Abstract
Objectives
To assess current pathology resident training in genomic and molecular pathology.
Methods
The Training Residents in Genomics (TRIG) Working Group has developed survey questions for the pathology Resident In-Service Examination (RISE) since 2012. Responses to these questions, as well as knowledge questions, were analyzed.
Results
A total of 2,529 residents took the 2019 RISE. Since 2013, there has been an increase in postgraduate year 4 (PGY4) respondents indicating training in genomic medicine (58% to approximately 80%) but still less than almost 100% each year for molecular pathology. In 2019, PGY4 residents indicated less perceived knowledge and ability related to both genomic and traditional molecular pathology topics compared with control areas. Knowledge question results supported this subjective self-appraisal.
Conclusions
The RISE is a powerful tool for assessing the current state and also trends related to resident training in genomic pathology. The results show progress but also the need for improvement in not only genomic pathology but traditional molecular pathology training as well.
Keywords: Genomics, Next-generation sequencing, Medical education, Pathology, Residency training
Key Points.
Pathologists play a key role in molecular and genomic testing, and trainees must acquire knowledge and skills in this area.
The pathology Resident In-Service Examination (RISE) is a powerful tool for assessing resident training in molecular and genomic pathology and shows over time increased training in genomic medicine.
Results from the 2019 RISE also show a need for continued improvement in not only genomic but traditional molecular pathology training as well.
Whether in academia or private practice, pathologists play a critical role in molecular and genomic testing. As such, pathologists-in-training have to acquire an understanding of key genetic concepts and their application to patient care. A published expert-developed list of competencies for practicing pathologists includes high-priority items such as determining sample adequacy, ensuring the appropriate molecular test was ordered, and effectively communicating genomic testing results through pathology reports and in interdisciplinary team settings.1 An Association for Molecular Pathology resident curriculum also includes “essential” topics such as describing how the clinical impact of a sequence variant can be determined and listing and describing the utility of online tools for molecular analysis.2
To further address the need for resources to improve genomic pathology education, in 2010 the Training Residents in Genomics (TRIG) Working Group was formed through the Program Directors Section of the Association of Pathology Chairs.3,4 TRIG is made up of experts in education, molecular pathology, and genetics and includes representatives of major pathology and genetics organizations.
Survey questions developed by the TRIG Working Group have been included in the American Society for Clinical Pathology (ASCP) Resident In-Service Examination (RISE) since 2012. Administered in every pathology residency program in the United States, the RISE provides a unique large-scale opportunity to gauge the current level of pathology resident molecular and genomics training.
The TRIG Working Group published the results of the 2013 RISE assessment.5 At that time, 42% of residents responded they had no training in genomic medicine. The TRIG Working Group has since further refined RISE survey questions. Graded genomic medicine and molecular pathology knowledge questions are also now included in the exam.
In this article, we report the responses for the survey questions developed by the TRIG Working Group for the 2019 RISE. We also report the scores for knowledge questions in genomic medicine and molecular pathology. The original report on the 2013 exam provided a baseline level of genomic training of pathology residents in the United States. The current results demonstrate that progress has been made, but there are also areas for improvement.
Materials and Methods
RISE
The ASCP RISE has been offered to pathology residency programs since 1983 and is administered over a 2-week period in the spring. Residency program directors use the exam to assess the knowledge gained over the course of residency, and the RISE has been shown to correlate with board examination performance.6 All US pathology residency programs, as well as a number of international programs, participated in the 2019 RISE.
The RISE includes an untimed demographic/survey section and a timed 6-hour examination. The exam has more than 350 multiple-choice questions in a “one-best-answer” format. There are three RISE sections: Anatomic Pathology (AP), Clinical Pathology (CP), and Special Topics (ST) common to AP and CP, including hematopathology and molecular pathology. At the beginning of the ST section, there is the untimed demographic portion, which includes the survey questions developed by the TRIG Working Group, followed by the timed knowledge questions. The knowledge questions are solicited from practicing pathologists and then vetted by a RISE committee consisting of experts in all tested areas. Each year, psychometric analysis is used to remove poorly performing questions from the exam.
Survey Question Design Process
Since 2012, the TRIG Working Group, using e-mails and monthly conference calls, has developed survey questions for the RISE. These questions are included in the demographic portion of the RISE and assess attitudes and perceived ability related to molecular pathology and genomic medicine. In 2019, the questions were revised by the TRIG Working Group to account for changes in current pathology practice and allow better assessment of specific topic areas and use of genomics online tools. There are also “control” questions related to gastrointestinal pathology and immunohistochemistry testing, as almost all residents are expected to have significant training in these subdisciplines. The questions as they appeared on the 2019 RISE are included in the supplemental materials (all supplemental materials can be found at American Journal of Clinical Pathology online).
Statistical Analysis
For survey questions, the response percentages are reported. The knowledge questions were a priori categorized by the RISE committee as hematopathology or molecular topic areas. For the 2019 RISE, a molecular pathologist member of the RISE committee, familiar with the TRIG curriculum but not a member of the TRIG Working Group, further divided the molecular questions into molecular pathology and genomic pathology prior to any viewing of scores. The distinction between molecular pathology and genomic pathology for question classification was based on the extent of the human genome involved. Questions focused on traditional methodologies, such as Sanger sequencing and polymerase chain reaction (PCR), were classified as molecular pathology, as were questions that involved a limited number of genes. Conversely, questions that explored the multiplexed analysis of many genes through next-generation sequencing (NGS), high-complexity bioinformatic analyses, and the algorithms required for genome data review were classified as genomic pathology. A two-tailed t test was used to compare the percentages of correct responses in the different topic areas and of postgraduate year 1 (PGY1) vs PGY4 residents. A P value of less than .05 was considered statistically significant.
Results
Demographics of RISE Participants
In 2019, of the 2,529 residents who took the RISE and survey, 662 were PGY1, 668 were PGY2, 625 were PGY3, and 574 were PGY4. The exam takers represented 168 programs, with 84% from the United States, 10% from Canada, and the remaining from other countries. Regarding the training pathway, 85% were AP/CP residents, 11% were AP only, 4% were CP only, and 1% were AP-neuropathology.
Trends in Resident Training
Since 2013, individuals taking the RISE have been asked to “please indicate how much training you have completed during your residency” in “molecular pathology” and “genomic medicine.” Figure 1 shows the trend for PGY4 residents. For “molecular pathology,” the results are similar in 2013 through 2019, with over 90% indicating training. For “genomic medicine,” in 2013, 58% reported any training in this area with, over the subsequent years, an increase to approximately 80%. Beginning in 2016, training in gastrointestinal (GI) pathology was also queried. Each year, over 90% of PGY4 residents answered they had such training.
Figure 1.
Postgraduate year 4 residents’ reported training in molecular pathology (MP) and genomic medicine (GM) since 2013. Examinees were asked, “Please indicate how much training you have completed during your residency in (molecular pathology/genomic medicine).” The y-axis shows the percent reporting any training in the different areas. Answer options included none, ≤1 week, >1 week and ≤4 weeks, >4 weeks and ≤6 months, and >6 months.
The 2019 results are shown by PGY in Figure 2. For PGY1 residents, 82% responded that they had GI pathology training rising to 96% by PGY4. In contrast, only 23% and 13% of PGY1 residents indicated training in “molecular pathology” and “genomic medicine,” respectively. Although 95% of residents had training in “molecular pathology” by PGY4, 78% indicated they had received “genomic medicine” training. Regarding the amount of training, 95% of PGY4 residents had more than 1 week of training in GI pathology, with 84% having more than 1 month. For “molecular pathology,” 92% had more than 1 week of training, with 42% having over 1 month. For “genomic medicine,” these numbers were lower, with 64% having more than 1 week of training and 18% having over 1 month. Only 1% of PGY4 residents reported more than 6 months of training in “molecular pathology” or “genomic medicine” compared with 23% for GI pathology.
Figure 2.
Residents’ (2019 Resident In-Service Examination) reporting training in gastrointestinal pathology (GI), molecular pathology (MP), and genomic medicine (GM). Examinees were asked, “Please indicate how much training you have completed during your residency in (gastrointestinal pathology/molecular pathology/genomic medicine).” The y-axis shows the percent reporting any training in the different areas by postgraduate year (PGY). Answer options included none, ≤1 week, >1 week and ≤4 weeks, >4 weeks and ≤6 months, and >6 months.
Molecular and Genomic Pathology Topics
Additional questions were asked to better delineate training and comfort level in specific topic areas. Although Sanger sequencing and PCR are traditional molecular pathology techniques, almost 16% of PGY4 residents reported no training in this area. This value was similar to the 19% that reported no training in NGS. Only 6% of PGY4 residents said they had no training in immunohistochemistry testing. With regard to self-rated knowledge, approximately 50% of residents indicated it was “good” or better for either Sanger sequencing/PCR or NGS-based testing in contrast to 87% for immunohistochemistry testing Figure 3.
Figure 3.
Postgraduate year 4 residents’ (2019 Resident In-Service Examination) reported knowledge in Sanger sequencing/polymerase chain reaction (PCR), next-generation sequencing (NGS), and immunohistochemistry testing. Examinees were asked to “rate your training during residency” for the different testing modalities.
Residents were also asked to indicate their comfort level with various tasks Figure 4. For discussing a colon biopsy report with a clinician, 60% of PGY4 residents felt very comfortable and 29% felt somewhat comfortable. For selecting the best tissue block for tumor genetic testing, the values were 52% for very comfortable and approximately 36% somewhat comfortable. PGY4 residents were less comfortable discussing either the results of an NGS-based or Sanger sequencing/PCR-based tumor genetic testing report with a clinician. Only approximately 10% felt very comfortable and 40% felt somewhat comfortable. There was greater comfort with determining the significance of a genetic variant, with 13% feeling very comfortable and 55% feeling somewhat comfortable.
Figure 4.
Postgraduate year 4 residents’ (2019 Resident In-Service Examination) reported comfort with pathology tasks. Examinees were asked, “Please rate how comfortable you are in your ability to do the following: knowledgably discussing the results of a next-generation sequencing (NGS)–based tumor genetic testing report with a clinician; knowledgably discussing results of a PCR or Sanger sequencing-based tumor genetic testing report with a clinician; determining the clinical significance of a genetic variant; selecting the best tissue block for tumor genetic testing; knowledgably discussing the results of a colon polyp biopsy report with a clinician.” PCR, polymerase chain reaction.
Familiarity With Online Genomics Tools
PGY4 residents were asked whether they had heard of or used several genomic-pathology related online tools7-15Figure 5. As expected as a control, almost all residents had used PubMed. Less than 50% had used each of the other online tools, with less than 25% having used important resources such as PharmGKB, PolyPhen, ClinGen, and cBioPortal. For all tools except OMIM and PubMed, 25% or more had not even heard of the resource.
Figure 5.
Postgraduate year 4 residents’ (2019 Resident In-Service Examination) reported use of online genomics resources. Examinees were asked, “Have you heard of or used any of the following websites?” including PharmGKB, PolyPhen, cBioPortal, ClinGen, ClinVar, ClinicalTrials.gov, COSMIC, OMIM, and PubMed.7-15
Knowledge Questions
There were 15 hematopathology, 17 molecular, and six genomics questions in the ST section on the 2019 RISE. PGY4 residents answered, on average, 81% of hematopathology questions correctly, compared with only 56% and 44% of molecular and genomics questions, respectively. Although the difference between hematopathology and genomics or molecular was significant (P < .001), the difference between molecular and genomics was not (P = .13).
As knowledge is expected to increase over residency training, the average percent correct in each topic area was compared for PGY1 vs PGY4 residents. For PGY1 residents, the averages were 50%, 39%, and 36% for hematopathology, molecular, and genomics questions, respectively. The differences between topic areas for PGY1 residents were not statistically significant (hematopathology vs molecular [P = .12] or genomics [P = .06] or molecular vs genomics [P = .66]). PGY4 residents answered a significantly higher percentage of questions correctly than PGY1 residents for hematopathology questions (P < .0001) and molecular questions (P = .02), with an average difference of 31% and 16%, respectively. There was no statistically significant difference for genomics questions (P = .65) Figure 6.
Figure 6.
Postgraduate year (PGY) 1 and 4 residents’ (2019 Resident In-Service Examination) performance on knowledge questions. The average percent correct for the 15 hematopathology (HP), 17 molecular pathology (MP), and six genomic medicine (GM) knowledge questions. Error bars represent the standard deviation.
Discussion
All pathologists, regardless of subspecialty or practice site, need a working knowledge of molecular and genomic pathology. Trainees must learn how to recommend appropriate molecular testing, select appropriate blocks for testing, and integrate test results with histologic findings to generate effective reports.1,2 The results of the 2019 RISE demonstrate progress and also areas for improvement regarding both molecular and genomic pathology training. Although more residents are receiving training in “genomic medicine” compared with 2013, the 80% level is still below the nearly 100% for molecular pathology and GI pathology. As would be expected based on current curricula, the amount of molecular and genomic training was less than for GI pathology training, and the training occurred later in residency. This training also does not appear to include use of several important online genomics tools, with over 50% never having used any of the queried resources besides PubMed. A significant number had not even heard of most of the listed online tools.
As the first question was worded, the definition of “molecular pathology” and “genomic medicine” was left up to the respondent. This question, since it had been asked since 2013, was kept in the survey to allow some tracking of resident perception of genomic and molecular pathology education. As noted above, there is some evidence of improvement, with an increase in those PGY4 residents reporting training in “genomic medicine” from 58% in 2013 to approximately 80% currently. However, recognizing the need for greater detail, questions were developed for the 2019 RISE to allow for a better assessment based on specific testing and tasks. With a longer history of incorporation into residency curricula, it was expected that there would be increased training as well as knowledge and comfort levels with traditional molecular methods (Sanger sequencing and PCR) than genomics-based NGS. Surprisingly, the numbers were very similar, with 15% to 20% of PGY4 residents reporting no training and relatively low knowledge and comfort levels in both areas compared with controls.
The knowledge question results support the resident perceptions of limited knowledge related to both molecular and genomics topics. For hematopathology, PGY4 residents answered, on average, 81% of questions correctly. In contrast, the results for molecular pathology and genomic pathology were significantly lower at less than 60%. There was also a 31-point difference in the percentage of hematopathology questions answered correctly between PGY1 and PGY4 residents but less of an increase for both molecular and genomics questions.
Taken together, it is encouraging that residents now perceive they are receiving more training in “genomic medicine” than in 2013. There is also overall comfort with selecting appropriate tissue blocks for tumor genetic testing, which is a key task for practicing pathologists.1 The more sobering and remarkable finding is the similarity in the perceived and actual lack of knowledge for both traditional molecular methods and more modern NGS genomics-related testing. Almost all residents should be exposed to PCR and Sanger sequencing methods, but 16% indicated they have not had training, and a large percentage rated their knowledge as poor and were not comfortable explaining results to clinicians. There was also the limited exposure to critical genomics online tools, with a significant number never having even heard of many of these resources.
There are two possible explanations for these results. First, residents are truly not exposed to PCR and Sanger sequencing methods, and such topics need to be included in training. Given the long existence of molecular pathology rotations, however, this possibility seems unlikely. Another explanation is that the residents are exposed to a single molecular pathology rotation, but it is not of adequate length, and there is little continuity and longitudinal learning. As a result, the residents may not recall different components, including specific molecular methods, of a single 2- to 4-week block or consider this minimal exposure equivalent to no training. In contrast, exposure to GI pathology runs throughout residency, which helps solidify understanding and ability. Hematopathology is also typically at least a 3-month block, which may be reinforced with required related conferences throughout training. This theory is backed by the data demonstrating over 80% of residents having training GI pathology beginning in PGY1 compared with only 23% for molecular pathology with a more gradual increase throughout residency. There is also the significantly better performance on the RISE for hematopathology compared with molecular and genomics questions.
As such, our results argue for increasing the length of molecular and genomic pathology rotations and better integration of molecular and genomics training throughout residency. A longer rotation period, more than the typical 2 to 4 weeks, may help ensure appropriate knowledge and skill acquisition. The rotation may be continuous or, perhaps even more useful, an introductory rotation early in training with an advanced rotation later to reinforce concepts in the context of greater overall pathology knowledge. As GI pathology and hematopathology are likely reinforced longitudinally through conferences and other activities, possible similar approaches for molecular and genomic pathology involve regular molecular/genomic tumor boards and routinely including discussion of genetic findings at pathology case conferences with reference to the results using online tools. In addition, to help reinforce the importance in patient care, discussion of molecular and genomic testing should become a standard part of the anatomic pathology signout. Residents should assist in selecting the blocks for molecular testing, and as some genomic and molecular testing results may be returned after a resident has rotated off a particular service, there should be mechanisms to forward and explain these results to the trainees. Local tools, such as teaching sets, can be created. TRIG resources and the Association for Molecular Pathology resident curriculum can also help in this regard.2,16
There are some limitations to our study. As there was a major survey revision in 2019 and we did not use the same questions all years, we cannot track results in every area longitudinally. As noted above, the definition of “molecular pathology” and “genomic medicine” was left up to the respondent, but the more specific survey and knowledge questions backed up the residents’ reported training in the different areas. Given space limitations, we were also not able to ask about every aspect of molecular and genomic education, and additional questions would have been useful. For example, “training” could be interpreted as didactics, hands-on experience and/or more broadly to include the overall educational experiences during rotations. While greater detail would have been informative, other questions show, regardless of training method, important findings in perceived and actual knowledge. The expertise of the TRIG Working Group was also used to hone in on critical concepts. We were also able to use only two control areas. However, these included the major formats for pathology education with both an area that is extensively covered in training (GI pathology) and one that is usually more of a focused rotation (hematopathology). The fact that both these formats, which involve generally at least 3 months of training, were associated with greater perceived or actual knowledge argues for greater exposure to molecular and genomic pathology during residency training. Last, due to the need to keep the RISE questions confidential, the differentiation of genomics and molecular questions was based on the review of a single expert, and a limited number of questions were classified as genomic. This limitation is of less concern given there was similar uniformly worse performance in both molecular and genomics topic areas compared with hematopathology.
The RISE is a very powerful tool for education research. Our work is one of the few examples of querying an entire group of trainees to determine perceived and actual knowledge in a subject area. We hope to continue to use the RISE to assess molecular and genomic pathology training. The 2019 results suggest there is still work to be done to ensure our trainees have the needed knowledge and abilities to practice modern pathology. Consideration should be given to improving not just training in genomic approaches but all aspects of molecular pathology. While increasing time on rotations is one approach, we recognize it is already difficult to cover everything a resident needs to know in only 3 or 4 years. Future studies should investigate ways to better integrate these topics across residency and other novel educational approaches.
Supplementary Material
Acknowledgments
The TRIG Working Group includes the following contributors: Adekunle M. Adesina, Department of Pathology, Texas Children’s Hospital, Baylor College of Medicine, Houston, TX; Lisa Browning, Department of Cellular Pathology, Oxford University Hospitals NHS Foundation Trust, Oxford, UK; Devon Chabot-Richards, Department of Pathology, TriCore Reference Laboratories and the University of New Mexico, Albuquerque; Allison M. Cushman-Vokoun, Department of Pathology and Microbiology, University of Nebraska Medical Center, Omaha; Alix R. D’Angelo, Department of Genetics, Louisiana State University Health Sciences Center, New Orleans; Marie C. DeFrances, Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA; Siddhartha Devarakonda, Division of Oncology, Washington University School of Medicine, St Louis, MO; Helen Fernandes, Department of Pathology and Cell Biology, Columbia University Medical Center, New York, NY; Pedro Fernandez, Department of Anatomical Pathology, Hospital Germans Trias I Pujol, Badalona, Spain; Ruta Gupta, NSW Health Pathology, Department of Tissue Pathology and Diagnostic Oncology, Royal Prince Alfred Hospital, Sydney, Australia; Michael E. Hurwitz, Yale Comprehensive Cancer Center, New Haven, CT; Neal I. Lindeman, Department of Pathology, Brigham and Women’s Hospital, Boston, MA; Alexander Nobori, Department of Pathology and Laboratory Medicine, David Geffen School of Medicine, Los Angeles, CA; Erik Nohr, Department of Pathology and Laboratory Medicine, Cumming School of Medicine, Calgary, Canada; Jaqueline Payton, Department of Pathology and Immunology, Washington University School of Medicine, St Louis, MO; Benjamin Saylor, Department of Pathology, University of Alabama at Birmingham; Mark E. Sobel, American Society for Investigative Pathology, Bethesda, MD; Keith F. Stringer, Department of Pathology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH; Chad M. Vanderbilt, Department of Pathology, Memorial Sloan Kettering Cancer Center, New York, NY; and Martin Young, Cytopathology Department, Royal Free Hospital, London, UK.
This work was supported by the National Institutes of Health (R25CA168544).
Contributor Information
Training Residents in Genomics (TRIG) Working Group:
Adekunle M Adesina, Lisa Browning, Devon Chabot-Richards, Allison M Cushman-Vokoun, Alix R D’Angelo, Marie C DeFrances, Siddhartha Devarakonda, Helen Fernandes, Pedro Fernandez, Ruta Gupta, Michael E Hurwitz, Neal I Lindeman, Alexander Nobori, Erik Nohr, Jaqueline Payton, Benjamin Saylor, Mark E Sobel, Keith F Stringer, Chad M Vanderbilt, and Martin Young
References
- 1. Laudadio J, McNeal JL, Boyd SD, et al. Design of a genomics curriculum: competencies for practicing pathologists. Arch Pathol Lab Med. 2015;139:894-900. [DOI] [PubMed] [Google Scholar]
- 2. Aisner DL, Berry A, Dawson DB, et al. A suggested molecular pathology curriculum for residents: a report of the Association for Molecular Pathology. J Mol Diagn. 2016;18:153-162. [DOI] [PubMed] [Google Scholar]
- 3. Haspel RL, Ali AM, Huang GC, et al. Teaching genomic pathology: translating team-based learning to a virtual environment using computer-based simulation. Arch Pathol Lab Med. 2019;143:513-517. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Haspel RL, Ali AM, Huang GC. Using a team-based learning approach at national meetings to teach residents genomic pathology. J Grad Med Educ. 2016;8:80-84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Haspel RL, Rinder HM, Frank KM, et al. The current state of resident training in genomic pathology: a comprehensive analysis using the resident in-service examination. Am J Clin Path. 2014;142:445-451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Rinder HM, Grimes MM, Wagner J, et al. Senior pathology Resident In-Service Examination (RISE) scores correlate with outcomes of the American Board of Pathology (ABP) certifying examinations. Am J Clin Pathol. 2011;136:499-506. [DOI] [PubMed] [Google Scholar]
- 7. Whirl-Carrillo M, McDonagh EM, Hebert JM, et al. Pharmacogenomics knowledge for personalized medicine. Clin Pharmacol Ther. 2012;92:414-417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Adzhubei IA, Schmidt S, Peshkin L, et al. A method and server for predicting damaging missense mutations. Nat Methods. 2010;7:248-249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Rehm HL, Berg JS, Brooks LD, et al. ClinGen: the clinical genome resource. N Engl J Med. 2015;372:2235-2242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Cerami E, Gao J, Dogrusozet U, et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2012;2:401-404. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Landrum MJ, Lee JM, Benson M, et al. ClinVar: improving access to variant interpretations and supporting evidence. Nucleic Acids Res. 2018;4:46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. US National Library of Medicine. https://clinicaltrials.gov/. Accessed January 31, 2020.
- 13. Tate JG, Bamford S, Jubb HC, et al. COSMIC: the Catalogue Of Somatic Mutations In Cancer. Nucleic Acids Res. 2019;47:D941-D947. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. McKusick-Nathans Institute of Genetic Medicine, Johns Hopkins University. Online Mendelian Inheritance in Man (OMIM) https://omim.org/. Accessed January 31, 2020.
- 15. US National Library of Medicine. https://pubmed.ncbi.nlm.nih.gov/. Accessed January 31, 2020.
- 16. Training Residents in Genomics. Resources http://www.pathologylearning.org/trig/resources. Accessed January 31, 2020.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.