Survival among patients with successful in-hospital cardiopulmonary resuscitation (CPR) and Marshfield Epidemiologic Study Area (MESA) residents who did not have CPR, by age.
Zoch TW, Desbiens NA, DeStefano F, Stueland DT, Layde PM. Short- and Long-term Survival After Cardiopulmonary Resuscitation. Arch Intern Med. 2000;160(13):1969–1973. doi:10.1001/archinte.160.13.1969
The objective was to evaluate the effect of patient characteristics and other factors on cardiopulmonary resuscitation (CPR) survival, hospital discharge survival and function, and long-term survival.
All patients 18 years and older experiencing in-hospital CPR from December 1983 through November 1991 at Marshfield Medical Center (Marshfield Clinic and adjoining St Joseph's Hospital), Marshfield, Wis, were selected. We performed a retrospective medical record review and augmented these data with updated vital status information.
Main Outcome Measures
Cardiopulmonary resuscitation survival, hospital discharge survival and function, and long-term survival.
Of 948 admissions during which CPR was performed, 61.2% of patients survived the arrest and 32.2% survived to hospital discharge. Mechanism of arrest was the most important variable associated with hospital discharge. Patients with pulseless electrical activity had the worst chance of hospital discharge, followed by those with asystole and bradycardia. Follow-up information was available for 298 patients who survived to discharge. One year after hospital discharge, 24.5% of patients, regardless of age, had died. Survival was 18.5% at 7 years in those 70 years or older, compared with 45.4% in those aged 18 to 69 years. Heart rhythm at the time of arrest strongly influenced long-term survival. Bradyarrhythmias produced a nearly 2-fold increased mortality risk compared with normal sinus rhythm.
Survival until hospital discharge after CPR at our institution during an 8-year period was higher than previously reported for other institutions. Long-term survival after discharge was equal to or higher than reported estimates from other institutions. Hospital admission practices and selection of patients receiving CPR may account for these findings.
CARDIOPULMONARY resuscitation (CPR) has received wide acceptance since the initial description of closed-chest cardiac massage by Kouwenhoven and colleagues.1 Initially, CPR was administered primarily to otherwise healthy individuals who experienced cardiac or respiratory arrest in monitored situations, such as in intensive care units or during surgery, or as a result of near drowning. Today, however, CPR may be attempted on any individual who experiences a cessation of cardiac or respiratory function. It has achieved wide application as a standardized team-based medical intervention.2 In the hospital setting, CPR is commonly viewed as an emergency procedure that is routinely administered to patients who experience cardiopulmonary arrest. As with other emergency procedures, consent for CPR is presumed unless otherwise indicated since patients have often not stated their preference for CPR and may be too sick to state their preference at the time of admission.3 A better understanding of the efficacy of CPR, the long-term outcomes after CPR, and factors that may affect successful resuscitation and eventual prognosis would help institutions make informed decisions and policies regarding CPR.
A recent review of studies of CPR outcomes reported that an average of 15% of patients experiencing arrest survive until hospital discharge (range, 3%-27%).4,5 This long-term success rate has remained stable for the last 30 years. The quality and comparability of the initial studies, diverse hospital settings, different population groups, and varying outcome measurements may explain the inability to show improvement in CPR success.6 Few studies have published long-term survival rates after successful in-hospital CPR and the results have been inconsistent.4,7- 9 Moreover, no studies have examined factors that may affect survival among discharged patients who have been resuscitated during their hospital stay.
We studied the outcomes of CPR performed at the Marshfield Medical Center, Marshfield, Wis, from December 1983 through November 1991 to determine in-hospital and long-term survival (up to 8 years after hospital discharge) and to identify patient characteristics affecting long-term survival.
We received institutional review board approval with waived informed consent to perform a retrospective medical record review on all patients who received CPR at the Marshfield Medical Center between December 1, 1983, and November 30, 1991. The study was restricted to individuals 18 years of age and older. Cardiopulmonary arrest was considered to have occurred if the code team performed a therapeutic intervention. The Marshfield Medical Center (Marshfield Clinic and the adjoining St Joseph's Hospital) is a major provider of primary and secondary medical care for central Wisconsin, as well as tertiary care for central and northern Wisconsin. The total number of hospital beds during the study ranged from 516 to 523, with 36 in intensive care units. The annual hospital occupancy rate during the time frame of the study ranged from 65% to 69%. Cardiopulmonary resuscitation was conducted by a code team consisting of a senior medical resident or staff emergency physician, an anesthetist or anesthesiologist, and other paramedical personnel, including nurses, respiratory therapists, electrocardiogram technicians, pharmacists, and phlebotomists. Each team member was certified in advanced cardiac life support according to the current American Heart Association standards.
A CPR log was kept each time a resuscitation team was called. The critical care committee reviewed this log on a monthly basis. It was possible to retrieve the medical records for all patients for whom the code team was called. Trained chart abstractors extracted all chart data. The following information was obtained for all patients: (1) demographics; (2) patient characteristics, including mental status, number of dependencies in activities of daily living (ADL), diagnoses (respiratory, cardiac, other), and home or nursing care requirements at time of admission; and (3) arrest characteristics, including time of day, physical location of patient at time of arrest, whether the arrest was witnessed or monitored, whether it was primarily respiratory or cardiac arrest, pulse and electrocardiogram-determined mechanism of arrest, and days in the hospital prior to arrest.
A computerized database contains appointments and hospital discharge data on all Marshfield Clinic and St Joseph's Hospital patients, as well as date of death, if applicable. The computer database was reviewed to update vital status information and date of death for all study patients who were discharged alive from the hospital. For analysis of long-term survival, patients not known to be dead were censored as of the last date that the patient had contact with the clinic or the hospital. All patients censored had at least one contact with the clinic or hospital after discharge.
Logistic regression was used to study the associations between independent variables and survival after the resuscitation attempt and survival after hospital discharge, using SAS software (SAS Institute, Cary, NC).10 Continuous variables (age and days in the hospital prior to arrest) were modeled as cubic spline functions to check for nonlinearity of their associations with mortality.11 Other variables were modeled dichotomously or with indicator variables. Both models contained all collected patient demographics and arrest characteristics. The area under the receiver operating characteristic curve was used as the measure of discrimination of the models. Kaplan-Meier survival curves were generated to describe the survival experience after discharge.12 Although we had patient follow-up to 10 years, the survival analysis was truncated at 8 years because few patients had longer follow-up and the results were unstable. Proportional hazards survival models were also developed to assess determinants of survival after discharge.13 To identify the independent variables to include in the proportional hazards model, we first generated univariate Kaplan-Meier curves. All variables that had an association with survival duration with a P value less than 0.20 using log-rank tests were included in the model.
To compare the mortality of patients who had CPR during their hospitalization with patients discharged from the hospital who did not have CPR, we used data from the Marshfield Epidemiologic Study Area (MESA). MESA is an epidemiologic catchment area that includes Marshfield and nearby communities.14 Marshfield Clinic uses several different mechanisms to routinely and systematically ascertain deaths among its patients, including review of all obituaries published in the Medical Center's area. We have shown in a previous study that mortality data is 100% complete among residents of MESA.14
A total of 1066 resuscitation attempts were performed on 948 hospital admissions (43 arrests/100 filled hospital beds per year). Resuscitation attempts were performed on 1% of admissions during the study period. Most of the arrested patients were married, male, and admitted from home. The primary arrest diagnosis was cardiac in 65.0% of the patients. The most common rhythm at the time of arrest was a bradyarrhythmia. Most of the individuals had a normal mental status before admission and no marked dependencies in their ADL. Most of the arrests were witnessed and monitored. The median age of arrested patients was 69 years. The median number of days in hospital before arrest was 4 (25th percentile, 1; 75th percentile, 10).
Of the 948 admissions during which resuscitation efforts occurred, 580 (61.2%) resulted in the patient surviving the arrest, and 467 (49.3%) surviving more than 24 hours. Three hundred five admissions resulted in the patient (32.2%) surviving to discharge. Most hospital discharges resulted in the patient returning home neurologically intact (93.3%).
Table 1 summarizes the association of selected characteristics with survival to hospital discharge. After adjustment for patient and arrest characteristics, the mechanism of arrest was the most important variable associated with hospital survival. Compared with patients in normal sinus rhythm, patients with pulseless electrical activity had the worst chance of hospital discharge (odds ratio [OR] for death in hospital, 6.12; 95% confidence interval [CI], 2.06-18.19) followed by those with asystole (OR, 4.88; 95% CI, 2.56-9.16) and bradycardia (OR, 2.84; 95% CI, 1.64-4.93). The chance of surviving the hospitalization for patients with sinus tachycardia, ventricular tachycardia with a pulse, atrial fibrillation or flutter, ventricular fibrillation or pulseless ventricular tachycardia, and unknown or other rhythms was similar to that of patients who had normal sinus rhythm at the time the first electrocardiogram was performed during the resuscitation effort. Compared with monitored patients, unmonitored patients had a greater chance of dying in the hospital (OR, 1.79; 95% CI, 1.18-2.73). Patients who had primarily a cardiac arrest had a greater chance of dying in the hospital (OR, 1.73; 95% CI, 1.03-2.91) than those with respiratory arrest. Finally, compared with patients who arrested in the winter, those who arrested in the spring were less likely to die in the hospital following their arrest (OR, 0.56; 95% CI, 0.36-0.86). Age was not significantly associated with survival to hospital discharge.
The 305 admissions (32.2%) resulting in survival to discharge occurred among 298 unique patients who were followed up at our facility. The first admission of the study period was selected for analysis. Among these 298 patients, mortality was steep during the first year following hospital discharge (Figure 1). Within 1 year after hospital discharge, about a quarter of the patients in each age group had died. The 70-year-and-older age group had the worst survival rate in subsequent years: 44.3% survival at 3 years, 35.5% survival at 5 years, and 18.5% survival at 7 years. Among the 18- to 69-year age group, survival declined more gradually: 62.2% survival at 3 years, 56.0% survival at 5 years, and 45.4% survival at 7 years. The survival experience of both age groups was substantially worse than the expected survival after hospital discharge of patients who did not require CPR, as determined from the MESA data.
The determinants of long-term survival after discharge among patients who had successful CPR included age, whether the patient had required nursing care prior to hospital admission, reason for hospital admission (cardiac vs noncardiac), cardiac rhythm at the time of cardiopulmonary arrest, and impairment in ADL at discharge (Table 2). Compared with patients who had a normal sinus rhythm at the time of their arrest, patients who had a bradyarrhythmia had nearly 2-fold increased risk of dying, and those with atrial fibrillation or atrial flutter experienced 223 higher risk of dying.
The proportional hazards model analysis on long-term survival suggested that women had a lower mortality risk than men, that patients who had not worked outside the home prior to arrest had a higher mortality rate than those who did work outside the home, and that patients who had been intubated prior to arrest had a higher mortality rate than nonintubated patients. Those who had a noncardiac arrest diagnosis were at increased risk of dying after hospital discharge compared with those with a cardiac arrest diagnosis. However, none of these results were statistically significant.
Our experience revealed a survival rate to hospital discharge after CPR of 32.2%. In a recent 25-year review of in-hospital CPR, overall survival to discharge was 14.6% (N=12961; range, 3%-27%).5 A 30-year review of in-hospital CPR reported an average survival to discharge of 15.0% (N=19955).4 The proportion of CPR procedures at our facility was similar to that previously reported in the literature: 1.0% of admissions compared with 1.1% to 1.5% nationally.7 By assuming that all arrested patients would have died without resuscitation, Coskey has estimated the hospital mortality reduction (calculated as the [number of CPR patients discharged home / hospital deaths]+[number of CPR patients discharged home] × 100) attributed to CPR to range from 5.9% to 6.4%.15 Using Coskey's formula, we estimated the mortality reduction due to CPR in our institution as 7.7%.
The higher proportion of patients surviving to discharge at our institution may have several explanations. We noted that a "do not resuscitate" (DNR) order was used with increasing frequency during the course of this study. If DNR orders were written for patients with the worst prognoses, those with the least chance of response to CPR would not have been included in our study. The lack of DNR orders in earlier studies may explain the lower survival to discharge percentages quoted. DNR orders were also used with increasing frequency at our surrounding nursing homes, especially in patients with advanced dementia, cancer, or other terminal illnesses. In addition, there was an active hospice movement with a home-visit hospice program, a hospital-based hospice unit, and a freestanding hospice home, which were developed during the course of this study.16 The low rate of dependencies in ADL seen in our arrested patients supports these hypotheses.
The pulse and electrocardiogram-determined mechanism of arrest was the most important variable associated with survival to hospital discharge. Compared with patients with normal sinus rhythm, those with pulseless electrical activity, asystole, and bradycardia at the time of the arrest had a greatly reduced chance of leaving the hospital. These findings are consistent with those of other studies that suggest that patients with ventricular fibrillation or tachycardia are more likely to survive the resuscitation effort.4,17- 22 The study by Rozenbaum and Shenkman23 noted that the mechanism of arrest was significantly associated with early successful resuscitation outcome (especially in patients whose arrest was caused by ventricular fibrillation or tachycardia), but not with death or discharge.
After adjusting for other variables, unmonitored patients had a poorer hospital survival than monitored patients. This is potentially an important finding, since the efficacy of this technology has never been assessed by randomized trials. However, this effect of monitoring may be explained by residual confounding, since the categorization of diagnoses that we used in this retrospective study is limited.11 Our finding of a seasonal effect on survival to hospital discharge may have a similar explanation.
Age was not a significant predictor for survival to discharge in our study. Our elderly patients (>70 years) had a higher rate of survival to discharge (29.3%) than previously published data would indicate (12.4%).4 Age has been noted to be a significant factor in predicting outcome of in-hospital CPR attempts in prior studies.24- 26 However, more recent studies corroborate our results that age is not a significant predictor of survival for in-hospital arrests.27,28 Our improved survival to hospital discharge after CPR for elderly patients could be explained by an increased rate of DNR orders in those with poor prognoses.
Most of our patients who survived to hospital discharge were discharged neurologically intact and functional. There were no patients who remained in a vegetative state for a significant time. Some studies describe patients who survive for a considerable time in the vegetative state after CPR attempts.29,30 However, most studies report that the long-term outcome for patients who survived to leave the hospital is good.29- 31
We found that long-term survival after hospital discharge following successful in-hospital CPR may be estimated from patient characteristics and other prearrest factors. Mortality increased with age and was higher among patients who required nursing care prior to admission and among those with impairment in ADL at discharge. Patients who had a cardiac admission diagnosis had lower long-term mortality risks. Heart rhythm at the time of the arrest was the one factor specific to the cardiac arrest with a clear effect on survival after hospital discharge.
Of the few studies that have reported survival for 1 year or longer after arrest or hospital discharge, none has analyzed factors that influence long-term survival.
DeBard7 reported follow-up of 201 patients who survived CPR in a community hospital in Ohio. Survival after hospital discharge was 75% at 1 year, 50% at 3 years, and 20% at 5 years. Peatfield et al8 reported a 10-year follow-up of 80 patients resuscitated in a district general hospital in England. They found that mortality was about 7% per year over the first 5 to 7 years and then leveled off at about 65% survival. Lemire and Johnson9 reported a 10-year follow-up of 230 cardiac arrest survivors discharged from the Royal Victoria Hospital in Montreal. Survival was 74% at 1 year, 51% at 3 years, 38% at 5 years, and about 28% at 9 years. Our results are most consistent with those of Lemire and Johnson.
Although our study is the largest of the long-term follow-up studies of CPR survivors, it has some limitations. Even though the abstraction form had been piloted to ensure that the data were available for the majority of patients and that they could be abstracted in an accurate, reproducible fashion, there are certainly limitations inherent to a retrospective chart review. We do not have information on long-term functional status or mental status after CPR. In addition, we do not have markers of severity of illness, detailed diagnostic categories, or comorbidities. All resuscitation attempts were included in the study when a member of the code team performed an act of therapeutic intervention after the code was called. However, we could not control for the possibility that a different threshold for declaring a code was used in our institution than in other studies. Finally, our institution primarily serves a rural, white population, so our results may not be generalizable to other geographical regions or racial minorities.
In conclusion, survival until hospital discharge after CPR at our institution during a recent 8-year period was higher than previously reported for other institutions. Long-term survival after discharge was comparable to or higher than reported from other institutions. Hospital admission practices and selection of patients who receive CPR may account for these findings. The arrhythmia leading to cardiopulmonary arrest was an important determinant of both in-hospital and postdischarge survival. Age and indices of disability were important influences on postdischarge, but not in-hospital, mortality.
Accepted for publication January 6, 2000.
Funding for this study was provided by Marshfield Medical Research and Education Foundation, Marshfield, Wis.
We would like to thank Sharon Wilke, LPN, and Debra Hein for data collection; June Kusta, Catherine Reinhart, and Angela Bruesewitz for secretarial assistance; Laura Wittman for manuscript editing and preparation of tables and figures; and Craig Gale for statistical review.
Reprints: Thomas W. Zoch, MD, Theda Clark Regional Medical Center, 130 Second St, PO Box 2021, Neenah, WI 54957-2021.