Triage of high-risk surgical patients for intensive care

Patients who undergo high-risk non-cardiac surgical procedures represent a large proportion of admissions to intensive care units (ICUs) in the developed world [1]. Ideally, surgeons, anesthesiologists, and intensivists admitting surgical patients to ICUs target the patients who will benefi t most from this highest level of postoperative care. However, accurately identifying which patients are at high risk of complications or death after major surgery remains diffi cult. For example, a recent study in the United Kingdom demonstrated that patients undergoing high-risk general surgical procedures comprised only 12.5% of surgical admissions to hospitals but over 80% of deaths, with less than 15% of these high-risk patients admitted to the ICU postoperatively [2]. Postoperative outcomes are a result of the complex interplay between the exact general surgical procedure performed, the previous health of the patient, and specifi c intra- and postoperative events. Outcomes may also be infl uenced by aspects of the particular healthcare system, such as the surgical procedure volume at diff erent hospitals [3], as well as care options, such as the availability and suitable use of intensive care beds. Appropriate triage of patients to intensive care postoperatively may have a large impact on outcomes after non-cardiac surgery. Th is chapter reviews the patient factors and scoring systems developed to help with triage, describes current ICU triage recommendations for postsurgical patients, and identifi es potential ways to improve evaluation and management of high-risk postoperative patients. Prediction of postoperative outcomes Predictors of postoperative outcomes may be divided into three categories: Known preoperative risk factors; the risk associated with the specifi c surgical procedure; and the unique aspects of each operative case that may contribute to a particular patient being at high risk for complications or death after surgery. Preoperative evaluation


Introduction
Patients who undergo high-risk non-cardiac surgical procedures represent a large proportion of admissions to intensive care units (ICUs) in the developed world [1]. Ideally, surgeons, anesthesiologists, and intensivists admitting surgical patients to ICUs target the patients who will benefi t most from this highest level of postoperative care. However, accurately identifying which patients are at high risk of complications or death after major surgery remains diffi cult. For example, a recent study in the United Kingdom demonstrated that patients undergoing high-risk general surgical procedures comprised only 12.5% of surgical admissions to hospitals but over 80% of deaths, with less than 15% of these high-risk patients admitted to the ICU postoperatively [2].
Postoperative outcomes are a result of the complex interplay between the exact general surgical procedure performed, the previous health of the patient, and specifi c intra-and postoperative events. Outcomes may also be infl uenced by aspects of the particular healthcare system, such as the surgical procedure volume at diff erent hospitals [3], as well as care options, such as the availability and suitable use of intensive care beds. Appropriate triage of patients to intensive care postoperatively may have a large impact on outcomes after non-cardiac surgery. Th is chapter reviews the patient factors and scoring systems developed to help with triage, describes current ICU triage recommendations for postsurgical patients, and identifi es potential ways to improve evaluation and management of high-risk postoperative patients.

Prediction of postoperative outcomes
Predictors of postoperative outcomes may be divided into three categories: Known preoperative risk factors; the risk associated with the specifi c surgical procedure; and the unique aspects of each operative case that may contribute to a particular patient being at high risk for complications or death after surgery.

Preoperative evaluation
Many preoperative risk factors can help distinguish which patients are most likely to experience poor postoperative outcomes. In particular, preoperative comorbidi ties are well-established as predictors of both morbidity and mortality after surgery and can be measured in diff erent ways (Table 1). Perhaps the best known is the American Society of Anesthesiologists' (ASA) physical status classifi cation system. Th is is a widely used preoperative scoring system that describes the overall health of the patient and burden of comorbidities. Th e score is ideal in being simple to apply and requiring no laboratory data; however, it is also subject to substantial interobserver variation in score assignment [4]. Despite this inherent subjectivity, the ASA classifi cation has been recognized as a helpful predictor of potential postoperative morbidity and mortality [5]. Another scoring system of preoperative comorbidities is the Charlson Comorbidity Index, which assigns weights to a variety of systemic diseases and predicts long-term survival [6]. While more detailed than the ASA classifi cation, this score is usually considered more useful for research purposes than for risk stratifi cation in real time. Th ere are also organ system-specifi c scores, most notably the Revised Cardiac Risk Index (RCRI), a scoring system that incorporates six factors to predict the risk of major cardiac events after non-cardiac surgery [7]. None of these scoring systems alone is generally suffi cient to provide adequate information regarding the risk for an individual patient. Th e RCRI predicts only cardiac risk, while the ASA and Charlson Comorbidity Index do not incorporate variables specifi c to the surgical procedure. Th ese preoperative scoring systems encompass a wide range of patient variables to produce a composite score. Individual patient factors have also been identifi ed in large studies as independently predicting increased risk of morbidity and mortality after surgery, including age and poor nutritional and functional status [8,9]. Functional capacity at baseline may indicate how a patient will respond to surgical stress in the perioperative period. Recent guidelines from the American College of Cardiology Foundation (ACCF) and the American Heart Association (AHA) for preoperative cardiovascular risk evaluation in non-cardiac surgery rely on a modifi ed version of the Duke Activity Status Index. Th is questionnaire estimates energy requirements for various patient activities. Depending on a patient's functional capacity, clinical risk factors, and type of surgery, the ACCF/AHA guidelines recommend either proceeding with surgery or further cardiac evaluation prior to the planned procedure. Clinical risk factors include active cardiac conditions and elements of the RCRI, and surgical proce dures are divided into those that are low, intermediate, or high risk [10]. While these guidelines only address cardiovascular risk, they illustrate the importance of including both the patient's preoperative status and the particular surgical procedure for evaluation of the risk of poor postoperative outcomes.

Intraoperative events
Surgical duration and urgency have both been shown to impact postoperative outcomes, with longer duration and emergent procedures associated with worse outcomes [2,9,11]. Preoperative patient characteristics were the most important predictors of postoperative mortality in a large study of the Veterans Aff airs database, but operative complexity became a signifi cant predictor with highly complex procedures [12]. Pearse et al. defi ned high-risk surgical procedures as those with at least a 5% risk of mortality and found that these procedures, in combination with advanced age, comorbidities, and emergency surgery, were highly predictive of increased risk of death postoperatively [2]. While the weight of the eff ects of patient-and surgery-specifi c factors on postoperative outcomes varies across diff erent studies, it is clear that it is ultimately the combination of the two that contributes to morbidity and mortality after surgery.
Surgical patients are unique as hospital patients in having a prolonged period of time in the operating room when they are closely monitored, providing detailed information on the physiologic perturbations associated with the anesthesia and surgery itself. Th e individual experience of a patient during surgery may have a large impact on outcomes. In particular, hemodynamic stresses to the body manifested as extremes of vital signs may infl uence the postoperative course. One study demonstrated that intraoperative tachycardia in long, complicated surgical procedures was independently associated with a composite measure of poor outcome [13]. In another case-control study, intraoperative tachycardia requiring treatment was associated with a signifi cantly increased risk of ICU admission, prolonged hospital stay, and hospital mortality [14]. A prospective cohort study in elderly patients showed that ASA class, emergency surgery, and intraoperative tachycardia were the most important predictors of adverse outcomes after surgery [11]. Th ese three studies indicate that intraoperative tachy cardia is associated with a wide spectrum of poor postoperative outcomes. Unanswered questions raised by these studies are whether tachycardia represents inadequate preoperative beta-blockade for high-risk patients or whether better control of intraoperative tachycardia would improve postoperative outcomes.
In addition to heart rate, blood pressure extremes may also predict outcomes after surgery. A prospective evaluation of over 1,000 patients revealed only three signifi cant independent predictors of one-year mortality after major general surgery: Th e Charlson Comorbidity Index, the cumulative deep hypnotic time, and intra operative hypotension [15]. Other researchers found that the risk of one-year mortality in elderly patients may be infl uenced not only by the presence or absence of hypotension but by its duration as well. Th e cut-off value associated with an increased risk of one-year mortality was only 5 minutes for mean blood pressure less than 50 mmHg but 30 minutes for mean blood pressure less than 60 mmHg [16]. It is important to note, however, that intraoperative hypertension has also been associated with poor outcome [13], suggesting that it is not just a single hemodynamic response but many diff erent types of physiologic stress that may aff ect postoperative morbidity and mortality in surgical patients.

Perioperative scoring systems
A number of scores aid in prediction of death specifi cally for patients admitted to the ICU (Table 1). While not developed solely for surgical patients, all of these scores account for postsurgical patients and provide risk prediction. Th e most commonly used scores are the Acute Physiology and Chronic Health Evaluation (APACHE) score, the Simplifi ed Acute Physiology Score (SAPS), and the Mortality Probability Model (MPM) [17,18]. Th e Sequential Organ Failure Assessment (SOFA) and the Multiple Organ Dysfunction Score (MODS) are two other ICU scoring systems used to describe organ dysfunc tion over the course of the ICU stay [19,20]. Th e main problem with all of these scoring systems for triage of patients is that they were developed and validated on patients already admitted to the ICU, and they do not necessarily provide adequate prediction for a broader range of patients. Moreover, none of the ICU scoring systems takes into account specifi c intraoperative data from the surgical procedure itself.
Other scoring systems were created specifi cally for surgical populations to aid in postoperative predictions of morbidity and mortality, regardless of admission to the ICU. Th ese scores incorporate patient and surgical factors to predict postoperative outcomes ( Table 1). One of the most widely used is the Portsmouth Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (P-POSSUM), comprised of twelve physiological variables and six operative variables. Th e Estimation of Physiologic Ability and Surgical Stress (E-PASS) is another system calculated from six pre operative variables and three intraoperative variables [17]. In the United States, the National Surgical Quality Improvement Program (NSQIP) was developed to allow comparison of risk-adjusted surgical outcome data between hospitals. Th e risk prediction requires collection of 97 peri operative variables [21]. Finally, the Surgical Apgar Score (SAS) was derived to provide a simple postoperative assessment of patients at the end of a surgical procedure. Th e score is calculated from three intraoperative variables and predicts those patients at highest risk of postoperative complications and mortality [22]. All of these surgery-specifi c scoring systems can be applied to patients postoperatively irrespective of patient location. Indeed, none use ICU admission as a variable or as an outcome measure.
A major concern is that most of these scoring systems were created and evaluated for the purpose of audit, benchmarking, and assessment of quality of care. As such, they are valid for predicting outcomes in groups of patients, but they may not be useful or accurate for individual patients [18]. In addition, some of the scores, such as SOFA, MODS, and SAS, are much easier to implement in real time compared with others (such as NSQIP) to allow for an understanding of a patient's current status and to potentially assist in individual triage decisions.

Postoperative complications
Despite increasing recognition of the importance of postoperative complications as outcomes, previous research and scoring systems primarily focused on short-term (hospital or 30-day) mortality as the main postoperative outcome measure [18]. However, complications potentially result in greater morbidity for patients, increased length of hospital stay, and higher hospital costs. Th e focus in U.S. healthcare has recently shifted towards an increased emphasis on measuring complications because of a change in hospital reimbursements in 2008 by the Centers for Medicaid and Medicare Services (CMS). Th e CMS implemented a pay-for-performance initiative designed to improve the quality of health care delivered to patients. Hospitals no longer receive additional reimbursement when Medicare patients experience certain 'preventable' complications, some of which are clearly post operative concerns, such as mediastinitis after cardiac surgery or an object retained inside the patient after surgery [23]. While increasing attention to these out comes, it is unclear whether these economic incentives will result in a reduced incidence of postoperative complications.
Data also suggest a link between the development of postoperative complications and long-term outcomes. In recognition of the potential importance of complications, NSQIP provides prospective data collection on postoperative morbidity as well as mortality. Using the NSQIP dataset merged with a Veterans Benefi ts Administration database, a multicenter study of over 100,000 Veterans Aff airs hospital patients undergoing eight diff erent surgical procedures demonstrated that the presence of any complication within the fi rst 30 postoperative days was an important predictor of 5-year survival, independent of preoperative risk [24]. Recognition and avoidance of postoperative complications may thus reduce shortterm morbidity and costs of care, and also impact longterm outcomes.

"Failure to rescue"
In two large studies of U.S. hospitals, Ghaferi et al. found that patients undergoing general and vascular surgical procedures had similar postoperative complication rates across hospitals but a very wide range of postoperative 30-day mortality rates [25,26]. Th e rates of "failure to rescue" (proportion of deaths in patients who developed a postoperative complication out of the total number of patients who developed a postoperative complication) were much higher in high-mortality hospitals. Th e authors suggest that rather than diff erences in complication rates, it is diff erences in rates of "failure to rescue" that better explain hospital variation in post-surgical mortality; those hospitals with higher mortality may not properly recognize and manage postoperative complications when they occur [25,26]. Th is concept of "failure to rescue" shifts some of the focus from preventing the development of complications to improving the care provided to manage complications.

Improved outcomes with intensive care
Routine postoperative care in an ICU after high-risk surgical procedures may allow for greater recognition and correct management of postoperative complications, thereby reducing long-term morbidity and mortality. While there is an assumption that ICU admission may improve postoperative outcomes, no randomized trials have addressed this issue. Given the limitations and variability of ICU bed availability worldwide [27], many patients referred for ICU admission must ultimately be refused. Th e potential benefi t of intensive care in reducing mortality for some groups of high-risk surgical patients is, therefore, suggested by multiple observational studies primarily examining patients refused admission to an ICU. In one prospective study that took place in seven countries, ICU admission of mixed medical and surgical patients was associated with a substantial reduction in 28-and 90-day mortality compared with patients refused admission. However, surgical patients were signifi cantly less likely to be refused admission to an ICU than medical patients [28]. Th is discrepancy may be explained by the fact that many patients with severe systemic illness who undergo surgical procedures for acute problems are still deemed to have a reasonable chance of recovery. Moreover, curative surgery often repre sents an investment of resources in a patient that may carry over to priority admission to an ICU.
Until recently, few studies emphasized the potentially diff erent nature and practice of triage decisions for postoperative patients compared with medical patients. One recent study of patients in two British hospitals that did focus solely on postoperative admission practices for high-risk surgical patients found that only one-third of these patients received intensive care. Furthermore, patients admitted to the ICU immediately after surgery had greatly improved survival compared with patients who were re-admitted or had delayed admission to the ICU postoperatively [29]. Th is study showed the potential drawbacks of ICU underuse, but it did not include details of the triage decision-making process for postoperative ICU admission. Objective, evidence-based criteria for ICU admission after surgery may facilitate resource allocation and potentially help identify patients who would benefi t most from admission to the ICU, with the ultimate goal of improving postoperative outcomes.

ICU admission guidelines
In 1994, the European Society of Intensive Care Medicine (ESICM) issued guidelines stating that patients should be admitted to the ICU if they have an unstable condition or if they are at high risk of developing a severe complication [30]. Th e American College of Critical Care Medicine then published broad guidelines for ICU admission in 1999 [31]. A prioritization model was described to aid with decision-making, ranging from Priority 1 (patients who will derive the most benefi t from admission) to Priority 4 (those who will not benefi t at all). Priority 1 patients include hemodynamically unstable or ventilatordependent postoperative patients; Priority 2 patients are those who may need immediate intervention, such as patients with chronic comorbidities with an acute surgical problem [31]. Th e Italian intensivist group SIAARTI (Società Italiana di Anestesia Analgesia Rianimazione e Terapia Intensiva) published similar guidelines with specifi c recommendations for prioritiza tion. A patient currently requiring intensive care should have priority for ICU admission over a patient who only needs intensive monitoring. Th is group also outlined a scale from Priority 1 (maximum benefi t expected) to Priority 4 (little or no benefi t expected) for determining which patients should take precedence for ICU admission [32]. Israeli consensus guidelines recommend ICU admis sion for all postoperative vascular or major general surgery patients with severe underlying systemic disease [33].
None of these sets of criteria, however, off ers guidance as to how to triage patients of similar acuity. One method to allocate scarce ICU resources might be on a fi rst-come, fi rst-served basis, as suggested by the American Th oracic Society [34]. Current guidelines and studies also do not specifi cally target surgical populations. Th ere are no universal criteria for postoperative admission of patients to the ICU. In a recent editorial, Goldhill and Down observed that postoperative critical care is highly inconsistent, with some groups of patients routinely admitted to the ICU postoperatively and others rarely admitted [35]. In some hospitals, patients are admitted postoperatively simply for monitoring, but it is unclear whether using the ICU for this purpose actually improves postoperative outcomes. While one early study found that a shortage of medical ICU beds led to restriction of admissions solely for monitoring purposes with no eff ect on mortality [36], no such study has been performed in surgical patients.

Alternatives to ICU admission
Mandatory ICU admission could potentially lead to overuse of intensive care for postsurgical patients, and a shift has occurred away from this practice. As an alternative to intensive care, the high-dependency unit (HDU) provides a location for patients with a potential but low risk of major complications who require more care than the ward but less than that available in the ICU [37]. Over a 14-year period from 1991 to 2004, postoperative care of major vascular surgery patients in one center changed from mandatory ICU admission for all patients to care for more than two-thirds of postoperative patients in an HDU and ward [38]. Another group retrospectively examined patients who had undergone radical cystectomy with urinary diversion followed by mandatory admission to the ICU postoperatively. Th ey identifi ed a stratifi cation system for triage of these patients to the ICU, HDU, or ward depending on the risk of requiring active treatment such as invasive monitoring, vasoactive medications, or mechanical ventilation. However, there was no prospective evaluation of this postoperative risk stratifi cation system [39]. Finally, an eight-year observational study compared ICU admissions for aortic abdominal surgery and surgery for lung cancer, two high-risk surgical procedures, before and after expansion of the hospital's post-anesthesia care unit (PACU). Admission rates to the ICU dropped significantly for both types of procedures, with no concomitant increase in postoperative morbidity or mortality [40]. Th ese studies illustrate the possibility of caring for highrisk surgical patients in locations other than the ICU and highlight the potential for development of procedurespecifi c criteria for postoperative ICU admission.

Post-anesthesia care units
Individual hospital resources are extremely variable and not all centers have access to HDU facilities or profi cient ward services. Th e PACU may often be used as an 'overfl ow' location for postoperative surgical patients who are not stable enough for ward care. Several issues may arise when a PACU is used to care for more severely ill patients. Physician coverage of these patients may be unclear, as anesthesia personnel, the primary surgical team, and/or an intensivist may all be involved in caring for such patients. In addition, PACU nurses must be fully competent and trained to care for ICU-level patients. Staffi ng patterns may need to be augmented, with increases in services from diff erent provider types, such as respiratory therapists. Responsibility for recordkeeping, daily progress notes, ordering systems, and fl owsheets must all be clarifi ed [41]. While the PACU can serve as a safety net for ICU patients during times of limited bed availability, patients who would most benefi t from postoperative ICU care would ideally gain admission to the ICU immediately after surgery, as outcomes may be improved with immediate rather than delayed ICU admission [29]. Th erefore, for high-risk patients, bypassing the PACU for immediate transfer to intensive care may ensure the most effi cient use of resources with the best outcomes.

Improvement of postoperative outcomes
A number of interventions have the potential to improve postoperative outcomes, especially in high-risk patients.

Patient interventions
Several studies show that hemodynamic optimization of high-risk surgical patients improves outcomes after surgery. Th e goal of hemodynamic optimization is to most effi ciently match oxygen delivery to oxygen consumption. Tissue hypoxia due to surgical stress may not be adequately monitored by heart rate, blood pressure, or central venous pressure. Various invasive and noninvasive devices provide alternate means of assessing and optimizing oxygen delivery, including mixed or central venous oxygen saturation, cardiac index, pulse contour analysis, or other parameters indicating preload. Many of the studies using goal-directed hemodynamic therapy show decreased rates of postoperative complications, mortality, and hospital length-of-stay [42]. Individualized goal-directed therapy for high-risk surgical patients may improve outcome, but large, multicenter trials are necessary before widespread adoption can occur. British consensus guidelines, however, do recommend preoperative therapy with intravenous fl uids and inotropes to attain predetermined cardiac output and oxygen delivery targets in high-risk surgical patients as a way to reduce postoperative mortality [43].

Hospital interventions
A simple way of improving surgical outcomes may include hospital-mandated use of a preoperative surgical safety checklist. A prospective study in eight countries looked at the rate of postoperative complications before and after implementing a checklist in all operating rooms. Th e checklist encompassed 19 items to ensure patient safety before induction of anesthesia, before surgical incision, and before exit from the operating room. All sites had reductions in major complication and mortality rates after implementation of the checklist [44].
Another method to reduce morbidity and mortality may be to improve ICU triage after surgery by creating and adhering to hospital guidelines that most effi ciently use intensive care resources. Th ese guidelines could include patient-and surgery-specifi c indications for admission to the ICU, HDU, or ward, and they could also delineate an ICU prioritization system for patients of similar acuity. `Fast-track' surgery is a recent concept encompassing multimodal pre-, intra-, and postoperative interventions to improve outcomes after surgery. Each of these approaches must be specifi c to the surgical procedure and focus on enhancing postoperative recovery. Data are accumulating in a variety of surgical subspecialties that support the feasibility of `fast-track' surgery [45]. Although practices vary, `fast-track' surgery may mean more rapid progression through intensive care or admission to a specialized PACU.
Use of dedicated intensivists to staff ICUs may also help improve outcomes for high-risk surgical patients. In particular, intensivists may be the most appropriate practi tioners to assist in ICU triage and bed allocation. Furthermore, dedicated intensivists may provide better care to ICU patients than non-specialized physicians. A survey of Maryland ICUs that provide postoperative care for patients undergoing abdominal aortic surgery showed that after adjusting for patient and hospital characteristics, the absence of daily rounds by an intensivist was associated with signifi cantly increased morbidity and mortality after this high-risk surgery [46]. Similarly, rapid response teams with critical care-trained personnel may evaluate patients with complications on the ward to better identify those who might benefi t from transfer to an ICU. Such teams may improve "failure-to-rescue" rates. Based on this concept, many hospitals have created rapid response teams [47], although their utility remains unproven in large multicenter studies.

Health care system interventions
On a health care system level, high volume hospitals and regionalization may also decrease the risk of post operative morbidity and mortality. In an analysis of Medicare data, hospitals that performed a greater number of specifi c high-risk surgical procedures annually had lower hospital mortality rates [3]. A meta-analysis of the eff ects of surgical volume and specialization recently showed that high individual surgeon volume and surgeon specialization were independently associated with im proved postoperative outcomes, but the relationship between high hospital volume overall and outcome remained unclear [48]. Regionalization to high-volume centers or to individual specially trained surgeons for certain high-risk procedures may be another way to improve outcomes after surgery, as is now the model for trauma care [49].

Conclusion
High-risk surgical patients continue to make up a substantial proportion of ICU admissions in most developed countries. Identifi cation and optimization of these patients prior to surgical interventions remains diffi cult. Pre-, intra-, and postoperative variables all play a role in the development of considerable morbidity and mortality for high-risk patients. Appropriate decisions relating to the need for intensive care after surgery are key for highquality patient care, yet adequate systems for triage remain elusive. Th e potential exists for underuse of critical care resources, with inappropriate lack of admission to an ICU for high-risk patients, as well as overuse, with unnecessary admissions leading to increased length of stay and costs. With the wealth of data now available regarding the perioperative status of many of these patients, further research is needed to help facilitate optimal care for postsurgical patients.