A Big Data and Learning Analytics Approach to Process-Level Feedback in Cognitive Simulations

Validating instructional design and predicting student performance in histology education: Using machine learning via virtual microscopy

As a part of modern technological environments, virtual microscopy enriches histological learning, with support from large institutional investments. However, existing literature does not supply empirical evidence of its role in improving pedagogy. Virtual microscopy provides fresh opportunities for investigating user behavior during the histology learning process, through digitized histological slides. This study establishes how students' perceptions and user behavior data can be processed and analyzed using machine learning algorithms. These also provide predictive data called learning analytics that enable predicting students' performance and behavior favorable for academic success. This information can be interpreted and used for validating instructional designs. Data on the perceptions, performances, and user behavior of 552 students enrolled in a histology course were collected from the virtual microscope, Cytomine®. These data were analyzed using an ensemble of machine learning algorithms, the extra-tree regression method, and predictive statistics. The predictive algorithms identified the most pertinent histological slides and descriptive tags, alongside 10 types of student behavior conducive to academic success. We used these data to validate our instructional design, and align the educational purpose, learning outcomes, and evaluation methods of digitized histological slides on Cytomine®. This model also predicts students' examination scores, with an error margin of <0.5 out of 20 points. The results empirically demonstrate the value of a digital learning environment for both students and teachers of histology.

Using Learning Analytics to Evaluate the Clinical Education Podcast Format

Learning analytics has been rigorously applied to various forms of e-learning, but not to the evaluation of clinical education podcasts. We created a podcast series using the Anchor by Spotify platform, uploading an episode every 2 weeks starting on March 30, 2021. We examined analytics data using a censoring date of November 1, 2022. Based on 29,747 total plays, average audience retention declined 10%/minute until an inflection point at 2.5 minutes, followed by a steady decline of 1.8%/minute. With a maximum episode length of 17 minutes, we did not identify a limit on learner attention for short-form podcasts.

Twelve Tips for using Learning Curves in Health Professions Education Research

Learning curves can be used to design, implement, and evaluate educational interventions. Attention to key aspects of the method can improve the fidelity of this representation of learning as well as its suitability for education and research purposes. This paper addresses when to use a learning curve, which graphical properties to consider, how to use learning curves quantitatively, and how to use observed thresholds to communicate meaning. We also address the associated ethics and policy considerations. We conclude with a best practices checklist for both educators and researchers seeking to use learning curves in their work.

Interactive computer-assisted learning as an educational method for learning pediatric interproximal dental caries identification

OBJECTIVE

We developed a web-based tool to measure the amount and rate of skill acquisition in pediatric interproximal caries diagnosis among pre- and postdoctoral dental students and identified variables predictive for greater image interpretation difficulty.

STUDY DESIGN

In this multicenter prospective cohort study, a convenience sample of pre- and postdoctoral dental students participated in computer-assisted learning in the interpretation of bitewing radiographs of 193 children. Participants were asked to identify the presence or absence of interproximal caries and, where applicable, locate the lesions. After every case, participants received specific visual and text feedback on their diagnostic performance. They were requested to complete the 193-case set but could complete enough cases to achieve a competency performance standard of 75% accuracy, sensitivity, and specificity.

RESULTS

Of 130 participants, 62 (47.7%) completed all cases. The mean change from initial to maximal diagnostic accuracy was +15.3% (95% CI, 13.0-17.7), sensitivity was +10.8% (95% CI, 9.0-12.7), and specificity was +15.5% (95% CI, 12.9-18.1). The median number of cases completed to achieve competency was 173 (interquartile range, 82-363). Of these 62 participants, 45 (72.6%) showed overall improvement in diagnostic accuracy. Greater numbers of interproximal lesions (P < .001) and the presence of noninterproximal caries (P < .001) predicted greater interpretation difficulty.

CONCLUSIONS

Computer-assisted learning led to improved diagnosis of interproximal caries on bitewing radiographs among pre- and postdoctoral dental students.

Empowering Health Care Education Through Learning Analytics: In-depth Scoping Review

BACKGROUND

Digital education has expanded since the COVID-19 pandemic began. A substantial amount of recent data on how students learn has become available for learning analytics (LA). LA denotes the "measurement, collection, analysis, and reporting of data about learners and their contexts, for purposes of understanding and optimizing learning and the environments in which it occurs."

OBJECTIVE

This scoping review aimed to examine the use of LA in health care professions education and propose a framework for the LA life cycle.

METHODS

We performed a comprehensive literature search of 10 databases: MEDLINE, Embase, Web of Science, ERIC, Cochrane Library, PsycINFO, CINAHL, ICTP, Scopus, and IEEE Explore. In total, 6 reviewers worked in pairs and performed title, abstract, and full-text screening. We resolved disagreements on study selection by consensus and discussion with other reviewers. We included papers if they met the following criteria: papers on health care professions education, papers on digital education, and papers that collected LA data from any type of digital education platform.

RESULTS

We retrieved 1238 papers, of which 65 met the inclusion criteria. From those papers, we extracted some typical characteristics of the LA process and proposed a framework for the LA life cycle, including digital education content creation, data collection, data analytics, and the purposes of LA. Assignment materials were the most popular type of digital education content (47/65, 72%), whereas the most commonly collected data types were the number of connections to the learning materials (53/65, 82%). Descriptive statistics was mostly used in data analytics in 89% (58/65) of studies. Finally, among the purposes for LA, understanding learners' interactions with the digital education platform was cited most often in 86% (56/65) of papers and understanding the relationship between interactions and student performance was cited in 63% (41/65) of papers. Far less common were the purposes of optimizing learning: the provision of at-risk intervention, feedback, and adaptive learning was found in 11, 5, and 3 papers, respectively.

CONCLUSIONS

We identified gaps for each of the 4 components of the LA life cycle, with the lack of an iterative approach while designing courses for health care professions being the most prevalent. We identified only 1 instance in which the authors used knowledge from a previous course to improve the next course. Only 2 studies reported that LA was used to detect at-risk students during the course's run, compared with the overwhelming majority of other studies in which data analysis was performed only after the course was completed.

Physician Training for Electrocardiogram Interpretation: A Systematic Review and Meta-Analysis

PURPOSE

Using electrocardiogram (ECG) interpretation as an example of a widely taught diagnostic skill, the authors conducted a systematic review and meta-analysis to demonstrate how research evidence on instruction in diagnosis can be synthesized to facilitate improvement of educational activities (instructional modalities, instructional methods, and interpretation approaches), guide the content and specificity of such activities, and provide direction for research.

METHOD

The authors searched PubMed/MEDLINE, Embase, Cochrane CENTRAL, PsycInfo, CINAHL, ERIC, and Web of Science databases through February 21, 2020, for empirical investigations of ECG interpretation training enrolling medical students, residents, or practicing physicians. They appraised study quality with the Medical Education Research Study Quality Instrument and pooled standardized mean differences (SMDs) using random effects meta-analysis.

RESULTS

Of 1,002 articles identified, 59 were included (enrolling 17,251 participants). Among 10 studies comparing instructional modalities, 8 compared computer-assisted and face-to-face instruction, with pooled SMD 0.23 (95% CI, 0.09, 0.36) indicating a small, statistically significant difference favoring computer-assisted instruction. Among 19 studies comparing instructional methods, 5 evaluated individual versus group training (pooled SMD -0.35 favoring group study [95% CI, -0.06, -0.63]), 4 evaluated peer-led versus faculty-led instruction (pooled SMD 0.38 favoring peer instruction [95% CI, 0.01, 0.74]), and 4 evaluated contrasting ECG features (e.g., QRS width) from 2 or more diagnostic categories versus routine examination of features within a single ECG or diagnosis (pooled SMD 0.23 not significantly favoring contrasting features [95% CI, -0.30, 0.76]). Eight studies compared ECG interpretation approaches, with pooled SMD 0.92 (95% CI, 0.48, 1.37) indicating a large, statistically significant effect favoring more systematic interpretation approaches.

CONCLUSIONS

Some instructional interventions appear to improve learning in ECG interpretation; however, many evidence-based instructional strategies are insufficiently investigated. The findings may have implications for future research and design of training to improve skills in ECG interpretation and other types of visual diagnosis.

Learning Pediatric Point-of-Care Ultrasound: How Many Cases Does Mastery of Image Interpretation Take?

OBJECTIVES

Using an education and assessment tool, we examined the number of cases necessary to achieve a performance benchmark in image interpretation of pediatric soft tissue, cardiac, lung, and focused assessment with sonography for trauma (FAST) point-of-care ultrasound (POCUS) applications. We also determined interpretation difficulty scores to derive which cases provided the greatest diagnostic challenges.

METHODS

Pediatric emergency physicians participated in web-based pediatric POCUS courses sponsored by their institution as a credentialing priority. Participants deliberately practiced cases until they achieved diagnostic interpretation scores of combined 90% accuracy, sensitivity, and specificity.

RESULTS

Of the 463 who enrolled, 379 (81.9%) completed cases. The median (interquartile range) number of cases required to achieve the performance benchmark for soft tissue was 94 (68-128); cardiac, 128 (86-201); lung, 87 (25-118); and FAST, 93 (68-133) (P < 0001). Specifically, cases completed to achieve benchmark were higher for cardiac relative to other applications (P < 0.0001 for all comparisons). In soft tissue cases, a foreign body was more difficult to diagnose than cobblestoning and hypoechoic collections (P = 0.036). Poor cardiac function and abnormal ventricles were more difficult to interpret with accuracy than normal (P < 0.0001) or pericardial effusion cases (P = 0.01). The absence of lung sliding was significantly more difficult to interpret than normal lung cases (P = 0.028). The interpretation difficulty of various FAST imaging findings was not significantly different.

CONCLUSIONS

There was a significant variation in number of cases required to reach a performance benchmark. We also identified the specific applications and imaging findings that demonstrated the greatest diagnostic challenges. These data may inform future credentialing guidelines and POCUS learning interventions.

Adopting Learning Analytics in a First-Year Veterinarian Professional Program: What We Could Know in Advance about Student Learning Progress

Over the past decade, the field of education has seen stark changes in the way that data are collected and leveraged to support high-stakes decision-making. Utilizing big data as a meaningful lens to inform teaching and learning can increase academic success. Data-driven research has been conducted to understand student learning performance, such as predicting at-risk students at an early stage and recommending tailored interventions to support services. However, few studies in veterinary education have adopted Learning Analytics. This article examines the adoption of Learning Analytics by using the retrospective data from the first-year professional Doctor of Veterinary Medicine program. The article gives detailed examples of predicting six courses from week 0 (i.e., before the classes started) to week 14 in the semester of Spring 2018. The weekly models for each course showed the change of prediction results as well as the comparison between the prediction results and students' actual performance. From the prediction models, at-risk students were successfully identified at the early stage, which would help inform instructors to pay more attention to them at this point.

Multi-level longitudinal learning curve regression models integrated with item difficulty metrics for deliberate practice of visual diagnosis: groundwork for adaptive learning

Visual diagnosis of radiographs, histology and electrocardiograms lends itself to deliberate practice, facilitated by large online banks of cases. Which cases to supply to which learners in which order is still to be worked out, with there being considerable potential for adapting the learning. Advances in statistical modeling, based on an accumulating learning curve, offer methods for more effectively pairing learners with cases of known calibrations. Using demonstration radiograph and electrocardiogram datasets, the advantages of moving from traditional regression to multilevel methods for modeling growth in ability or performance are demonstrated, with a final step of integrating case-level item-response information based on diagnostic grouping. This produces more precise individual-level estimates that can eventually support learner adaptive case selection. The progressive increase in model sophistication is not simply statistical but rather brings the models into alignment with core learning principles including the importance of taking into account individual differences in baseline skill and learning rate as well as the differential interaction with cases of varying diagnosis and difficulty. The developed approach can thus give researchers and educators a better basis on which to anticipate learners' pathways and individually adapt their future learning.

Expertise development in volumetric image interpretation of radiology residents: what do longitudinal scroll data reveal?

The current study used theories on expertise development (the holistic model of image perception and the information reduction hypothesis) as a starting point to identify and explore potentially relevant process measures to monitor and evaluate expertise development in radiology residency training. It is the first to examine expertise development in volumetric image interpretation (i.e., CT scans) within radiology residents using scroll data collected longitudinally over five years of residency training. Consistent with the holistic model of image perception, the percentage of time spent on full runs, i.e. scrolling through more than 50% of the CT-scan slices (global search), decreased within residents over residency training years. Furthermore, the percentage of time spent on question-relevant areas in the CT scans increased within residents over residency training years, consistent with the information reduction hypothesis. Second, we examined if scroll patterns can predict diagnostic accuracy. The percentage of time spent on full runs and the percentage of time spent on question-relevant areas did not predict diagnostic accuracy. Thus, although scroll patterns over training years are consistent with visual expertise theories, they could not be used as predictors of diagnostic accuracy in the current study. Therefore, the relation between scroll patterns and performance needs to be further examined, before process measures can be used to monitor and evaluate expertise development in radiology residency training.

Image interpretation: Learning analytics-informed education opportunities

OBJECTIVES

Using a sample of pediatric chest radiographs (pCXR) taken to rule out pneumonia, we obtained diagnostic interpretations from physicians and used learning analytics to determine the radiographic variables and participant review processes that predicted for an incorrect diagnostic interpretation.

METHODS

This was a prospective cross-sectional study. A convenience sample of frontline physicians with a range of experience levels interpreted 200 pCXR presented using a customized online radiograph presentation platform. Participants were asked to determine absence or presence (with respective location) of pneumonia. The pCXR were categorized for specific image-based variables potentially associated with interpretation difficulty. We also generated heat maps displaying the locations of diagnostic error among normal pCXR. Finally, we compared image review processes in participants with higher versus lower levels of clinical experience.

RESULTS

We enrolled 83 participants (20 medical students, 40 postgraduate trainees, and 23 faculty) and obtained 12,178 case interpretations. Variables that predicted for increased pCXR interpretation difficulty were pneumonia versus no pneumonia (β = 8.7, 95% confidence interval [CI] = 7.4 to 10.0), low versus higher visibility of pneumonia (β = -2.2, 95% CI = -2.7 to -1.7), nonspecific lung pathology (β = 0.9, 95% CI = 0.40 to 1.5), localized versus multifocal pneumonia (β = -0.5, 95% CI = -0.8 to -0.1), and one versus two views (β = 0.9, 95% CI = 0.01 to 1.9). A review of diagnostic errors identified that bony structures, vessels in the perihilar region, peribronchial thickening, and thymus were often mistaken for pneumonia. Participants with lower experience were less accurate when they reviewed one of two available views (p < 0.0001), and accuracy of those with higher experience increased with increased confidence in their response (p < 0.0001).

CONCLUSIONS

Using learning analytics, we identified actionable learning opportunities for pCXR interpretation, which can be used to allow for a customized weighting of which cases to practice. Furthermore, experienced-novice comparisons revealed image review processes that were associated with greater diagnostic accuracy, providing additional insight into skill development of image interpretation.

Child Abuse Recognition Training for Prehospital Providers Using Deliberate Practice

BACKGROUND

In most states, prehospital professionals (PHPs) are mandated reporters of suspected abuse but cite a lack of training as a challenge to recognizing and reporting physical abuse. We developed a learning platform for the visual diagnosis of pediatric abusive versus non-abusive burn and bruise injuries and examined the amount and rate of skill acquisition.

METHODS

This was a prospective cross-sectional study of PHPs participating in an online educational intervention containing 114 case vignettes. PHPs indicated whether they believed a case was concerning for abuse and would report a case to child protection services. Participants received feedback after submitting a response, permitting deliberate practice of the cases. We describe learning curves, overall accuracy, sensitivity (diagnosis of abusive injuries) and specificity (diagnosis of non-abusive injuries) to determine the amount of learning. We performed multivariable regression analysis to identify specific demographic and case variables associated with a correct case interpretation. After completing the educational intervention, PHPs completed a self-efficacy survey on perceived gains in their ability to recognize cutaneous signs of abuse and report to social services.

RESULTS

We enrolled 253 PHPs who completed all the cases; 158 (63.6%) emergency medical technicians (EMT), 95 (36.4%) advanced EMT and paramedics. Learning curves demonstrated that, with one exception, there was an increase in learning for participants throughout the educational intervention. Mean diagnostic accuracy increased by 4.9% (95% CI 3.2, 6.7), and the mean final diagnostic accuracy, sensitivity, and specificity were 82.1%, 75.4%, and 85.2%, respectively. There was an increased odds of getting a case correct for bruise versus burn cases (OR = 1.4; 95% CI 1.3, 1.5); if the PHP was an Advanced EMT/Paramedic (OR = 1.3; 95% CI 1.1, 1.4) ; and, if the learner indicated prior training in child abuse (OR = 1.2; 95% CI 1.0, 1.3). Learners indicated increased comfort in knowing which cases should be reported and interpreting exams in children with cutaneous injuries with a median Likert score of 5 out of 6 (IQR 5, 6).

CONCLUSION

An online module utilizing deliberate practice led to measurable skill improvement among PHPs for differentiating abusive from non-abusive burn and bruise injuries.

A think-aloud study to inform the design of radiograph interpretation practice

Models for diagnostic reasoning in radiology have been based on the observed behaviors of experienced radiologists but have not directly focused on the thought processes of novices as they improve their accuracy of image interpretation. By collecting think-aloud verbal reports, the current study was designed to investigate differences in specific thought processes between medical students (novices) as they learn and radiologists (experts), so that we can better design future instructional environments. Seven medical students and four physicians with radiology training were asked to interpret and diagnose pediatric elbow radiographs where fracture is suspected. After reporting their diagnosis of a case, they were given immediate feedback. Participants were asked to verbalize their thoughts while completing the diagnosis and while they reflected on the provided feedback. The protocol analysis of their verbalizations showed that participants used some combination of four processes to interpret the case: gestalt interpretation, purposeful search, rule application, and reasoning from a prior case. All types of processes except reasoning from a prior case were applied significantly more frequently by experts. Further, gestalt interpretation was used with higher frequency in abnormal cases while purposeful search was used more often for normal cases. Our assessment of processes could help guide the design of instructional environments with well-curated image banks and analytics to facilitate the novice's journey to expertise in image interpretation.

The Variable Journey in Learning to Interpret Pediatric Point-of-care Ultrasound Images: A Multicenter Prospective Cohort Study

OBJECTIVES

To complement bedside learning of point-of-care ultrasound (POCUS), we developed an online learning assessment platform for the visual interpretation component of this skill. This study examined the amount and rate of skill acquisition in POCUS image interpretation in a cohort of pediatric emergency medicine (PEM) physician learners.

METHODS

This was a multicenter prospective cohort study. PEM physicians learned POCUS using a computer-based image repository and learning assessment system that allowed participants to deliberately practice image interpretation of 400 images from four pediatric POCUS applications (soft tissue, lung, cardiac, and focused assessment sonography for trauma [FAST]). Participants completed at least one application (100 cases) over a 4-week period.

RESULTS

We enrolled 172 PEM physicians (114 attendings, 65 fellows). The increase in accuracy from the initial to final 25 cases was 11.6%, 9.8%, 7.4%, and 8.6% for soft tissue, lung, cardiac, and FAST, respectively. For all applications, the average learners (50th percentile) required 0 to 45, 25 to 97, 66 to 175, and 141 to 290 cases to reach 80, 85, 90, and 95% accuracy, respectively. The least efficient (95th percentile) learners required 60 to 288, 109 to 456, 160 to 666, and 243 to 1040 cases to reach these same accuracy benchmarks. Generally, the soft tissue application required participants to complete the least number of cases to reach a given proficiency level, while the cardiac application required the most.

CONCLUSIONS

Deliberate practice of pediatric POCUS image cases using an online learning and assessment platform may lead to skill improvement in POCUS image interpretation. Importantly, there was a highly variable rate of achievement across learners and applications. These data inform our understanding of POCUS image interpretation skill development and could complement bedside learning and performance assessments.

Deliberate practice as an educational method for learning to interpret the prepubescent female genital examination

BACKGROUND

Correct interpretation of the prepubescent female genital examination is a critical skill; however, physician skill in this area is limited.

OBJECTIVE

To complement the bedside learning of this examination, we developed a learning platform for the visual diagnosis of the prepubescent female genital examination and examined the amount and rate of skill acquisition.

PARTICIPANTS AND SETTING

Medical students, residents, and fellows and attendings participated in an on-line learning platform.

METHODS

This was a multicenter prospective cross-sectional study. Study participants deliberately practiced 158 prepubescent female genital examination cases hosted on a computer-based learning and assessment platform. Participants assigned the case normal or abnormal; if abnormal, they identified the location of the abnormality and the specific diagnosis. Participants received feedback after every case.

RESULTS

We enrolled 107 participants (26 students, 31 residents, 24 fellows and 26 attendings). Accuracy (95 % CI) increased by 10.3 % (7.8, 12.8), Cohen's d-effect size of 1.17 (1.14, 1.19). The change in specificity was +16.8 (14.1, 19.5) and sensitivity +2.4 (-0.9, 5.6). It took a mean (SD) 46.3 (32.2) minutes to complete cases. There was no difference between learner types with respect to initial (p = 0.2) or final accuracy (p = 0.4) scores.

CONCLUSIONS

This study's learning intervention led to effective and feasible skill improvement. However, while participants improved significantly with normal cases, which has relevance in reducing unnecessary referrals to child protection teams, learning gains were not as evident in abnormal cases. All levels of learners demonstrated a similar performance, emphasizing the need for this education even among experienced clinicians.

The Role of Active Engagement of Peer Observation in the Acquisition of Surgical Skills in Virtual Reality Tasks for Novices

OBJECTIVE

Peer-assisted learning has been regarded as an adjunct to teaching modalities. It remains inconclusive regarding the benefits of peer observation in skills learning. Hence, we investigated whether the active engagement (AE) of peer observation in addition to expert demonstration would facilitate the performance in the virtual reality (VR) tasks.

SETTING/DESIGN

The programs involved 4 VR tasks including basic (camera targeting), intermediate (energy dissection and energy switching), and advanced (suture sponge) tasks in the da Vinci Skills Simulators, which were set up in the operating room at Taipei Medical University Hospital. Fifty medical students participated in the study. The AE of the participants was defined as the total number of peer observations in addition to expert observation before their performance. We assessed the correlations between AE and surgical task performance using Pearson correlation and the concept of learning analytics.

PARTICIPANTS

Medical students (sixth-year students in Taiwan, equivalent to fourth-year students in the US system) from Taipei Medical University were recruited.

RESULTS

AE was correlated with the energy dissection task (r = 0.329, p = 0.02) and marginally associated with the energy switching task (r = 0.271, p = 0.057). However, AE was not correlated with either task scores for camera targeting (r = 0.096, p = 0.509) or task scores for suture sponge (r = -0.091, p = 0.529).

CONCLUSIONS

Our findings suggest that AE of peer observation may facilitate learning energy dissection task, which is an intermediate-level task, but not in other basic or advanced tasks in a VR context. The study highlights the potential effect of AE of peer observation on surgical learning based on a distinct level of tasks. Tasks that fit the learners' level are recommended. Nevertheless, the effectiveness of peer observation on surgical training still has to be explored to ensure favorable results and optimal learning outcomes.

Building Emergency Medicine Trainee Competency in Pediatric Musculoskeletal Radiograph Interpretation: A Multicenter Prospective Cohort Study

OBJECTIVES

As residency programs transition from time- to performance-based competency standards, validated tools are needed to measure performance-based learning outcomes and studies are required to characterize the learning experience for residents. Since pediatric musculoskeletal (MSK) radiograph interpretation can be challenging for emergency medicine trainees, we introduced Web-based pediatric MSK radiograph learning system with performance endpoints into pediatric emergency medicine (PEM) fellowships and determined the feasibility and effectiveness of implementing this intervention.

METHODS

This was a multicenter prospective cohort study conducted over 12 months. The course offered 2,100 pediatric MSK radiographs organized into seven body regions. PEM fellows diagnosed each case and received feedback after each interpretation. Participants completed cases until they achieved a performance benchmark of at least 80% accuracy, sensitivity, and specificity. The main outcome measure was the median number of cases completed by participants to achieve the performance benchmark.

RESULTS

Fifty PEM fellows from nine programs in the US and Canada participated. There were 301 of 350 (86%) modules started and 250 of 350 (71%) completed to the predefined performance benchmark during the study period. The median (interquartile range [IQR]) number of cases to performance benchmark per participant was 78 (60-104; min = 56, max = 1,333). Between modules, the median number of cases to achieve the performance benchmark was different for the ankle versus other modules (ankle 366 vs. other 76; difference = 290, 95% confidence interval [CI] = 245 to 335). The performance benchmark was achieved for 90.7% of participants in all modules except the ankle/foot, where 34.9% achieved this goal (difference = 55.8%, 95% CI = 45.3 to 66.3). The mean (95% CI) change in accuracy, sensitivity, and specificity from baseline to performance benchmark was +14.6% (13.4 to 15.8), +16.5% (14.8 to 18.1), and +12.6% (10.7 to 14.5), respectively. Median (IQR) time on each case was 31.0 (21.0-45.3) seconds.

CONCLUSIONS

Most participants completed the modules to the performance benchmark within 1 hour and demonstrated significant skill improvement. Further, there was a large variation in the number of cases completed to achieve the performance endpoint in any given module, and this impacted the feasibility of completing specific modules.

How well is each learner learning? Validity investigation of a learning curve-based assessment approach for ECG interpretation

Learning curves can support a competency-based approach to assessment for learning. When interpreting repeated assessment data displayed as learning curves, a key assessment question is: "How well is each learner learning?" We outline the validity argument and investigation relevant to this question, for a computer-based repeated assessment of competence in electrocardiogram (ECG) interpretation. We developed an on-line ECG learning program based on 292 anonymized ECGs collected from an electronic patient database. After diagnosing each ECG, participants received feedback including the computer interpretation, cardiologist's annotation, and correct diagnosis. In 2015, participants from a single institution, across a range of ECG skill levels, diagnosed at least 60 ECGs. We planned, collected and evaluated validity evidence under each inference of Kane's validity framework. For Scoring, three cardiologists' kappa for agreement on correct diagnosis was 0.92. There was a range of ECG difficulty across and within each diagnostic category. For Generalization, appropriate sampling was reflected in the inclusion of a typical clinical base rate of 39% normal ECGs. Applying generalizability theory presented unique challenges. Under the Extrapolation inference, group learning curves demonstrated expert-novice differences, performance increased with practice and the incremental phase of the learning curve reflected ongoing, effortful learning. A minority of learners had atypical learning curves. We did not collect Implications evidence. Our results support a preliminary validity argument for a learning curve assessment approach for repeated ECG interpretation with deliberate and mixed practice. This approach holds promise for providing educators and researchers, in collaboration with their learners, with deeper insights into how well each learner is learning.

Identifying error types in visual diagnostic skill assessment

BACKGROUND

Misinterpretation of medical images is an important source of diagnostic error. Errors can occur in different phases of the diagnostic process. Insight in the error types made by learners is crucial for training and giving effective feedback. Most diagnostic skill tests however penalize diagnostic mistakes without an eye for the diagnostic process and the type of error. A radiology test with stepwise reasoning questions was used to distinguish error types in the visual diagnostic process. We evaluated the additional value of a stepwise question-format, in comparison with only diagnostic questions in radiology tests.

METHODS

Medical students in a radiology elective (n=109) took a radiology test including 11-13 cases in stepwise question-format: marking an abnormality, describing the abnormality and giving a diagnosis. Errors were coded by two independent researchers as perception, analysis, diagnosis, or undefined. Erroneous cases were further evaluated for the presence of latent errors or partial knowledge. Inter-rater reliabilities and percentages of cases with latent errors and partial knowledge were calculated.

RESULTS

The stepwise question-format procedure applied to 1351 cases completed by 109 medical students revealed 828 errors. Mean inter-rater reliability of error type coding was Cohen's κ=0.79. Six hundred and fifty errors (79%) could be coded as perception, analysis or diagnosis errors. The stepwise question-format revealed latent errors in 9% and partial knowledge in 18% of cases.

CONCLUSIONS

A stepwise question-format can reliably distinguish error types in the visual diagnostic process, and reveals latent errors and partial knowledge.

Competency milestones for medical students: Design, implementation, and analysis at one medical school

Competency-based assessment seeks to align measures of performance directly with desired learning outcomes based upon the needs of patients and the healthcare system. Recognizing that assessment methods profoundly influence student motivation and effort, it is critical to measure all desired aspects of performance throughout an individual's medical training. The Accreditation Council for Graduate Medical Education (ACGME) defined domains of competency for residency; the subsequent Milestones Project seeks to describe each learner's progress toward competence within each domain. Because the various clinical disciplines defined unique competencies and milestones within each domain, it is difficult for undergraduate medical education to adopt existing GME milestones language. This paper outlines the process undertaken by one medical school to design, implement and improve competency milestones for medical students. A team of assessment experts developed milestones for a set of focus competencies; these have now been monitored in medical students over two years. A unique digital dashboard enables individual, aggregate and longitudinal views of student progress by domain. Validation and continuous quality improvement cycles are based upon expert review, user feedback, and analysis of variation between students and between assessors. Experience to date indicates that milestone-based assessment has significant potential to guide the development of medical students.