In this two-part series, reducing medical errors can be aided by individual efforts, system design, and the use of quality initiatives to evaluate trends and coach towards improvement.
As part of our quality assurance program, I’ve seen an uptick in return visits where it looks like our docs could have done better the first time. It seems like we’re always rushing to get to the next patient and sometimes just checking boxes to help dispo the current ones. What advice do you have to improve the care we’re providing?
Early on in my med school intro to clinical medicine class, I remember the attending presenting a case for us to create a differential diagnosis. We didn’t know much medicine, so it wasn’t easy. On the other hand, the attending said that within years, we’d have a differential diagnosis (ddx) in mind within minutes of hearing the chief complaint.
All of us can relate to that statement and all of us know exactly what our ddx (and plan is) as soon as the patient says chest pain, abdominal pain, headache, etc. That’s what we do. We evaluate for life threatening processes with the tools that we have based on the signs and symptoms.
In many of our current practices, where we’re seeing patients in the waiting room or rushing through the simpler patients to find the time to spend with the sicker patients, we’re often trying to get patients to fit into our order sets.
Part of leadership and management is coaching, teaching, and providing feedback. There are a variety of techniques that physicians should individually apply in the clinical area. There are also a few ways that medical leaders can intervene to help understand and reduce the risk of an error.
Avoid tunnel vision and cognitive biases— There was a great article published on cognitive biases a few years back. Bias is a huge issue and contributes to medical errors. The authors described four biases that can impact medical decision making and contribute to medical errors.
We’ve all seen examples of cognitive bias in our practice. For instance, take the patient referred in by the primary care doctor for treatment of pneumonia who gets IV antibiotics and admitted, only to be diagnosed with CHF later that night because the ER doc had a narrow vision based on the referral. Or the patient with chest pain referred for a rule out MI who has a PE or aortic dissection.
These are real cases and happen more than you think. We fit the symptoms into our check box list, hit the easy button on the admission, and move on. This is typically confirmation bias, which is when we selectively choose the data that fits our diagnosis, often ignoring contradictory data, and not considering alternative diagnoses. The hospitalist is also busy and trusts our judgement, and clicks the order set that fits.
This is an example of diagnostic momentum, where each doc builds on to the current diagnosis without questioning its validity. Typically, the diagnosis is ultimately made but it may delay care or lead to unnecessary morbidity and/or mortality. Tunnel vision has been described as the “failure to see the big picture and gather all of the evidence.” This is similar to anchoring bias, which is when physicians prioritize information to confirm their initial impression.
Affect heuristic “describes when a physician’s actions are swayed by emotional reactions instead of rational deliberation about risks and benefits.”
For some of us it might be the complaining patient or the patient who insists on ranking their pain as a 12 on a 1-10 scale. Finally, we get to outcome bias which refers “to the practice of believing that clinical results—good or bad—are always attributable to prior decisions.” This is the type of bias that prevents us from accurately assigning blame or finding fault in a clinical decision because we are impacted by knowing the outcome.
We are typically over-critical if there is a bad outcome and under-critical if there is a good outcome. There is no way to eliminate cognitive biases – we just need to develop strategies to mitigate their impacts on our decision-making. Most authors agree that an important first step is becoming aware of our cognitive biases and how they can land us in hot water.
Pause before clicking on discharge and make sure your chart is aligned: I realize many of us are charting after the patient has left due to all of the various time constraints I’ve already mentioned. I do my charting in phases where I try to write in the majority of the HPI right after I see the patient when it’s fresh in my mind. I’ll usually fill in some of the medical decision making and plan after I get some data back and update the patient on next steps. Finally, prior to clicking on discharge, I’ll review the entire chart to make sure it aligns with the data, my medical decision making, final plan, and discharge diagnosis.
I was taught that the chart needs to tell a clear story. Your chart shouldn’t start with “severe stabbing and tearing chest pain radiating to the shoulder blades,” with physical exam “remarkable for unequal radial pulses” and then use a diagnosis of abdominal pain of unclear etiology without addressing why it’s not an aortic dissection. Usually, my charts make sense and I feel discharging the patient is the appropriate decision. Very rarely, but it definitely happens, I go back and look at my chart and data (including vitals), and realize I didn’t cover all of my bases and need to clarify something or even get another test. It’s not ideal from a flow perspective, but this overall review of the patient care, has definitely appropriately changed a few dispositions in my career.
Consider the serial exams we used to do on possible appendicitis patients. If you were pretty convinced the patient has appendicitis, even if the CT is negative, it doesn’t mean they don’t have appendicitis. Reexamine them. Maybe the exam has changed. Maybe they’re now febrile and tachycardic which makes it easy. If not, it may not mean you need a surgical consult, but the conversation you have with the patient becomes insisting on them returning in 12-24 hours for a reexam, and that may be a game changer.
Attention to detail matters. We are always in a rush and typically we face frequent interruptions. However, being obsessive about the little things matters. There are subtle clues that when missed can lead to bad outcomes. Bumps in creatinine and bilirubin, or a decrease in platelets can be the difference between identifying severe sepsis and TTP or lead one to make an incorrect disposition. As rushed as we are during a shift, there are brief periods of time that need intense focus. Thinking about how quickly we review and sign off on EKGs is an opportunity to make sure you’re focusing on the details. This could include the QTc on an alcoholic patient, which could indicate hypomagnesemia or subtle ST changes when comparing them to prior tracings.
Be a STAR—You may be ordering labs and meds on an unstable patient, answering the phone to take a patient referral, and have the tech waiting for you to review/sign an EKG at the same time (and several times a day). Each of these is time sensitive and requires concentration. I’m more likely to leave out key orders if I don’t complete this task in its entirety before clicking “sign.”
Therefore, the other two tasks wait. EKG reviews are typically quick and negative but unfortunately, we may have to ask the tech to wait a sec while we complete orders before turning out full attention to EKG. And if you need to compare it to an old EKG, take the time to do it. Finally, the referral. I think having a formal referral process is critical for several reasons, including hearing the backstory from the referring provider and it’s good for volume growth, though it may be the least urgent of the three tasks at hand. When you’re feeling “taskus interruptus,” be a STAR—Stop, Think, Act, Review. Give each action the intense focus it deserves and make sure to get it right the first time.
Optimize your EMR. This is a broad statement and can be a never-ending battle. However, easy wins come with optimizing order sets, having notifications for pediatric patients that require weight-based medication dosing or patients who are pregnant (to make sure you’re aware of the impact to work up and/or the final diagnosis), and the use order sets for disease processes (like sepsis) or complex medications like coagulation reversal. If designed properly, order sets can keep us from forgetting a critical order, such as the lactate when we order blood cultures or the Rh on pregnant vaginal bleeding. If used properly, order sets can reduce our misses.
On the other hand, between seeing patients in the waiting room and clicking a chief complaint-based order set, we may rush through the nuance of the individual patient’s history. If you know your order sets inside and out, take a second to reflect after you’ve made that click to see if there is anything outside of the order that you need. Typically, the answer is no. And if there are other tests that should be included in your order set, work with your team to make it happen.
Your EMR can also be used to aid medical decision making. You can build decision making algorithms into your dot phrases and even build entire algorithms into your chart. I find that docs are pretty happy to click on the phrase if it can help them make a decision while at the same time provide necessary documentation. If you want to try something easy, build phrases for the PECARN rules (with options to not need a CT and also to require one) as well as the HEART Score. These are really simple and if you need help, send me a DM.
Design your quality reviews with purpose—Although in part 2 we’ll take a deep dive into reviewing bounce backs and how to provide feedback, I’d like to reintroduce something many of us take for granted. Reviewing 72-hour returns has been a mainstay of quality review for decades. I remember the lesson I learned at my site 15 years ago. This was really before i-stat lactates and the emphasis on sepsis, but we clearly saw trends. We saw a lot of cellulitis in this facility and probably to no one’s surprise, febrile diabetics who were tachycardic bounced back with a high frequency.
The cases were more subtle than it sounds, but it led to education, feedback, and a change to management. This led to a separate project on tachycardia and discharge vital signs. The concern at the time was trying to reduce the risk and better identify patients with sepsis and pulmonary emboli. The project led to nursing education about specifically communicating to the docs when patients had tachycardia on discharge vital signs.
This also involved the docs recognizing that these patients deserve another look. This improved our team’s situational awareness. Febrile tonsillitis patients who are taking PO but have a mild tachycardia are probably fine. So are asthmatics after nebs. However, if you pay attention, you’ll find patients that presented with chest pain, had a negative trop and normal EKG, get put up for discharge, and tachycardia might be the only clue that helps you diagnose their PE.
Next time, we’ll put a new twist on 72 hour returns and discuss properly providing feedback to your group.
Doherty T, Carroll A. Believing in Overcoming Cognitive Biases. AMA J Ethics. 2020;22(9):E773-778. doi: 10.1001/amajethics.2020.773.
Weinstock MB, Longstreth R, Henry G. Bouncebacks! Emergency Department Cases: ED Returns. 2nd ed. 2007. Anadem Publishing. Columbus, Ohio.
Chartier LB, Ovens H, Hayes E, et al. Improving Quality of Care Through a Mandatory Provincial Audit Program: Ontario’s Emergency Department Return Visit Quality Program. Ann Emerg Med. 2021; 77:193-202.
Shy BD, Shapiro JS, Shearer PL, et al. A conceptual framework for improved analyses of 72-hour return cases. Am J Emerg Med. 2015;33:104-107.
Shy BD, Loo GT, Lowry T, et al. Bouncing back elsewhere: multilevel analysis of return visits to the same or a different hospital after initial emergency department presentation. Ann Emerg Med. 2018;71:555-563.e1.
Hartigan S, Brooks M, Hartley S, et al. Review of Cognitive Error in Emergency Medicine: Still No Easy Answers. West J Emerg Med. 2022 No; 21 (6z): 125-131.