Hong Kong Med J 2015 Oct;21(5):455–61 | Epub 28 Aug 2015
© Hong Kong Academy of Medicine. CC BY-NC-ND 4.0
Using the script concordance test to assess clinical reasoning skills in undergraduate and postgraduate medicine
SH Wan, MB, ChB, MRCP (Edin)
School of Medicine Sydney, University of Notre Dame, 160 Oxford Street, Darlinghurst, NSW 2010, Australia
Corresponding author: Dr SH Wan (firstname.lastname@example.org)
The script concordance test is a relatively new format of written assessment that is used to assess higher-order clinical reasoning and data interpretation skills in medicine. Candidates are presented with a clinical scenario, followed by the reveal of a new piece of information. The candidates are then asked to assess whether this additional information increases or decreases the probability or likelihood of a particular diagnostic, investigative, or management decision. To score these questions, the candidate’s decision in each question is compared with that of a reference panel of expert clinicians. This review focuses on the development of quality script concordance questions, using expert panellists to score the items and set the passing score standard, and the challenges in the practical implementation (including pitfalls to avoid) of the written assessment.
Script concordance test (SCT) is a relatively new format of written assessment to assess higher-order clinical reasoning and data interpretation skills of medical candidates.1
In recent years, universities and postgraduate colleges worldwide have used SCT for both formative and summative assessment of clinical reasoning in various medical disciplines including paediatric medicine, paediatric emergency medicine, neurology, orthopaedics, surgery, and radiology.2 3 4 5 6 7 8 In the classic written assessment, multiple-choice questions (MCQ) and short-answer questions (SAQ) usually examine the candidates’ simple knowledge recall at the lowest ‘knows’ level of the Miller’s Pyramid (Fig 1).9 10 Questions in SCT are able to test candidates at the higher order of thinking at the ‘knows how’ and even ‘shows how’ level. It is a unique assessment tool that targets the essential clinical reasoning and data interpretation skills in a very authentic way that reflects the element of ‘uncertainty’ in real-world clinical scenarios prevalent in clinical practice. This is the key aspect of clinical competency that enables medical graduates or fellows in training to link and transfer their mastery of declarative clinical knowledge and skills into clinical practice in a real clinical setting. Recent literature reports the value of using SCT to assess other areas of disciplines where classic questions are difficult to develop, for example, in assessing medical ethical principles and professionalism.11
The structure and format of script concordance test
In SCT, candidates are presented with a clinical vignette/scenario, followed by the reveal of a new piece of information. The candidates are then asked to assess whether this additional information increases or decreases the probability or likelihood of the suggested provisional diagnosis, and increases or decreases the usefulness/appropriateness of a proposed investigation or management option. The process reflects everyday real-world decision-making processes where clinicians retrieve their ‘illness scripts’ or network of knowledge (about similar patient problems and presentations stored in their memory) when faced with uncertainty in a clinical presentation. This enables them to determine the follow-on diagnosis and management options most appropriate to the situation. As further clinical encounters are experienced, the scripts are updated and refined.12 Script concordance test assesses the candidates’ clinical reasoning and data interpretation ability in the context of uncertainty, particularly involving ill-defined patient problems in clinical practice.13 Sample SCT questions in Table 1 illustrate the structure and format of the SCT questions. As the clinical scenario unfolds, additional data such as clinical photos, radiological images, or audiovisual material can also be presented to enhance the authenticity of the scenarios.5 14 15
In scenario A in Table 1, the ‘clinical vignette’ is that of a 22-year-old woman who presents to the Emergency Department with severe abdominal pain. A piece of ‘new information’ is then revealed that her serum beta–human chorionic gonadotropin (β-HCG) is normal. The candidate is asked whether this additional information makes the ‘diagnosis’ of ectopic pregnancy: much less likely (-2), less likely (-1), neither more nor less likely (0: no effect on the likelihood), more likely (+1), or much more likely (+2). The next piece of new information (independent of the first one) is that the examination shows marked guarding and rigidity of the abdomen and the candidate is asked to determine the likelihood of a diagnosis of acute appendicitis.
In scenario B in Table 1, a similar format is used to assess the appropriateness of ordering an investigation in relation to the respective piece of additional information. The first question asks for the appropriateness of ordering a computed tomographic scan of the abdomen for a 16-year-old girl who presents with acute abdominal pain if her last menstrual period was 8 weeks ago.
In scenario C in Table 1, the focus is on the usefulness of different management options after being presented with different pieces of new information related to the clinical vignette.
In preparing candidates to answer the questions, it is crucial to emphasise that each piece of new information is independent of the previous piece but in the same clinical setting. For example, in scenario A, when answering the second question given that she has guarding and rigidity in the abdomen, she does NOT have a serum β-HCG test done.
With respect to the likelihood descriptors used in the SCT questions for the diagnosis type of scenario, the preference is to use the option of “much less likely (-2)” rather than “ruling out the diagnosis”; and “much more likely (+2)” rather than “almost certain/definite diagnosis”. This will allow candidates to use the full range of the five options. In the practice of medicine, there are usually few situations wherein a diagnosis can be confidently excluded or definitely diagnosed with a few pieces of information provided.3
There are nonetheless limitations to the design and format of SCT. Candidates cannot seek additional information to that given in the question; the scenario is only a snapshot of the clinical encounter without the comprehensive history, physical examination, and investigations that would be particularly desirable in an ambiguous clinical situation.16
Scoring script concordance test
To score these questions, the candidate’s decision in each question is compared with that of a reference panel of expert clinicians. Each member of the panel attempts the same set of questions and the answers are recorded. As there is no single best correct answer to the question, a full (1) mark will be awarded if the candidate’s decision concurs (hence the name ‘concordance’) with the majority of the expert panel. A proportional (partially credited or weighted) score (<1) will be given if the candidate’s decision concurs with the minority of the panel. The candidate will score a ‘0’ if no panellist chooses this option.3 The formula to calculate the weighted scores is shown in Table 2.
There are other scoring methods reported in the literature where a consensus-based single-answer scoring method or 3-point Likert scale scoring method is employed to determine the candidate scores.4 17
Selecting the reference panel
In general, a panel of 10 to 15 expert members relevant to the discipline is recommended to produce credible and reliable scores.18 The inter- and intra-rater reliability in the SCT panel have been shown to be good.19
The composition of the panel should include clinical teachers and academics who are familiar with the curriculum and experts in the field relevant to the discipline tested. Studies have shown that using general practitioners (GPs) in the panel may produce similar mean scores to specialists but with a wider standard deviation.3
A recent study, however, raised concerns about the reference standard and judgement of the expert panel. The study compared 15 emergency medicine consultants’ judgement scores with evidence-based likelihood ratios. The results showed that 73.3% of the mean judgement was significantly different to the corresponding likelihood ratios, with 30% overestimation, 30% underestimation, and 13.3% with diagnostic values in the opposite direction.20 Other studies raised concerns about the possibility of outdated clinical knowledge and cognitive bias in the experts’ decision-making.21 22 Evidence of context specificity has also been highlighted whereby the agreement between SCT scores derived using different scoring keys with expert reference panels from a different context (hospitals and specialty) was poor.23
Implementation of script concordance test in formative and summative assessments
The structure and layout of the SCT questions can easily be implemented in the usual pen and paper-based or online electronic format. Candidates answer each question (with five options) using a standardised answer sheet to facilitate computer scanning and scoring or directly online using the computer.
It is often difficult to get busy clinicians to meet together face-to-face to answer the questions. By uploading the questions online, the panellists can attempt them anytime and make the questions available through a secure online platform. The response data can then be collated and the weighted scores for responses on each score scale calculated.3
After capturing the candidates’ responses for all items, scoring of responses for each question can then be performed using the formula described above. This will ensure a rapid turnaround time that will be very effective in the assessment process.
For formative assessment purposes, expert panel consensus scores are provided to the candidates, followed by expert clinicians explaining and discussing the options in each scenario with the candidates for constructive feedback. Script concordance test can also be used to identify borderline students with suboptimal clinical reasoning skills for appropriate remedial measures such as bedside teaching, tutorials, or clinical simulations.24
For summative assessment purposes, particularly where there is not a large pool of SCT items, it is important to avoid constructing irrelevant variance in SCT scores, by not releasing or discussing post-examination, the expected responses (based on expert panel’s responses), and the associated score for each of the answer options in SCT items.
Unlike MCQ where there is only one single best answer that candidates could memorise and disseminate after the examination, the partial credit scoring model applied in SCT, where multiple answer options are accepted and each carries a fraction or all of the allocated mark, has to a certain extent rendered sharing of ‘correct’ answers after the examination difficult.
Developing quality script concordance test questions
Each clinical scenario has to be authentic and the presentation represents a realistic clinical encounter that is relevant to the specific discipline, preferably with a certain degree of uncertainty. The new information presented needs to stimulate the candidate to re-consider and re-evaluate how that particular piece of new information will affect the likelihood of the initial diagnosis, or appropriateness of initial planned investigation or management option. This will ensure the content validity in the SCT questions.
Particular care should be taken to develop options that will attract the full range of the five options available for the candidate to choose from. In other words, the additional pieces of new information should result in the consideration of -2 and +2 as well as -1, 0, and +1 options. A test-wise candidate might choose to consider only the options of -1, 0 and +1 if they notice that most panel consensus answers with a full score of 1 mark usually fall within these three options rather than also covering the -2 and +2.25 As a result, developing good-quality SCT questions is not easy. Care should be taken to develop clinical scenarios that do not focus solely on factual recall but involve a reasoning process with elements of uncertainty that will likely attract responses that spread across the 5-point Likert scale.26
Reliability and validity of script concordance test as an assessment tool
The reliability of SCT as an assessment tool has been investigated.2 6 A 60- to 90-minute examination will produce a Cronbach’s alpha of 0.70 to 0.85.7 25 27 28 There are concerns, however, about inter-panellist errors in SCT; the use of Cronbach’s alpha in measuring the reliability of the test where partial credit model of scoring is used, ie multiple options/responses are awarded either a full or fraction of allocated mark; and case scenarios that could create inconsistencies among items.
As an assessment tool, SCT has been shown to be valid in assessing clinical reasoning.13 14 19 28 Studies have shown that SCT scores correlate well with other assessment scores from the clinical years of the candidates.2
The construct validity of SCT questions can be examined by correlating the scores with the level of training to predict future performance on clinical reasoning. A recent study has compared the progression of clinical reasoning skills of medical students with those of a group of practising GPs who are also their Problem Based Learning group tutors.29 Another study showed that there was a statistically significant gain in SCT performance over a 2-year period in two different cohorts of medical students using the same set of 75-item SCT.26 There was significant progression of clinical reasoning skills from medical students at the novice level through to practising GP clinicians, reflected by the higher scores in the GP group attempting the SCT questions. Empirical evidence supporting the construct validity based on progression of SCT scores with clinical experience from undergraduate students to postgraduate training has also been reported.2 5 24 30 31 The construct validity of SCT has been questioned because of the logical inconsistencies as a result of partial credit scoring methodology making it possible for a hypothesis to be simultaneously more likely and less likely.32 Nonetheless, a certain degree of variability in panel scores has been shown to be a key determinant of the discriminatory power of the test and allows richness of thinking about clinical cases.33 34 Another study found that 27% of residents in one SCT administration scored above the expert panel’s mean, which may indicate issues with the construct validity, particularly in the credibility and validity of the scoring key and hence the resulting SCT scores interpretation.33
Test-wise candidates would select the answers to be around ‘0’ rather than ‘-2’ or ‘+2’ if they noticed that most panellist scores did not fall in the ‘extreme’ (-2 or +2) range due to the construct of the SCT questions and options. This could be avoided by first using the option descriptor of “much less likely (-2)” and “much more likely (+2)” rather than “ruling out the diagnosis” and “almost certain/definite diagnosis” as described in the format of SCT section above.19 Second, when collating the SCT questions into an examination paper, one could select a relatively equal number of items with both ‘extreme’ answers as well as median answers. Recent data have shown that by employing the above strategies in developing the paper, candidates who chose ‘0’ for all the questions would score only around 25% in the SCT examination and would gain no advantage (unpublished data). This is in contrast to the finding of another study wherein candidates who chose the midpoint scale (‘0’) performed better than the average candidate.32
The correlation of SCT scores with other modalities of assessment would be expected to be low as SCT is designed to measure clinical reasoning rather than factual or knowledge recall. The correlation coefficient between SCT and MCQ was poor (r=0.22), and that between SCT and extended matching questions (EMQ) was 0.46.4
Collating and moderating the expert reference panellist responses
In collating the SCT questions for use in a summative examination, appropriate clinical scenarios/vignettes with the related diagnoses, investigations, and management should be selected according to the blueprint of the assessment. The clinical topics should have a good spread and represent core areas of learning that are relevant to the curriculum and appropriate to the level of training of the candidates.
In reviewing the expert panel responses to each question, bi-modal and uniform divergence responses should trigger a detailed scrutiny of the clinical vignette and the options. In the case of bi-modal response (Fig 2a), the panel has an equal split of the best option between -2 and +2. This usually results from an error in the question or a controversial investigation or management option with discordant ‘expert opinions’. A modification of the question and re-scoring will usually solve this issue. If re-scoring results in the same bi-modal response, the question should be discarded for scoring in the examination. In the case of uniform divergence responses (Fig 2b), there is an equal spread in the number of members choosing all the five options. This usually signifies a non-discriminating question and the item should again be discarded. A discrete outlier response (Fig 2c) usually represents an error in the particular panellist’s decision or ‘clicking the wrong answer accidentally’ when the member should have answered -2 instead of +2. The ideal pattern would be relatively close convergence with some variation (Fig 2d).3
Figure 2. Number of responses from the expert panel to script concordance test questions
(a) Bi-modal response, (b) uniform divergence response, (c) discrete outlier response, and (d) ideal response
As mentioned previously, the set of questions in the SCT examination should be selected in such a way that there are similar numbers of full marks in each option across the five options. This will avoid the test-wise candidates being advantaged by selecting only the -1, 0, or +1 options and avoiding the extreme options of -2 and +2.3 By employing this strategy to select questions that cover the full 5-point Likert scale, test-wise students will only score 25% in the SCT examination if they choose the response of ‘0’ for all questions (unpublished data) compared with 57.6% in another cohort sitting a SCT test without the specific question selection process.32
Standard setting the pass/fail cutoff score
In setting the pass/fail cutoff score of the SCT questions, the expert panels’ mean scores and standard deviations are chosen to guide the process. This is calculated by asking all the members of the panel to attempt the same set of SCT questions and their responses are then scored accordingly. The borderline score of the undergraduate students is usually set at 3 to 4 standard deviations below the expert panel’s mean score.3 35 Studies have shown that using recent graduates or fellows in training might result in a mean score that is closer to the students’ mean and therefore a smaller number of standard deviations would be more appropriate.3
Other methods of standard setting include using the single correct answer method.29 36 Standard setting of a pass/fail cutoff score is an area that warrants ongoing research to inform and improve the practice of using SCT as a summative assessment tool for clinical data interpretation and decision-making skills.
The use of script concordance test in the Asia-Pacific region and its limitations
Examinations using SCT have been successfully implemented in the school-entry medical schools in Indonesia, Singapore, Taiwan, and Australia3 7 36 37; and in graduate-entry medical schools in Australia.29 38 Such test has the potential to supplement MCQ and SAQ to test the higher-order thinking of medical candidates to allow a more robust overall written assessment in the assessment programme. In fact, SCT is one of the few currently available assessment tools for clinical reasoning in a written format.28 It can be implemented relatively easily in the paper-based format or online. Initial pilot examinations can be set as a formative exercise to enhance candidates’ feedback and learning.24 Further collaboration with other institutions to develop, score, and share question items can ensure effective and efficient delivery of such examinations.
Limitations to the widespread usage of SCT could be due to: difficulties in developing good-quality SCT clinical scenarios, concerns about the validity of the test, recruiting a sufficient number of appropriate expert clinicians for the reference panel, lack of a general consensus in setting the borderline pass mark, and the candidates’ familiarity with the question format.3 24 25 28 32 34
This article attempts to review the current use of SCT in assessing clinical reasoning and data interpretation skills in undergraduate and postgraduate medicine. The empirical evidence reported for the reliability and validity of SCT scores from existing literature seems encouraging. Approaches to develop quality items, moderation of expert panel scoring and these post-hoc quality assurance measures, and optimisation of scoring scale will to a certain extent mitigate the threat to the validity of SCT score interpretation and its use for summative examination purposes. Combining SCT (testing the clinical reasoning and data interpretation skills with authentic written simulations of ill-defined clinical problems set at the ‘knows how’ level) with MCQ/SAQ/EMQ (testing the ‘knows’ and ‘knows how’), objective structured clinical examination (testing ‘shows how’), and workplace-based assessment (testing the ‘does’) in the medical curriculum will enhance the robustness and the credibility of the assessment programme.
Further research into the use of SCT in both undergraduate and postgraduate medical education is warranted, particularly on standard setting for the pass/fail cutoff score and best practices that may help reduce the threat to the validity of SCT scores.
1. Charlin B, Roy L, Brailovsky C, Goulet F, van der Vleuten C. The Script Concordance test: a tool to assess the reflective clinician. Teach Learn Med 2000;12:189-95. Crossref
2. Carrière B, Gagnon R, Charlin B, Downing S, Bordage G. Assessing clinical reasoning in pediatric emergency medicine: validity evidence for a Script Concordance Test. Ann Emerg Med 2009;53:647-52. Crossref
3. Duggan P, Charlin B. Summative assessment of 5th year medical students’ clinical reasoning by Script Concordance Test: requirements and challenges. BMC Med Educ 2012;12:29. Crossref
4. Kelly W, Durning S, Denton G. Comparing a script concordance examination to a multiple-choice examination on a core internal medicine clerkship. Teach Learn Med 2012;24:187-93. Crossref
5. Lubarsky S, Chalk C, Kazitani D, Gagnon R, Charlin B. The Script Concordance Test: a new tool assessing clinical judgement in neurology. Can J Neurol Sci 2009;36:326-31. Crossref
6. Meterissian S, Zabolotny B, Gagnon R, Charlin B. Is the script concordance test a valid instrument for assessment of intraoperative decision-making skills? Am J Surg 2007;193:248-51. Crossref
7. Soon D, Tan N, Heng D, Chiu L, Madhevan M. Neurologists vs emergency physicians: reliability of a neurological script concordance test in a multi-centre, cross-disciplinary setting. Neurology 2014;82(10 Suppl):327.
8. Talvard M, Olives JP, Mas E. Assessment of medical students using a script concordance test at the end of their internship in pediatric gastroenterology [in French]. Arch Pediatr 2014;21:372-6. Crossref
9. Miller GE. The assessment of clinical skills/competence/performance. Acad Med 1990;65(9 Suppl):S63-7. Crossref
10. Mehay R. The essential handbook for GP training and education. Chapter 29. Assessment and competence. Available from: http://www.essentialgptrainingbook.com/chapter-29.php. Accessed 10 May 2015.
11. Foucault A, Dubé S, Fernandez N, Gagnon R, Charlin B. Learning medical professionalism with the online concordance-of-judgment learning tool (CJLT): A pilot study. Med Teach 2014 Oct 22:1-6. Epub ahead of print. Crossref
12. Schmidt HG, Norman GR, Boshuizen HP. A cognitive perspective on medical expertise: theory and implications. Acad Med 1990;65:611-21. Crossref
13. Lubarsky S, Charlin B, Cook DA, Chalk C, van der Vleuten CP. Script concordance testing: a review of published validity evidence. Med Educ 2011;45:329-38. Crossref
14. Brazeau-Lamontagne L, Charlin B, Gagnon R, Samson L, Van Der Vleuten C. Measurement of perception and interpretation skills during radiology training: utility of the script concordance approach. Med Teach 2004;26:326-32. Crossref
15. Collard A, Gelaes S, Vanbelle S, et al. Reasoning versus knowledge retention and ascertainment throughout a problem-based learning curriculum. Med Educ 2009;43:854-65. Crossref
16. Lineberry M, Kreiter CD, Bordage G. Script concordance tests: strong inferences about examinees require stronger evidence. Med Educ 2014;48:452-3. Crossref
17. Williams RG, Klamen DL, White CB, et al. Tracking development of clinical reasoning ability across five medical schools using a progress test. Acad Med 2011;9:1148-54. Crossref
18. Gagnon R, Charlin B, Coletti M, Sauvé E, van der Vleuten C. Assessment in the context of uncertainty: how many members are needed on the panel of reference of a script concordance test? Med Educ 2005;39:284-91. Crossref
19. Dory V, Gagnon R, Vanpee D, Charlin B. How to construct and implement script concordance tests: a systematic review. Med Educ 2012;46:552-63.
20. Ahmadi SF, Khoshkish S, Soltani-Arabshahi K, et al. Challenging script concordance test reference standard by evidence: do judgments by emergency medicine consultants agree with likelihood ratios? Int J Emerg Med 2014;7:34. Crossref
21. Ramos K, Linscheid R, Schafer S. Real-time information-seeking behavior of residency physicians. Fam Med 2003;35:257-60.
22. Norman GR, Eva KW. Diagnostic error and clinical reasoning. Med Educ 2010;44:94-100. Crossref
23. Tan N, Tan K, Ponnamperuma G. Expert clinical reasoning is not just local but hyperlocal—insights into context specificity from a multicentre neurology script concordance test. Neurology 2015;84(14P4):191.
24. Ducos G, Lejus C, Sztark F, et al. The Script Concordance Test in anesthesiology: Validation of a new tool for assessing clinical reasoning. Anaesth Crit Care Pain Med 2015;34:11-5. Crossref
25. See KC, Tan KL, Lim TK. The script concordance test for clinical reasoning: re-examining its utility and potential weakness. Med Educ 2014;48:1069-77. Crossref
26. Humbert AJ, Miech EJ. Measuring gains in the clinical reasoning of medical students: longitudinal results from a school-wide script concordance test. Acad Med 2014;89:1046-50. Crossref
27. Gagnon R, Charlin B, Lambert C, Carrière B, Van der Vleuten C. Script concordance testing: more cases or more questions? Adv Health Sci Educ Theory Pract 2009;14:367-75. Crossref
28. Nouh T, Boutros M, Gagnon R, et al. The script concordance test as a measure of clinical reasoning: a national validation study. Am J Surg 2012;203:530-4. Crossref
29. Wan SH. Using Script Concordance Testing (SCT) to assess clinical reasoning—the progression from novice to practising general practitioner. Proceedings of the 11th Asia Pacific Medical Education Conference; 2014 Jan 15-19; Singapore.
30. Charlin B, van der Vleuten C. Standardized assessment in contexts of uncertainty: The script concordance approach. Eval Health Prof 2004;27:304-19. Crossref
31. Lambert C, Gagnon R, Nguyen D, Charlin B. The script concordance test in radiation oncology: validation study of a new tool to assess clinical reasoning. Radiat Oncol 2009;4:7. Crossref
32. Lineberry M, Kreiter C, Bordage G. Threats to validity in the use and interpretation of script concordance test scores. Med Educ 2013;47:1175-83. Crossref
33. Charlin B, Gagnon R, Lubarsky S, et al. Assessment in the context of uncertainty using the script concordance test: more meaning for scores. Teach Learn Med 2010;22:180-6. Crossref
34. Lubarsky S, Gagnon R, Charlin B. Scoring the Script Concordance Test: not a black and white issue. Med Educ 2013;47:1159-61. Crossref
35. Wan SH, Clarke R. Using a clinician panel to set the borderline mark for Script Concordance Testing (SCT) to assess clinical reasoning for graduating medical candidates. Proceedings of the 8th International Medical Education Conference; 2013 March 13-15; Kuala Lumpur, Malaysia.
36. Irfannuddin I. Knowledge and critical thinking skills increase clinical reasoning ability in urogenital disorders: a Universitas Sriwijaya Medical Faculty experience. Med J Indonesia 2009;18:53-9. Crossref
37. Tsai TC, Chen DF, Lei SM. The ethics script concordance test in assessing ethical reasoning. Med Educ 2012;46:527. Crossref
38. Ingham AI. The great wall of medical school: a comparison of barrier examinations across Australian medical schools. Aus Med Student J 2011;2:5-8.