Appropriately conducted peer review of medical practices provides the greatest opportunity for health care professionals to learn from their mistakes and improve the quality and safety of health care. But in practice, peer review has not been an effective learning tool because it is subjective and irreproducible. Physicians reviewing the same cases disagree over the cause(s) of adverse outcomes and the quality and appropriateness of care, and agreement is not improved by training, use of objective review criteria, or having the reviewers discuss the cases. The underlying reason is a general lack of understanding and an oversimplified view of the causes of medical errors in complex, high-risk organization and a preoccupation with attributing medical errors to particular individuals. This approach leads to judgments, not understanding, and creates a culture of blame that stops learning and undermines the potential for improvement. For peer review to have an impact on the quality of care and patient safety, it must be standardized to remove cognitive biases and subjectivity from the process.
“The human understanding when it has once adopted an opinion draws all things else to support and agree with it. And though there be a greater number and weight of instances to be found on the other side, yet these it either neglects and despises, or … sets aside and rejects, in order that … the authority of its former conclusion may remain inviolate.”
We all learn from experience, and “experience”, Oscar Wilde noted, “is the name everyone gives to their mistakes.” The renowned colorectal surgeon, John Golligher, is said to have greeted visitors to his unit in Leeds, England, saying, “You have not come here all this distance to hear about my successes; let me tell you about my failures because that is what you learn from.” According to information gain theory, the expected gain of information is directly proportional to its improbability (ie, the extent to which the information is unexpected or surprising). That is why we learn more when things go awry than as expected.
Peer review is the name given to the process by which the medical profession oversees the quality of medical care. The keystone of peer review is a retrospective review of complications, adverse outcomes, and incident reports by one’s peers. Ongoing and Focused Professional Practice Evaluations that the Joint Committee on the Accreditation of Healthcare Organizations now requires can identify physicians who are outliers with respect to certain outcomes in a statistical sense but not why they are outliers or what can be done about it.
Appropriately conducted peer review provides the greatest opportunity for health care professionals to learn from their mistakes. Nevertheless, peer review has not been an effective learning tool because physicians are not informed about the cognitive processes that underlie medical errors and causal judgments. Therefore, even under optimal, nonadversarial, research conditions, peer review is largely subjective, and its reproducibility is low, barely more than what would be expected by chance.
Physicians reviewing the same cases disagree over the cause(s) of adverse outcomes and the quality and appropriateness of care, and agreement is not improved by training, use of objective review criteria, or having the reviewers discuss the cases. In real-world settings, in which motivational factors are often operative, evaluations can be almost completely arbitrary, even when external reviewers are used to try to ensure objectivity and avoid bias.
For example, 2 out-of-town board-certified obstetrician-gynecologists were retained by 2 unaffiliated hospitals to review the same 26 cases of the same physician who had privileges at each hospital. They were unaware of each other’s review. Three of the 26 cases had complications, each of which was criticized by each reviewer but 2 of them for different reasons. More tellingly, at least one of the reviewers criticized 14 of the 23 uncomplicated cases no one at the hospital had raised concerns about (61%), but, with one exception, they criticized completely different cases, and the one case they both criticized, they criticized for different reasons. Thus, only 4 of the 17 cases criticized by at least one of the reviewers were criticized by both of them, and they criticized 3 of the 4 cases for different reasons.
Motivational factors are likely operative in such real-world situations, as they are in the medical malpractice context, in which courts are well aware that physicians are willing to “express opinions [in a courtroom] that they might not be willing to express in an article submitted to a refereed journal of their discipline … .”
Unless one tries to understand the causes of this irreproducibility, peer review will remain ineffective, be susceptible to misuse, and cause inappropriate sanctioning of physicians. Although hospitals and professional organizations balk at the suggestion, bad faith peer review is a fact of life. For example, the enactment of the Health Care Quality Improvement Act, a federal statute that immunizes physicians engaged in peer review that meets certain criteria from liability, was triggered by peer review that was described by the Ninth Circuit as “shabby, unprincipled, and unprofessional,” and the court found “substantial evidence that the defendants acted in bad faith in the hospital’s peer review process.”
The potential misuse has a chilling effect on peer review because it is not always conducted when it should be and important opportunities for improvement are lost. For example, a case in which a physician abandoned a hysteroscopic procedure because the nursing staff did not have the wherewithal to monitor fluid balance was not reviewed, only to lead to fluid overload in a patient undergoing hysteroscopic myomectomy 6 months later (see the following text). This situation can be remedied only if physicians familiarize themselves about the cognitive processes and biases involved in peer review and causal attributions and structure peer review to enhance its reproducibility.
Cause attribution
People engage in causal reasoning to make sense of the world. Unusual, unexpected, abnormal, or unlikely events spontaneously engender causal reasoning to explain them. But causation cannot be observed and must be inferred. How people infer causation from observations continues to be debated by philosophers, psychologists, and experts in computer learning and artificial intelligence, but for the purposes of this discussion, there are 2 fundamentally different types of causes: necessary and sufficient.
A necessary cause (C) is one that is insufficient by itself to cause an event (E) but if removed will prevent the event from happening (ie, C can occur without E but E cannot occur without C). A sufficient cause (C) is one that is enough to bring the event about but that is not necessary for the event to occur (ie, C cannot occur without E, but E can occur without C). Necessary causes are generally referred to as enabling conditions, rather than causes. For example, lightning would be considered a (sufficient) cause of a forest fire, but oxygen, a necessary cause, would be considered an enabling condition.
Few, if any, outcomes in medicine have a single sufficient and necessary cause. Most outcomes have many necessary, insufficient causes that are only jointly sufficient to bring the outcome about. Nevertheless, people discriminate between causes, and usually select–from equally necessary conditions–the one condition they deem to be the cause of the event, which is tantamount to erroneously treating as independent causes that are in fact jointly dependent. Which member of a sufficient set is selected as the cause of the outcome depends on context and is influenced by subjective perspectives and cognitive biases, especially the tendency of causal beliefs to dominate judgment, and explains the irreproducibility of expert opinions.
For example, sexual intercourse, lack of birth control, and fertility of the man and woman are all equally necessary conditions for an unwanted teenage pregnancy to occur, but religious parents would likely view illicit sexual intercourse as the cause of an unwanted teenage pregnancy, whereas policy makers would likely consider insufficient sex education and birth control as the causes. But neither would likely consider fertility, an equally necessary condition, as the cause of an unwanted pregnancy. In general, by focusing on causes to attribute to individuals rather than enabling conditions, decision makers lose valuable opportunities to prevent adverse outcomes from recurring because only removal of enabling conditions (necessary causes) can prevent events from occurring.
To determine whether a condition is necessary for an event to occur or what caused an adverse outcome, people engage in counterfactual thinking, which is a form of mental simulation. That is, they create a mental representation of the facts (ie, what actually happened) and then mentally change or undo some aspects of that representation and imagine how the change would affect the outcome, specifically whether the event could occur without its putative cause. This mental process is automatic and spontaneous but it is not random, and events are mentally altered in a biased, not a systematic, manner.
Most people tend to have the same kind of counterfactual thoughts that focus on the same types of events. People tend to alter actions rather than inactions; controllable events (eg, route of surgery) rather than uncontrollable ones (eg, pathology); unusual, abnormal, or socially unacceptable events; and events that come last in a sequence. Knowledge that an adverse outcome occurred (hindsight bias), and the personal preferences and beliefs of the reviewer, both influence the factors that will be considered unusual or abnormal.
For example, root cause analysis was conducted after a physician’s privileges were summarily suspended because a patient of his sustained a cardiac arrest during hysteroscopic myomectomy because of fluid overload. The outpatient facility at which the procedure was performed did not have a closed fluid monitoring system or even a form on which fluid input-output could be recorded. Those conducting the root cause analysis did not consider either of these deficiencies as causes of the mishap, notwithstanding that a simple form on which fluid intake and output could be recorded every 15 minutes would very likely have avoided the physician’s error in assessing fluid balance (ie, they were enabling conditions).
Counterfactual thinking and judgments about causation, like most decisions people make, are made in an automatic, effortless way and is biased by prior knowledge and understanding of the subject matter, context, and other heuristics. People usually accept the first cause(s) that come to mind and do not continue to seek alternative causes (a characteristic referred to as satisficing). What first comes to mind then biases how the facts of the case and evidence bearing on causation are reviewed. This is called confirmation bias. Evidence consistent with this initial judgment will be sought out, given more weight, and will be better remembered, whereas inconsistent evidence will be ignored or discounted or else scrutinized hypercritically in an effort to invalidate it, and ambiguous evidence will be construed to be consistent with this initial judgment.
People tend to attribute the actions of others to their personal characteristics rather than situational factors but to attribute their own decisions to external factors rather than idiosyncratic personal experiences or character traits. This is called the fundamental attribution error. It leads physicians to blame colleagues by attributing adverse outcomes to their lack of skill or poor judgment rather than the pathology being treated, the inherent risks of treatment, or other technical factors. The related false consensus effect causes people to consider their own opinions and viewpoints as more widely shared than they actually are. This creates unwarranted overconfidence in causal attributions and may explain why discussion among disagreeing reviewers does not produce greater agreement.
Causes of medical errors
Medical errors are the result of errors in cognition of the person most proximate in time and place to the error, the person traditionally held accountable for the error. Human beings can process information rapidly and make effective, error-free judgments and decisions under most circumstances. However, the semiautomated aspects of this process that endow it with these capabilities also create biases that under some circumstances cause people to make erroneous judgments and incorrect decisions ( Table 1 ).
| Bias or heuristic | Effect of bias or heuristic |
|---|---|
| Anchoring | Tendency to fix on most salient features of a patient’s initial presentation. This initial impression may not be properly adjusted to take account of later information. |
| Ascertainment bias | Some information is inherently more readily obtainable than others. There is also a tendency to mentally register information that conforms to expectations than other available information. |
| Availability heuristic | What readily comes to mind is judged to be more frequent than it may actually be. |
| Base rate neglect | Tendency to ignore the prevalence of disease. |
| Cognitive lock-up | Failure to revise decision or action as a situation changes or develops. |
| Confirmation bias | Tendency to preferentially seek out information that confirms the hypothesis one is testing. |
| Expectation bias | Tendency to give more weight to evidence that confirms one’s expectations. It is not quite same as tendency to seek out (confirmation bias) or notice (ascertainment bias) what one expects. |
| Hindsight bias | The likelihood of a known outcome is judged to be much more likely than it actually was and hence to be much more foreseeable/predictable than it really was. |
| Inert knowledge | Person has requisite knowledge but it is not activated and used when required. |
| Loss of situational awareness | Failure to appreciate that circumstances are changing and that action(s) have to be changed or taken in light of the changed circumstances to avoid an error, which leads to cognitive lock-up. |
| Outcome bias | Favoring diagnoses that lead to good outcomes, which leads to delays in diagnosis of postoperative complications, for example. |
| Overconfidence | General tendency of people to overrate their own knowledge, abilities, and performance. The Lake Wobegon phenomenon, in which everyone is above average. |
| Premature closure | Search for evidence stops when a diagnosis is reached, which precludes considering alternative diagnoses. This is a common reason diagnoses are missed. |
| Search satisficing | Results in premature closure as search stops when a diagnosis is found. |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
