Professional Documents
Culture Documents
PII: S1876-2859(17)30561-2
DOI: https://doi.org/10.1016/j.acap.2017.10.008
Reference: ACAP 1112
Please cite this article as: Erika L. Abramson, Caroline R. Paul, Jean Petershack, Janet Serwint,
Janet E. Fischel, Mary Rocha, Meghan Treitz, Heather McPhillips, Tai Lockspeiser, Patricia
Hicks, Linda Tewksbury, Margarita Vasquez, Daniel J. Tancredi, Su-Ting T. Li, From Design to
Dissemination: Conducting Quantitative Medical Education Research, Academic Pediatrics
(2017), https://doi.org/10.1016/j.acap.2017.10.008.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will
undergo copyediting, typesetting, and review of the resulting proof before it is published in its
final form. Please note that during the production process errors may be discovered which could
affect the content, and all legal disclaimers that apply to the journal pertain.
1 From Design to Dissemination: Conducting Quantitative
2 Medical Education Research
3
4 Erika L. Abramson, MD MSa, Caroline R. Paul, MDb, Jean Petershack, MDc, Janet Serwint, MDd,
5 Janet E. Fischel, PhDe, Mary Rocha, MDf, Meghan Treitz, MDg, Heather McPhillips MD MPHh,
6 Tai Lockspeiser MD MHPEi, Patricia Hicks, MD MPHEj, Linda Tewksbury, MDk, Margarita
7 Vasquez, MDl, Daniel J. Tancredi, PhDm, Su-Ting T. Li, MD MPHn
8
9
10 Corresponding Author Information:
11 Erika L. Abramson, MD MS
12 Weill Cornell Medicine
13 Departments of Pediatrics and Healthcare Policy & Research
14 525 E. 68th Street, Rm M610A
15 New York, NY 10065
16 Phone: 212 746-3929
17 Fax: 212-746-3140
18 Email: err9009@med.cornell.edu
19
20 b Caroline R. Paul, MD
21 University of Wisconsin School of Medicine and Public Health
22 Department of Pediatrics
23 Email:crpaul@wisc.edu
24
25 c Janet E. Fischel, PhD
26 Stony Brook University School of Medicine
27 Department of Pediatrics
28 Email: Janet.Fischel@stonybrookmedicine.edu
29
30 d Jean Petershack, MD
31 University of Texas Health Science Center at San Antonio
32 Department of Pediatrics
33 Email: Petershack@uthscsa.edu
34
35 e Janet Serwint, MD
36 Johns Hopkins University School of Medicine
37 Department of Pediatrics
38 Email: jserwint@jhmi.edu
39
40
41 f Mary Rocha, MD
42 Baylor College of Medicine
43 Department of Pediatrics
44 Email: Mary.Rocha@bcm.edu
45
46 g Meghan Treitz, MD
Page 1 of 38
1 University of Colorado School of Medicine
2 Department of Pediatrics
3 Email: meghan.treitz@childrenscolorado.org
4
5 h Heather McPhillips, MD MPH
6 University of Washington, Seattle Children’s Hospital
7 Email: heather.mcphillips@seattlechildrens.org
8
9 i Tai Lockspeiser MD MHPE
10 University of Colorado School of Medicine
11 Department of Pediatrics
12 Email: tai.lockspeiser@childrenscolorado.org; tai.lockspeiser@ucdenver.edu
13
14 j Patricia Hicks, MD, MHPE
15 Perelman School of Medicine at the University of Pennsylvania
16 Department of Pediatrics
17 Email: pjhpeds@gmail.com
18
19 k Linda Tewksbury, MD
20 New York University School of Medicine
21 Department of Pediatrics
22 Email: Linda.Tewksbury@nyumc.org
23
24 l Margarita Vasquez, MD
25 University of Texas Health Science Center at San Antonio
26 Department of Pediatrics
27 Email: LONDONOM@uthscsa.edu
28
29 m Daniel J. Tancredi, PhD
30 University of California, Davis
31 Department of Pediatrics and the Center for Healthcare Policy and Research
32 Email: djtancredi@ucdavis.edu
33
34 n Su-Ting T. Li, MD, MPH
35 University of California, Davis
36 Department of Pediatrics
37 Email: sutli@ucdavis.edu
38
39 Key Words: medical education, research
40
41 Running title: Conducting quantitative medical education research
42
43 Abstract Word Count: 221
44
45 Main Text Word Count: 4969
46
Page 2 of 38
1 Funding: This work was not supported by any funding source.
2
3 Conflicts of Interest: The authors have no conflicts of interest to disclose.
4
5
6
Page 3 of 38
1 Abstract
2 Rigorous medical education research is critical to effectively develop and evaluate the training
3 we provide our learners. Yet, many clinical medical educators lack the training and skills needed
4 to conduct high quality medical education research. This paper offers guidance on conducting
5 sound quantitative medical education research. Our aim is to equip readers with the key skills
6 and strategies necessary to conduct successful research projects, highlighting new concepts and
8 discuss strategies to ensure that the research question of interest is worthy of further study and
9 how to use existing literature and conceptual frameworks to strengthen a research study.
10 Through discussions of the strengths and limitations of commonly used study designs, we expose
11 the reader to particular nuances of these decisions in medical education research and discuss
12 outcomes generally focused upon, as well as strategies for determining the significance of
14 preparing results for dissemination to a broad audience. Practical planning worksheets and
15 comprehensive tables illustrating key concepts are provided in order to guide researchers through
16 each step of the process. Medical education research provides wonderful opportunities to
17 improve how we teach our learners, to satisfy our own intellectual curiosity and ultimately, to
19
Page 4 of 38
1 Introduction
2 Rigorous medical education research provides evidence for approaches to improve the
3 education of our learners, and shares with clinical, basic and translational research the ultimate
4 goal of improved patient outcomes. Medical education research can provide data to affirm that
6 education research, can guide changes in the delivery of medical education in order to assure that
7 the educational “product,” the trainee, is best prepared for the practice of medicine.1
9 study requires attention to factors quite different from traditional clinical research. For example,
10 the effectiveness of educational interventions often results from a complex interplay between the
11 learner, the educator, and the educational and clinical learning environment. As these factors
12 may vary significantly across programs, conducting large-scale studies and demonstrating
14 contamination across learners who have and have not received an intervention can threaten one’s
15 ability to demonstrate an intervention’s effectiveness, as can the natural maturation that tends to
16 occur as learners gain experience and knowledge over time. Also important are institutional
17 review board (IRB) considerations when administering an educational intervention that may
18 impact only a subset of learners, or when trainees are the study subjects and may fear negative
20 Unfortunately, many clinical medical educators lack the training needed to conduct high-
21 quality medical education research and clear reporting of experimental medical education
22 research has been modest at best.2-4 It is imperative to fill this skill gap because quality medical
23 education research has the potential to improve several facets of the complex educational
Page 5 of 38
1 process: teaching and learning strategies, curriculum development and evaluation methods,
3 To address these challenges, we have developed this paper on how to conduct meaningful
4 and rigorous quantitative medical education research. We use Glassick’s criteria for scholarship,
5 which includes clear goals, adequate preparation, appropriate methods, significant results,
6 effective presentation, and reflective critique, as essential components of a framework that can
7 be used to answer any research question.5 This paper fills important gaps through its discussion
8 of conceptual frameworks, focus on the nuances and challenges that govern methodologic and
9 ethical considerations in medical education research, and provision of practical tools and
10 suggestions for those embarking on a research project.6, 7 Qualitative research and studies with
11 mixed methodologies are increasingly being recognized for their considerable value in medical
12 education scholarship for many reasons, including with regard to identifying worthwhile
13 hypotheses or most relevant measures for further studies. Coverage of both quantitative and
14 qualitative research methods in a single paper would be prohibitively expansive. This paper
16 excellent primer on qualitative medical education research for guidance with such designs.8
17
19 Prior to beginning any research endeavor, every researcher must reflect on his or her own
20 skill set and that of potential collaborators. Consider working with content and methodologic
21 mentors within and outside your institution to facilitate planning and conducting your work.
23 Societies (PAS), Association for Pediatric Program Directors (APPD), or Council on Medical
Page 6 of 38
1 Student Education in Pediatrics (COMSEP) offer practical approaches to designing and
2 conducting medical education research. Programs such as the Academic Pediatric Association
3 (APA) Educational Scholars Program, the Association of American Medical Colleges (AAMC)
4 Medical Education Research Certificate Program, and the APPD Leadership in Educational
5 Academic Development program offer more formal training.9-11 People interested in more
6 rigorous training may benefit from pursuing a Masters or a PhD in medical education, which can
10 Sound scholarship starts with a sound idea. We outline below an approach for developing an
11 idea into a compelling research argument, transforming it into a research question guided by
12 conceptual frameworks, and developing a study design and analysis guided by the question.
13 While we present this as a linear process for the sake of clarity, this process is highly iterative.
14
17 recognized problem. For example, suppose you want to improve the rates of successful neonatal
20 outcomes 2) this is true despite conventional neonatal resuscitation program participation; and 3)
22 successfully addresses these issues will likely be of interest to many residency programs.
23
Page 7 of 38
1 Conceptual Frameworks
2 With a compelling problem and idea in hand, a key step is to consider the educational
3 philosophies which may illuminate the approach to addressing this problem. Georges Bordage
5 problem or a study, or ways of representing how complex things work the way they do.”13
7 recognized as key to advancing the rigor of quantitative studies.14 One or multiple conceptual
8 frameworks may be used and should be made explicit in order to afford readers greater
9 understanding of what guided particular decisions and how findings might translate to other
10 educational contexts. See Table 1 for examples of conceptual frameworks often used to frame
11 teaching interventions.
12 Consider the neonatal resuscitation example. The theory of deliberate practice may
16 conceptual model may help you refine and frame your question – What is the optimal frequency
18 As an example from the literature, authors Hunt et al were interested in using a novel
20 (CPR). In traditional simulation interventions, learners are given a scenario, progress through
21 the entire scenario, and then debrief afterwards about what was done effectively or ineffectively.
22 In this study, however, the authors explicitly drew upon the theory of deliberate practice to test
23 the idea that a simulation-based intervention would be more effective if mistakes were corrected
Page 8 of 38
1 in real-time in order to maximize the time spent deliberately practicing a skill performed in the
2 correct way. Thus, they designed their educational simulation intervention as “rapid cycle
3 deliberate practice training” in which residents were given a scenario, performed a hands-on
4 skill, got immediate coaching to correct mistakes, and continued performing the hands-on skill
5 correctly to create muscle memory. This novel approach, guided by the use of a common
10 Adapting the PICOTS format that has been successfully used in evidence-based
11 medicine can help you develop a clear and focused research question. PICOTS includes
12 population, intervention, relevant comparison group(s), outcome(s) assessed, timing and setting
13 (of the outcome assessment).16 For our example, the research question might be: Among
15 weekly 30-minute neonatal intubation simulation during neonatology rotations, compared with
16 participation in an annual standard neonatal resuscitation training program, result in higher rates
17 of successful intubation of patients at the end of the academic year? Our hypothesis may be:
19 rotations will lead to mastery of neonatal intubation more effectively than annual standard
20 neonatal resuscitation programs. Hypothesis testing, a hallmark of clinical and bench research,
22 The strength of a research question can be measured with the I-SMART criteria (an
23 acronym for important, specific, measurable, achievable, relevant, and timely), which connects
Page 9 of 38
1 well to the first Glassick criterion: Clear Goals (Table 2).17 Given rigorous publication
3 question.” A classic paradigm often utilized for educational outcomes is Kirkpatrick’s Learning
4 Evaluation Model, which consists of 4 levels of educational outcomes: Level 1: reaction (learner
5 satisfaction), Level 2: learning (attitude, knowledge, and skill acquisition), Level 3: behavior
6 change, and Level 4: patient impact (Table 3).18 Aim as high on the pyramid as possible.
7 Research with lower outcome levels may be publishable if highly innovative, while less
8 innovative studies generally require higher outcome levels. Importantly, one cannot assume that
9 changes in knowledge, skills, and attitudes in a training or research context will have a direct
10 impact on behavior in the clinical setting or on patient outcomes. Quality improvement, medical
11 education, and perspectives from behavioral economics all support this assertion.19-22 Measuring
12 behavior change in situ and measuring relevant patient outcomes are challenging but critical for
14
16 Glassick’s second criterion, adequate preparation, ensures that the research question
17 satisfies the relevant element of I-SMART. Adequate preparation includes: (1) understanding
18 what has already been studied and where the gaps lie through literature review; (2) acquiring
20 approval/exemption.
21
22 Literature Review
10
Page 10 of 38
1 Before embarking on your study, you must understand what is known and unknown
2 about a particular problem and how well your question addresses these gaps. A medical librarian
3 is a valuable resource to assist with this process. Typical databases include MEDLINE,
4 PubMed, Google Scholar, Scopus, Cumulative Index to Nursing and Allied Health Literature
6 Information Center (ERIC) [articles, books, theses, guidelines in education], EdIT Library
8 reviewed educational tools and curricular materials], MERLOT [a repository of peer reviewed
9 educational materials for all of higher education], Association of Pediatric Program Directors
10 (APPD) Share Warehouse [available to APPD members], the British Educational Index, and Best
12 determine whether existing studies are current and comprehensive, methodologically sound, and
13 broadly generalizable. If your idea has already been studied, consider if your project offers a
14 different perspective, such as a different level of learner, an expanded sample size, newer and
16
18 Building a research team with the necessary skills and expertise is key to any project’s
19 success. Statisticians, content and methodological mentors, physician scientists, and non-clinical
20 research faculty can all serve as outstanding partners. Trainees or junior faculty who can
21 contribute in a substantial manner and, in turn, benefit from collaborative efforts, can also be
22 valuable. It is also important to identify the costs to complete your project and potential funding
23 sources if necessary. These sources may include local funding opportunities, educational grants
11
Page 11 of 38
1 available through national organizations such as the American Academy of Pediatrics (AAP),
2 APA, APPD, COMSEP or larger foundation and federal grants (particularly for projects
6 It is essential to receive formal approval through the IRB prior to beginning any study.
7 However, there are several unique considerations for educational research.23 First, educational
9 educational settings, involves normal educational strategies, and the participant information is
10 de-identified.24 Thus comparing different instructional techniques and curricula through test
11 scores, performance evaluations, or de-identified surveys is often IRB exempt. Some institutions
12 may even have a blanket IRB exemption for all activities considered education research.
13 On the other hand, learners are a vulnerable population at risk for coercion for many
14 reasons including that they often interact with or are evaluated by those conducting the research
15 study. Therefore, protocols must clearly describe the process of informed consent and detail how
16 learners can refuse to participate or feel free to give honest opinions without fear of retribution,
17 while still being able to participate in all other aspects of any educational intervention that will
18 be implemented broadly. Consideration must also be given as to who will be collecting data so
19 that learners do not feel unfairly scrutinized by those in a position of authority. In addition,
20 concerns about fairness and equity in education may prompt some IRBs to have difficulty
21 approving studies where only some trainees receive an educational intervention, and may be
23 program, or crossover study designs.25 An additional challenge may occur when multiple
12
Page 12 of 38
1 institutions participate in the same educational research study. IRB approval at one institution
2 may suffice; however, often most institutions require their own individual approval.
4 Appropriate Methods – How will you design your project to achieve your specific aims?
5 The saying “begin with the end in mind” applies to the selection of methods for a
6 research study. Some projects aim to generate hypotheses; others to test them. In this section,
7 we will introduce some of the more common terminology and quantitative designs used in
8 educational research, using our neonatal resuscitation example (Table 4). We will focus
9 primarily on methodologies used for hypothesis testing where the goal is to obtain measurable
10 outcomes to answer a question (often described as a positivist paradigm). While this paradigm
12 in which subjective experiences are emphasized, are also well suited to medical education
13 research.
14 Let us begin with this scenario: A program director, Dr. Smith, is concerned about the
15 effectiveness of the training her residents receive in neonatal resuscitation. While eager to
16 design an actual intervention to address this concern, she recognizes that she does not have a
17 good understanding of the true nature of the underlying problem and wants to explore the
18 problem through a descriptive study. Her research question might be: What are the facilitators
19 and barriers affecting pediatric residents’ comfort with and skill in performing neonatal
20 resuscitation? To answer this question, she may perform a cross-sectional study in which all data
21 are collected at one time point, such as through a survey with close-ended questions. Her a priori
22 hypotheses can be explored by having residents self-assess comfort and indicate agreement or
23 disagreement with pre-identified barriers and facilitators listed in the survey instrument.
13
Page 13 of 38
1 The results of a descriptive study often provide useful background to generate hypotheses
2 for conducting further explanatory studies. Perhaps the descriptive research revealed that
3 residents are predominantly struggling with the inter-professional teamwork aspect of conducting
4 a neonatal resuscitation. To address this challenge, Dr. Smith decides to introduce a new
5 curriculum that focuses on facilitating the teamwork of residents and inter-professional staff
6 during high-fidelity simulation training with mannequins. Her research question might be: Does
8 neonatal resuscitation training improve resident resuscitation skills and ability to work in inter-
10 Several different explanatory study designs could be utilized to answer this question,
11 each with distinct strengths and limitations. Cohort studies follow one or more populations of
12 learners longitudinally with respect to an outcome of interest. They are observational rather than
14 interest. Cohort studies cannot show causality, and thus, in this case, a cohort study would be
16 In this case, there are quasi-experimental or experimental designs that Dr. Smith can
17 consider. Experimental designs are characterized by the investigator being able to assign
20 attributes of experimental designs, but typically lack a randomized control group, which make
21 them vulnerable to internal validity threats. One such quasi-experimental design is the pre-post
23 learner are conducted. The main limitation to this design is the lack of a concurrent comparison
14
Page 14 of 38
1 group. While you may demonstrate changes within each learner with a pre-post design, it is
2 difficult to attribute these to the intervention, because other confounding factors, such as natural
6 “intervention” and “control” groups respectively. However, these groups are naturally existing,
7 such as residents in different academic years or at different institutions. They are not randomly
8 allocated. Again returning to our example, if Dr. Smith opts for this quasi-experimental
9 approach, she might have all interns complete the new curriculum, and compare their
10 performance at baseline and the end of intern year to the performance of external NICU rotators
11 who experience the same resuscitation training but are not exposed to the new curriculum. To
12 address one of the main limitations of this study design – that the two groups are fundamentally
13 different from each other -- she might collect information on potential confounding factors such
14 as baseline comfort with and exposure to intubations and resuscitations, which could be used in
15 restricting the sample or in statistical analysis procedures to minimize bias from measured
17 “bleeding” between the control and intervention groups that often occurs when residents work
18 together, which can be limited by choosing historical controls or controls at separate sites that
20 A third option is the randomized control trial, in which subjects are randomly allocated
21 to an “intervention” or “control” arm. While this design is often considered the gold standard in
22 clinical research, it can be particularly challenging in medical education as sample sizes are
23 typically small and random assignment may be difficult or impossible due to deliberate
15
Page 15 of 38
1 placement of trainees on specific rotations. Randomization in a small cohort of students working
2 closely together may also result in exchange of information, which can be difficult to prevent but
3 important to note in any limitation section. In addition, with all experimental study designs,
4 educators may need to consider how to offer an educational intervention to all learners if proven
5 beneficial; one can consider designing a study that includes a plan to train or teach the “control”
6 participants the newly studied skill at the conclusion of the trial, or, with IRB approval, train the
7 “control” participants at the conclusion of the trial and measure their performance at some end
9 Sample Size Planning. Sample size justifications typically consider the necessary
10 precision to satisfactorily estimate the parameters of interest or the sample size necessary to
11 provide good statistical power (i.e. a high probability of detecting meaningful effects). Ensuring
12 you will have adequate sample size is critical to obtaining meaningful results. One of the major
13 challenges in educational research is that settings such as residency programs are often limited in
14 terms of size and representativeness. As a result, research projects should have built in, from the
15 beginning, a long-term plan for scaling up beyond a simple pilot to extend either to multiple sites
18 study registry (e.g. clinicaltrials.gov) can be an essential element to promote the rigor and
19 transparency of the study and for some study designs (e.g. clinical trials, including educational
21
16
Page 16 of 38
1 Meaningful results are both internally valid (include appropriate research design and analysis
2 to answer the important research question) and externally valid (replicable and generalizable to
3 other populations and settings). Design your analysis to answer your research question. For
4 quantitative studies, choose your study design, outcome variables (often called dependent
5 variables), measurement approaches, and analysis methods in collaboration with your statistician
7 significance may have multiple explanations: 1) there truly is no difference between the groups;
8 2) there is a difference between the groups but it cannot be detected with your sample’s size and
9 the variability on the outcome measure of interest (Type 1 and Type 2 errors); and 3) there are
10 confounding factors that are not controlled or not equivalent across groups. To properly interpret
11 your results, your study must be appropriately powered to detect a difference between groups.
12 Collect outcome variables and potential confounding variables appropriate to your research
13 question. Confounding variables are variables that could correlate with both the outcomes you
15 variable). For example, if you are studying whether a just-in-time neonatal intubation simulation
16 is associated with increased success in intubations, you may want to include potential
18 Measure data using instruments with validity evidence to address your research question.26
19 Validity evidence includes content validity, response process, internal structure, relationship to
20 other variables, and consequences.26 If you conduct a research project that does not yield the
21 outcomes you are seeking, it could be that the outcome you sought did not actually occur in your
22 study. Alternatively, it could be that your measurement instrument was insensitive to those
23 outcomes or changes that did indeed take place. You will not be able to untangle those two
17
Page 17 of 38
1 explanations, so it is important to use instruments with validity evidence.
2 When a measurement instrument with validity evidence does not already exist, you might
3 consider developing and validating a measurement instrument to disseminate for further use. For
4 example, if designing a survey, consider whether the survey questions are based on the literature
5 and agreed upon by content experts in the field (content validity), examined via cognitive
6 interviews to ensure the survey items were clear, relevant, and correctly interpreted (response
7 process), and piloted to ensure survey scores measuring the same construct are correlated
8 (internal structure) and relate to other measures in ways that make sense (relationship to other
9 variables)27 These concepts should be assessed for pre-existing as well as de novo surveys.
10 Returning to the example detailed in the methods section, it is key to note that Dr. Smith could
11 only perform her research effectively if she had a valid instrument with which to assess resident
12 performance in resuscitation skills and resident ability to work effectively as part of an inter-
13 professional team.
14 Utilize data analysis appropriate for your research (Table 5).28 In most instances, instead of
15 or in addition to reporting p-values, one should report parameter estimates (including treatment
16 effect estimates) along with 95% confidence intervals or other measures of uncertainty.29, 30
17 When making multiple comparisons, consider the possibility that as the number of comparisons
18 increase, so too does the risk of incorrectly concluding that a comparison is statistically
19 significant (Type 1 error). Hence, it may be desirable to use methods for simultaneous statistical
20 inference that can limit the overall probability of such an error for a family of comparisons.31
21 The more the study population represents the population at large, the more generalizable the
22 results of the study. Single site studies can be published if they add new information to the
23 literature and address important innovative questions. They may lead to multi-institutional
18
Page 18 of 38
1 studies, which can enhance generalizability of the results. The APPD Longitudinal Educational
2 Assessment Research Network (LEARN) and APA Continuity Research Network (CORNET),
5 Describe your sample in sufficient detail so that others can determine whether your study
6 sample is similar to their learners in their setting. If you have information about the population
7 at large, compare your study sample to that population. For example, if you are surveying
9 nationally in terms of program size and location.33 When making inferences about a population,
10 sample the population in a manner that promotes generalizability and minimizes selection bias.
11 For example, if you are studying whether graduating pediatric residents perceive themselves to
12 be prepared to counsel patients on obesity prevention, sample from a population of all graduating
13 pediatric residents, such as the AAP Annual Survey of Graduating Residents.34 However, if your
14 study question is to determine whether resident characteristics and experiences are related to
15 practicing in underserved areas and your hypothesis is that their background may affect their
17 residents.35
18
20 Effective presentation and dissemination are critical to advancing the science of medical
21 education. Furthermore, academic success and promotion are typically measured, in part, by the
23 work builds a track record and creates a niche reflective of your research interests and expertise.
19
Page 19 of 38
1 There are a number of ways to share your research findings. Initially you may present your
2 work at the departmental level as research-in-progress. Once you have results to share, consider
3 presenting at local, regional or national meetings. Through these venues, the feedback and
4 potential mentoring you receive can help frame your writing plan and forge potential multisite
5 collaborations that propel your work to a higher level of importance. Also, consider if your
6 study lends itself to workshop development and delivery. Evaluations from such national
7 workshops that demonstrate the effectiveness of the presentation are critical to include for
8 promotional capital.
9 Identification of the optimal journal for publication requires a critical look into the
10 educational literature. Which journals are interested in this type of work? What journals have
11 published articles using similar methods? Which audience do you wish to reach? What is the
12 impact factor of the journal? The impact factor is a proxy measure of a journal’s relative
13 importance in a field and measures the yearly average number of citations to recent articles
15 assessment tool, you can consider submitting your work to a resource portal such as
16 MedEdPORTAL or APPD Share Warehouse. Recognize that it may take several attempts to
17 publish a single paper. Incorporating reviewer comments can strengthen a paper prior to
18 resubmission. (See Appendix 2, checklist for manuscript submission, to facilitate the writing
19 process).36, 37 Also of note, for a variety of study designs, there are evidence-based reporting
21 recommend that these guidelines be used both in the planning and reporting of studies.
22
20
Page 20 of 38
1 Glassick’s final criterion requires scholars to critically evaluate their own work. This
2 involves reflecting on the literature review, critiquing the findings, and considering next steps to
3 advance your research. Your results should be viewed in the context of the gap identified for the
4 study and should be carefully linked to previous studies and/or the conceptual framework(s) that
5 you utilized. You must display a clear understanding not only of the results of the study but also
6 of the strengths and limitations. Simply listing limitations does not suffice. It is important to
7 consider how the limitations impact your results and their implications. Finally, a primary goal
8 of educational scholarship is to create the foundation for future scholarship. Since no single
9 study solves or resolves all issues around a topic, reflective critique must discuss the implications
10 for educational practice as well as future research to move the field forward.
11
13 Quality educational research benefits our trainees as well as the faculty who teach them,
14 our institutions, the medical education community at large and ultimately, the patients we serve.
15 When educators identify and disseminate best educational practices, they have the opportunity to
16 greatly influence the next generation of physicians. Because of the continually changing
17 healthcare environment, and the challenges that educational institutions face as they prepare,
18 promote and assess the knowledge, skills and attitudes of their learners, educators need to
19 continually evaluate educational programs and share best practices among institutions and across
21 Building the skills of clinical medical educators to conduct high quality quantitative
22 research is therefore critical to advancing education at the pace we need. Educational research
23 has made great advances in the past 10 years, as demonstrated in part by an increase in the
21
Page 21 of 38
1 number of articles published as well as journals devoted specifically to medical education.14
2 However, there remain important gaps. Indeed, weak study questions and designs were the
3 primary reasons for manuscripts not being sent for review by the editors of Academic
4 Medicine.38
7 so that the rigor of the study can be optimized at the outset. Second, conceptual frameworks
8 should guide the development of educational interventions and their assessments. Quality
9 checklists have been proposed that can be used by authors, editors, and consumers of medical
10 education research to help ensure the rigor of research studies.14, 39 In the past these did not
11 explicitly include conceptual frameworks, while more current checklists do. Third, in addition to
12 assessing the effects of interventions on learner knowledge, skills, attitudes, and behaviors, more
13 research must focus on the processes that facilitate application of learning and behavior, as well
14 as the impact of interventions on patients and organizational goals and outcomes. These
15 principles have recently been endorsed as part of the New World Kirkpatrick Model.40 Fourth,
16 within pediatrics, there is growing infrastructure to allow for multi-site studies, such as the
18 Network (APPD LEARN) and CORNET.32 Researchers should take advantage of such
19 resources in order to ensure that research findings are generalizable across settings. Lastly,
20 journal readers, authors, and editors alike must continue to demand these high standards of
21 medical education research in order to ensure that these standards become routine and that we
23
22
Page 22 of 38
1 References
2 1. Klein MD, Li ST. Building on the Shoulders of Giants: A Model for Developing Medical
3 Education Scholarship Using I-PASS. Academic pediatrics. 2016;16:499-500.
4 2. Levinson W, Rubenstein A. Integrating clinician-educators into Academic Medical
5 Centers: challenges and potential solutions. Acad Med. 2000;75:906-912.
6 3. Cook DA, Beckman TJ, Bordage G. Quality of reporting of experimental studies in
7 medical education: a systematic review. Medical education. 2007;41:737-745.
8 4. Castiglioni A, Aagaard E, Spencer A, et al. Succeeding as a Clinician Educator: useful
9 tips and resources. J Gen Intern Med. 2013;28:136-140.
10 5. Glassick CE HM, Maeroff GI. Scholarship assessed – evaluation of the professoriate.
11 San Francisco, CA: Jossey-Bass; 1997.
12 6. Yarris LM, Deiorio NM. Education research: a primer for educators in emergency
13 medicine. Acad Emerg Med. 2011;18 Suppl 2:S27-35.
14 7. Beckman TJ, Cook DA. Developing scholarly projects in education: a primer for
15 medical teachers. Medical teacher. 2007;29:210-218.
16 8. Hanson JL, Balmer DF, Giardino AP. Qualitative research methods for medical
17 educators. Academic pediatrics. 2011;11:375-386.
18 9. Jerardi KE, Mogilner L, Turner T, Chandran L, Baldwin CD, Klein M. Investment in
19 Faculty as Educational Scholars: Outcomes from the National Educational Scholars
20 Program. J Pediatr. Vol 171. 2016/03/29 ed2016:4-5 e1.
21 10. Medical Education Research Certificate Program. In: Colleges AoAM, ed.
22 11. Trimm F, Caputo G, Bostwick S, et al. Developing leaders in pediatric graduate
23 medical education: the APPD LEAD Program. Academic pediatrics. 2015;15:143-146.
24 12. Campbell DM, Barozzino T, Farrugia M, Sgro M. High-fidelity simulation in neonatal
25 resuscitation. Paediatrics & child health. 2009;14:19-23.
26 13. Bordage G. Conceptual frameworks to illuminate and magnify. Medical education.
27 2009;43:312-319.
28 14. Sullivan GM, Simpson D, Cook DA, et al. Redefining Quality in Medical Education
29 Research: A Consumer's View. Journal of graduate medical education. 2014;6:424-
30 429.
31 15. Ericsson KA KR, Tesch-Römer C. The role of deliberate practice in the acquisition of
32 expert performance. Psychology Review. 1993;100:363-406.
33 16. Medicine CfEB. Asking Focused Questions. Nuffield Department of Primary Care
34 Health Science, University of Oxford.
35 17. Doran GT. There’s a SMART way to write management’s goals and objectives.
36 Manage Rev. 1981;70:35-36.
37 18. Kirkpatrick D. Great ideas revisited. Techniques for evaluating training programs.
38 Revisiting Kirkpatrick's four-level model. . Technology and Development.
39 1996;50:54-59.
40 19. Hostetter M, and Klein S;. In Focus: Using Behavioral Economics to Advance
41 Population Health and Improve the Quality of Health Care Services. Quality Matters
42 Archive: The Commonwealth Fund; 2013.
43 20. Davis D, O'Brien MA, Freemantle N, Wolf FM, Mazmanian P, Taylor-Vaisey A. Impact
44 of formal continuing medical education: do conferences, workshops, rounds, and
23
Page 23 of 38
1 other traditional continuing education activities change physician behavior or
2 health care outcomes? Jama. 1999;282:867-874.
3 21. Team; IE. In: Exchange AHI, ed. Sustaining and Spreading Quality Improvement. Vol
4 20172014.
5 22. Mayer-Mihalski NaD, M;. Effective Education Leading to Behavior Change. In:
6 ParagonRx, ed. Vol 20172009.
7 23. Keune JD, Brunsvold ME, Hohmann E, Korndorffer JR, Jr., Weinstein DF, Smink DS.
8 The ethics of conducting graduate medical education research on residents. Acad
9 Med. 2013;88:449-453.
10 24. Miser WF. Educational research--to IRB, or not to IRB? Family medicine.
11 2005;37:168-173.
12 25. Kraus CK, Guth T, Richardson D, Kane B, Marco CA. Ethical considerations in
13 education research in emergency medicine. Acad Emerg Med. 2012;19:1328-1332.
14 26. Downing SM. Validity: on meaningful interpretation of assessment data. Medical
15 education. 2003;37:830-837.
16 27. Rickards G, Magee C, Artino AR, Jr. You Can't Fix by Analysis What You've Spoiled by
17 Design: Developing Survey Instruments and Collecting Validity Evidence. Journal of
18 graduate medical education. 2012;4:407-410.
19 28. Windish DM, Diener-West M. A clinician-educator's roadmap to choosing and
20 interpreting statistical tests. J Gen Intern Med. 2006;21:656-660.
21 29. du Prel JB, Hommel G, Rohrig B, Blettner M. Confidence interval or p-value?: part 4
22 of a series on evaluation of scientific publications. Deutsches Arzteblatt international.
23 2009;106:335-339.
24 30. Cummings P, Rivara FP. Reporting statistical information in medical journal articles.
25 Arch Pediatr Adolesc Med. 2003;157:321-324.
26 31. Gelman A, Hill J, Yajima, M. . Methodological Studies: Why we (usually) don’t have to
27 worry about multiple comparisons. Journal of Research on Educational Effectiveness.
28 2012;5:189-211.
29 32. Schwartz A, Young R, Hicks PJ. Medical education practice-based research networks:
30 Facilitating collaborative research. Medical teacher. 2016;38:64-74.
31 33. Abramson EL, Naifeh MM, Stevenson MD, et al. Research training among pediatric
32 residency programs: a national assessment. Acad Med. 2014;89:1674-1680.
33 34. Frintner MP, Liebhart JL, Lindros J, Baker A, Hassink SG. Are Graduating Pediatric
34 Residents Prepared to Engage in Obesity Prevention and Treatment? Academic
35 pediatrics. 2016;16:394-400.
36 35. Laraque-Arena D, Frintner MP, Cull WL. Underserved Areas and Pediatric Resident
37 Characteristics: Is There Reason for Optimism? Academic pediatrics. 2016;16:401-
38 410.
39 36. AAMC Review Criteria for Research Manuscripts. In: Durning SJ CJ, ed2015.
40 37. Li ST KM, Gusic M, Vinci R, Szilagyi P. . Crossing the Finish Line: Getting your
41 Medical Education Work Published. . Pediatric Academic Societies Workshop2016.
42 38. Meyer HS, Durning SJ, Sklar D, Maggio LA. Making the First Cut: An Analysis of
43 Academic Medicine Editors' Reasons for Not Sending Manuscripts Out for External
44 Peer Review. Acad Med. 2017.
45 39. Reed DA, Beckman TJ, Wright SM, Levine RB, Kern DE, Cook DA. Predictive validity
46 evidence for medical education research study quality instrument scores: quality of
24
Page 24 of 38
1 submissions to JGIM's Medical Education Special Issue. J Gen Intern Med.
2 2008;23:903-907.
3 40. Moreau KA. Has the new Kirkpatrick generation built a better hammer for our
4 evaluation toolbox? Medical teacher. 2017;39:999-1001.
5 41. Bandura A. Social foundations of thought and action: A social cognitive theory.
6 Englewood Cliffs, NJ1986.
7 42. Young HN, Schumacher JB, Moreno MA, et al. Medical student self-efficacy with
8 family-centered care during bedside rounds. Acad Med. 2012;87:767-775.
9 43. Ericsson KA. Acquisition and maintenance of medical expertise: a perspective from
10 the expert-performance approach with deliberate practice. Acad Med.
11 2015;90:1471-1486.
12 44. Hunt EA, Duval-Arnould JM, Nelson-McMillan KL, et al. Pediatric resident
13 resuscitation skills improve after "rapid cycle deliberate practice" training.
14 Resuscitation. 2014;85:945-951.
15 45. Kolb DA. Experiential Learning: Experience as the Source of Learning and
16 Development. Upper Saddle River, NJ: Prentice Hall, Inc.; 1984.
17 46. Klein M, Vaughn LM. Teaching social determinants of child health in a pediatric
18 advocacy rotation: small intervention, big impact. Medical teacher. 2010;32:754-
19 759.
20
21
22
25
Page 25 of 38
1
26
Page 26 of 38
1 Table 1: Examples of Conceptual Frameworks Used in Medical Education
2
Conceptual Description Example of use of conceptual framework
Framework
27
Page 27 of 38
Table 2: Elements of an I-Smart Research Question
28
Page 28 of 38
Table 3. Examples of Kirkpatrick’s Pyramid of Educational Outcomes
29
Page 29 of 38
Table 4: Common Quantitative Study Designs Used in Medical Education Research
Study Goal Study Type Example Research Question Example Methods Study Advantages Study Disadvantages
Describe/Explore Descriptive, What are the most common Cross-sectional survey Provides As data collected at
a group or Quantitative facilitators and barriers affecting administered to residents descriptive data one point in time,
phenomenon pediatric resident comfort with Presents preset choices of barriers from one point in often relies on recall
performing neonatal resuscitation and facilitators which residents time of information which
from the point of view of the have to select or rank Less demand on is subject to bias
residents? Answers will include percentages, resources as does Cannot demonstrate
rank order lists, counts not require causality
follow-up
Test a hypothesis Cohort Study How likely is it that residents exposed to Follow one group of residents Takes advantage Can only demonstrate
(Explanatory) a neonatal simulation-based curriculum longitudinally after exposure to of naturalistic associations, not
will chose a career in neonatology? the simulation-curriculum to see setting causality
how many go on to choose a Avoids learning
career in neonatology from pretest
Can compare rates to a cohort of
residents not exposed to the
curriculum
Intervention only, Does introduction of a novel, Conduct a pretest of residents No need for Cannot attribute any
pre/posttest design (a simulation-based neonatal resuscitation prior to exposure to new control change/gains to
quasi-experimental program involving residents and curriculum assessing resuscitation Can demonstrate intervention alone
design because of the interprofessional staff improve resident skills (such as number of attempts change/gains (time, practices, and
absence of a resuscitation skills and ability to work in per successful intubation, time to other educational
randomized comparison interprofessional teams in the delivery successful intubation) and experiences may also
group) room? teamwork skills (using a validated factor in)
teamwork assessment scale)
Deliver curriculum
Administer same assessments
post-curriculum exposure to
assess change in performance
Pre/post design with Does introduction of a novel, Pre-intervention and Post- Able to compare Need to control for
nonequivalent parallel simulation-based neonatal resuscitation intervention outcome data on two to control; yet baseline differences
groups (a quasi- program involving residents and groups: Control and still more feasible Impact of different
experimental design interprofessional staff improve resident Intervention non-randomized than randomized sites/times
because of the absence resuscitation skills and ability to work in groups. study Subject to immersion
of assignment of interprofessional teams in the delivery Conduct a pretest of all interns and ecological effects
individuals to study room better than a traditional neonatal assessing resuscitation skills
groups) resuscitation program? (such as number of attempts per
successful intubation, time to
30
Page 30 of 38
successful intubation) and
teamwork skills (using a validated
teamwork assessment scale)
First three blocks of interns get
novel curriculum, second three
blocks of interns get traditional
neonatal resuscitation program
Conduct similar post-curriculum
assessments and compare results
for the two groups
Does introduction of a novel, Random allocation for control Reduces Difficult to do
simulation-based neonatal resuscitation and intervention group. allocation bias Resource intensive
Equivalence parallel program involving residents and Conduct a pretest of all interns (minimizes Does not address non-
groups pre-post design interprofessional staff improve resident assessing resuscitation skills baseline uniform intervention
(an Experimental resuscitation skills and ability to work in (such as number of attempts per differences) Intervention “bleed”
design characterized by interprofessional teams in the delivery successful intubation, time to Subject to immersion
Random Assignment of room better than a traditional neonatal successful intubation) and and ecological effects
individuals to study resuscitation program? teamwork skills (using a Education Ethics
groups) validated teamwork assessment concerns
scale)
Use a random number generator
to allocate interns to two
groups: one group receives
traditional neonatal resuscitation
program at orientation, the other
group receives the novel
simulation-based curriculum
Conduct similar post-curriculum
assessments and compare results
for the two groups
31
Page 31 of 38
Table 5. Common Statistical Tests Used in Medical Education Research
What are you trying to Example Question Statistical How should results
determine? Methods be reported?
Summary values for a random What is the average number Estimated Mean (95% CI) and
variable with a bell-shaped of patients that third-year mean and standard deviation
distribution medical students take care standard estimates
of on the general inpatient deviation
wards? from sample
Midpoint value of a rank- What is the median medical Median Median (often
ordered list –minimizes school educational debt of accompanied by the
influence of extreme values; pediatric residents? minimum and
useful for ordinal data maximum value
and/or the 25th and
75th percentiles)
Most common value – useful for What is the most common Mode Mode
nominal or ordinal data subspecialty pediatric
residents enter after
residency?
Compare observed vs. expected Is the study population of Chi-Square p-values
values – categorical variables pediatric residents similar tests
to all pediatric residents in
the United States in regards
to gender?
Compare means of 2 Does an interactive web- Unpaired T- Differences in
independent groups – data based module on EKG test means (95% CI) and
normally distributed interpretation improve p-value, possible
residents’ ability to adjusted for multiple
accurately interpret EKGs comparisons
compared to a lecture on
EKG interpretation?
Compare means of 2 paired Does an EKG module Paired t-test Differences in
groups (e.g., pre- and post-test) improve residents’ ability to means (95% CI) and
– data normally distributed accurately interpret EKGs? p-value, possibly
adjusted for multiple
comparisons
Compare means of 3 or more Is there a difference in Analysis of (Adjusted) mean
groups medical school debt for Variance differences (95% CI)
residents who choose to (ANOVA) or and p-value,
practice in rural, urban, or multiple possibly adjusted for
suburban areas? regression multiple
comparisons
Correlation – data normally How well do resident self- Pearson Correlation
distributed (parametric) assessments of intubation product- coefficient and 95%
skills correlate with faculty moment CI
assessment? correlation
32
Page 32 of 38
coefficient
Correlation – data not normally How well does resident Spearman’s Correlation
distributed (non-parametric) performance on the In- rank coefficient with 95%
Training-Examination correlation CI
correlate with their coefficient
performance on the
American Board of
Pediatrics certifying exam?
Association – interval and What are factors associated Linear Regression
ordinal data with USMLE Step 1 regression coefficient with 95%
scores? CI
Association – binary data What are factors associated Logistic Odds Ratio with
with passing the American regression 95% CI
Board of Pediatrics
certifying exam on the first
attempt?
CI: Confidence Interval
Interval – Data where the difference between two values is meaningful (i.e., Age)
Ordinal – Data where there is a sense of order, but consecutive values may not be
equally spaced (i.e., Likert scales: Strongly Disagree – 1; Disagree – 2; Neither agree nor
disagree – 3; Agree – 4; Strongly Agree – 5)
Nominal – Categorical - Data in which there is no inherent order (i.e., cardiology,
pulmonary, general pediatrics)
33
Page 33 of 38
Appendix 1: Educational Research Scholarship Guide/Timeline
1. Clear Goals (I-SMART) - Specific Aim: What are you trying to do?
a. Important, interesting
b.Specific, simple to
understand
c. Measurable outcome
d.Achievable
e. Relevant and not
rehashing
f. Timely
What conceptual
framework(s) are you
utilizing?
2. Adequate preparation – Are you ready to do this project? Deadline
a. Literature review –
Sources, keywords
c. Acquire necessary
resources
(i.e. collaborators,
statistical support, etc.)
d.IRB Considerations and
Submission
e. Registration in study
registry (e.g.
clinicaltrials.gov)
f. Selection of relevant
reporting guidelines
(from EQUATOR-
NETWORK)
3. Appropriate methods – How are you going to do it?
a. Study Design, including
sample size justification
and selection of
measurement approaches
and observation schedule
b.Analysis/Evaluation
4. Significant results – So
34
Page 34 of 38
what?
35
Page 35 of 38
Appendix 2: Checklist for Authors Prior to Educational Manuscript Submission
Title/Abstract
1. Title is clear and representative of content
2. Abstract concisely describes study and key findings
3. Conclusions in abstract are justified given information provided in abstract
4. All information provided in abstract are presented in text
5. All information in abstract/ text/figures/tables are consistent
Introduction
1. Builds a convincing case why this problem is important with literature
review
2. Identifies gaps in literature and addresses how this study will fill the gaps
3. Conceptual framework is explicit and justified (and/or in Discussion)
4. Specific aim of the study (and hypothesis where applicable) is clearly
stated
Methods
For ALL Studies
1. Research design appropriate to address research question
2. Research design clearly stated (i.e., cross-sectional cohort study)
3. Methods clearly described in sufficient detail to permit study to be
replicated
3a. Study population (sampling, selection bias)
3b. Study intervention (objectives, activities, time allocation, training)
3c. Study instrument validity evidence (instrument development, content,
preparation of observers/interviewers/raters, scoring method, psychometric
properties)
3d. Study outcomes clearly defined (and high on Kirkpatrick’s pyramid –
may be inversely related to level of innovation, with less innovative ideas
requiring higher outcome levels)
4. Data analysis appropriate for research design and research question
5. Data analysis procedures clearly described in sufficient detail to be
replicated
6. IRB approval/exemption and consent clearly stated
For Quantitative Studies:
1. Study is generalizable due to selection of participants, setting, educational
intervention/materials (external validity – less innovative studies require
higher generalizability with more sites, etc.)
2. Potential confounding variables addressed and adjusted for in analysis
(internal validity)
3. Statistical tests appropriate. Effect size, functional significance discussed
when appropriate. When making multiple comparisons, adjustment for
significance level for multiple tests/comparisons are considered.
4. Power issues are considered in studies that make statistical inferences
(particularly if results not significant)
For Qualitative Studies:
36
Page 36 of 38
1. Study offers concepts or theories that are transferable to other settings and
methods described in sufficient detail (setting, sample)
2. Philosophical framework clearly stated (i.e., grounded theory)
3. Study design incorporates techniques to ensure trustworthiness (i.e.,
triangulation, prolonged observation)
4. Characteristics of the researchers that may influence the research are
described and accounted for during data collection/analysis
5. Describe how members of the research team contribute to coding,
identifying themes, and/or drawing inferences (dependability,
confirmability)
For Mixed-Methods (Quantitative and Qualitative) Studies:
1. Justify use of mixed-methods (Study must do justice to both methodology)
2. Justify order of quantitative and qualitative study
Results
1. All results are presented and align with study question and methods. All
results are presented in Results section (and not in other sections)
2. Sufficient data is presented to support inferences/themes
3. Tables, graphs, figures used judiciously to illustrate main points in text
Discussion
1. Key findings clearly stated. Conclusions follow from design, methods,
results
2. Findings placed in context of relevant literature, including conceptual
framework. Alternative interpretations of findings are considered as
needed
3. Study Limitations and Study Strengths discussed
4. Practical significance or implications for medical education are discussed.
Guidance for future studies is offered.
References
1. Literature review is comprehensive, relevant, and up-to-date
2. Ideas and materials of others are appropriately attributed (No plagiarism)
Final Journal Check
1. Study is relevant to mission of journal and journal audience
2. Author guidelines are followed (including word count)
3. Prior publication(s) by author(s) of substantial portions of the data are
appropriately acknowledged
4. Conflicts of interest are disclosed
5. Text is well written and easy to follow
6. Manuscript is well organized
Note: This table was adapted from AAMC Review Criteria for Research Manuscripts, 2nd Edition. Eds:
Durning SJ, Carline JD. 2015.
37
Page 37 of 38
38
Page 38 of 38