PA-PSRS has received several reports in which patients were discharged too soon after presenting to the Emergency Department (ED) with poisoning.

Case #1: A four year-old patient presents to the ED with the mother after having ingested a large number of anticonvulsant pills prescribed for an adult; the patient appears asymptomatic at presentation. The ED physician orders a test which measures the level of the drug in the patient’s blood; this test comes back in the normal therapeutic range. The patient is then discharged but returns to the ED unresponsive several hours later. The patient ultimately recovered without sequelae.

Case #2: A patient presents to the ED following a suicide attempt in which he ingested ethylene glycol (i.e., antifreeze) along with several alcoholic beverages. He is asymptomatic on presentation, and initial laboratory tests are negative. The patient is transferred to a psychiatric facility for evaluation before the physician receives the result of another diagnostic test showing a very high ethylene glycol level. The patient required urgent dialysis.

In both cases, the patients presented as asymptomatic and presumably remained asymptomatic at discharge. Also, both involve scenarios in which the pharmacokinetics of the ingested substances result in a delay in presentation of the signs and symptoms of toxicity.

In Case #1, the drug the patient ingested has a variable rate of metabolism and a long half-life. The Patient Safety Officer at this facility indicates that the pharmacokinetics of the particular drug ingested were not taken into account. Without holding the child for observation and taking at least a second blood level, there was no way to know whether the drug level was increasing or decreasing at the time of examination.

In Case #2, the patient likely failed to develop classic signs of toxicity on some standard lab tests because he ingested alcoholic beverages along with the ethylene glycol. Ethanol (the alcohol found in alcoholic beverages) is one treatment for ethylene glycol poisoning. The ethanol in the alcoholic beverages likely suppressed metabolism of the ethylene glycol enough for the patient to test negative for acidosis and fail to develop an anion gap characteristic of ethylene glycol poisoning.^1,2 The physician may not have considered that the clinical signs and symptoms usually associated with ethylene glycol poisoning would be tempered by the concomitant alcohol ingestion.

All diagnostic tests, including the physical exam, represent a “snapshot” of the patient’s condition at a single moment in time. That condition may change quickly and dramatically. While clinicians may understand this at an intellectual level, they may, like any person, be victims of well-known cognitive biases. One form of cognitive bias described by Cook and Woods that may have played a role in both of these cases is treating a dynamic situation as static.³

Other forms of cognitive bias may also have played a role. For example, in the first case, the physician may have developed a false sense of security from the blood test result’s being “in the normal therapeutic range.” While the test suggested that at that time the patient did not have toxic levels of the drug circulating in the blood stream, the concept of the therapeutic range was irrelevant for this patient, who was not prescribed this drug and was not taking a stable daily dose. The physician surely reviews dozens of test results every day, which are typically presented with reference values that constitute the normal range for each parameter tested. That the lab test was interpreted as “normal” may have framed the clinical situation in a way that downplayed its critical nature. This is consistent with the “framing effect,” in which one’s interpretation of information is influenced by the form in which it is presented.⁴

The ethylene glycol case is suggestive of confirmation bias—a tendency to selectively perceive information consistent with our prior beliefs and to discount or avoid information that is inconsistent with those beliefs.⁵ The physician seems to have ordered a range of tests during the initial evaluation. Presumably, he or she believed those tests to be necessary for an adequate diagnosis at that time. That the patient appeared asymptomatic was surely considered as part of the diagnosis, and when initial lab tests returned negative they may have “confirmed” a prior belief based on the lack of expected symptomatology. Not waiting for the ethylene glycol level suggests a discounting of potentially relevant and disconfirming information.

Cognitive biases are errors in human decision-making resulting from the heuristics (or “rules of thumb”) we use in everyday life to process information. These errors have been shown in individuals at all levels of education.⁶ When performing a root cause analysis of your facility’s Serious Events or Incidents, consider whether cognitive biases may have played a role and whether any systems changes may help to mitigate the risk these biases pose.  Protocols or algorithms based on evidence-based rational decision-making processes can provide structure that may minimize these biases.  In poisoning cases, poison control centers are at the core of these protocols and algorithms.

Notes

1. Ammar KA, Heckerling PS. Ethylene glycol poisoning with a normal anion gap caused by concurrent alcohol ingestion: Importance of the osmolal gap. Am J Kidney Dis. 1996 Jan;27(1):130-3.
2. Eder AF, McGrath CM, Dowdy YG, et al. Ethylene glycol poisoning: Toxicokinetic and analytical factors affecting laboratory diagnosis. Clin Chem. 1998 Jan;44(1):168-77.
3. Cook RI, Woods DD. Operating at the sharp end: The complexity of human error, 13. In: Bogner, MS, Ed. Human Error in Medicine. Hillsdale (NJ): Lawrence Erlbaum Associates; 1994.
4. Tversky A, Kahneman D. Judgment under uncertainty: Heuristics and biases. Science. 1974;185:1124-31.
5. Nickerson RS. Confirmation bias: a ubiquitous phenomenon in many guises. Rev Gen Psychol 1998;2:175-220.
6. Clarke JR. Clinical Surgical Decision Making. In: Rutkow, I.M., ed. Socioeconomics of Surgery. St. Louis: Mosby, 1989:315-331.