Clinical reasoning: how to build it in the clinical cycle
Doing well on written exams and still feeling that clinical reasoning is missing is a common turning point in the first year of internship. On the exam, the case comes summarized and the answer choices point the way. On the ward, the patient tells the story out of order, test results arrive little by little, and someone expects you to say what you think. The knowledge is the same; what changes is how you use it.
NEXO is a daily medical diagnosis game: one clinical case a day, six progressive clues, and ICD-10 proximity feedback across five levels.
This text shows how to develop clinical reasoning in the clinical cycle with four tools that studies describe: illness scripts, problem summary with semantic qualifiers, prioritization of hypotheses and practice with feedback. Cognitive biases and the step by step of differential diagnosis are in the text about differential diagnosis with clues; here the focus is the foundation that comes before.
Main points
- Clinical reasoning combines fast pattern recognition and slow analysis; both modes have been shown to be equally effective in studies.
- Illness scripts are how clinical knowledge is organized in memory: who gets it, how it starts, what appears and why.
- A good problem summary uses semantic qualifiers, such as acute or chronic, and has a three part structure: who the patient is, how the condition evolved and what the syndrome is.
- Prioritizing means ordering hypotheses by probability and by severity, without leaving out those that require immediate action.
- Deliberate practice with immediate feedback and repetition is among the most supported paths to gaining experience.
Summary
- Fast thinking, slow thinking and illness scripts
- Summarizing the problem
- Prioritizing hypotheses
- Integrating theory and case
- Practice with feedback
- A roadmap for the internship
- Frequently asked questions
- Clinical reasoning checklist
Fast thinking, slow thinking and illness scripts
Clinical reasoning is the process of gathering a patient's data, arriving at hypotheses and deciding what to do. Studies describe two modes that work together in this process.
Two modes, both useful
The dual process model describes one fast, intuitive system and another slow, analytical one (Croskerry, 2009). A review of clinical reasoning studies concluded that the two processes are equally effective, and that the fast mode, despite depending on each person's experience, is not more error prone nor more subject to biases than the analytical one (Norman, 2009). The same review reports that guiding clinicians to use both modes explicitly consistently reduced errors (Norman, 2009).
For the student, this takes a weight off: intuition is not the enemy of reasoning. What is missing at the start of the internship is material for it to work with, that is, cases and patterns stored in memory. The biases that can hinder both modes are in the text about differential diagnosis.
What an illness script is
Script theory explains how diagnostic knowledge is organized: they are networks of knowledge stored in memory, assembled for clinical tasks, with the expected and unexpected values for each feature of a disease (Charlin et al., 2000). When a new case appears, the activated script creates expectations: it tells you which findings to look for to confirm or rule out the hypothesis and what to do next (Charlin et al., 2007).
In practice, an illness script usually answers four questions:
- Who gets it? Risk factors, age, context.
- How does it start and evolve? Time of onset, course, pattern.
- What appears? Typical symptoms, signs and exam findings.
- Why does it happen? The mechanism that links the cause to the findings.
How scripts form
According to the review on the development of expertise, biomedical knowledge becomes encapsulated with clinical experience and, gradually, integrated into illness scripts (Schmidt and Rikers, 2007). The authors recommend teaching basic sciences in a clinical context and presenting patient problems early in the course to support this process (Schmidt and Rikers, 2007). Thirty years after the introduction of the concept, a historical review concluded that it has a solid basis and serves both to understand expertise and to guide clinical teaching (Custers, 2015).
Does teaching with scripts work?
In a randomized trial in Hong Kong, fourth-year students who took part in a clinical reasoning workshop with illness scripts, web-based problems, individual feedback and tutors reasoning out loud scored 14% higher on clinical problem solving than the control group (Lee et al., 2010). On a questionnaire about diagnostic thinking style, the groups were the same (Lee et al., 2010). The gain showed up in performance with cases, not in the way students described their own thinking.
From the exam case to the patient
On the written exam, the case usually comes with the typical findings already selected, which activates the script easily. In the patient, the findings come mixed with irrelevant information and sometimes the most typical ones are exactly what is missing. So a good script also includes the variations: how the disease appears in older adults, in those taking certain medications, at the start of the picture. Every atypical patient you follow is a chance to add those variations.
Build your scripts
A simple way to start is to write, for each disease you study in the clerkship, a short card with the four questions. Then compare the card with the patient you saw: what was typical, what strayed from the pattern. Each patient adds a variation to the script, and it is that collection of variations that forms clinical intuition.
Summarizing the problem
Before thinking about hypotheses, you need to say, in one or two sentences, what the problem is. This summary, called the problem representation or summary statement, is the point where loose data become a picture.
Semantic qualifiers
Semantic qualifiers are pairs of abstract, opposite terms that turn a specific finding into a characteristic comparable across diseases: acute or chronic, continuous or intermittent, localized or diffuse, first episode or recurrent. The problem representation built with semantic qualifiers was linked to better diagnostic results in earlier studies, and a group of researchers tested whether it could be taught (Nendaz and Bordage, 2002).
In the study, second-year students trained to turn findings into qualifiers and use them to compare hypotheses used more than twice as many qualifiers in their summaries (1.40 versus 0.63) and remembered the collected findings better (Nendaz and Bordage, 2002). Diagnostic accuracy, however, was the same between the groups, and the authors concluded that qualifiers are a necessary tool, but not enough when used alone (Nendaz and Bordage, 2002).
The three-part structure
In a study that compared summaries by beginning residents and by internal medicine attendings, summaries with a three-part structure (patient data, course over time and clinical syndrome) and with time qualifiers were linked to more diagnostic accuracy, regardless of experience (McQuade et al., 2024). The experts included fewer irrelevant comorbidities and more exam findings, and synthesized the data into higher-level concepts, such as sepsis, instead of listing each finding (McQuade et al., 2024).
Examples of a summary
The examples below are editorial guidance, to show the shape of the summary. They do not replace clinical assessment.
| Beginner version | Version with qualifiers and three parts |
|---|---|
| 58-year-old man, hypertensive, diabetic, with dyslipidemia, who smokes, came in with chest pain that started 40 minutes earlier when climbing the stairs, with sweating, and the pain did not go away | 58-year-old man with several cardiovascular risk factors, with acute chest pain, pressure-like, triggered by exertion and persistent, with sweating |
| 25-year-old woman with cough for 3 days, fever, yellow sputum, had a chest X-ray | Young woman, previously healthy, with an acute picture of fever and productive cough |
| Older adult with tiredness, weight loss and anemia on tests | Older adult with a chronic picture, worsening over the months, of asthenia and weight loss, with anemia |
Notice what changes: the data become fewer and more abstract, the time course appears as a qualifier and the syndrome becomes clear. A good summary already suggests which illness scripts to activate.
Common mistakes in the summary
Some errors appear often when you start summarizing cases. The first is repeating the whole story, in the order it was told, instead of selecting what matters. The second is including every comorbidity, even those unrelated to the problem; the experts, in the study of summaries, included fewer irrelevant comorbidities than the beginners (McQuade et al., 2024). The third is putting a diagnosis in the summary, which closes the door to alternatives too early; the experts also did this less often (McQuade et al., 2024).
A quick test helps: if someone else reads your summary, can they picture the patient and think of the same three hypotheses as you? If so, the summary did its job.
Evaluating your own summary
A rubric developed to evaluate student summaries in virtual patient cases uses five components: accuracy of the facts, appropriate narrowing of the differential, conversion of information into abstract terms, use of semantic qualifiers and an overall score (Smith et al., 2016). You can use the same criteria to review your summaries after each case.
An honest caveat is warranted: in an OSCE with ten stations, asking students to write a summary before the diagnosis did not increase accuracy, and better summaries did not correlate with more accuracy (Valentine et al., 2022). The summary organizes the reasoning, but accuracy also depends on knowledge of the diseases it should activate.
Prioritizing hypotheses
With the problem summarized, the hypotheses appear. The next step is to order them. The order decides what you investigate first and what needs immediate action.
The most likely
The first question is which hypothesis best explains the set of findings, taking into account the frequency of the disease in that context. A common disease with a typical presentation usually comes before a rare one with a similar presentation. The text on differential diagnosis shows how each finding raises or lowers that probability.
The most serious
The second question is which hypothesis, if not recognized now, can cause serious harm. A reasoning model for common problems in general practice explicitly includes a safety strategy to consider life-threatening conditions and other serious diseases during diagnosis (Murtagh, 1990). Checklists for diagnosis follow the same logic and were proposed to avoid the most common cause of error, which is not considering the correct diagnosis; they have not yet been rigorously evaluated (Ely et al., 2011).
The treatable one
The third question is which hypothesis has a treatment that changes the outcome if started early. A treatable disease with a short window deserves a high place on the list, even when it is less likely. This third question is editorial guidance, common in clinical teaching, and combines with the two previous ones: the goal is not to miss what you can resolve.
When the three questions point in different directions
Often the most likely hypothesis is not the most serious. In that case, the list keeps both at the top: the most likely one guides the main investigation, and the most serious one gets a question or a finding that confirms or rules it out soon. Saying this out loud in the case presentation, for example "I think first of X, but I need to rule out Y", shows the preceptor that you considered both.
Putting the three questions together
| Question | What it protects | Example of reasoning (editorial) |
|---|---|---|
| Which is the most likely? | Efficient investigation | In exertional chest pain in a patient with risk factors, acute coronary syndrome comes at the top |
| Which is the most serious? | Patient safety | Aortic dissection and pulmonary embolism go on the list even when less likely |
| Which is treatable and has a short window? | Outcome | Hypotheses with time-dependent treatment require rapid confirmation or exclusion |
Integrating theory and case
The basic cycle gave the mechanism; the clinical cycle asks for the name of the condition. Connecting the two is part of the reasoning, not an extra.
Why basic science helps
A review on the role of biomedical knowledge in clinical reasoning concluded that it helps the beginner form a coherent and stable representation of disease categories (Woods, 2007). With this representation, the student retains clinical knowledge for longer and maintains diagnostic accuracy in the face of difficult cases (Woods, 2007). The review's recommendation is that clinical teaching make explicit links between biomedical knowledge and clinical facts (Woods, 2007).
How to make the link
When building the script for a disease, write the fourth question, "why does it happen?", with the mechanism that explains each finding. If heart failure causes edema, write why. If a medication causes an adverse effect, link it to the mechanism. The mechanism is what helps you recognize the disease when it appears atypically.
Cases early, mechanism always
Presenting patient problems early in the course and teaching basic sciences in a clinical context support the formation of scripts (Schmidt and Rikers, 2007). For those already in the clinical cycle, the reverse path also works: with each patient, go back to the mechanism. The text on learning based on clinical cases shows how to study from cases.
Practice with feedback
Experience alone does not guarantee good performance. What makes a difference is the type of practice.
Deliberate practice
Studies on expertise show a weak relationship between time of experience and observed performance; the performance of experts is linked to deliberate practice (Ericsson, 2008). This practice focuses on improving specific tasks and includes immediate feedback, time to solve problems and evaluate, and opportunities to repeat to refine performance (Ericsson, 2008).
Feedback changes performance
In a systematic review on assessment and feedback in the clinical performance of physicians, 32 of 41 studies, or 74%, showed a positive impact of feedback (Veloski et al., 2006). Feedback was more effective when it came from a source with authority and was maintained over time (Veloski et al., 2006).
How to ask for feedback during the clerkship
The most useful feedback is specific. Instead of asking the preceptor whether the case was right, ask about one step: "did my summary leave out something important?", "does the order of my hypotheses make sense?", "which finding would you have looked for first?". Questions like these are quick to answer in the middle of the routine and show exactly where to adjust. Write the answer in the disease script, so that today's feedback becomes knowledge with the next patient.
Reflect on the case
With clinical cases, reflecting in a structured way on the findings for and against each hypothesis yielded more correct answers one week later than giving a diagnosis alone, in fourth-year students (Mamede et al., 2014). The case feedback, combined with this reflection, closes the cycle of deliberate practice.
Train with cases in clues
In NEXO, each case arrives in up to six clues, and each guess shows the proximity in ICD-10 at five levels. This immediate feedback tells you whether you are in the right group of diseases and how far off the guess was, which helps separate a script error, in which you confused neighboring diseases, from a representation error, in which the problem was summarized in the wrong direction. In the third clue the area of the case appears and, in the fifth, the ICD range; you can skip a clue whenever you want to move forward.
The Daily mode brings one case per day, the same for everyone, which allows you to discuss the reasoning with the class. The Archive gathers previous cases, Multimedia brings cases with image, audio and other exams, and Continents, cases linked to Africa, the Americas, Asia, Europe and Oceania. Adrenaline, with the dose chosen beforehand, trains reasoning with time. The Anki button creates the card straight from NEXO and opens it in the Anki app, searches for the cards you already have and, when the case has an image, takes the image to the card. To understand the limits of each mode, see the NEXO economy page.
A roadmap for the clerkship
The four tools work best together. The roadmap below is an editorial guide to use with each patient you follow and each case you train.
With each patient
- After the history and the exam, write the summary of the problem in one or two sentences, with semantic qualifiers and the three parts.
- List three hypotheses and mark which is the most likely, which is the most serious and which is treatable with a short window.
- For each hypothesis, note the finding that supports it and the one that contradicts it.
- Present the case to the preceptor in that order and ask for feedback on the summary and on the list.
- At the end of the day, update the scripts of the diseases that appeared.
With training cases
Use short cases to practice the same script with immediate feedback. For each case, compare your summary with the final diagnosis and ask which qualifier would have led to the right script. One case per day, every day, adds up to many patterns over the course of the clerkship.
In the discussion with the class
Discussing the same case with colleagues shows other paths to the same answer. The NEXO ranking has the Global, Universities and Group tabs, the latter with an invite code, which makes it easier to gather the class or the league around the case of the day. If you coordinate a medical school, league or residency and want to use cases like this in teaching, get to know NEXO Closer. Medical education researchers can get to know NEXO Researcher, and doctors who want to review cases, NEXO Curator. Those who produce Medicine content and want to bring cases like this to their audience can get to know NEXO Creators.
Frequently asked questions
What is clinical reasoning?
It is the process of gathering a patient's data, forming hypotheses and deciding what to do. It combines a fast mode, of pattern recognition, and a slow, analytical mode, which have proven equally effective in studies (Norman, 2009).
How to develop clinical reasoning in the clinical cycle?
Build scripts of the diseases you see, summarize each case with semantic qualifiers, prioritize hypotheses by probability, severity and treatment and practice many cases with immediate feedback. Deliberate practice, with feedback and repetition, is what expertise studies associate with good performance (Ericsson, 2008).
What are illness scripts?
They are knowledge structures stored in memory that gather who usually has the disease, how it begins and evolves, what appears and why (Charlin et al., 2000). They create expectations that guide what to look for in a new case (Charlin et al., 2007).
What are semantic qualifiers?
They are pairs of opposite and abstract terms, such as acute or chronic and continuous or intermittent, that convert findings into characteristics comparable across diseases. Training their use increased the number of qualifiers in summaries and the recall of findings (Nendaz and Bordage, 2002).
How to write a good summary of the problem?
Use three parts: who the patient is, how the condition evolved over time and what the syndrome is. Summaries with this structure and with time qualifiers were associated with more diagnostic accuracy (McQuade et al., 2024).
Is experience in the clerkship enough to develop reasoning?
Experience helps, but time in practice, by itself, had a weak relationship with the performance observed in expertise studies (Ericsson, 2008). What matters most is practicing specific tasks, receiving immediate feedback and repeating. In the clerkship, this means summarizing each patient, prioritizing hypotheses and asking for feedback on these steps, and not just waiting for cases to pile up.
Does basic science still matter in the clinical cycle?
It does. Biomedical knowledge helps the beginner form a coherent representation of diseases and maintain accuracy in the face of difficult cases (Woods, 2007).
How does NEXO help train clinical reasoning?
Each case arrives in up to six clues, and the guess shows the proximity in ICD-10, an immediate feedback on the distance between your reasoning and the diagnosis. With one case per day, you practice summary, prioritization and review of the hypothesis. To get to know the game, read what NEXO is; to fit it into your study, the text on study routine.
Clinical reasoning checklist
- Write the summary of the problem before listing hypotheses.
- Use semantic qualifiers: acute or chronic, continuous or intermittent, localized or diffuse.
- Follow the three parts: patient, time and syndrome.
- List three hypotheses and mark the most likely, the most serious and the treatable one.
- Note the finding for and against each hypothesis.
- Link each finding to the mechanism that explains it.
- Update the scripts for the diseases you saw during the day.
- Ask for feedback on the summary and on the list, not just on the diagnosis.
- Practice short cases with immediate feedback.
- Discuss with the group the path, not just the answer.
Want to practice today? Play the case of the day on NEXO.
References
- Croskerry P. A universal model of diagnostic reasoning. Acad Med. 2009. https://pubmed.ncbi.nlm.nih.gov/19638766/
- Norman G. Dual processing and diagnostic errors. Adv Health Sci Educ Theory Pract. 2009. https://pubmed.ncbi.nlm.nih.gov/19669921/
- Charlin B, Tardif J, Boshuizen HP. Scripts and medical diagnostic knowledge: theory and applications for clinical reasoning instruction and research. Acad Med. 2000. https://pubmed.ncbi.nlm.nih.gov/10693854/
- Schmidt HG, Rikers RM. How expertise develops in medicine: knowledge encapsulation and illness script formation. Med Educ. 2007. https://pubmed.ncbi.nlm.nih.gov/18004989/
- Woods NN. Science is fundamental: the role of biomedical knowledge in clinical reasoning. Med Educ. 2007. https://pubmed.ncbi.nlm.nih.gov/18045369/
- Charlin B, Boshuizen HP, Custers EJ, Feltovich PJ. Scripts and clinical reasoning. Med Educ. 2007. https://pubmed.ncbi.nlm.nih.gov/18045370/
- Custers EJ. Thirty years of illness scripts: Theoretical origins and practical applications. Med Teach. 2015. https://pubmed.ncbi.nlm.nih.gov/25180878/
- Lee A, Joynt GM, Lee AK, et al. Using illness scripts to teach clinical reasoning skills to medical students. Fam Med. 2010. https://pubmed.ncbi.nlm.nih.gov/20373168/
- Nendaz MR, Bordage G. Promoting diagnostic problem representation. Med Educ. 2002. https://pubmed.ncbi.nlm.nih.gov/12191059/
- McQuade CN, Simonson MG, Lister J, et al. Characteristics differentiating problem representation synthesis between novices and experts. J Hosp Med. 2024. https://pubmed.ncbi.nlm.nih.gov/38528679/
- Smith S, Kogan JR, Berman NB, et al. The Development and Preliminary Validation of a Rubric to Assess Medical Students' Written Summary Statements in Virtual Patient Cases. Acad Med. 2016. https://pubmed.ncbi.nlm.nih.gov/26726864/
- Valentine J, Tackett S, Bord S, et al. Problem representation and diagnostic accuracy during an OSCE. Adv Health Sci Educ Theory Pract. 2022. https://pubmed.ncbi.nlm.nih.gov/35859246/
- Murtagh J. Common problems: a safe diagnostic strategy. Aust Fam Physician. 1990. https://pubmed.ncbi.nlm.nih.gov/2346427/
- Ely JW, Graber ML, Croskerry P. Checklists to reduce diagnostic errors. Acad Med. 2011. https://pubmed.ncbi.nlm.nih.gov/21248608/
- Ericsson KA. Deliberate practice and acquisition of expert performance: a general overview. Acad Emerg Med. 2008. https://pubmed.ncbi.nlm.nih.gov/18778378/
- Veloski J, Boex JR, Grasberger MJ, et al. Systematic review of the literature on assessment, feedback and physicians' clinical performance: BEME Guide No. 7. Med Teach. 2006. https://pubmed.ncbi.nlm.nih.gov/16707292/
- Mamede S, van Gog T, Sampaio AM, et al. How can students' diagnostic competence benefit most from practice with clinical cases? The effects of structured reflection on future diagnosis of the same and novel diseases. Acad Med. 2014. https://pubmed.ncbi.nlm.nih.gov/24280855/