Making a diagnosis is one of my most common challenges. Families and patients are understandably more interested in the therapy to alleviate illness and suffering than exact diagnosis. We are swept along with this laudable therapeutic objective in the current of prompt empiric therapy bringing relief and hopefully confirming an empiric diagnosis and buying time before more definitive testing. I object to this approach even as I engage in it. Necessary in many situations, we become accustomed to “symptom, pattern, therapy” thought process leaving diagnosis for a second pass, if things do not work out. Syndromic diagnosis is not diagnosing a specific syndrome. Consider use of the pneumonia protocol. There is perfect adherence to the protocol including cultures, CXR, approved antibiotic and short door-to-needle time except the patient doesn’t have pneumonia. I am not criticizing. It is easy for the consultant to arrive later and say, “Not this, not that.” I strive primarily for diagnosis because having the diagnosis, therapy is either evident or easily obtained. Let’s examine Bayes’ theorem, Ockham’s razor, tempo, intuition, the management of uncertainty and the contributions of history, physical exam and imaging to differential diagnosis. 60 percent of diagnoses come from the history. “Given enough time, the patient will tell you the diagnosis” is an adage. The patient may not state an ICD-10 qualified diagnosis but will say something is tipping the experienced observer to a diagnosis. My patient had a high fever, severe malaise, joint inflammation with negative tap, high WBC and no response to a variety of antibiotics. I had no options left after an extensive workup except to go back to the history. “Tell me again. When did this all start?” “It began after my pet rat bit my finger…” Fever? Rat Bite? Rat-bite fever: Streptobacillus moniliformis! How could I have missed it? I learned about the Rat Fanciers Society, their website and strange things that I shall not relate. “The patient is always right” is another saying if we can see the portion of truth in their assertions. One patient arrived in the ER with drooping eyelids and said: “I have botulism.” I asked: “How do you know?” “I had it once before. We make a delicacy by shaking cayenne pepper flakes and salt with raw fish leaving it in a ziplock bag on the windowsill for a month. Occasionally there is a problem. Get the antiserum from Kennedy Airport.” Can it always be that easy? Other patients need forceful encouragement. “Do you have any medical problems?” “No.” “Do you have diabetes?” “Yes.” “How long have you had diabetes?” “Since my heart transplant.” “Why did you have a heart transplant?” “I went to Long Island and got a virus.” I borrow a classification for grading diagnoses from our colleagues in law. First are purely conjectural diagnoses. Next, likely diagnoses require a civil standard — “the preponderance of the evidence.” A confirmed diagnosis rests on: “beyond a reasonable doubt” the standard of a criminal trial. 15 percent of the diagnoses come from the physical exam, so we are under water if we don’t have a good differential after the H&P. Another 15 percent of diagnoses come from lab and imaging even though patients (and sometimes doctors) believe that the most definitive part of the diagnostic process is resulted, cleanly printed, in a standard format, on paper or computer screen. “What is bothering you?” I may ask. “My MRI says that I have…” 5 percent of diagnoses are confirmed by a trial of empiric therapy, hopefully in conjunction with a vigorous workup with robust differential. 5 percent of diagnoses are never discovered. The patient either mysteriously recovers or dies. We think in terms of differential diagnosis, listing the diagnoses from most likely to least likely, featuring those that are common, dangerous or treatable. This process references Bayes’ theorem. The Reverend Thomas Bayes (1701 to 1761) described a method by which a subjective belief could be modified by additional evidence. This conceptual approach was later extended to a “Frequentist Model” where the prior probability of an event or diagnosis is modified by subsequent information to get a posterior probability. This concept is useful for statistical analysis in many fields. It is used in medicine to examine the performance characteristics of a laboratory test for the population tested. A common example of Bayesian analysis is that a positive malaria smear is likely to be a false positive in an individual who has never visited a malarious zone. A high CT coronary calcium score is likely to indicate obstructive disease (true positive) in a patient with multiple coronary risk factors. We are constantly using Bayes’ theorem unconsciously when we reorder our differential diagnosis with each answer from the history and with each positive or negative finding in the physical exam and when evaluating the results of lab, imaging or response to therapy. We try to choose the test with the highest predictive value given our prior subjective beliefs. Consciously and methodically working through the diagnostic process, diagnoses rise and fall like presidential candidates in opinion polls before the nominating convention. In examining our differential, keep in mind Ockham’s razor. When confronted by a complex collection of clinical findings, a single diagnosis accounting for all of them is more likely to be correct. If multiple diagnoses are needed, we may be more facile than insightful. This principle should not be taken too far. Complex elderly usually have multiple interacting problems that should not lead to conceptual oversimplification. Ockham’s razor was repealed in pre-HAART AIDS when multiple diagnoses were common. Einstein said, “A theory should be as simple as possible, but not simpler.” If the clinical history is the gold mine of diagnosis, how can we find the mother lode? I must approach the patient with a blank but alert consciousness ready for any story and any diagnosis. I pretend that I am the patient. As the process unfolds, diagnoses emerge unbidden from my unconscious mind to dance and spar for top billing. They are consciously re-ordered with each symptom, physical finding, test result or feature of disease progression in a stew of Bayesian analysis. Only one diagnosis can be first and one can be second. Only those diagnoses with which I am established can come forth. I have heard of many diagnoses in my career but am established in those seen or carefully studied and brought back for regular exercise. This intuitive process is the way that an idea can come from nothing. I treat my intuition as I treat my teenage children. Trust but verify. Seek the confirmatory element of history, physical exam or testing. Use Bayesian statistics and Ockham’s razor and evidence-based medicine to keep intuition from flying off to fantasy. One of the most underutilized elements of clinical reasoning is a careful analysis of the tempo of an illness. Patients are usually vague: “a while ago.” Memory is rarely rooted in time unless referenced. It is time-consuming and frustrating to construct a reliable time course of an illness but most unfold with the temporal symptoms and signs that strongly re-order the differential. Adhering to this diagnostic process with integrity, a meaningful differential diagnosis will emerge. Strive to become established in a rich array of diagnoses. Know that we are doing God’s work and will not be lead astray. Own the process, not the result. Do not let anxiety over the result impair the search. This is the management of uncertainty. Reduce uncertainty with a steady hand avoiding daily management changes even as the clinical picture is constantly changing. As long as we cannot simultaneously measure the velocity and position of the electron, there will always be uncertainty. Fidelity to the process offers confidence when trekking the wilderness. Soon you will have more experience being confused than anyone. You can brag about it too. Source
The history thing reminds me of a patient I asked about her past medical history , she denied any, but she had DM, HCV and high blood pressure
