Evaluating Medical Decision-Making: Using the EHR to Assess and Structure Resident Autonomy Free

Resident autonomy has long been a topic of discussion, especially as it relates to clinical decision-making on nights when in-person faculty supervision is less common. In response to changes in the Accreditation Council for Graduate Medical Education requirements, there has been an increase in the presence of attending oversight.^1,2  On one hand, researchers have suggested that having an attending overnight has led to an increase in the education provided for residents during those hours. However, other data have revealed that decreasing resident autonomy with an in-house attending may make residents less prepared for independent practice and does not result in significant changes in patient outcomes.^3,4  One way to evaluate resident medical decision-making (MDM) is to evaluate the orders that are being placed, specifically during a period of limited attending oversight. In this issue of Hospital Pediatrics, Freiman et al⁵  share a study that was designed to assess the types of orders placed by residents during an overnight shift. They evaluated daytime changes to orders placed by residents overnight as a way of assessing accurate MDM.

In this retrospective study, the authors looked at orders placed by residents on new patients overnight that were discontinued after resident handoff along with what new orders were placed on these patients during morning rounds. Orders were organized into several different types, including medicine, diagnostics, diet, patient care, and other. In addition, they took note of the reason for the addition or discontinuation of said orders and further classified them into other subtypes, including change in medical trajectory, medical error, and MDM. Orders were classified as discontinued if they were placed by overnight residents and then discontinued during daytime rounding hours. Orders were classified as new if they were first placed during daytime rounding hours. The study initially revealed 819 discontinued orders and 5108 new orders across its duration. Of these, 269 discontinued orders met inclusion criteria, and subsequently, new orders were screened until 269 met inclusion criteria. In the end, 66.4% of order discontinuation or addition was due to MDM. The remaining orders were characterized as a change in either patient trajectory (28.1%) or medical error (5.6%). When subcategorizing the changes within medical errors, the authors noted omission of home medication as the most common error (30%).

Freiman et al⁵  used the electronic health record (EHR) in a unique way, leveraging it to audit a large number of orders and further classify them to better understand their rationale. Additionally, the EHR allowed for a way to measure potential medical errors that may have been made while providing patient care overnight. The decision to perform chart reviews also helped to establish the justification for the orders placed. However, the study has a few limitations. One limitation is that the authors could not identify all orders that affected MDM, especially if the documentation was ambiguous; however, we believe that if the order change was significant, some form of documentation usually would be made. Although the authors mention an underestimation of resident errors as a result of only capturing changes to resident orders on morning rounds, errors can be made that are not linked to orders in the EHR. Furthermore, this thought can be applied to orders that were modified in the EHR; thus, because the authors did not assess modified orders, it reveals a future opportunity to assess resident MDM. An example of orders that might typically be modified rather than discontinued include the escalation of albuterol frequency for asthma exacerbations. In the same vein, the authors also potentially missed MDM orders that were made and discontinued overnight.

In the overall discussion of resident autonomy, few researchers have spoken on how increased attending supervision affects the personal perception of resident autonomy. Haber et al⁶  described that having an attending physician overnight increases the educational opportunities for residents but has no real significant impact on autonomy; the residents surveyed in this study did not report that their autonomy had decreased. With regard to the medical errors noted, most errors were seen with medication reconciliation. Climente-Martí et al⁷  noted that errors with medical reconciliation were more likely to occur at discharge than on admission to an inpatient adult unit. They specifically looked at errors in medical reconciliation but found that only 3.4% were true errors and the remaining 96.6% were intended omissions because of a medication not being necessary for the admission or because of a formulary substitution. Additionally, Climente-Martí et al⁷  revealed that the true errors were with medical reconciliation, most commonly because of omission of a drug, and were more likely to occur on discharge rather than on admission. Although “omitted intended order” was not specifically defined, our understanding is that these were medications that were not ordered because of the patient not needing them. As a result, the finding that the majority of medical errors were actually due to omitted home medications reveals a contrast to the aforementioned study.⁷  Unfortunately, because of the small sample size of the Freiman et al⁵  study, it is difficult to define the true incidence of medical errors, especially during a period of increased resident autonomy.

The night shift offers residents the benefit of increased autonomy in conjunction with decreased in-person attending staffing at some institutions. As such, one might assume that in a period of increased resident autonomy, more medical errors would be noted. However, Freiman et al⁵  suggest that the rate of medical errors is actually low. We agree with their next step of using the EHR to provide feedback to residents with regard to their MDM overnight. We believe that this feedback can inform residents on any potential changes that could have been made earlier in the hospital course. However, some additional next steps to take would be to look at the rates of medical errors between day shifts and night shifts, weekdays and weekends, and patient admission and discharge. We suggest these as next steps to help to elucidate the rates of medical errors during periods where there is decreased medical staffing (ie, weekends). In response to the authors’ suggestion of using the EHR to track adherence to clinical practice guidelines, we believe that it will be challenging to assess the particular context or clinical changes that may dictate a deviation from a clinical pathway. Furthermore, this study was performed in only 1 hospital, limiting its generalizability and applicability to other attending-resident setups. It would be interesting to see how the results might vary in different hospital systems, particularly with regard to institutions that have a different model of attending supervision overnight.

In summary, we applaud Freiman et al⁵  for clarifying and critiquing the types of orders placed by residents on a night shift. As the landscape of residency education continues to evolve, we believe that this information will be useful while programs attempt to best balance resident autonomy and supervision. We look forward to future studies that can find other ways to incorporate use of the EHR to further assess resident clinical aptitude and development of entrustable professional activities.