Postoperative Mortality

Preventable intraoperative mortality decreased 100-fold during the initial century after anesthesia was discovered in 1846. Since my residency, ending in 1985, it has decreased another order of magnitude. Intraoperative mortality is now so low that it is hard to quantify. This is a remarkable accomplishment; no other specialty has remotely reduced mortality by a factor of 1,000. Previous generations of anesthesiologists can justifiably be proud of solving intraoperative mortality.The trouble is that postoperative mortality remains high – so high that the 30 days after surgery is the world's third-leading cause of death, just behind heart disease and cancer.1 About 500 patients per hour die within 30 days after surgery. Furthermore, postoperative mortality is not obviously improving, although, in fairness, we now perform procedures on much sicker patients. Some postoperative mortality is predictable and unpreventable (after major trauma, for example). But surgeons do not normally operate on patients they expect to die within a month. Cardiovascular complications, rather than surgical ones, are the leading cause of postoperative death – and many are presumably preventable and treatable. Postoperative mortality is the major perioperative problem. Half of 30-day postoperative mortality occurs during the initial hospitalization; that is, under our care in our highest-level health care facilities.2 It is likely that much in-hospital and post-discharge mortality is preventable. It is now up to us to solve postoperative mortality. Doing so will give the next generation of anesthesiologists the glorious legacy of lives saved that was deservedly enjoyed by prior generations, who reduced operative mortality 1,000-fold. The ball is in our court. Postoperative mortality incidence The most recent broad evaluation of intraoperative mortality was based on data from the entire Hubei Chinese province and incorporated a full range of facilities – from outstanding tertiary care to less-skilled rural hospitals.3 Nonetheless, the incidence of preventable intraoperative anesthetic mortality among more than 9 million surgical cases was just 6.4 per 100,000. Presumably, it is even less in western countries, where health care is more homogeneous. In marked contrast, the most recent estimate for in-hospital mortality after inpatient surgery, based on 28 million German cases, is 1.4%.4 Thirty-day mortality is ≈2% in Western hospitals5,6 and substantially higher in lower-income countries.7 Reported values depend on where studies are conducted and who is included in the denominator. For example, the fraction of inpatient versus outpatient surgery varies enormously from country-to-country, making direct comparisons challenging. It is nonetheless clear that the 30 days after surgery is a high-risk period and contributes substantially to mortality from all causes. Mortality is not randomly distributed; instead, it is most likely to occur in patients with serious underlying conditions.8 Causes of postoperative mortality Some deaths have obvious causes, such as cancer, major stroke, or a pulmonary embolus. But in many patients, the cause of death is notoriously hard to attribute, even for the responsible clinicians. Most patients die with many overlapping problems, often presenting as multiorgan failure. Concomitant complications make it hard to identify specific causes of death, much less what triggered the ultimately fatal sequence. Attribution to triggering causes is even less certain when dealing with the large observational registries that are needed for generalizable results. Another problem with registry data is that potentially lethal complications are themselves poorly characterized. For example, 90% of myocardial injury is asymptomatic9 and therefore mostly missed without scheduled screening, which remains uncommon. Registry searches for myocardial injury thus typically identify about a fifth of the cases that presumably occurred, leading to false mortality attributions. Similarly, the apparent incidence of acute kidney injury is much higher if the diagnosis is based on stage 1 injury (50% preoperative-to-postoperative increases in creatinine) rather than renal injury sufficiently severe to generate a billing code. Yet, both are used in various analyses, generating wildly different mortality attributions. A consequence is that we are left to evaluate associations between proximal events such as sepsis or acute kidney injury – themselves often poorly characterized – and mortality, with the causal links remaining unclear. For example, there are complications associated with death that are almost surely not causal, such as in delirium. Causality matters, though, because preventing a putative inciting event such as myocardial injury will reduce mortality only if the relationship is causal. Pending the very large trials necessary to convincingly demonstrate that specific interventions reduce mortality, it nonetheless seems worth considering interventions that reduce major postoperative complications, such as acute kidney injury and infections, which are themselves serious. Postoperative complications and mortality are highly predictable. For example, about 20% of serious postoperative complications occur in the 5% of patients with the highest preoperative risk; similarly, about half of all serious complications occur in the 20% of patients designated preoperatively to be at highest risk.10,11 Surprisingly, the type of surgery contributes much less than might be expected. Instead, risk of various complications and mortality is largely determined by patients' underlying risk12 – which is largely unmodifiable. Thirty-day postoperative mortality is most strongly associated with myocardial injury and bleeding. Sepsis is a distant third.9 (Sepsis is more deadly than myocardial injury, but much less common.) The contribution of subtle renal injury remains unclear but may be substantial. Respiratory deaths are even less common, but of particular interest since nearly all are presumably preventable. Myocardial injury Myocardial injury after noncardiac surgery (MINS) is thought to result from a supply-demand mismatch and is thus a type 2 infarction. MINS is of particular interest because it is common, occurring in about 10% of surgical inpatients ≥45 years old.13 The syndrome is defined by postoperative troponin elevation thought to be due to cardiac ischemia, with or without symptoms or signs.14 For reference, troponin elevation with signs (e.g., EKG consistent with myocardial ischemia) or symptoms (e.g., chest pain or shortness of breath) defines myocardial infarction.15 More than 90% of MINS is asymptomatic, and thus undetected without routine troponin monitoring.9 Consequently, four major groups, including the American Heart Association, recommend troponin screening preoperatively and for two to three postoperative days while patients remain hospitalized in moderate-to-high-risk surgical inpatients.16,17 It is natural to assume that asymptomatic MINS is much less serious than myocardial infarctions, which require troponin elevation and an abnormal EKG or symptoms. However, that assumption is wrong: 30-day mortality is only slightly less for asymptomatic MINS patients than for patients with full myocardial infarctions.9 In fact, the mortality for an asymptomatic postoperative patient with troponin elevation is similar to that of a patient presenting in the emergency department with shortness of breath and crushing chest pain. Even slight troponin elevations triple 30-day postoperative mortality, and mortality increases to 30% at the highest troponin concentrations.9 However, the major risk to MINS patients is not during the initial postoperative month. The more serious risk is reinfarction that occurs over subsequent years. A remarkable one in seven MINS patients will have a reinfarction within just 16 months,18 and risk presumably continues in subsequent years. Methods to prevent nonoperative reinfarctions are well established. In appropriate patients, they include aspirin, statins, angiotensin-converting enzyme inhibitors and receptor blockers, blood pressure and heart rate control, and counseling on healthful eating, exercise, and smoking cessation. Furthermore, anticoagulation reduces risk of reinfarction in MINS patients by 28%.18 Transitioning MINS patients to effective long-term care is thus likely to save many lives. There is currently no known safe prophylaxis for postoperative myocardial infarction. Perioperative beta blockers reduce risk, but at the expense of hypotension and strokes, which increase overall mortality.19 Prophylactic aspirin increases serious and life-threatening bleeding risk without reducing the risk of myocardial infarction.20 Clonidine similarly increases bradycardia and hypotension, but without reducing infarction risk.21 Avoiding nitrous oxide22 and administration of tranexamic acid23 do not alter infarction risk. Intraoperative and postoperative hypotension are associated with myocardial injury, myocardial infarction, acute kidney injury, and death.24 These associations are unsurprising since at some level hypotension surely causes organ injury. Observational data suggest that the population harm threshold is about 60-65 mmHg and is similar in patients with baseline hypertension. On the other hand, two recent trials in which patients were randomized to tight versus routine blood pressure targets reported no difference in composite complications.25,26 (Two other trials, each of which randomized more than 1,000 patients, also report that tight blood pressure management does not reduce serious complications.) However, none of the trials included substantive exposures to mean arterial pressures <65 mmHg. Trial results are thus consistent with observational data but leave us uncertain about what the actual harm threshold is and the extent to which it might differ among individuals. However, we can reasonably conclude that intraoperative mean arterial pressures exceeding 65 mmHg are safe in most patients. Predicting cardiovascular complications Remarkably, the type and duration of surgery contribute relatively little to postoperative mortality risk. Instead, cardiovascular complications reflect underlying cardiac disease. Consequently, the risk of cardiovascular complications and death can be predicted reasonably well from baseline patient characteristics. For example, the Revised Cardiac Risk Index is a widely used prediction system based on just six patient characteristics. While simple to use and requiring only readily available information, the calculator assigns most patients to “intermediate risk,” which provides only marginal clinical guidance. Furthermore, the Revised Cardiac Risk Index substantially underestimates cardiovascular risk because it was developed before MINS was identified as the most common perioperative cardiovascular complication.27 Stress echocardiography rarely provides actionable clinical guidance and appears to be vastly overused.28,29 Remarkably, computerized coronary artery tomography – an expensive high-radiation test – worsens Revised Cardiac Risk Index assessments by incorrectly reclassifying many patients deemed at intermediate risk as high risk.30 The best preoperative risk-assessment test appears to be NT-ProBNP, which correctly reclassifies about a quarter of the patients deemed intermediate risk by the Revised Cardiac Risk Index to both higher and lower risk designations, thus much improving predictive accuracy.31 The test is inexpensive, available on a point-of-care basis, and probably should be used far more often than it currently is. Continuous ward monitoring Vital signs on surgical wards are monitored much as they were half a century ago, typically at four- to six-hour intervals. However, surgical inpatients are far older and sicker than when the system was established, and we now routinely do much larger and more invasive procedures. Ward vital sign abnormalities are consequently common. Intermittent ward vital sign monitoring, even at four-hour intervals, misses many instances of hypoxemia and hypotension.32-34 Even prolonged abnormal episodes are often missed by meticulous nurses, because many episodes occur between observation periods.35 There are now alternatives. Untethered, wearable, continuous vital sign monitors are now available, and many companies are developing new ones. Soon there will be many options that optimize various combinations of comfort, battery life, wireless communication, and tracked variables. Some already evaluate patient position, mobility, sleep quality, and location within a hospital – all of which might be considered new vital signs. Continuous monitoring prompts nursing interventions that reduce vital sign abnormalities, especially desaturation.36 However, we do not yet have evidence that continuous postoperative ward monitoring reduces serious complications such as naloxone administration, rapid response team activation, and unplanned ICU admission. Very large trials will be necessary to determine whether continuous ward monitoring reduces hard outcomes. A challenge for such trials is that monitoring alone cannot improve outcomes. Monitoring needs to be combined with signal recognition and skilled intervention – an implementation challenge. It is nonetheless likely that continuous ward monitoring, at the very least, detects respiratory events before they become critical, allowing timely intervention. It therefore seems likely that continuous ward monitoring will soon be routine, because it will be hard to avoid implementing technology that detects potentially serious instability far better than the current system. The obvious benefit of continuous monitoring is that clinicians will have access to everyone's vital signs in real time. The disadvantage is that nurses will be overwhelmed if we simply feed artifact-laden, unfiltered, continuous vital signs to them – especially since a single nurse might easily care for more than 10 patients. Nurses are not trained to deal with the quantity of complex data that continuous ward monitoring will generate; and of course, they already have full-time jobs. To be useful, continuous data needs to be cleaned and interpreted in context. Machine learning systems seem ideal for this task, but they have yet to be developed and validated.37 Better ward care Most surgical inpatients have complex underlying conditions that are often poorly managed. These conditions are aggravated by the stress and inflammation consequent to surgery, which sometimes kill them. Many patients on surgical wards are so sick that they would have been sent to critical units just a couple decades ago. Instead, they get care on surgical wards with sparse, intermittent vital sign monitoring. An additional concern is that surgeons visit postoperative patients daily, usually briefly and early before starting the day's operations. Postoperative patients are also sometimes visited by anesthesiologists who tend to focus on nausea and vomiting, pain, and related anesthetic issues. But few postoperative patients get the sort of intense medical care necessary to prevent underlying conditions, aggravated by surgery, from spiraling out of control. Anesthesiologists are well positioned to provide such care, since we have a good understanding of both surgical and medical issues. But we're not the only ones. Conclusions Intraoperative mortality was solved two decades ago. In contrast, postoperative mortality remains among the world's leading causes of death – and is clearly the major perioperative problem. For two decades, our societies have asserted that anesthesiologists are perioperative physicians. But brief postoperative visits and the occasional pain consult do not constitute meaningful postoperative care. With some notable exceptions, few anesthesiologists truly function as perioperative physicians. Some specialty will solve postoperative mortality. It's a problem of interest to all physicians who care for surgical patients, including surgeons, internists, hospitalists, and intensivists. For example, there are already fellowships in perioperative care – in internal medicine departments! But, arguably, anesthesiologists are best positioned to understand and prevent postoperative mortality. Uniquely, we have good understanding of both surgical and medical issues, and how they interact. We are also experts in handling the sorts of complex and dense streams of artifact-laden data that continuous ward monitoring generates. Nonetheless, too many of us define our responsibility as largely ending when patients leave the postanesthesia care unit. That is exactly the same as declaring anesthesia to be irrelevant to the major perioperative problem. We should do the opposite and instead grasp this opportunity to provide intense postoperative management and save lives. Specifically, I propose establishing a fourth branch of anesthesia, one that will supplement OR anesthesia, critical care, and pain medicine.38 Our opportunity is short-lived, because other specialties are eying this clinical opportunity; once established, they will be hard or impossible to dislodge. Letting others solve postoperative mortality and establish themselves in this new clinical space is not a recipe for our specialty's continued influence – or even our long-term survival. Carpe diem. Disclosure: Dr. Sessler is a cofounder, consultant, and shareholder for the Health Data Analytics Institute, Dedham, MA, and holds stock in Perceptive Medical.Daniel I. Sessler, MD, Professor and Vice President for Clinical and Outcomes Research, Center for Outcomes Research, and Department of Anesthesiology, UTHealth, Houston, Texas.