The Autonomy Toolbox: A Multicenter Collaborative to Promote Resident Autonomy

This QI initiative occurred at 5 tertiary care children’s hospitals from July 2020 to June 2022 (Supplemental Table 3). We formed a multidisciplinary team of Pediatric Hospital Medicine (PHM) faculty, PHM fellows, chief residents, pediatric residents, educational staff, and QI analysts. The interventions were performed on PHM inpatient teams. Although each institution had unique team structures (e.g., presence of fellows, number of residents and students), all teams had at least 1 SR and an attending physician. Site leaders had flexibility to adapt interventions to fit with the culture of their institutions.

A 22-item needs assessment survey (Supplemental Fig 4) was adapted from the validated survey by Biondi et al.³  We implemented minimal modifications to the questions to fit our specific program design and edited our survey using an iterative consensus process with pediatric chief residents, PHM fellows, PHM faculty, and SRs. Using Kane’s validity framework,²⁰  scoring and generalizability inferences can be drawn given our adaptation of the Biondi et al³  survey using similar syntax and intention to survey a comparable population. Internal review of our survey was performed for consistency and additional review with surrogate survey respondents (ie, chief residents, Infectious Diseases faculty) provided increased reliability. We sent our survey to 427 pediatric SRs and 220 PHM faculty to determine baseline perception of SR autonomy. SR surveyed included any upper level pediatric or combined pediatrics resident in summer or fall of 2020; survey respondents were instructed to reflect specifically on their time on PHM service. We identified 10 questions in which faculty and SR had statistically significant discordant perceptions of SR autonomy. Of these, 6 were chosen by group consensus to comprise a holistic picture of SR perceptions of autonomy. These were combined to form the RAS and were targets of improvement. The 6 questions incorporated into our RAS were:

1. I feel that my input is taken seriously by my attending when it comes to making important medical decisions.

2. I feel that I am encouraged to make medical decisions for straightforward cases on my own.

3. I feel that my attending physicians allow my plan to be followed, even if they prefer an equivalent alternative

4. I feel that I receive constructive feedback from my attending to help me improve my medical decision-making skills.

5. I feel like I am the leader of my medical team.

6. I feel that the level of attending oversight regarding patient care activities is just right.

Responses to these questions, as well as free-text survey responses, allowed us to identify key drivers to improving the perception of SR autonomy (Fig 1).

We created an Autonomy Toolbox change package informed by our key drivers. This included SR and faculty development sessions, SR and attending shared expectation-setting, and SR-led independent rounding.

In the first plan-do-study-act cycle, SRs and PHM faculty participated in professional development sessions on the SREA-21 framework.¹⁶  This curriculum focuses on empowering the SR to lead the team during patient- and family-centered rounds (PFCR). Attending physicians were encouraged to remain silent, enter the patient room after the SR, create a safe learning environment, and optimize teaching and communication to set the SR up for success. Dissemination of this curriculum was individualized at each site. Some sites supplemented instructional workshops with educational videos about SREA-21. One site developed sessions for faculty focused on potential barriers to increasing SR autonomy, like having a struggling SR or PHM fellow. Others provided visual reminders of SREA-21 tenants via “badge buddies.”

We formalized expectation-setting huddles between the attending and SR to outline roles and responsibilities of team members at the start of a service week, including specific expectations regarding daily operations of the team (teaching, staffing, and communication with families and staff). Some sites used standardized preservice agreement forms for the SR and attending to sign, whereas others used just-in-time e-mail reminders to discuss expectations (Supplemental Fig 5).

We asked PHM faculty to allow SRs to lead rounds without an attending present at least once weekly for a minimum of 2 patients. Creating this broad definition of independent rounding allowed each institution to implement this change at a level that worked for them.

Regular e-mail communication to SR and PHM faculty at the beginning of service weeks provided just-in-time reminders of our initiative and interventions. We shared resources, such as onboarding videos and a 1-page handout for expectation-setting, across the collaborative. Interventions were reinforced periodically during faculty meetings. Our project was supported by the pediatric residency programs at all 5 sites.

Answers to our 6 questions of interest were converted to a 5-point scale corresponding to the percentage of the time the SR reported each question was true (1 = 0% to 20% of the time, 2 = 21% to 40% of the time, etc). We averaged scores from each question to generate the composite RAS. Data from the needs assessment was used to determine the baseline RAS. All SRs who completed at least 1 week on PHM service during the study period were sent surveys with the 6 component questions of the RAS. Surveys were disseminated to SRs via e-mail monthly. We managed surveys and data via REDCap. Site leaders met virtually monthly to discuss institution specific results and pitfalls, and collaboratively solve problems.

Our primary outcome, SR perception of autonomy, was measured via the RAS. We selected a goal RAS of 4.50 by consensus, representing a 25% increase from the baseline. For our process measures, we embedded additional survey questions to determine the degree of uptake of our expectation-setting and independent rounding interventions each month. We gathered feedback from surveys of families and nurses regarding the quality and efficiency of PFCR, if family and nursing issues were adequately addressed, and if families were satisfied with the discussion on PFCR with their team as our balancing measures. Questions were decided by group consensus as to what is important to families and nursing staff. Surveys were collected at time of service either by paper or via quick response code. Additionally, PHM attending physicians were surveyed via e-mail after every service block regarding their perspectives on how efforts to improve resident autonomy affected rounding efficiency, patient safety, and quality of teaching.

We compared attending and SR responses using Fisher’s exact test (R statistical software, version 3.5.2, https://r-project.org). Free text responses regarding perceived barriers to SR autonomy and suggestions to improve autonomy were reviewed, with common themes identified and tallied.

The monthly average RAS was plotted using an X-bar statistical process control chart (Microsoft Excel for Microsoft 365 MSO). The X-bar chart enabled the QI team to define sequential stage means. Calculation of control limits was based on process averages and the within-subgroup estimate of the SD, with 3 σ limits based on a normal distribution of the data.²¹  statistical process control chart principles were used to determine if there was special cause variation; 8 consecutive points below or above the centerline led to a centerline shift.²²  As start dates from the 5 institutions were asynchronous, data were standardized to time from the first intervention at each institution.

This study was approved by the institutional review board at all sites.

At least 1 question of the needs assessment survey was answered by 59% (129 of 220) of PHM faculty and 46% (198 of 427) of SRs. Six questions from the needs assessment were identified as targets of improvement as described above (Fig 2). SR answers to these questions were averaged and combined to produce a baseline RAS of 3.67.

SR and PHM attending physicians identified similar but distinct barriers to autonomy in their free-text responses (Table 1). SR identified attending qualities (early career, micromanages, or “too available”) as biggest barriers to SR autonomy, whereas attending physicians identified a lack of SR skill, knowledge, or motivation as their biggest barriers to providing autonomy. Both groups noted that patient acuity or complexity limited SR autonomy.

Attending physicians and SRs provided comparable suggestions for improving autonomy. Both groups noted expectation-setting between SR and attending and SR-led independent rounding would improve autonomy. SRs also reported that attending physicians designating them as the leader of the team was important for their autonomy. Attending physicians suggested that personal changes, like deliberate positioning or less frequent interrupting on PFCR, could improve SR autonomy.

Forty percent (209 of 522) of SR surveyed answered all 6 questions comprising the RAS during the study period. Special cause variation was noted in the average RAS beginning 1 month after implementing professional development (Fig 3). The new centerline of 4.36 was sustained for the remainder of the study period.

Scores for 5 of the 6 RAS components improved during our study (Supplemental Table 4). The score for the question “I feel that I am encouraged to make medical decisions for straightforward cases on my own” did not change. Scores for questions regarding attending consideration of SR input, attending physicians following SR plans even if preferring an equivalent alternative, and attending oversight being “just right” increased 1 month after implementing professional development sessions. Resident perception as the leader of the team improved 2 months after our first intervention, whereas the reception of faculty feedback improved 7 months postintervention.

During the study period, 72% (144 of 200) of SRs reported their attending set expectations with them at least half of the time, whereas 46% (92 of 200) of SRs reported this was happening 76% to 100% of the time. Of those surveyed, 59% (78 of 132) of SRs reported they were given the opportunity to independently round.

Overall, 98% of families surveyed were satisfied with the discussion with their medical team and felt their concerns were adequately addressed (Table 2). The majority of nursing staff reported that the quality of PFCR was “excellent” or “above average” (60%), SR communication was “excellent” or “above average” (53%), patient and family concerns were adequately addressed (98%), and nursing concerns were adequately addressed (91%). Forty-two percent of nurses also reported that improved SR autonomy positively affected rounding efficiency.

More attending physicians surveyed noted rounding efficiency had been positively affected than negatively affected (24% vs 13%) by efforts to increase SR autonomy. Eighty-two percent of attending physicians reported that our interventions did not affect patient safety, 3% reported interventions negatively impacted patient safety. Twenty-eight percent of attending physicians reported teaching on PFCR was improved by efforts to increase SR autonomy, and 10% reported a decrease in the quality of teaching.

In this multi-institutional QI initiative, we created and implemented an Autonomy Toolbox that led to a sustained increase in SR perception of autonomy on PHM services across 5 pediatric academic centers. Although it is known that individual changes like independent rounding,¹⁹  implementing SREA-21,^16,17  or expectation setting¹⁸  can improve resident autonomy, this is the first study to implement combined interventions using QI methods. The use of the Autonomy Toolbox allowed for institution-specific adaptation and led to an improvement in the perception of SR autonomy, as demonstrated by a 19% increase in the RAS.

We gained an understanding of the drivers of and barriers to SR autonomy through our needs assessment. SR reported barriers included having fellows on service and the culture of pediatric medicine to involve supervisory physicians in nearly all medical decisions. However, the most cited barriers among SRs were attending-specific factors, such as micromanaging, being overly conservative, or being “too available.” Conversely, attending physicians often identified SR-specific factors, such as perceived lack of knowledge, competence, or confidence, as limiting SR autonomy. Similar to previous studies, residents and attending physicians had discordant perceptions of the level of resident autonomy afforded to SRs.^3,9  Surveyed SRs and attending physicians provided suggestions that laid the framework for our interventions. SRs wanted to be designated as the leader of the team; both SRs and attending physicians wanted SRs to independently round and to formally set expectations. The Autonomy Toolbox was created to fit these needs.

The increase in the RAS occurred 1 month after SR and attending physician professional development sessions. We speculate that increasing awareness of the importance of providing SR autonomy can greatly improve the level of autonomy afforded to SR. Framing learning activities in terms of intrinsic goals in an autonomy-supportive environment can increase dedication to learning, as well as learner competence.²³  Our development sessions helped to frame SR autonomy as an estimable goal on the wards. Additionally, the initiation of this QI work may have incentivized participants to increase SR autonomy via the Hawthorne effect.²⁴  Direct observation of SR and PHM faculty dyads at some sites may have heightened this effect and created momentum for our initiative. Neither expectation-setting nor independent rounding directly led to significant increase in the RAS. As education alone typically does not produce a prolonged change in behavior, it is likely that subsequent interventions were necessary to sustain the initial increase in the perception of SR autonomy that otherwise would have faltered.

Some factors likely blunted the response to our interventions. Not all PHM faculty-SR dyads participated in expectation-setting or independent rounds. Independent rounds have been shown to promote critical thinking, self-reflection, and patient ownership and SRs report them as beneficial to their learning.⁷  At institutions where independent rounding was a new practice, uptake was low and hesitation to change was high. Conversely, those institutions with a form of independent rounding before our initiative more readily accepted SRs independently rounding. Attending physicians may be able to decrease the perceived burden by rounding in parallel a few patients behind the SR-led team (also called “shadow rounding,” Supplemental Fig 6) or by seeing early discharges independently.

All but 1 of the domains of the RAS improved during our study. SR perception of faculty feedback improved, despite having no targeted intervention. We hypothesize that because SRs were encouraged to be primarily responsible for team education, family communication, and patient care decisions, attending physicians were better able to provide quality feedback on how SRs performed in these areas. Many SRs particularly noted that expectation-setting at the onset of their service week improved their perception of autonomy; they reported feeling more confident in their decisions and in their ability to voice opinions after expectation-setting.

SRs described their autonomy on PHM as higher than on other pediatric inpatient services at baseline. Previous literature supports the idea that hospitalists typically provide more autonomy than subspecialists.²⁵  We theorize the Autonomy Toolbox may further increase SR perception of autonomy in areas that would not have as high of a baseline RAS. The adaptability of the tool allows it to be applied in a variety of settings and institutions.

This study is not without limitations. First, this study was performed at large academic children’s hospitals and may not be generalizable to other hospital settings. Additionally, the focus of this work was to increase SR autonomy and may not be as applicable to other learners. Our findings relied on voluntary survey responses and may not reflect the perspectives of all SR, especially given the response rate of 40%, which overrepresented 2 of our sites. Our primary measurement was of SR perception, which is subjective. Although comprised of many participants across institutions and collected across several months, our baseline RAS was formed by 1 data point. Our RAS does not reflect all relevant domains of resident autonomy, but highlighted areas for improvement. Further validation of the RAS and its parent questions is warranted.

Our interventions spanned 2 academic years. “New” SRs were likely granted less autonomy in July, which may have decreased the response to our interventions. Post graduate year (PGY) affected the perception of SR autonomy. On analysis of the RAS by PGY year, PGY-2 SR had a higher baseline RAS than PGY-3 and 4 SR, though both groups had a similar trajectory and endpoint that mirrored the total RAS. Additionally, low patient volumes at the start of the 2019 coronavirus pandemic and high patient volumes because of surges in pediatric viral illnesses during the summer of 2021 may have confounded the response to our interventions.

Although we obtained nurse, patient, and attending survey data about how SR-led rounds affected care, we were unable to collect baseline data for these measures and our survey tools were not validated. Family and nurse response rates were low, therefore these efforts were not continued at every institution throughout the entire study period. Survey questions did not have alternate options to “yes” or “no” regarding family satisfaction; this may have positively skewed our results. However, all families included in our survey participated in PFCR, which may lead to higher family satisfaction.²⁶ 

We created a novel and adaptable change package to sustainably increase the perception of SR autonomy. We identified drivers of and barriers to SR autonomy. When implemented together, professional development, expectation-setting, and independent rounding are effective tools to increase the perception of resident autonomy. The Autonomy Toolbox can be adapted to a variety of needs to help SR become more autonomous and better prepare SRs for independent practice. Possible areas of future study include further examination of how increasing SR autonomy impacts patient safety, family communication, and medical education.

We thank Kelsey Porada MA, CCRP for her instrumental help with survey distribution, collection, and organization.