Academic Medicine最新文献

Problem: Although the use of artificial intelligence (AI) as a diagnostic aid is increasing in clinical practice, medical education provides little training on how to incorporate AI-generated information into diagnosis and use it effectively in shared decision-making (SDM) with patients.

Approach: The authors developed and piloted a simulation-based course to train AI-assisted SDM to final-year medical students preparing for residency. Conducted between June and October 2023, the course combined online prelearning with onsite simulations using clinically approved AI tools (Lunit INSIGHT CXR, version 3.1.4.1 and MMG, version 1.1.4.3; Lunit Inc, Seoul, South Korea; used November 16 and 27, 2023). Scenarios portrayed asymptomatic patients with incidental findings (eg, pulmonary nodules, breast microcalcifications). Students engaged in two 12-minute simulated patient encounters featuring SDM with 2 management options. Sessions concluded with simulated patient-written feedback and expert-facilitated debriefing. Twenty-seven students from 3 medical schools participated.

Outcomes: Program evaluation showed significant improvements in participants' comprehension and confidence in SDM (t = 6.51 and t = 7.56, P < .001, respectively) and AI-assisted SDM (t = 5.72 and t = 5.80, P < .001, respectively). Students found AI tools helpful for facilitating SDM and patient communication. Thematic analysis of interviews highlighted strengths, such as structured course design and reflective debriefing. Participants noted that prior education focused on diagnostic algorithms, whereas this course emphasized patient communication and preference-based decisions. They found AI tools useful for diagnosis and supporting discussion with patients through visual outputs. However, they identified limitations, including their own clinical knowledge gaps and lack of explainability in AI tool shortage. They suggested integrating SDM and AI-assisted diagnosis training into formal curricula to better prepare students for clinical practice.

Next steps: Future efforts should focus on integrating this course into undergraduate curricula or transition training programs to provide experiential learning opportunities in AI-assisted clinical practice.

Purpose: Shame is a deeply personal, complex, and underexplored emotion in medical training; however, how medical learners engage with shame (ie, how they process, recover from, and/or address shame) and the environmental factors affecting this process are currently unknown. This study used hermeneutic phenomenology to explore how medical learners (ie, resident physicians and medical students) engage with shame once it has occurred and the factors that influence this engagement.

Method: This study, which is part of a qualitative research program addressing shame in medical learners, used data collected from 12 residents (2016 and 2017) and 16 medical students (2018) from a residency program and a medical school in the United States. Data collection occurred via semistructured interviews during which participants reflected on shame experiences, including their engagement with it. The authors selected 14 transcripts (7 medical students, 7 residents) to achieve a range of shame experiences and impacts. Data were analyzed using Ajjawi and Higgs' 6 steps of hermeneutic analysis.

Results: Internal scaffolding (thought processes, self-evaluative tendencies, and position relative to others that informed participants' self-concept) was central to shame engagement. Learners' internal scaffoldings shaped and were shaped by distressing shame-integrating engagement (ie, hiding the self, deflecting shame, and transferring shame) and constructive shame-disintegrating engagement (ie, orienting toward others, exerting agency over self-evaluation, and reorienting to a core sense of self). Learning environments influenced shame engagement; environmental values that promoted shame-disintegrating engagement included learner centeredness, inclusivity, vulnerability, and respect.

Conclusions: Although struggle in medical training is inevitable, how learners respond to the shame that can follow is not. The divergent nature of shame engagement highlights the importance of learner agency and environmental response to shame. The authors provide specific suggestions for learners, faculty, and leaders to advance constructive shame engagement and the growth, connection, and belonging it can inspire.

Purpose: Entrustable professional activities (EPAs) detail essential activities within a given specialty. Although 17 general pediatrics EPAs have been defined, it is not known how many are needed to make high-reliability overall entrustment decisions about resident readiness for practice at the time of graduation and initial certification. This study sought to determine how many general pediatrics EPAs are needed.

Method: During the 2021 to 2022, 2022 to 2023, and 2023 to 2024 academic years, the authors collected entrustment-supervision levels, determined by clinical competency committees biannually, for the 17 general pediatrics EPAs for residents at 48 U.S. pediatric residency training programs. Midyear reports were collected between November and January of each year, and end-of-year reports were collected between May and July. The authors conducted generalizability and decision studies to determine the number of EPAs needed to make a reliable overall entrustment decision.

Results: A total of 166,077 individual entrustment-supervision levels were collected for 4,250 pediatric residents across the 17 general pediatrics EPAs. Across all data reporting cycles, the authors found that assessing 6 EPAs yields a generalizability coefficient of 0.8 and assessing 12 EPAs yields a generalizability coefficient of 0.9. However, results differed for midyear compared with end-of-year data collection timepoints as well as by postgraduate year. At graduation, 9 to 13 EPAs are needed to make a highly reliable (generalizability coefficient of 0.9) overall decision about degree of entrustment for unsupervised practice.

Conclusions: This study provides rich insight into the number of EPAs needed to make reliable entrustment decisions about resident readiness to provide patient care. Although readiness can be determined with as few as 9 general pediatrics EPAs (an assessment task), more may be needed to inform a comprehensive curriculum that ensures focus in all areas important to developing general pediatricians during residency training (a curricular task).Teaser text: This study sought to determine how many entrustable professional activities are necessary to make high reliability overall entrustment decisions about pediatric resident readiness for unsupervised practice.

Purpose: In 2003, the Indiana University School of Medicine launched the Leadership in Academic Medicine Program (LAMP), a cohort-based leadership program for early-career faculty. This study evaluated the program's effectiveness in faculty retention, promotion, and attainment of leadership roles.

Method: The authors identified early-career faculty hired at the assistant professor rank between 2000 and 2017 who were eligible for LAMP (2003-2017). Propensity score matching was used to create comparable groups of LAMP participants and matched controls (nonparticipants) based on track, department type, age at hire, hire year, sex, and underrepresented in medicine status. Survival analyses, including Kaplan-Meier curves and Cox proportional hazards models, were conducted to compare time to promotion and faculty departure between the matched groups. Logistic regression was used to evaluate differences between groups and their likelihood of achieving leadership roles.

Results: Among the 1,329 early-career faculty included in the study, 443 LAMP participants demonstrated higher rates of promotion (hazard ratio [HR], 1.57; 95% CI, 1.31-1.88; P < .001) and significantly lower rates of departure (HR, 0.72; 95% CI, 0.60-0.85; P < .001) compared with 886 matched controls. Additionally, LAMP participants were significantly more likely to attain a leadership role (odds ratio [OR], 2.02; 95% CI, 1.49-2.73; P < .001). The positive impact of LAMP was especially pronounced among nontenure-track faculty, with notable improvements observed in promotion, retention, and leadership attainment. Although both male and female faculty benefited from participation, women had particularly increased odds of securing leadership positions (OR, 2.49; 95% CI, 1.59-3.87; P < .001).

Conclusions: Participation in LAMP significantly improved promotion rates, retention, and leadership attainment, particularly for nontenure-track and female faculty. These findings demonstrate the program's effectiveness in supporting faculty development and career advancement.

Problem: Medical educators pay increasing attention to the potential utility of narrative data for assessment, but lack of efficient and standardized ways of interpreting the data have limited its use. Natural language processing (NLP) algorithms could provide new insights for using narrative data within assessment especially identifying at-risk or low-performing students.

Approach: Assessment data were reviewed from 16 cohorts of medical students from the University of Cincinnati College of Medicine (graduating classes of 2006-2022). A T-score Average (TSA) was calculated for each student based on clerkship assessment data. Narrative data from core clerkship evaluations in responses to the prompt "any opportunities for improvement" were utilized for analysis. The narrative data and calculated TSA were then used to train and test 4 NLP models with the goal of utilizing NLP to identify at-risk students as defined by the bottom 10% average TSA score.

Outcomes: Based on typical NLP methods, the developed NLP models all performed adequately in identifying student performance with an overall accuracy of 0.8 across all 4 models. However, none of the NLP models was able to identify students within the bottom 10% of performance. During this process, we uncovered the presence of "copy/paste" behavior, a previously undocumented phenomenon within narrative data where preceptors duplicated comments from 1 student to the next. Training NLP models including "copy/paste" comments improved NLP ability to identify students within the bottom and top 10% of performance.

Next steps: NLP models were unable to accurately identify at-risk students. Model accuracy increased with the inclusion of "copy/paste" comments, indicating there may be some discriminatory functionality within aspects of narrative data beyond keywords such as lexical diversity and word quantity. Future work will explore how these other aspects of narrative data associate with performance and utilize more advanced large language models of analysis.Based on typical natural language processing (NLP) methods, the tested NLP models performed adequately in identifying student performance with an overall accuracy of 0.8 across all models. However, none of the NLP models was able to identify students within the bottom 10% of performance, revealing widespread "copy/paste" behavior in evaluations that may limit narrative data's discriminatory utility.