The purpose of this study was to determine the relationship of four postanesthesia care unit (PACU) cardiovascular events to long-term outcomes (unplanned critical care admission or mortality) and to evaluate the contribution of anesthetic management compared with other perioperative factors in predicting these events.
For patients admitted to the PACU after receiving general anesthesia (n = 18,380), the risk of long-term outcomes was examined for patients in the PACU with hypertension, tachycardia, bradycardia, or hypotension. Using logistic regression (P < 0.05), risk factors (grouped as patients, surgical, anesthetic, operating room observations, and other PACU observations) for each cardiovascular event were determined. For each factor grouping, the relative contributions to each cardiovascular event were compared using maximum likelihood chi-square analysis.
Patients in the PACU with hypertension or tachycardia had more unplanned critical care admissions (2.6% and 4.0% vs. 0.2% for patients with no events) and greater mortality (1.9% and 2.3% vs. 0.3% and 0.4%) (P < 0.01). For PACU hypertension (rate 2.0%), age, smoking, renal disease, female gender, and angina were significant risk factors. For PACU tachycardia (0.9%), intraoperative tachycardia and dysrhythmia were the major contributors. Patient factors also increased the risk of bradycardia (2.5%); namely age, ASA physical status 1 or 2, and preoperative beta blocker therapy. For hypotension (2.2%), duration of surgery > 2 h, completion after 6 PM, and gynecologic intraabdominal procedures were significant risk factors. Compared to patient, surgical, intraoperative, or PACU observations, anesthetic factors studied (premedication, induction agent, ventilation, use of opioids) provided only a small contribution in predicting these events.
Hypertension and tachycardia in the PACU, although infrequent, are associated with increased risk of unplanned critical care admission and mortality. Patient, surgical, intraoperative, or PACU observations contribute more to cardiovascular events in the PACU than do differences in anesthetic management identified in this study.
Key words: Anesthesia: quality assurance. Complications: bradycardia; hypertension; hypotension; tachycardia. Postanesthesia care unit: outcomes.
ASSESSING outcomes of clinical practice is an important component in determining the quality of anesthesia care. However, is it meaningful to determine only the frequency of common anesthetic-related problems in the postanesthesia care unit (PACU)? It may be more important to determine the long-term ramifications when patients experience problems in the PACU and if anesthetic management plays a role in the occurrence of these events.
Other studies have highlighted the frequency of cardiovascular problems in the PACU, but the definitions of cardiovascular events were imprecise and many were forced into single categories. [1-3]These studies surveyed patient problems in the PACU but did not delineate risk factors (patient, surgical, or anesthetic) or other outcomes associated with these events.
In this study, we focus on the contribution of anesthetic management to cardiovascular events in the PACU as compared with other perioperative factors to determine which have the most impact in predicting cardiovascular events. The study addresses three inter-related questions. First, do cardiovascular events occurring in the PACU show a clinically significant association with long-term adverse outcomes (i.e., unplanned critical care admission and mortality)? Second, if these associations do exist, are they linked to direct aspects of anesthetic care (e.g., anesthetic technique, intraoperative anesthetic observations)? Third, are there other important factors beyond direct anesthetic care (e.g., patient or surgical factors) that contribute to these cardiovascular events and what are their relative contributions as compared with anesthetic care?
Methods
Patient Population and Data Collection
After obtaining institutional and ethical approval for the study, we prospectively collected preoperative, intraoperative, and postoperative information on all patients admitted to the PACU who had received general anesthesia (n = 18,380) during a 2-yr period (January 1991-December 1992). Patients who had postoperative mechanical ventilation were transferred directly from the operating room (OR) to a critical care unit, so they were not included in the study.
The PACU record at the hospital was redesigned to log cardiovascular events (hypertension, tachycardia, bradycardia, hypotension), as well as other observations of patient problems with the definitions clearly displayed on the form (Table 6). These events and observations of patient problems were based on physiologic parameters outside a normal range or specific treatment interventions. Similarly, the anesthetic record (OR record) was reconfigured in a tick-off format with definitions so that patient demographics, illnesses, current medications, anesthetic management strategies, and intraoperative patient observations were included. Copies of both records were reviewed daily by a research nurse and a clinical anesthesiologist to ensure consistency and completeness of data. Surgical procedures were encoded according to the International Classification of Diseases codes (ninth revision) and further subdivided into 18 categories based on physiologic trespass. All data, including information on preoperative, intraoperative, and PACU drugs and dosages, were entered into an anesthetic database and were analyzed with SAS statistical software (version 6.04, SAS, Cary, NC). [4]As a basis for subsequent analysis, the rates of four PACU cardiovascular events (hypertension, tachycardia, bradycardia, and hypotension) were determined from the database.
Statistical Analysis
Our initial analysis sought to determine if patients having one of the four cardiovascular events were at risk for long-term morbidity or mortality. We used unplanned critical care admission from the PACU (admission to any postoperative unit with increased nursing:patient ratio compared with standard ward care, not anticipated preoperatively) and in-hospital mortality for each cardiovascular event as a measure of long-term outcome. Only the initial operative visit for patients with multiple visits to the OR during a single hospital admission was included in the mortality analysis. The crude rates of unplanned critical care admissions and of in-hospital mortality were determined for patients with specific cardiovascular events as compared with patients who did not have the specific cardiovascular event with statistical significance determined at the P < 0.01 level.
We next determined if the cardiovascular events (hypertension, tachycardia, bradycardia, and hypotension examined separately) were directly associated with aspects of anesthetic care. This entailed the use of statistical modeling involving three stages.
For stage one, a list of potential variables that might be related to the four events was generated from the literature, clinical practice, and from our own previous research. [5,6]Individual items were then grouped into patient, surgical, anesthetic, intraoperative observations, or other PACU observations (see Table 6for a list of variables and their definitions). Using chi-square analysis, each variable (grouped categorically) was tested to determine if it was significantly related to the event (P < 0.05).
Patient factors included age, gender, weight, ASA physical status, smoking, medical illnesses, and current medications with each characteristic compared to a reference group. For example, were males more likely than females to be at risk for hypertension, or were patients who currently received beta blockers at greatest risk for bradycardia compared with patients who received no medications? Surgical variables examined were outpatient status, emergency procedure, OR duration, OR completion time, and the type of procedure.
Specific anesthetic risk factors tested for an association with the four cardiovascular events were type of premedication, induction agent, ventilation, inhalational agent, and intraoperative opioids at different dosage ranges. Patients with noteworthy intraoperative or PACU observations were compared with patients with no intraoperative or PACU observations.
For stage two, only variables significant in stage one were used in logistic regression models. Because some patients had more than one significant factor, we used hierarchical groupings to ensure mutually exclusive categories. For example, a patient in the PACU may have experienced inadequate ventilation, excessive pain, and nausea/vomiting. The hierarchical grouping worked as follows: The variable with the greatest statistical significance was examined first so that for this example, all patients who had inadequate ventilation and any other problem were grouped. The next grouping was any remaining patients who had excessive pain and any other problem. Finally, remaining patients with nausea/vomiting were counted. Patients with none of these problems in the PACU formed the reference group. Hierarchical categories were developed for preoperative illnesses, current medications, and OR and PACU observations.
Variables that were highly correlated with one another were also identified (for example, physical status and history of angina), and only those with the most clinical relevance were retained in the model (determined on a case by case basis). Once hierarchical categories were developed and correlated variables were eliminated, we created logistic regression models for each of the five factor groupings (i.e., patient, surgical, anesthetic, OR observations, and other PACU observations) and thus ran five logistic regressions for each of the four cardiovascular events (20 models in all).
As a third step, only those factors that were significant (P < 0.05) in the step two regression models were included in a final overall logistic regression analysis. Thus, for each of the four cardiovascular events, a single logistic regression model included the remaining significant variables from the patient, surgical, anesthetic, intraoperative observations, and other PACU observations groupings. The relative odds (and 95% confidence intervals) of the event occurring were determined for each variable.
The final analysis was the determination of the relative contribution of patient, surgical, OR observations, and PACU observations to the four cardiovascular events as compared with anesthetic care. This was determined using an extension of the logistic regression method. Whereas multiple linear regression uses least squares, logistic regression analysis uses a technique known as the "maximum likelihood estimation" to determine the final parameter estimates. [7]To calculate if the overall model is significant, the statistic D = -2 ln (likelihood ratio) is tested against a chi-square distribution.
To determine the relative contribution of specific variables to predict the event, the model is rerun leaving out the variables of interest. The value of D is recalculated and the difference in value of D between the model with all the variables in and the value of D without the variables of interest is determined. This numeric "difference" also can be tested for statistical significance using a chi square test with the number of degrees of freedom derived from the number of variables in the model. In other words, this method tested whether a grouping of variables contributed significantly to the model. For example, the model is first run with all significant variables in the five factor groupings then recalculated leaving out patient-related variables; a third run restores patient variables but leaves out surgical variables and so forth.
This allowed us to determine the relative contribution of the five factor groupings (i.e., patient, surgical, anesthetic, intraoperative observations, and other PACU observations), "removing" each factor grouping and deriving the chi-square values for the remaining factor groupings. The magnitude of the chi-square values were compared; the larger the difference, the greater the relative contribution of that variable grouping. These differences were displayed as histograms for the four cardiovascular events. The four events cannot be directly compared to one another because they are derived from different models, but the relative size of the differences for anesthetic factors as compared to patient, surgical, OR observations, and other PACU observations can be evaluated.
Results
Rate of Cardiovascular Events in the Postanesthesia Care Unit
Most patients admitted to the PACU were healthy (87.2% were ASA physical status 1-2, 46.1% had no preoperative illnesses, and 57.0% had no current medications); outpatients composed 32.0% of the population, 91.2% of procedures were done on an elective basis, and 88.9% were completed on weekdays between 8 AM and 6 PM. The most common procedures involved intraabdominal gynecologic surgery (17.8%) and major extremity orthopedic surgery (13.7%).
One or more cardiovascular events occurred in 7.2% of total admissions during the 2 yr (hypertension (2.0%), tachycardia (0.9%), bradycardia (2.5%), and hypotension (2.2%)). Medication given in the PACU for hypertension included opioid analgesics for 28.3% of patients and vasodilators (nitroglycerin or nifedipine) for 20.6%. Those with tachycardia received beta blocking agents (7.5%) and with bradycardia, atropine (26.8%). Crystalloid fluid boluses were given to all hypotensive patients.
Long-term Outcomes
Morbidity and mortality for patients who experienced cardiovascular events in the PACU are displayed in Table 1. Patients who had PACU hypertension or tachycardia were at increased risk of an unplanned critical care admission and had a greater hospital mortality rate than patients who did not (P < 0.01). Neither bradycardia nor hypotension in the PACU was associated with increased rate of unplanned critical care admission or mortality.
Risk Factors for Cardiovascular Events in the Postanesthesia Care Unit
Significant patient-related factors for increased risk of PACU hypertension included increasing age, smoking, and renal disease (Table 2). Patients with a preoperative history of angina and female patients were less likely to be hypertensive. Long duration (greater or equal to 2 h), late OR finish (> 6 PM), and intracranial procedures placed patients at increased risk of hypertension in the PACU, whereas outpatients had a lower risk. When hypertension or tachycardia was documented in the OR, patients were more likely to be hypertensive in the PACU. Patients with inadequate ventilation, excessive pain, and nausea or vomiting in the PACU were also at increased risk for hypertension.
Tachycardia in the PACU occurred less frequently among older patients, smokers, and those with lower physical status scores (Table 3). Emergency surgical procedures as well as those longer than 4 h were more likely to be associated with tachycardia in the PACU. Significant observations predictive of PACU tachycardia included intraoperative tachycardia and dysrhythmias as well as other problems in the PACU, namely, inadequate ventilation, presence of an endotracheal tube on arrival at the PACU, agitation, and shivering.
Only a few factors were related to bradycardia in the PACU (Table 4). These included increasing age, ASA physical status scores 1 or 2, current use of beta blockers, bradycardia observed in the OR, and in the PACU, nausea and vomiting. Patients whose operation ended late in the day had lower risk of PACU bradycardia.
Hypotension in the PACU was more commonly seen among older and female patients (Table 5). For operations lasting longer than 1 h and for intraabdominal gynecologic procedures, hypotension in the PACU was more common. When hypotension or tachycardia were documented in the OR, the rate of PACU hypotension was increased as well as when shivering or nausea or vomiting was noted during the PACU stay.
Different anesthetic choices were associated with the four cardiovascular events (Table 2, Table 3, Table 4, Table 5). For PACU hypertension or tachycardia, the one significant anesthetic factor was choice of opioid: fentanyl with alfentanil was associated with an increased risk of PACU hypertension and fentanyl with morphine was associated with PACU tachycardia. For PACU bradycardia, patients receiving propofol had a greater risk of bradycardia than patients receiving thiopental. Those having spontaneous compared with mechanical ventilation had lower risk. For hypotension, the use of opioids or sedatives for premedication raised the risk. Choice of inhalational agent (isoflurane or enflurane) showed no relationship with the four cardiovascular events in the PACU.
The Relative Contribution of Anesthetic Choices in Predicting Cardiovascular Events
The contributions of the five risk-factor groupings to the total model chi-square value associated with the cardiovascular events were displayed as histograms (Figure 1). All factor groups contributed significantly (P < 0.05); however, the relative contributions of patient, surgical, anesthetic, OR observations, and other PACU observations differed among the four cardiovascular events. Anesthetic factors provided only small contributions toward predicting risk for the four events. For hypertension and bradycardia, patient factors had the greatest influence; for tachycardia, OR observations were the most important, and for hypotension, surgical factors had by far the greatest contribution.
Discussion
Our study found that two cardiovascular events in the PACU (hypertension and tachycardia) were associated with long-term morbidity and mortality; two other events (bradycardia and hypotension) did not show a clinically significant association with outcomes in the long term. A current debate in the anesthesia literature surrounds what is (are) the best outcome measure(s) for evaluating the quality of anesthesia care. Using intermediate events (such as cardiovascular events in the PACU) has some advantages in that they can be clearly defined, occur frequently, and were thought to be more directly attributable to what the anesthesiologist does than are major outcomes such as death. [8]However, some have cogently argued that the so-called "intermediate" or "surrogate" events on which most anesthesia studies rely are problematic. While the occurrence of these intermediate events may be undesirable, if they do not lead to serious true outcomes such as mortality, longer recovery time, or unanticipated admission to hospital, are they appropriate as an endpoint? [9,10]Furthermore, abnormal physiologic readings may not necessarily be pathologic. [8]Others also contest that ultimately what happens to patients should be the true endpoint to assess quality of anesthesia care. [3,9,11].
Having shown that patients experiencing two cardiovascular events (hypertension and tachycardia) were at greater risk for unplanned admission to the critical care unit and at greater risk of dying illustrates that the occurrence of these two intermediate events are markers of increased risk for long-term outcomes (i.e., true outcomes). However, our analysis of the relative contribution of the five factor groupings (patient, surgical, anesthetic, OR observations, and other PACU observations) revealed that specific anesthetic choices contributed little toward these cardiovascular events. Thus, while experiencing two of these cardiovascular events places patients at risk for true adverse health outcomes (critical care admission or mortality), specific choices (drugs and technique) by the anesthesiologist included in this study had only a small impact on the occurrence of these cardiovascular events. This is in contrast to critical respiratory events in the PACU where we found that selection of anesthetic drugs was an important etiologic factor. [6]Other clinically important characteristics beyond direct anesthetic care were noted in the study. The most important of these were patient factors (predicting PACU hypertension and bradycardia), surgical factors (predicting PACU hypotension), and OR observations (predicting PACU tachycardia); these factors showed more impact on the prediction of the event than did anesthetic factors. These results are similar to those of an earlier study, which showed that anesthesia did not contribute significantly to operative mortality when compared to patient or surgical factors. [12].
Other studies also have documented cardiovascular complications in the PACU, but the frequency depended on the definition used. The study by Hines et al. of 18,473 patients showed complication rates similar to those in our study. [2]A survey of 112,000 patients between 1975 and 1983 identified PACU hypotension in 1.7% and hypertension in 0.9% of patients, but these complications were poorly defined. [5]A more recent study by the same authors recorded PACU hypertension in 0.9-7.9% of patients varying across four hospitals. [3].
There have been few studies identifying risk factors for early cardiovascular events. Hines et al. found that patients with ASA physical status 2 or 3 had greater risk of both hypertension and hypotension in the PACU. [2]Forrest et al. noted severe perioperative cardiovascular problems in 3.5% of 17,201 patients, and associated risk factors were history of cardiac failure, cardiovascular surgery, increasing age, and unlike our study, the use of isoflurane. [13]Moller et al. recorded hypertension in 0.6%, hypotension in 0.9%, and hypovolemia in 0.2% of 28,802 patients during a trial of pulse oximetry; bradycardia was more likely to occur in those having pulse oximetry. [14].
The OR and PACU environments are ideal for the detection of early perioperative cardiovascular problems because patients are intensively monitored and standardized protocols and treatment regimens are used. Nonetheless, this study, which relied on the anesthesiologists and PACU nurses to recognize and record physiologic parameters outside a defined range, may not have recorded all occurrences. Although definitions of PACU events were directly available from PACU records, we cannot rule out variations in interpretation, and recording of events stopped at the time of PACU discharge.
We noted a strong relationship between cardiovascular events in the PACU and similar cardiovascular problems in the OR, which suggests that for some patients, the management of the original problem may have been inadequate or unsuccessful. We were unable to distinguish if specific drug therapy in the OR was instituted as a preventive measure or as a treatment, nor were we able to determine if therapy was successful. Therefore, our data did not permit examination of whether patients with cardiovascular problems in the OR who were successfully treated were less likely to have similar cardiovascular problems in PACU than patients who were not successfully treated. As well, data related to preoperative medication were restricted to sedatives and opioids, and we were not able to examine the relationship between cardiac drugs given immediately before surgery and cardiovascular events in PACU. Furthermore, it is unknown if other occurrences in the PACU (e.g., inadequate ventilation and nausea and vomiting) that contributed to cardiovascular events might have been avoided with a change in anesthetic management.
In summary, cardiovascular events in the PACU and after PACU discharge remain a postoperative concern. While the occurrence of these cardiovascular events in the PACU was infrequent, hypertension and tachycardia were both associated with risk for long-term outcomes such as unplanned admission to a critical care unit and hospital mortality. Nonetheless, specific choices in anesthetic management identified in this study hospital did not make a major contribution to these cardiovascular events in the PACU. Their etiology is primarily related to patient and surgical factors beyond the control of the anesthesiologist.