Perspectives in Hospital Medicine

Discrepant Advanced Directives and Code Status Orders: A Preventable Medical Error

© 2019 Society of Hospital Medicine

The United States health system has been criticized for its overuse of aggressive and medically ineffective life-sustaining therapies (LST).1 Some professional societies have elevated dialog about end-of-life (EOL) care to a quality measure,2 expecting that more open discussion will achieve more “goal-concordant care”3 and appropriate use of LST. However, even when Advanced Directives (AD) or Physician Orders for Life-Sustaining Therapy (POLST) have been created, their directions are not always followed in the hospital. This perspective discusses how preventable errors allow for use of LST even when patients designated it as unwanted. Two cases, chosen from several similar ones, are highlighted, demonstrating both human and system errors.

During the time of these events, the hospital policy required admission orders to contain a “code status” designation in the electronic medical record (EMR). All active and historical code status orders were listed chronologically and all AD and POLST documents were scanned into a special section of the EMR. Hospital policy, consistent with professional society guidelines,4,5 stated that patients with AD/POLST limiting EOL support should have individualized discussion about resuscitation options in the event of a periprocedural critical event. Automatic suspension or reinstatement of limited code orders was not permitted.

CASE 1

A 62-year-old woman with refractory heart failure was admitted with recurrence. The admitting code order was “initiate CPR/intubation” even though a POLST order written 10 months earlier indicating “do not intubate” was visible in the EMR. A more recent POLST indicating “No CPR/No intubation” accompanied the patient in the ambulance and was placed at bedside, but not scanned. There was no documented discussion of code status that might have explained the POLST/code order disparity. Notably, during two prior admissions within the year, “full code” orders had also been placed. On the fifth hospital day, the patient was found in respiratory distress and unresponsive. A “code” was called. ICU staff, after confirming full code status, intubated the patient emergently and commenced other invasive ICU interventions. Family members brought the preexisting POLST to medical attention within hours of the code but could not agree on immediate extubation. Over the next week, multiple prognosis discussions were held with the patient (when responsive) and family. Ultimately, the patient failed to improve and indicated a desire to be extubated, dying a few hours later.

CASE 2

A 94-year-old woman was admitted from assisted living with a traumatic subcapital femur fracture. Admission code orders were “initiate CPR/intubation” despite the presence in the EMR of a POLST ordering “no CPR/no intubation.” The patient underwent hemiarthroplasty. There was no documented discussion of AD/POLST by the surgeon, anesthesiologist, or other operating room personnel even though the patient was alert and competent. On postoperative day one, she was found to be bradycardic and hypotensive. A code was called. After confirming full code status in the EMR, cardiac compressions were begun, followed by intubation. Immediately afterward, family members indicated that the patient had a POLST limiting EOL care. When the healthcare proxy was reached hours later, she directed the patient be extubated. The patient died 16 minutes later.

DISCUSSION

Data on the frequency of unwanted CPR/intubation due to medical error are scarce. In the US, several lawsuits arising from unwanted CPR and intubation have achieved notoriety, but registries of legal cases6 probably underestimate the frequency of this harm. In a study of incorrect code status orders at Canadian hospitals, 35% of 308 patients with limited care preferences had full code orders in the chart.7 It is unclear how many of these expressed preferences also had legal documents available. There was considerable variability among hospitals, suggesting that local practices and culture were important factors.

Spot audits of 121 of our own patient charts (median age 77 years) on oncology, geriatrics, and cardiac units at our institution found 36 (30%) with AD/POLST that clearly limited life-sustaining treatments. Of these, 14 (39%) had discrepant full code orders. A review of these discrepant orders showed no medical documentation to indicate that the discrepancy was purposeful.

A root cause analysis (RCA) of cases of unwanted resuscitation, including interviews with involved nurses, medical staff, and operating room, hospitalist, and medical informatics leadership, revealed several types of error, both human and system. These pitfalls are probably common to several hospitals, and the solutions developed may be helpful as well (Table).

ROOT CAUSE 1: HASTE

Haste leads to poor communication with the patient and family. Emergency departments and admitting services can be hectic. Clinicians facing time and acuity pressure may give short shrift to the essential activity of validating patient choices, regardless of whether an AD or POLST is available. Poor communication was the major factor allowing for discrepancy in the Canadian study.7 Avoiding prognostic frankness is a well-known coping strategy for both clinicians and patients8,9 but in all these cases, that obstacle had been overcome earlier in the clinical course of disease, leaving inattention or haste as the most likely culprit.

ROOT CAUSE 2: INADEQUATE COMMUNICATION

“It is not our hospital culture to surveille for code status discrepancies, discuss appropriateness on rounds or at sign out.”

In all reviewed cases of unwanted resuscitation, numerous admitting or attending physicians failed to discuss LST meaningfully despite clinical scenarios that were associated with poor prognosis and should have provoked discussion about medical ineffectiveness. The admitting hospitalist in case 2 stated later that she had listed code choices for the patient who chose full code despite having a POLST stating otherwise. However, that discussion was not in depth, not reviewed for match to her POLST, and not documented.

Moreover, all the cases of AD/POLST and code status discrepancy were on nursing units with daily multidisciplinary rounds and where there had been twice-daily nurse-to-nurse and medical staff–to–medical staff sign out. Queries about code status appropriateness and checks for discrepant AD/POLST and code orders were not standard work. Thus, the medical error was perpetuated.

Analysis of cases of unwanted intubation in postoperative cases indicated that contrary to guidelines,4,5 careful code status review was not part of the preoperative checklist or presurgical discussion.

ROOT CAUSE 3: DECEIVED BY THE EMR

The EMR is a well-recognized source of potential medical error.10,11 Clinicians may rely on the EMR for code status history or as a repository of relevant documents. These are important as a starting place for code status discussions, especially since patients and proxies often cannot accurately recall the existence of an AD/POLST or understand the options being presented.9,12 In case 1, clinicians partially relied upon the erroneous historical code status already in the chart from two prior admissions. This is a dangerous practice since code status choices have several options and depend upon the clinical situation. In the case of paper AD/POLST documents, the EMR is set up poorly to help the medical team find relevant documents. Furthermore, the EMR clinical decision support capabilities do not interact with paper documents, so no assistance in pointing out discrepancies is available. In addition, the scanning process itself can be problematic since scanning of paper documents was not performed until after the patient was discharged, thus hiding the most up-to-date documents from the personnel even if they had sought them. Moreover, our scanning process had been labeling documents with the date of scanning and not the date of completion, making it difficult to find the “active” order.

ROOT CAUSE 4: WE DID NOT KNOW

Interviews with different clinicians revealed widespread knowledge deficits, including appreciation of the POLST as durable across different medical institutions, effective differences between POLST and AD, location of POLST/AD within the EMR, recommendations of professional society guidelines on suspending DNR for procedures, hospital policy on same, the need to check for updates in bedside paper documents, and whether family members can overrule patients’ stated wishes. Education tends to be the most common form of recommendation after RCA and may be the least efficacious in risk mitigation,13 but in this case, education reinforced by new EMR capabilities was an essential part of the solutions bundle (Table).

AD/POLST and similar tools are complex, and the choices are not binary. They are subject to change depending upon the medical context and the patient status and may be poorly understood by patients and clinicians.14 Accordingly, writing a goal-concordant code status order demands time and attention and as much nuanced medical judgment as any other medical problem faced by hospital-based clinicians. Though time-consuming, discussion with the patient or the surrogate should be considered as “standard work.” To facilitate this, a mandatory affirmative statement about review of LST choices was added to admission templates, procedural areas, and clinician sign outs (Table).

Unwanted, and therefore unwarranted, resuscitation violates autonomy and creates distress, anger, and distrust among patients and families. The distress extends also to frontline clinicians who are committed to “do no harm” in every other aspect of their professional lives.

Respecting and translating patients’ AD/POLST or similar tools into goal-concordant code status order is an essential professional commitment. Respect for patient safety and autonomy demands that we do it well, teach it well, and hold each other accountable.

Disclosures

The authors have nothing disclose.

References

   Comments ()