Clinical Reasoning Skills

Detailed Steps in the Clinical Reasoning Process

1. Obtain and filter information.

   Information may be obtained primarily through reading, visual imagery, and listening.

   Other sensory input (e.g., tactile, olfactory) may be obtained.

2. Formulate an initial small set of hypotheses.

   This set of hypothesis is formulated in the context of identified questions and problems in the current case, as well as a knowledge base of prior cases (using schemas and pattern recognition).

   Experts quickly develop a small set of hypotheses with minimal clinical data to represent the problem to be solved. Short-term memory can actively handle only about 5 items at once.

   Experts will generally have the final diagnosis in this set within 5 minutes of starting.

   Novice and intermediate learners will take longer to develop a set of hypotheses.

3. Obtain additional information as directed by initial hypotheses.

   The initial small set of hypotheses forms a framework for additional focused information gathering. This process is repeated and refined. Novices and intermediates have more iterations of this process.

4. Use a reasoning strategy (deductive v. inductive) to process the information in the clinical context of the case.

   • Deductive reasoning works from general to specific. We develop hypotheses to explain a case problem and apply collected information to test the hypotheses in order to try and confirm or exclude a hypothesis. In a hypothetical-deductive process, a classic rank-ordered list of differential diagnoses is generated.

     The process goes: if - then - but - therefore (yes, no)

     If we have certain information, then certain hypotheses may be true, but we test against further information, and therefore it is true or not. This is akin to the scientific principle, in which one tries to prove a hypothesis.

   • Inductive reasoning works from specific to general. One starts with information from observations matched to an established pattern (algorithm) to come to a hypothesis. The hypothesis is then matched for fit to the problem in the case. Induction yields discoveries that are probable, but not proven. Inductive reasoning becomes powerful when an expert-derived algorithm is followed. The algorithms have been derived with statistical relevance to real cases.

   • The human body is very complex, and we cannot obtain all information we want, so that regardless of the reasoning process utilized, we can never absolutely prove or disprove most hypotheses in many cases. We derive the 'most likely' diagnosis, but we may need to eventually consider others if more information becomes available or the outcome is different than expected.

5. Perform an analysis of hypotheses by probabilistic and cause-effect means.

   May be utilized in either deductive or inductive reasoning processes:

   • Hypotheses are refined by Bayesian inference.

     Bayesian inference occurs when evidence or observations (such as population studies of disease) are used to determine the probability that a hypothesis may be correct. If tests are performed, such as laboratory tests, calculated results for test sensitivity, specificity, positive predictive value, and negative predictive value are useful in analysis.

   • Hypotheses are refined by cause-effect analysis to apply principles of pathophysiology (such as biomedical knowledge) and determine if a hypothesis is based upon a sound scientific basis.

   • 'Evidence-based medicine' is another description of this process.

6. Employ abstract ideas and concepts that are interpreted and used effectively.

   Avoid concrete thinking (child-like, literal interpretation; can't generalize).

   Avoid linear thinking (single unbranching series of cause and effect relationships).

7. Formulate a final diagnosis.

8. Test the final diagnosis.

   Test against positive and negative findings and standard criteria for description of a disease process.

   Working diagnoses for patient prognostic or therapeutic recommendations are finalized only after they are assessed for their adequacy in explaining all positive, negative, and normal clinical findings.

   The pathophysiologic reliability of the diagnosis is a check on the reasonableness of causal linkages between clinical events, ascertained from use of biomedical knowledge. Does the diagnosis fit with cause and effect? Is the diagnosis consistent with pathophysiologic principles?.

9. Consider other possible diagnoses.

   To diminish the possibility of premature closure, assume your working diagnosis is incorrect and then consider alternative diagnoses.

10. Evaluate the process. (Stop, Think, Act, Review)

11. Communicate the diagnosis.

12. Follow up.

    Clinical reasoning is improved when errors in information, judgment, and reasoning are discovered and discussed when reviewing the case. The quicker this happens, the greater the improvement.

Synopsis

The above detailed steps may not be immediately recognizable or flow in the same sequence in the context of actual clinical reasoning. Experts apply pattern recognition with non-analytic cognitive processing during the initial phases of considering a novel clinical case, then apply analytic processing in hypothesis testing. Novices may work the other way round. However, these two forms of reasoning can be interactive and not sequential. They are complementary contributors to the overall accuracy of the clinical reasoning process, each one influencing the other. Persons who use both (and start with non-analytic cognitive processing) perform better than persons using either non-analytic or analytic approaches alone.

References:

Eva KW. What every teacher needs to know about clinical reasoning. Med Educ. 2005;39(1):98-106.

Kassirer JP. Teaching clinical reasoning: case-based and coached. Acad Med. 2010; 85:1118-1124.