Resuscitation in the Hospital: Circadian Variation of Cardiopulmonary Arrest
Article Outline
Abstract
Purpose
Over 25 reports have found outpatient frequency of sudden cardiac death peaks between 6 am and noon; few studies, with inconsistent results, have examined circadian variation of death in hospitalized patients. This study assesses circadian variation in cardiopulmonary arrest of in-hospital patients across patient, hospital, and event variables and its effect on survival to discharge.
Methods
A retrospective, single institution registry included all admissions to the Medical Center of Central Georgia in which resuscitation was attempted between January 1987 and December 2000. The registry included 4692 admissions; only the first attempt was reported. Analyses of 1-, 2-, 4-, and 8-hour intervals were performed; 1- and 4-hour intervals are presented.
Results
Significant circadian variation was found at 1 hour (P
=.01), but not at 4-hour intervals. Significant circadian variation was found for initial rhythms that were perfusing (P=.03) and asystole (P=.01). A significantly higher percentage of unwitnessed events were found as asystole during the overnight hours (P=.002). Using simple logistic regression, time in 4-hour intervals and rhythm were each significantly related to patient survival until hospital discharge (P=.003 and P <.0001). In multivariate analysis, only rhythm remained significant.
Conclusions
Circadian variation of cardiopulmonary arrest in this hospital has several temporal versions and is related to survival. Late night variation in witnessed events and rhythm suggests a delay between onset of clinical death and discovery, which contributes to poorer outcomes.
Keywords: In-hospital resuscitation, Circadian variation, Cardiopulmonary arrest, Resuscitation, Diagnosis, Registry
Since the initial report of closed chest cardiopulmonary resuscitation in 1960,1 refinement of resuscitative technique and efforts to explain factors influencing cardiopulmonary arrest have been at the forefront of the medical literature. One area that has received attention is the circadian variation of acute illness and sudden death. Multiple clinical and epidemiological studies have focused on circadian variation in acute coronary syndrome,2, 3, 4, 5 sudden cardiac death,2, 3, 6, 7, 8, 9, 10, 11, 12, 13 pulmonary thromboembolism,14, 15 and ischemic stroke16, 17 in various clinical settings and patient subgroups. Many have shown an increased morning incidence of ischemic and thromboembolic events.
In an attempt to create uniformity within a vast and increasing body of data on cardiopulmonary arrest, a stratification of convenience and practicality has emerged that subdivides resuscitative efforts into 2 main locations: events in a prehospital setting and events confined to emergency departments, hospital wards, and intensive care units. This organization allows for more uniform comparison of data collected in each venue, as factors influencing both events and outcome vary substantially based on location, expertise of responders, and availability of monitors/defibrillators and medications. Protocols for standardized reporting in each venue have been established.18, 19
Circadian variation in outpatient frequency of sudden cardiac death was first reported in 198720 and revealed a primary peak between 10 and 11 am and a second peak between 5 and 6 pm. Subsequently, over 25 reports on various aspects of circadian variation in out-of-hospital cardiopulmonary arrest have been published. Seven reports, with a cumulative total of more than 78,000 patients, demonstrated a peak of sudden cardiac death between 6 am and noon.6, 11, 21, 22, 23, 24, 25
Literature dedicated to inpatient circadian variation of cardiopulmonary arrest is limited and consistently nonuniform in overall findings. Only 7 single-institution, inpatient studies exist, with a cumulative total of only 1140 patients.26, 27, 28, 29, 30, 31, 32 All studies predate the Utstein in-hospital recommendations.19 Variations in findings may result from small sample sizes, variable time periods, and a wide variability in populations; inpatient studies are clearly under-represented.
This is one of a series of reports from an ongoing study of in-hospital resuscitation designed to include an extensive range of variables for analysis over time. As suggested by the Utstein style,19 hospital, patient demographics, arrest, and outcome variables have been incorporated into the database. The current report extends the previously reported data set33, 34 to encompass a 14-year time span. The objective of this study was to demonstrate the presence or absence of circadian variation in frequency of cardiopulmonary arrest in a hospital setting, across sex, age, diagnosis, witness of event, monitoring of an event, year of study, and rhythm. A second objective was to delineate any influence of circadian variation of cardiopulmonary arrest on survival to discharge.
Materials and methods
Data Collection
A more detailed description of methods has been reported.33, 34 This longitudinal registry (retrospective cohort) includes all resuscitation attempts, excluding those occurring in operating rooms and the neonatal intensive care unit, at the Medical Center of Central Georgia (MCCG), from January 1, 1987 through December 31, 2000. This study was reviewed and approved by the Institutional Review Boards of MCCG and Mercer University School of Medicine (MUSM). Events are identified through review of standardized resuscitation records (code sheets), hospital financial and diagnostic records, quality appraisal review, and hospital death records. Information from each code sheet is entered into a database and cross-referenced. Detailed chart review is completed for all questionable events. All registry entries are thoroughly reviewed.
Hospital Variables
MCCG is a 634-bed tertiary care center located in Macon, Ga; it is the major teaching affiliate of MUSM. Resuscitation attempts are conducted by multidisciplinary teams trained in advanced cardiac life support (ACLS) and include residents, attending physicians, nurses, respiratory therapists, pharmacists, and ancillary staff. Code teams are notified by intercom and personal pagers.
Patient Variables
Approximately 53% (n=2439) of patients were male, and 47% (n=2202) were female. Patients ranged in age from 0 to 103 years (mean age of 63 years); ∼3% (n=133) were <20 years old, 31% (n=1437) were between the ages of 20 and 59 years, and the remaining 66% (n=3068) were aged >59 years; age was not available for 3 events.
Arrest Variables
A resuscitation attempt was defined, following Bedell35 and Utstein,19 as any attempt to reverse clinical death using ACLS protocols.36 To consider an event a resuscitative effort, sudden changes in level of consciousness, respirations, or rhythm incompatible with spontaneous recovery were required. Initial rhythm for monitored patients was defined as rhythm displayed on the monitor when the code was called; initial rhythm for unmonitored patients was defined as the first identifiable rhythm after connecting a monitor. Events were considered witnessed if a change in rhythm was discovered on telemetry or a patient decompensated in the presence of a staff member or visitor. Rhythms are categorized into 6 groups, including supraventricular tachycardia, ventricular tachycardia, ventricular fibrillation, perfusing, pulseless electrical activity, and asystole, as previously cited.33
Outcome Variables
Analysis of circadian variation is reported in simple frequencies. The sole outcome variable is survival to discharge. To ensure accuracy, outcomes on code sheets, hospital discharge status, and death logs were cross-validated.
Statistical Analysis
Statistical analyses were conducted via SAS/STAT version 9.1.37 The frequencies of cardiopulmonary arrest were computed for the study population by 1-, 2-, 4-, and 8-hour intervals. We utilized χ2 tests for goodness-of-fit to detect differences in the frequency of arrest within these time intervals.
Focused analyses on 4-hour intervals were based on previous findings,32 avoiding small cell sizes, and stability of analyses. Data were combined into 6 4-hour intervals with time beginning at 7 am to capture change of shifts in our institution. Simple categorical predictor variables were converted to contrast codes; eg, for sex, male and female were respectively coded 1 and −1. Comparisons were assessed using simple effects by χ2 analysis. Variables included in focused analysis are shown in Table 1.
Table 1. χ2 Analyses of Event Frequency versus Other Variables
| Variable | χ2 | P Value |
|---|---|---|
| Sex | 2.9 | .72 |
| Age group | 7.5 | .67 |
| Year of study | 72.7 | .24 |
| Card Dis E Card Surg v Other Diag | 2.3 | .80 |
| Card Surg E Card Dis v Other Diag | 7.1 | .21 |
| All Card Dis v Other Diag | 4.8 | .44 |
| Monitored vs unmonitored | 4.5 | .48 |
| Witnessed vs unwitnessed | 22.9 | .0004 |
| Survival to discharge | 9.0 | .11 |
Frequencies of witnessed and unwitnessed events across rhythms were compared by χ2. Because unwitnessed supraventricular tachycardia and ventricular tachycardia were rare, these were excluded from this analysis.
Simple logistic regressions with survival to discharge as the outcome variable were completed for the following predictor variables: time in 1-hour and 4-hour intervals, and initial rhythm. Multiple logistic regression analyses were used to assess association between time and rhythm with survival as the outcome. Survival to discharge was treated as a dichotomous measure in all analyses.
Results
Data are reported according to Utstein style19 (Figure 1). From January 1, 1987 through December 31, 2000, there were 354,763 total admissions and 4692 admissions in which resuscitation was attempted. Fifty-one calls were determined to be nonarrests, yielding a total study group of 4641. There were 1435 patients discharged alive (31%). Supraventricular tachycardia accounted for 2% of events (n=80), ventricular tachycardia 3% (n=143), ventricular fibrillation 14% (n=641), perfusing 31% (n=1422), pulseless electrical activity 35% (n=1634), and asystole 15% (n=714).
Figure 1.
In-hospital Utstein style template. ED
=emergency department; SVT=supraventricular tachycardia; BLS=basic life support; ALS=advanced life support; VF=ventricular fibrillation; VT=ventricular tachycardia; PEA=pulseless electrical activity.
χ2 goodness-of-fit concerning frequency of events in 1-hour time periods was significant (χ2 [23]=41.2, P=.01) (Figure 2). The goodness-of-fit test for 4-hour time periods was not significant (χ2 [5]=7.2, P=.20), but the obvious variation exhibited in Figure 3 and previous reports of 4-hour periodicity30 convinced us to compare the late-morning peak to the late-evening trough in a separate test, which was significant (χ2 [1]=6.5, P <.025).
Figure 2.
Goodness of fit for hourly event frequency. χ2
=41.2; P=.01.
Figure 3.
Goodness of fit for frequency of events. χ2
=7.2; P=.20; χ2=6.5; P <.025 (comparing peak and trough).
χ2 tests did not detect any significant relationships between 4-hour periodicity and the following variables: sex, age group, year of study, and monitored versus unmonitored. There was also no statistically significant relationship between frequency in 4-hour time intervals and cardiac diseases versus other diagnoses (Figure 4). The only significant relationship between 4-hour periodicity and patient variables was with initial rhythm (χ2 [25]=45.2, P=.008), which is displayed in Table 2. Further analysis of initial rhythms indicated that only perfusing rhythms (χ2 [5]=12.4, P=.03) and asystole (χ2 [5]=14.6, P=.01) exhibited circadian variation with the 4-hour periods (Figure 5, Figure 6, respectively).
Figure 4.
Cardiac disease event frequency versus other diagnoses. χ2
=3.4; P <.75 for other diagnoses; χ2=8.9; .10<P <.25 for all cardiac disease.
Table 2. Frequency of Events by Rhythm
| Time Period | Rhythm | ||||||
|---|---|---|---|---|---|---|---|
| SVT | VT | PER | VF | PEA | ASYS | Total | |
| 7 am-11 am | 17(2.2) | 26(3.3) | 260(33.4) | 107(13.7) | 248(31.8) | 121(15.5) | 779 |
| 11 am-3 pm | 12(1.5) | 30(3.7) | 264(32.4) | 115(14.1) | 292(35.9) | 101(12.4) | 814 |
| 3 pm-7 pm | 16(2.0) | 20(2.6) | 241(30.7) | 114(14.5) | 296(37.8) | 97(12.4) | 784 |
| 7 pm-11 pm | 14(1.8) | 21(2.8) | 214(28.1) | 123(16.2) | 270(35.5) | 119(15.6) | 761 |
| 11 pm-3 am | 12(1.7) | 19(2.6) | 209(29.0) | 89(12.3) | 269(37.3) | 123(17.1) | 721 |
| 3 am-7 am | 8(1.1) | 26(3.5) | 218(29.4) | 87(11.7) | 253(34.1) | 149(20.1) | 741 |
| Total | 79(1.7) | 142(3.1) | 1406(30.6) | 635(13.8) | 1628(35.4) | 710(15.4) | 4600 |
Figure 5.
Goodness of fit for perfusing rhythm event frequency. χ2
=12.4; P=.03.
Figure 6.
Goodness of fit for asystolic event frequency. χ2
=14.6; P=.01.
Despite the fact that most events (84.6%) were witnessed, this institutional variable exhibited circadian variation within the 4-hour time periods (χ2 [5]=22.9, P=.0004) (Table 3). The greatest percentage of unwitnessed events occurs during 3:00-6:59 am (19.5%). The 4-hour periodicity of unwitnessed events was strongly related to initial rhythm (χ2 [15]=35.9, P=.002). As evidenced in Table 4, a significantly greater percentage of unwitnessed events were found as asystole during the time periods of 11:00 pm-2:59 am and 3:00-6:59 am. As depicted in Figure 7, there were more unwitnessed than witnessed events among those whose initial rhythm was asystole during the overnight hours (χ2 [5]=18.8, P=.002). There was also a significantly lower percentage of events in which the initial rhythm was ventricular fibrillation during the same 11:00 pm-2:59 am and 3:00-6:59 am time periods (χ2 [5]=11.1, P=.05).
Table 3. Frequency of Events by Witnessed versus Unwitnessed Arrest
| Time Period | Unwitnessed | Witnessed | Total |
|---|---|---|---|
| 7 am-11 am | 109(14.8) | 626(85.2) | 735 |
| 11 am-3 pm | 96(12.4) | 677(87.6) | 773 |
| 3 pm-7 pm | 102(13.9) | 632(86.1) | 734 |
| 7 pm-11 pm | 99(13.9) | 615(86.1) | 714 |
| 11 pm-3 am | 128(18.7) | 555(81.3) | 683 |
| 3 am-7 am | 135(19.5) | 556(80.5) | 691 |
| Total | 669(15.5) | 3661(84.6) | 4330 |
Table 4. Frequency of Unwitnessed Events by Rhythm
| Time Period | PER | VF | PEA | ASYS | Total |
|---|---|---|---|---|---|
| 7 am-11 am | 17(15.7) | 19(17.6) | 29(26.9) | 43(39.8) | 108 |
| 11 am-3 pm | 14(14.7) | 6(6.3) | 42(44.2) | 33(34.7) | 95 |
| 3 pm-7 pm | 12(11.8) | 11(10.8) | 45(44.1) | 34(33.3) | 102 |
| 7 pm-11 pm | 10(10.2) | 13(13.3) | 37(37.8) | 38(38.8) | 98 |
| 11 pm-3 am | 13(10.2) | 7(5.5) | 42(33.1) | 65(51.2) | 127 |
| 3 am-7 am | 16(12.0) | 7(5.3) | 38(28.6) | 72(54.1) | 133 |
| Total | 82(12.4) | 63(9.5) | 233(35.1) | 285(43.0) | 663 |
Figure 7.
Frequency of asystolic events witnessed versus unwitnessed. χ2
=18.8; P=.002.
Using simple logistic regression, we identified a significant relationship between time in 1-hour intervals and survival to discharge (χ2 [1]=8.2, P=.004), as well as between time in 4-hour intervals and survival to discharge (χ2 [1]=8.6, P=.003). As depicted in Figure 8, there is clearly decreased survival in the overnight hours. Initial rhythm was significantly, and much more strongly, related to patient survival until hospital discharge (χ2 [1]=680.1, P <.0001). When both rhythm and 4-hour periodicity were entered simultaneously into logistic regression on survival, initial rhythm was the only significant predictor (χ2 [1]=669.9, P <.0001); 4-hour periodicity became nonsignificant (χ2 [1]=2.4, P=.12).
Figure 8.
Frequency of events by survival to discharge. χ2
=9.0; P=.11.
Discussion
Our results demonstrate circadian variation of cardiopulmonary arrest in a hospital setting. Nonuniform distribution of resuscitation attempts was clearly obtained for 1-hour periods, but not so clearly for 4-hour periods. The 4-hour period is most prevalent in previous in-hospital literature.29, 32 Although minor peaks and troughs were lost when focusing on 4-hour time intervals, the relationships between 4-hour periodicity and witnessed/unwitnessed, initial rhythm, and survival lead us to conclude this is the more useful time period for understanding circadian rhythm in an in-hospital population.
In contrast to numerous studies that have shown a morning peak of cardiopulmonary arrest,6, 11, 21, 22, 23, 24 acute coronary syndrome,2, 3, 4, 5 and sudden cardiac death,2, 3, 6, 7, 8, 9, 10, 11, 12, 13 we obtained a broader distribution without a clear peak (Figure 2). Previous findings are consistent with physiology studies revealing circadian variation of factors influencing ischemic events such as cortisol secretion,38, 39 platelet aggregation,40, 41 catecholamine secretion,39 heart rate,42, 43 and blood pressure.42, 43 However, interventional studies with anti-ischemic medications, including beta-blockers and low-dose aspirin, have shown modification or abolition of the circadian distribution of cardiac arrhythmias, sudden cardiac death, and fatal myocardial infarction.7, 44, 45, 46, 47, 48, 49
Population studies, multi-center reports, and meta-analyses of the circadian variation of cardiopulmonary arrest also have been completed. Mortality data from Framingham revealed a significant circadian variation of sudden cardiac death, with a peak incidence from 7 to 9 am and a decreased incidence from 9 am to 1 pm.13 Analysis of the Multicenter Investigation of Limitation of Infarct Size database revealed a statistically greater incidence of myocardial infarction from 6 am to noon.3 A meta-analysis of acute myocardial infarction and sudden cardiac death, which included 30 studies and 66,635 patients, revealed a significant peak of acute myocardial infarction between 6 am and noon.2 The greatest advantage of meta-analyses is large sample sizes; however, each uses data collected in both inpatient and outpatient arenas. Disadvantages include lack of uniformity in definitions, reporting, statistical analysis, and population demographics across individual studies.
Resuscitation is a valid model to study sudden death; however, many studies that demonstrated diurnal variation restricted patients to sudden cardiac death or acute ischemia. We identified the events occurring among patients with cardiac disease and those with acute coronary syndromes. We found no significant diurnal variation among these groups. Comparison with subjects with other illnesses showed no differences.
There is significant circadian variation by rhythm. Subdivision into individual rhythms reveals significant circadian variation for only perfusing rhythms and asystole (Table 2). Asystole increases from a nadir of 12.4% during the 11:00 am-2:59 pm and 3:00-6:59 pm periods, to a peak of 20.1% from 3:00-6:59 am. Perfusing rhythms vary from a maximum of 33.4% during the 7:00-10:59 am period to a minimum of 28.1% during the 7:00-10:59 pm period. Perfusing rhythms represent an earlier stage of arrest. As discussed below, there are more witnessed arrests during the hours of 7:00-10:59 am, and as a result, patients are more likely to be found in earlier stages of arrest. Although circadian variation of ventricular fibrillation was not statistically significant, there was clearly a trend in its occurrence. The highest percentage of events (16.2%) recognized as ventricular fibrillation were during 7:00-10:59 pm and the lowest, only 11.7%, occurring during 3:00-6:59 am. The overall incidence of ventricular fibrillation in the hospital is low. Pulseless electrical activity, supraventricular tachycardia, and ventricular tachycardia were fairly uniformly distributed.
Previous research has emphasized the prognostic value of witnessed or unwitnessed events.50 The majority of all events (∼85%) at our institution are witnessed, but there is a statistically significant difference in the likelihood of witnessed events as a function of time. From 7:00 am-10:59 pm, approximately 86% of all events are witnessed; this falls to 81% from 11:00 pm-2:59 am. When unwitnessed events are further stratified into rhythm categories, the rhythm distribution is significantly different from expected. As represented in Table 4, the occurrence of asystole during the overnight time period is significantly higher than expected, whereas the percentage of events recorded as ventricular fibrillation is significantly lower.
Perhaps of greatest interest is that circadian variation exists for survival to discharge, with significantly lower survival from 7:00 pm-6:59 am (Figure 8). Initial rhythm was a significant predictor of survival, consistent with previous reports.33, 34 Multiple logistic regression modeling demonstrated that the difference in survival was explained by circadian changes in initial rhythm. Coupled with the variation in witnessed events, it appears that the circadian variation of survival may reflect a higher proportion of patients discovered later in the dying process during the overnight hours, when fewer staff members are available for early detection.
Limitations of this study include the fact that this was an exploratory analysis, and data were assessed in a variety of ways to determine significant relationships. Variables were added to preliminary analyses to better understand initial findings. Multiple analyses may have contributed to positive findings. Variations in time intervals and methods of analysis occasionally led to differing results in statistical significance. Although the chosen time periods were clinically reasonable, further replication of the relevance of 4-hour periodicity to in-hospital resuscitation is needed. Perhaps more importantly, we were unable to obtain hourly census information, so we cannot determine whether the increased frequency of events resulted from a higher number of patients in the hospital during those time periods. Additionally, this study was completed within a single institution that is a referral center. Our facility receives many acute coronary syndromes from outlying facilities and completes a large number of open-heart surgeries. Medication effects were not assessable in this data set. This may limit generalizability of these findings to smaller institutions. Finally, this study follows the Utstein guidelines19 for reporting in-hospital resuscitation, which means that findings reported herein should consistently lend themselves to further study. However, because all previous in-hospital studies predated the standardized reporting recommendations, comparison warrants extreme caution.
We have found that circadian variation in hospital is strongly related to initial rhythm, which is different from previously reported population and outpatient studies. We obtained no circadian variation of cardiopulmonary arrest specific to patients with primary cardiac disease. The explanation for this is uncertain and merits further investigation. Clearly, as a referral center, with fluctuations in time of admissions and modification of acute coronary syndrome with medications, as well as interventions, differences should be expected. Finally, late night variation in witnessed events and initial rhythm suggest delay between onset of clinical death and discovery, which certainly contributes to negative outcome.
References
- . Closed-chest cardiac massage. JAMA. 1960;173:1064–1067
- Meta-analysis of the morning excess of acute myocardial infarction and sudden cardiac death. Am J Cardiol. 1997;79:1512–1516
- Circadian variation in the frequency of onset of acute myocardial infarction. N Engl J Med. 1985;313:1315–1322
- . Variation in the time and day of onset of myocardial infarction and sudden death. N Z Med J. 1995;108:332–334
- . Circadian variation and triggers of acute coronary syndromes. Eur Heart J. 1998;19(suppl C):C12–C23
- Circadian variation of sudden cardiac death reflects age-related variability in ventricular fibrillation. Circulation. 1993;88:2284–2289
- . Circadian variation of primary cardiac arrest or sudden cardiac death in patients aged 62 to 100 years (mean 82). Am J Cardiol. 1993;71:1455–1456
- . Sudden cardiac death, circadian rhythms and defibrillators. Eur Heart J. 1995;16:1162–1164
- . Individual circadian variation and sudden death. Circulation. 1999;100:447–448
- . Circadian patterns and triggers of sudden cardiac death. Cardiol Clin. 1996;14:185–194
- Circadian variation in sudden cardiac death: effects of age, sex, and initial rhythm. Ann Emerg Med. 1996;27:29–34
- . Epidemiologic studies demonstrating increased morning incidence of sudden cardiac death. Am J Cardiol. 1990;66:15G–17G
- Circadian variation in the incidence of sudden cardiac death in the Framingham Heart Study population. Am J Cardiol. 1987;60:801–806
- Circadian distribution in fatal pulmonary thromboembolism. Am J Cardiol. 1989;64:403–404
- Sudden death from pulmonary thromboembolism: chronobiological aspects. Eur Heart J. 1992;13:661–665
- Circadian variation in the frequency of ischemic stroke. Stroke. 1990;21:387–389
- . Circadian variation in the timing of stroke onset: a meta-analysis. Stroke. 1998;29:992–996
- Recommended guidelines for uniform reporting of data from out-of-hospital cardiac arrest: the Utstein style. Ann Emerg Med. 1991;20:861–874
- Recommended guidelines for reviewing, reporting, and conducting research on in-hospital resuscitation: the in-hospital “Utstein style”. Ann Emerg Med. 1997;29:650–679
- Circadian variation in the frequency of sudden cardiac death. Circulation. 1987;75:131–138
- . Further prospective evidence of a circadian variation in the frequency of call for sudden cardiac death. Intensive Care Med. 1995;21:45–49
- Diurnal, weekly and seasonal variation of sudden death (Population-based analysis of 24,061 consecutive cases). Eur Heart J. 2000;21:315–320
- Prospective evidence of a circadian rhythm for out of hospital cardiac arrests. JAMA. 1992;267:2935–2937
- Circadian variation of cardiopulmonary disease onset in the general population: an emergency care system perspective from Berlin. Ann Emerg Med. 1994;23:281–285
- . Circadian variation in the time of request for helicopter transport of cardiac patients. Ann Emerg Med. 1992;21:1196–1199
- . Sudden cardiac arrest: clinical characteristics and predictors of survival. Jpn Circ J. 1989;53:1536–1540
- . Circadian variation in unexpected postoperative death. Br J Surg. 1992;79:1300–1302
- . Cardiopulmonary resuscitation of old people. Lancet. 1983;2:267–269
- . Cardiac resuscitation in 512 hospitalized patients. Penn Med. 1970;73:45–48
- . Cardiopulmonary resuscitation in a teaching hospital. Anaesthesia. 1981;36:526–530
- . Cardiorespiratory arrests—a diurnal variation?. Br J Anaesth. 1979;51:72
- Circadian distribution of in-hospital cardiopulmonary arrests on the general medial ward. Arch Intern Med. 1992;152:1282–1288
- Resuscitation in the hospital: differential relationships between age and survival across rhythms. Crit Care Med. 1999;27:2137–2141
- Resuscitation in the hospital: relationship of year and rhythm to outcome. Resuscitation. 2000;47:219–229
- . Survival after cardiopulmonary resuscitation in the hospital. N Engl J Med. 1983;309:569–576
- In: Textbook of Advanced Cardiac Life Support. Dallas, TX: American Heart Association; 2003;p. 1
- . SAS/STAT Users Guide. Cary, NC: SAS Institute, Inc; 2003;
- . Circadian rhythms and coronary artery disease. Am J Cardiol. 1987;59:13C–17C
- . Circadian variations of plasma catecholamine, cortisol and immunoreactive insulin concentrations in supine subjects. Clin Chim Acta. 1974;55:389–397
- Concurrent morning increase in platelet aggregability and the risk of myocardial infarction and sudden cardiac death. N Engl J Med. 1987;316:1514–1518
- . Circadian variation in circulating platelet aggregates. Ann Med. 1989;21:429–433
- . Circadian variation of blood-pressure. Lancet. 1978;1:795–797
- Ambulatory intra-arterial blood pressure in normal subjects. Am Heart J. 1990;120:160–166
- Propranolol and the morning increase in the frequency of sudden cardiac death (BHAT Study). Am J Cardiol. 1989;63:1518–1520
- Circadian pattern of arrhythmic death in patients receiving encainide, flecainide or moricizine in the Cardiac Arrhythmia Suppression Trial (CAST). J Am Coll Cardiol. 1994;23:283–289
- Modification of the circadian pattern of ventricular tachyarrhythmias by beta-blocker therapy. Clin Cardiol. 1997;20:253–257
- Increased morning incidence of myocardial infarction in the ISAM Study: absence with prior beta-adrenergic blockade. Circulation. 1989;80:853–858
- Circadian variation of acute myocardial infarction and the effect of low-dose aspirin in a randomized trial of physicians. Circulation. 1990;82:897–902
- . Circadian variation of the total ischemic burden and influence by beta-blocking agents. J Cardiovasc Pharmacol. 1990;16(suppl 5):S100–S104
- Validation of a clinical decision aid to discontinue in-hospital cardiac arrest resuscitations. JAMA. 2001;285:1602–1606
PII: S0002-9343(06)00781-9
doi:10.1016/j.amjmed.2006.06.032
© 2007 Elsevier Inc. All rights reserved.
