How does one become competent at making a diagnosis? Although a full understanding of cognitive factors leading to diagnostic competence remains incomplete, it is now possible to create more efficient and effective approaches to DDX instruction and assessment. These improvements in DDX education are based in part on seven particularly important cognitive and instructional sciences-based research findings describing the development of DDX competence.
A. Diagnostic competence is 'problem specific.' Competence at diagnosing one problem does not predict competence at diagnosing another problem. That is, observing that a student or physician correctly diagnoses a patient presenting with chest pain, does not enable one to predict whether that student/physician will correctly diagnose a second patient presenting with a different problem such as cephalgia, dyspnea, acute visual loss, etc. This is so because the development of diagnostic competence is dependent upon the acquisition of knowledge that is specific to the problem at hand. Without adequate knowledge relevant to the problem at hand, the likeliness that a student/physician will correctly diagnose patients presenting with that particular problem is markedly reduced.
B. Diagnostic competence is 'disease-specific.' Within a given problem (e.g., chest pain), a student/physician correctly diagnosing one patient presenting with disease 'x' does not predict competence in diagnosing a second patient with the same problem, but due to a different disease - disease 'y'. That is, the ability to correctly diagnose one disease that causes a particular problem does not convey information about the likelihood of a student/physician correctly diagnosing a different cause of the same problem.
For example, assume for a moment that two patients present with chest pain. A student/physician correctly diagnosing chest pain case #1 as due to myocardial infarction does not enable an observer to assume that the same student/physician will also correctly diagnose patient #2, a patient with chest pain due to a different disease/disorder such as pneumonia, pulmonary embolus, dissecting thoracic aneurysm, etc. Without disease-specific knowledge (i.e., knowledge of that disease's particular 'pattern' of both characteristic and distinguishing signs and symptoms), diagnostic competence is not possible.
Thus far we have described the fact that diagnostic competence depends upon both problem and disease specific knowledge. That is, competence against one problem (chest pain) does not predict competence in a different problem (dyspnea). Furthermore, the development of diagnostic competence for a given problem, and its associated disease differentials, requires knowledge of the pattern of signs and symptoms associated for each of the common and important disease differentials likely to cause a given problem.
C. Diagnostic competence is a function of 'case typicality.' Assume for a moment that a student/physician has correctly diagnosed a patient presenting with the problem with acute chest pain due to myocardial infarction. Could an independent observer safely assume that that same student/physician would also correctly diagnose a different patient who also presented with the problem of acute chest pain due to myocardial infarction? Are all cases of a given disease equally likely to be correctly diagnosed by a student/physician who has correctly diagnosed one previous case portrayal of that same disease? The simple answer is no.
For example, evidence that a student correctly diagnosed an easy (typical) presentation of a given disease can not be used as evidence that they would also correctly diagnose a more difficult (less typical) presentation of that same disease. Rather, research has demonstrated that diagnostic accuracy is a function of the case's 'typicality'. That is, the more typical the case (the more closely the pattern of signs and symptoms in a given case approximates the 'prototypical pattern' of signs and symptoms that characterize a given disease), the more likely it will be correctly diagnosed. The less typical the case presentation (the fewer the number of prototypical signs and symptoms in the case), the less likely the case will be correctly diagnosed.
Medical educators have long recognized that for each patient problem (e.g., chest pain, dyspnea, etc) and disease (myocardial infarction, appendicitis, etc), the development of diagnostic competence requires substantial exposure to multiple case portrayals representing the various ways (typical through atypical case presentations) with which any given disease might manifest itself. Unfortunately, both preclinical students and students on clinical rotations have an insufficient number and variety of case presentations of various: 1) clinical problems, 2) diseases, and 3) case portrayals (ranging from prototypical through atypical for any given disease) to develop a rudimentary set of problem and disease-specific DDX competencies.
D. Differential diagnosis involves 'pattern recognition.' Cognitive sciences research has established a clear distinction between knowing 'what' and knowing 'how.' Knowing 'what' is called declarative knowledge — knowledge that can be memorized and simply recalled. Knowing 'how' is called procedural knowledge. The development of procedural knowledge will not likely proceed in the absence of declarative knowledge, but nonetheless, procedural knowledge (e.g., how to act, how to proceed, how to apply, etc) requires lots of practice. Reading and memorization is not doing. One can read much about how to ski and thus talk skiing, but can one actually ski, and if so, how well?
The declarative knowledge base (the 'what') students must acquire in order to diagnose a given medical problem consist of knowledge of: 1) the most common and/or important disease differentials most likely to case the problem, and 2) the sign/symptom pattern associated with each of these disease differentials.
The 'what' component of DDX competence — knowledge of the disease differentials, and, the pattern of signs and symptoms characterizing a particular disease, represent basic facts that must be stored in memory. However, acquiring and memorizing the declarative component (the 'what') of diagnostic competence is very different from developing competence in 'how' to apply that knowledge during a diagnostic workup (developing procedural competence). Simply put, it is the ability to efficiently and effectively apply this declarative knowledge of differentials and disease patterns (knowing 'how' - the development of procedural knowledge and skills) that accelerates the development of rudimentary levels of diagnostic competence.
To better understand how clinicians apply their problem and disease-specific declarative knowledge, it is useful to first recognize that medical practitioners perform DDX in large part (but certainly not exclusively) via cognitive processes usually referred to as pattern recognition. In order to perform DDX via pattern recognition, the clinician must have a large declarative knowledge base consisting of knowledge of both the several diseases most likely to cause a given problem, and, for each of those disease differentials, knowledge of their associated signs and symptoms.
However, during the act of differential diagnosis, the clinician must be able to utilize both declarative knowledge, and, procedural knowledge, in order to identify the disease that best accounts for the signs and symptoms that are in the case at hand. Cognitive sciences research suggests that the act of differential diagnosis (or pattern recognition) appears to involve the simultaneous application of two distinct yet interrelated 'procedural' skills: pattern matching and pattern discrimination.
Pattern matching enables the creation of a 'rule-in list' consisting of those diseases most likely causing a problem. Pattern matching is the ability to 'scan for and recognize' those signs and symptoms in the patient case that best match the signs and symptoms associated with one or more diseases represented within their own, problem-specific declarative knowledge base. This scanning for matching signs and symptoms (S/S) enables the clinician to create a small list of 'rule-in' (R/I) disease differentials — a list representing the more likely causes of the patient's problem.
For example, during the history and physical exam (the data gathering/scanning phase of DDX), a number of positive signs and symptoms will be identified. These positive findings serve as signals that the clinician subsequently uses to probe their declarative knowledge base of diseases associated with the problem at hand. By recalling those disease that have some of the same S/S found during the history and physical, the clinician can begin to reduce their list of all known possible disease differentials for the problem, to a smaller list of more likely disease differentials. Thus begins the R/I process — the process of reducing a large list of all possible disease differentials, to a smaller list of more likely disease differentials.
Pattern discrimination starts with the rule-in list (the reduced list of more likely causes), but then enables its further reduction via a 'rule-out' (R/O) process. That is, pattern discrimination enables further differentiation among the reduced list of more likely causes to one or perhaps two most likely disease etiologies. Pattern discrimination involves using declarative knowledge representing those S/S that are not commonly associated with one or more of the reduced list of differentials.
For example, if disease A and B are both associated with S/S 1, 2, 3, 4, and, all four of those S/S are present in the given case, the clinician will create a R/I list of diseases A and B with which to begin their final differentiation. The clinician must then look to their declarative knowledge base to identify which additional S/S are associated with either disease A or B - but not the other. For example, if disease A is associated with S/S 5 and 6, and, disease B is associated with S/S 7 and 8, then the clinician will now look to the case to see whether S/S 5, 6, 7 or 8 are present. If S/S 5 and/or 6 are present while S/S 7and 8 are not present, the clinician will like decide disease A is the most likely cause. If S/S 7 and/or 8 are present while S/S 5 and 6 are not present, the clinician will likely decide that disease B is the most likely cause. Simply put, pattern discrimination involves looking for S/S that are more closely associated with one disease and not another.
E. The development of diagnostic competence requires 'Deliberate Practice'. These two procedural skills of pattern matching and pattern discrimination only develop via practice, practice and more practice. Like skiing or riding a bike, one can read endlessly (acquire more and more declarative knowledge) about how to ski or ride, but it is only through actual practice on the slopes or a road/trail that one becomes competent at the task.
Deliberate practice at, and training in, pattern matching and pattern discrimination appears to be one critical factor underlying the development of DDX competence. Practice at, and training in, pattern matching and pattern discrimination enables the student/physician to learn 'how' to use their declarative knowledge base — how to rapidly and reliably scan for, and identify the most important characteristic and discriminating elements contained within your declarative knowledge base — the S/S — the 'what' that is present and absent both in the case at hand and in your mind's list of disease-specific S/S.
Deliberate practice supports learning how to see the disease hidden within the patient's case presentation. Deliberate practice supports learning how to first match, and then further discriminate, among those possible matches, the most likely etiology. Deliberate practice provides the basis for developing increasingly reliable (accurate) diagnostic capabilities.
The bottom line is that the development of pattern recognition-oriented 'procedural' skills (knowing 'how') can only evolve via multiple opportunities to apply and reflect upon the utility and accuracy of your evolving, problem-specific 'declarative' knowledge base. One cannot develop without the other. More will be said about how KBIT uses multiple practice cases to support the development and refinement of pattern matching and pattern discrimination capabilities in the following.
F. The development of diagnostic competence requires immediate, individually-tailored formative feedback. How does a learner know if their evolving declarative knowledge and procedural skills are developing in the right direction? The simple answer is feedback.
The efficient and effective development of competence - in any endeavor - involves not only effort on the part of the learner but also effort on the part of the instructor. The more immediate the instructor's feedback, and, the more tailored it is to the learner's specific errors; the more efficient and effective the student's learning experience.
There is reason to believe that during the early developmental stages of DDX competence, diagnostic errors are largely the result of errors involving declarative knowledge. Errors involving: 1) the initial storage and/or retrieval of the S/S most closely associated with a given disease pattern (pattern matching-related declarative knowledge), and/or 2) appreciating and utilizing those S/S that best differentiate one disease from another (pattern discrimination-related declarative knowledge). While more will be said about how KBIT tutorials provide students with pattern recognition-oriented feedback later on, it is sufficient to simply affirm at this point that immediate, individually tailored feedback is critical to efficient and effective learning.
G. Evidence of the attainment of diagnostic competence requires the establishment of performance criteria — criteria defining a minimal level of acceptable performance. Traditional medical training programs have little, if any data or evidence, with which they might document that a given student has attained any minimal DDX performance criteria. There are many reasons why this is so. One reason is that for decades, training programs have assumed that DDX competence was dependent upon the development of 'general problem solving skills'. The inability to produce any evidence that such a skill set even plays a role in DDX, let alone that it might be measurable, has largely enabled training programs to avoid/evade measuring this alleged skill of general problem solving. If such a skill set existed and was measurable, then students would regularly receive feedback regarding their 'problem solving skills'. When was the last time a medical student received, or a faculty member provided students a grade for general problem solving?
Research findings consistently demonstrating that DDX competence is both knowledge-based, and, problem and disease-specific, make clear that DDX performance measures can not be based upon some 'generalized/global assessment of overall DDX performance' such as general problem solving. Rather, this body of research demonstrates that any attempt to produce reliable and valid measures of DDX capabilities must be predicated upon the use of some large number and variety of problem and disease-specific test cases. Furthermore, it is clear that DDX performance capabilities should be predicated upon and reported in terms of performance on a problem by problem basis. Only by attempting to produce problem-specific DDX performance measures can training programs be in a position to begin to establish problem-specific pass/fail competence criteria.
Many years and perhaps even decades of research must be conducted in order for medical training programs to be in position to establish 'national' DDX performance criteria on the 100 or so most common/important primary care problems. Until that day comes, KBIT provides faculty and students here and now with an approach that enables any training program to at very least, establish local/institutional DDX performance criteria on a problem by problem basis.