that resuscitations performed between 12 AM and 6 AM resulted in significantly lower survival rates.

The general hospital discharge survival rate after CPR has gradually improved over the years, from 10% in the decade 1952 to 1961 to more than 15% in the decade after 1982.³ These figures are more impressive than they might seem because in the 1950s and 1960s CPR was largely reserved for patients in whom a high probability of survival was anticipated, whereas today, unless a patient has do not resuscitate orders, all patients experiencing an arrest are resuscitated. A large pooled analysis by Schneider et al⁴ revealed that worldwide, 15% of 19 955 patients were successfully resuscitated to hospital discharge. Survival to hospital discharge rates in Western countries³ are consistent with the data from Schneider et al⁴: 15% in the United States, 16% in Canada, 17% in the United Kingdom, and 14% in other European countries. In general, survival in community hospitals (18%) is better than in teaching hospitals (15%), with the Veterans Affairs Hospital System having the lowest survival (9%).³

Our study concurs with others'³ showing that survival rates are not affected by patient age. We also did not find a difference in survival as a function of sex. This is in agreement with results of most studies,^5-8 although Makker et al⁹ and Gray et al¹⁰ reported better survival rates for women.

In our hospital, patients with primary cardiovascular disease had the highest survival to hospital discharge (30%) (Table 4). Patients whose primary comorbidity was infectious disease had 15% survival, and those with end-stage disease had 8% survival. Patients who did not fit into one of these categories had survival to discharge of 20%. Although other studies have not broken down patients into precisely the same categories, their results, in general, are similar. For example, Peatfield et al¹¹ found the best prognosis for survival when the cardiac arrest was associated with myocardial infarction.

Survival at each period was significantly better for witnessed arrests than for nonwitnessed arrests, with survival to discharge being 25% vs 7%, respectively. This observation is in concert with results of the study on resuscitation outcomes by Saklayen et al.³

In our study, hospital location played a significant role in survival to hospital discharge. Patients who had an arrest in an ICU or on a nursing unit had 25% survival, whereas those experiencing an arrest in miscellaneous other locations had survival of 16%. Other studies^12,13 offer similar results, although most demonstrated somewhat better survival rates for arrests occurring in ICUs and coronary care units vs nursing units. Marwick et al¹⁴ found the immediate survival rate to be significantly better if a cardiac arrest occurred in an ICU, coronary care unit, or operating room; however, the survival rate to hospital discharge was not significantly better. In our study, resuscitations in the emergency department had the lowest survival at 48 hours (19%) and at hospital discharge (5%). Several authors^6,10,15 report similar disappointing results with emergency department resuscitations. Others,⁸ however, report little or no difference between emergency department resuscitations and those taking place elsewhere in the hospital. There might be several explanations for this discrepancy. For our patients resuscitated in the emergency department, more than 90% experienced arrest outside the hospital, where resources and expertise were less than ideal, compared with some other studies in which a large proportion of patients were experiencing cardiac symptoms and arrest either en route to or in the emergency department. Furthermore, some studies include primarily respiratory arrests, often due to drug overdose, which can have a high rate of successful resuscitation. We report only those cases of cardiac arrest requiring chest compressions.

As in other studies,^3,16-18 we found an inverse correlation between the length of the resuscitation effort and the survival rate immediately, at 24 hours, at 48 hours, or until hospital discharge. Survival to discharge for resuscitations lasting less than 16 minutes was 40%, whereas survival for those lasting 16 minutes or longer was 19%. The median resuscitation time for our patients who survived to discharge was 20 minutes and for those who died was 29 minutes. Varon and Fromm¹⁹ demonstrated that the time of resuscitation averaged about 26 minutes for survivors and 33 minutes for nonsurvivors. Cooper and Cade²⁰ demonstrated that the most important factor affecting the immediate survival rate was a duration of resuscitation of less than 20 minutes. Indeed, the subgroup of our patients who did not require endotracheal intubation had a mean resuscitation time of 17 minutes and high survival (73%).

The ECG events during a resuscitation are significant predictors of survival. In our study, when ventricular fibrillation was the only event, 45% of patients survived to hospital discharge. Other authors^14,20-22 reported similar survival, ranging from 40% to 65%. Our results concur with the findings of Saklayen et al,³ demonstrating successively lower survival rates when the rhythm at the time of initiation of CPR was bradycardia/asystole, followed by pulseless electrical activity, with the worst results occurring when multiple dysrhythmias occur during resuscitation.

We found that the numbers of defibrillation attempts, epinephrine doses, and atropine doses were all independent predictors of a poor outcome. The need for a second defibrillation reduced chances of survival to hospital discharge by more than 70%. The administration of any atropine during the resuscitation cut the survival rate in half, and additional atropine doses resulted in survival to hospital discharge of less than 5%. The need for a single dose of epinephrine reduced survival by 20%, and each additional dose further halved the survival to hospital discharge rate. A prolonged duration of resuscitation and the attendant poor outcome correlated with multiple defibrillations and multiple doses of drugs. These observations have been made previously. van Walraven,²³ Schindler,²⁴ and Saklayen³ and their colleagues observed that the need for increasing doses of epinephrine was associated with lower survival rates. Saklayen et al³ and Cooper and Cade²⁰ noted that the fewer the number of defibrillations the better the chances of survival. Finally, Saklayen et al³ and van Walraven et al²³ showed, as we do, that the need for atropine administration is associated with a poor survival rate.

It is interesting that in a large, well-staffed teaching hospital the results of resuscitations occurring late at night were significantly worse than the results during the day and evening. Immediate survival for nighttime resuscitations was 60% of that for other times, and at 24 hours, 48 hours, and hospital discharge the survival rate was less than half that for day and evening resuscitations. Several others have looked at this issue. Saklayen et al,³ reviewing 340 resuscitations at a Veterans Administration hospital, found that the arrests were evenly distributed among the shifts and that there was no difference in the survival rate as a function of time of day. Cooper and Cade,²⁰ in a recent large study from a nonteaching hospital, found no difference in the immediate survival rate between resuscitations during the day and those at night; however, they did not give the results for survival to hospital discharge. On the other hand, Takeda et al,²² reporting 90 cardiac arrests occurring on their general ward, noted a decreased survival rate, both immediate and to hospital discharge, for resuscitations at night as opposed to during the day.

There are a variety of factors that could contribute to the poorer survival rate for nighttime resuscitations. One important factor, which has been stressed by others,³ is whether the arrest was witnessed. In our data, 21% of nighttime arrests were not witnessed as opposed to only 9% for the other times of day. Immediate survival for nonwitnessed arrests at night was 24% vs 39% for nonwitnessed arrests during the day and evening. Likewise, survival to hospital discharge for nonwitnessed arrests at night was 0% vs 13% for other times of the day; none of the 25 patients with nonwitnessed nighttime arrests survived, whereas 4 of the 31 during the day survived. Overall, hospital locations and primary preexisting medical conditions were similar for patients arresting during the day and during the night. However, considering only nonwitnessed arrests, 96% (24 of the 25) occurring at night were on regular nursing units compared with only 60% during other times. Furthermore, among these nonwitnessed arrests, 3 times as many than during the day were patients with end-stage organ failure. When nonwitnessed arrests were discovered at night, asystole was the initial ECG presentation 92% of the time vs 74% during the day or evening. Furthermore, at least 1 ampoule of atropine was given for 32% of nonwitnessed arrests at night compared with 10% of nonwitnessed arrests during the day, substantiating that asystole was observed more frequently at night. We interpret these data to indicate that when very sick patients in unmonitored beds arrest at night, there is a considerably longer delay to their discovery than during the day. The significantly higher incidence of asystole as the initial ECG manifestation of nighttime cardiac arrests tends to support this assumption. We suspect that other institutions have similar situations, but without such extensive detail in their analysis they might not be able to uncover the specific problem areas. However, late discovery and delayed initiation of resuscitation might not be the only explanation for the lower survival rate at night because there was also a disparity in survival to hospital discharge between night and day witnessed arrests, 17% vs 28%, respectively.

We conclude that in our setting and similar ones, very sick patients, in particular those with end-stage organ disease who are in unmonitored beds, are at significantly higher risk at night for an arrest that is not witnessed. Furthermore, because of the delay in discovery, resuscitation efforts for these patients are almost uniformly unsuccessful, and this factor, together with their underlying pathological condition, leads to a very low rate of survival to hospital discharge.

Our literature search did not uncover any other data on the outcome of resuscitations limited by advance directives. Although the number of patients in our limited code group was small, the outcome of the limited resuscitation efforts was uniformly poor. In particular, 17 of 37 patients with advance directives limiting resuscitation were in the cardiac disease group where, based on the results at our institution, approximately 5 (29%) would have been expected to survive to hospital discharge; however, none did. The futility of limited resuscitations has been argued by many authors, and, based on our data, advance directives for limited resuscitation are nearly equivalent to do not resuscitate orders. (Numerous discussions of this topic are listed in the National Library of Medicine under the following MESH terms: euthanasia, passive; resuscitation; ethics, medical; and resuscitation orders.)

Accepted for publication December 5, 2000.

Corresponding author: Denis L. Bourke, MD, Anesthesiology Service, Baltimore Veterans Affairs Medical Center, 13004 Gent Rd, Reisterstown, MD 21136-5717. Reprints: John A. Dumot, DO, Department of Gastroenterology, Cleveland Clinic Foundation, 9500 Euclid Ave, Cleveland, OH 44195.