Improving Perioperative Outcomes: My Journey into Risk, Patient Preferences, Guidelines, and Performance Measures

AS a clinical anesthesiologist and health services researcher, my goal is to improve patient outcomes both on an individual and a population level through direct patient care and research. In recent years, I have had the privilege of being involved in the development of numerous guidelines.1–7Guidelines are a set of recommendations for patient management that identifies a specific or range of management strategies based on the evidences. Guidelines are usually developed and promulgated by a specialty medical society or government agency (i.e. , National Institutes of Health or Agency for Healthcare Research and Quality). It is important to recognize that guidelines are the culmination of both research and expert opinion and that application of guidelines can improve care, but if done incorrectly, it can lead to worse outcomes. It is within this context that I would like to review my perspective on the research that underlies both guidelines and their implementation into clinical practice.The goal of improving perioperative outcomes comes under the general rubric of outcomes research. The agency for healthcare research and quality defines outcomes research as:When focusing on improving outcomes in anesthesiology, it is important to define those outcomes of interest including death; major organ dysfunctions, such as myocardial infarction, pneumonia, and renal failure; and minor morbidity, including pain, nausea, and vomiting. There is also increasing interest in patient-oriented outcomes such as quality of recovery, quality of life, and patient satisfaction. The time horizon of interest to the anesthesiologist has traditionally been 24 to 48 h, although several investigators have suggested that perioperative actions by the anesthesiologist may impact outcomes at 1 yr or longer.8In thinking about the influence of anesthesia on perioperative risk, it is important to recognize that there is a risk that is directly attributable to anesthesia (e.g. , loss of airway), whereas there are outcomes related to the stress of undergoing surgery, which can be mitigated by the care provided by an anesthesiologist. With respect to the former, anesthesiologists have made major advances in making anesthesia safer. It is estimated that the risk of death directly attributable to anesthesia may be in the order of 1 death in 185,000 anesthetics, although one investigator has questioned this conclusion.9,10However, perioperative organ dysfunction is common after surgery, and it is associated with an increased cost of care.11As perioperative physicians, our goal should be to mitigate perioperative and long-term organ dysfunction by application of the best evidence.One approach to achieve this goal is to begin by assessing or determining those patients at greatest risk for major and minor morbidity and mortality. Once the patients at risk are identified, it is important to study their actions or interventions that might influence risk and reduce complications. If an intervention is found to be either efficacious or effective (the difference is discussed later) then it is important to disseminate this information. If the level of evidence is sufficiently strong or there is a preponderance of expert opinion on a topic, then a guideline or practice parameter can be developed to better disseminate the information to the individual practitioner. However, applying the evidence to the individual patient must be done within the framework of understanding the benefits and risks associated with the intervention and how the individual patient views the potential trade-offs from differing interventions and outcomes. Finally, it is important to recognize that even if evidence is properly disseminated, the adoption of new and important evidence may be slow. One of the paradigms that the current government has adopted to increase the speed of diffusion and adoption is the use of performance measures. However, the value of measuring and paying for “correct” performance that leads to improved outcomes is a matter of debate. This article attempts to review the quest of anesthesiologists for improving perioperative and periprocedure outcomes within this proposed framework.An underlying assumption in medical care is that treatment should be individualized for a given patient if optimal outcomes are desired. In the perioperative period, anesthesiologists have multiple options with regard to the choice of anesthesia, monitors, and drugs used to achieve the desired outcome. A large portion of our patients will do well if they receive a standard anesthetic (e.g. , propofol, a muscle relaxant, an inhalational agent, and narcotic). There is a basic assumption that obtaining a preoperative history, physical examination, and appropriate laboratory testings will influence management choices, which will then lead to better outcomes. The concept of establishing a baseline database to assess risk of disease and risk of developing complications can be framed within a decision paradigm (fig. 1). By using our understanding of clinical findings, previous experience, and the application of the literature, we can assess the probability of disease within certain confidence intervals (CI). Within this paradigm, the clinician may decide that the probability of disease reaches some threshold for action. For patients with cardiovascular disease who are undergoing noncardiac surgery, these actions can include modification of medical management including initiation or continuation of β-blockers and statins, treatment for unstable coronary symptoms, and coronary revascularization among other interventions.4Therefore, risk assessment is a critical first step from both the clinical and research perspectives. One of the first and best known studies of assessing cardiovascular risk is the Cardiac Risk Index published by Goldman et al .12During the ensuing three decades, others have further refined the Index, and most recently, Lee et al .13published the revised Cardiac Risk Index. They identified six risk factors with an increasing probability of developing cardiac events with an increasing number of factors present. Important to any such development is the validation of these indices, and many such studies have been published recently.Another major area of interest of my group is admission after outpatient surgery. Given the low rates of admission, administrative date (e.g. , state discharge summaries and Medicare claims data) has been used to determine predictors of admission or death.14,15An outpatient surgery admission index was created that establishes the risk factors, and based on the number of risk factors, the probability of admission after outpatient surgery was defined.15Based on this probability, I believe that the patient and the physician can determine the optimal location of care (free standing vs. hospital-based outpatient surgery) to yield the best outcome.Importantly, the clinical evaluation and integration of knowledge does not always lead to a clear decision for action. In some cases, the probability of disease is so high that the optimal action is well defined because the probability is far above any threshold for action. Alternatively, if patients have only minor risk factors for coronary disease then they clearly do not require any further evaluation because no action would be taken. Similarly, patients with known and severe coronary artery disease who are already in their optimal medical condition also rarely need further evaluation because they are above the threshold for action but the action is likely already taken. A key question is the optimal action when the probability and extent of disease is less precise, and the threshold for any action lies within the CIs of the risk assessment. Testing can potentially have value in such situations because a negative test would decrease the probability and obviate any need for action, whereas a positive test would increase the probability above threshold and therefore lead to the action.Within this framework, cardiovascular testing has undergone an evolution, during the past two decades, from initially being used routinely in all patients to being used selectively based on an assessment of risk.16In addition, I and others have helped to refine interpretation of the test. For example, a larger effect or area of regional wall motion abnormality denotes greater risk in dipyridamole thallium imaging and dobutamine stress echocardiography when compared with a simple positive or negative scan.17,18This approach formed the basis of guidelines for perioperative cardiovascular evaluation.Randomized clinical trials form the strongest basis for determining the evidence supporting an action, and during the past several decades, there has been a marked increase in their number and quality in the perioperative period. Randomized clinical trials are designed to prove efficacy in determining whether an intervention works under ideal conditions. They have defined patient inclusion/exclusion criteria and usually have strict protocols of care. Although the internal validity of these trials is high, the external validity of the trials may be low. However, effectiveness refers to how an intervention works under real world conditions. In such situations, the intervention may behave identical or different from the randomized clinical trial.In the patient with cardiovascular disease, examples of randomized clinical trials include the study of β-blockers, statins, coronary revascularization, and thermal management. The recent series of studies of perioperative β-blockade in noncardiac surgery help illustrate both the importance of how the baseline risks of the patient and how study protocol can influence effectiveness.19For example, β-blockade started several weeks in advance of surgery has been shown to be efficacious in vascular surgery patients at high risk.20However, subsequent studies questioned the finding in patients at lower risk and in real world usage (i.e. , effectiveness).To study efficacy in lower risk populations, a larger sample size may be required. In DECREASE IV, β-blockers were studied in a randomized trial of 1,066 patients at intermediate risk.21Patients randomized to bisoprolol had a lower incidence of perioperative cardiac death and nonfatal myocardial infarction than those randomized to bisoprolol-control (2.1% vs. 6.0% events; hazard ratios: 0.34; 95% CI 0.17-0.67). Importantly, bisoprolol was started at least 7 days in advance of surgery. To study the effectiveness in low-risk patients, administrative datasets (e.g. , Medicare claims data) may prove useful. For example, Lindenauer et al. 22retrospectively reviewed the records of 782,969 patients and determined who received β-blocker treatment during the first 2 days of hospital stay. The relationship between perioperative β-blocker treatment and the risk of death varied directly with cardiac risk. Among the 580,665 patients with a revised Cardiac Risk Index score of 0 or 1, treatment was associated with no benefit and possible harm, whereas among the patients with a revised Cardiac Risk Index score of 2, 3, and 4 or more, the adjusted odds ratios for death in the hospital were 0.88 (95% CI 0.80-0.98), 0.71 (95% CI 0.63-0.80), and 0.58 (95% CI 0.50-0.67), respectively (fig. 2).The Perioperative Ischemic Evaluation Study (POISE) trial demonstrated the importance of study protocol with respect to the effectiveness of a drug.23A total of 8,351 patients were randomized to controlled-release oral metoprolol succinate or placebo. The primary endpoint of cardiac death, nonfatal myocardial infarction, or cardiac arrest was reduced in the metoprolol group compared with placebo (5.8% vs. 6.9%; hazard ratio 0.84; 95% CI 0.70-0.99; P = 0.04), driven by a reduction of nonfatal myocardial infarctions, however, at the costs of an increased incidence of total mortality and stroke. One of the major conjectures with regard to the difference between the results in the Perioperative Ischemic Evaluation Study and other trials is the differences in the protocol. Metoprolol succinate, a long-acting β-blocker, was used at much higher doses and was started on the morning of the surgery in contrast to lower doses titrated to effect and started weeks before surgery.24As new knowledge is acquired, it is important to disseminate the information to the individual practitioner. Presentations at national meetings and publication of studies in journals are the two primary modes of dissemination. Adoption and diffusion of innovation (such as procedures, drugs, and devices) have been well studied. Rogers25proposed an adoption or diffusion model based on agricultural studies. Initially, there is innovation followed by the early adopters. As seen in figure 3, this is followed by the early majority, the late majority, and finally the laggards. The innovators and early adopters can be used as change agents and influence others to adopt best practices. However, the overall speed of adoption may be slow. It is well known that many interventions that are found to be effective take years to be adopted. For example, thrombolytic therapy was not advocated in textbooks for nearly 20 yr after it was shown to be effective in a meta-analytic analysis26(fig. 4).Although there are large number of articles published annually, the ability to keep abreast of all of the new knowledge and to integrate the information into a coherent practice is difficult. Traditionally, review articles by experts offered a means of integrating knowledge, but these articles have many limitations. For example, the authors may not approach the review of the literature in a systematic way, and there may be selection bias in the studies included in the analysis. Therefore, these articles frequently include the bias of the author. During the past two decades, there has been a movement toward evidence-based medicine and the use of systematic approaches to the review of the literature and the study of medical interventions. Some of the techniques used in evidence-based medicine include systematic reviews, meta-analysis, and cost-effectiveness analysis.27One means of codifying these evidence-based reviews is through the development of standards and guidelines. It is important to recognize that a standard implies that a therapy or practice should be used for patients with a particular condition. Standards are only approved if an assessment of the probabilities and uses of the group indicate that the decision to choose the treatment or a strategy would be virtually unanimous and should be applied rigidly, although many standards do not have a body of literature to support their use. One such example is pulse oximetry monitoring for anesthesia.28However, guidelines are a set of recommendations for patient management that identifies a specific or range of management strategies based on the evidence. Guidelines are usually developed and promulgated by a specialty medical society or government agency (i.e. , National Institutes of Health or Agency for Healthcare Research and Quality). Formal recommendations are published, which include the strength of recommendation. For example, the American College of Cardiology/American Heart Association classifications of evidence that are used in their Guidelines to summarize indications for a particular therapy or treatment are as follows:The levels of evidence are as follows:The problem is that the quality of the evidence, even in the well-studied area of cardiovascular disease, is poor. Tricoci et al .29evaluated 16 current American College of Cardiology/American Heart Association Guidelines reporting levels of evidence and found that only 314 recommendations of 2,711 total are classified as level of evidence A, whereas 1,246 are level of evidence C. Recommendations with the level of evidence A are mostly concentrated in class I, but only 245 of 1,305 class I recommendations has level of evidence A.If an intervention is found to be effective or advocated in a Guideline, then should it always be implemented in every patient? Clearly, the answer is no because the patient of interest may not be similar to the original population studied or the processes of care, and the method of implementation is different at the local institution than those of the original studies. As described earlier, patients have different levels of risk, and in addition, patients have a complex characteristic of diseases for which the best practice may actually be in conflict. Boyd et al .30demonstrated that most clinical practice guidelines do not modify or discuss the applicability of their recommendations for older patients with multiple comorbidities. They also demonstrated that adverse interactions between drugs and diseases could result. Therefore, applying best practices (which may be defined in an understanding of the applicability of the results and the risks and benefits in the individual This is a further application of efficacy effectiveness and the importance of this information into guidelines and defined for the are an important of the application of evidence to the individual different patients value and different to different and the optimal decision for any given patient is to these can be using which has been applied to such as nausea, and choice of or outpatient approach to assess is the use of standard the is offered a choice of a known and also offered a probabilities of a of morbidity or mortality. example of this is used on the or in which the is offered a defined from the a that their a higher or lower techniques have frequently been applied to determine how such outcomes as and so the risk or risk of the research group has applied these techniques to perioperative and demonstrated that for to and for and most patients would after although the of patients to take the risk of is much for severe critical area in which patient impact on is in the area of preoperative cardiovascular testing in patients undergoing intermediate risk surgery with one to two risk As in the American College of Cardiology/American Heart Association to the with was a class preoperative testing was a class to determine the decision for the individual a for surgery management can the For example, if the patient is determined to be at high risk, then they may choose to a or no surgery at to healthcare and to reduce costs has interest on The for Medicare and as strategy that can help to the current by for high clinical The optimal strategy would be to and based on their however, the importance of baseline risk in assessment on is well numerous attempts at optimal have been adopted for means of the goal of is through the development of performance known to lead to improved and to achieve the performance The underlying assumption is that do not adopt best practices in the literature or in as by the low rates on practices such as appropriate continuation of β-blockers, and shown in the are two general of performance hospital based and physician of hospital-based are those advocated by There is no between performance and in the Medicare but between reporting and of with the can be found on the level performance are also used by the Medicare and by many are developed by the specialty for and of the American of and by the American Association or other national and the National Perioperative examples include of in a and of is critical to that such performance are to the evidence and that they change with in the evidences. For example, there were numerous including which initially set the standard for β-blocker therapy as in all patients at risk. The expert was about the results of the and therefore to only include continuation of β-blockers in those patients already these agents as the the trial was is the potential of practices For example, many of the of within a in the for may lead to of patients who have In addition, there is a of that the goal should not be because with preoperative for may lead to multiple preoperative Among in a the was not associated with a in quality of care or outcomes for myocardial the investigators not evidence that for performance had an adverse with in processes of myocardial infarction care that were not to are other about paying for Some investigators have questioned the value of paying for performance as to quality Lindenauer et al to and of quality during a of 2 yr at that information about the quality of care through a national reporting including that in a by the for Medicare and They that compared with the greater in all of including of care for myocardial infarction, and and a of measures. were made for differences in baseline performance and other hospital was associated with from to during the it is whether actually lead to improve outcome. and a study of hospital care between 1 and for myocardial infarction, and at care in the included in the and compared with hospital mortality which were using Medicare A claims all myocardial infarction performance the reduction in mortality rates between in the those in the was for for and for mortality. in mortality rates for in the on all within a condition those lower than the on all for myocardial infarction between = and = They that hospital performance differences in hospital mortality rates and that should be made to performance that are to patient if organ dysfunction and complications were it is whether mortality would also et al the concept of to as a of hospital investigators suggested that measuring mortality in patients who developed complications associated with improved to include higher ratios and the of of this research was developed using administrative Medicare mortality and morbidity from the American College of National investigators demonstrated that the of complications was similar between in the but mortality varied and to varied in this population (fig. the of an important may not reduce the past several decades, there has been a to improve perioperative outcomes through the performance of randomized trials and and of the into recommendations of care. can have impact on perioperative through the application of evidence to reduce organ To achieve those it is critical that anesthesiologists the evidence and help processes to be appropriate to the individual these the current approach to it is clear that quality that outcomes do in It is also that anesthesiologists can reduce the of and therefore can impact the of death in patients who complications. must to to improve the outcomes of our patients through a to