One of the major forms of waste in many health systems is excess and inappropriate diagnostic testing. Although, overall estimates for waste in health care are 30% of all spending, when it comes to diagnostic testing, I suspect the percentage of waste is far higher-- possibly 50 to 60% range. Why do physicians, nurse practitioners, and physician's assistants waste so much money on testing? Many inexperienced providers utilize what I call a "shotgun approach" to ordering tests. If they think about it, they order it. Too often after they develop bad habits, they continue this poor approach-- an approach that is expensive and can lead to confusion. Studies reveal large differences between the novice and the expert diagnosticians when it comes to the number of tests ordered and the accuracy of their tests. On the left side of this graph you see that the novices order large numbers of tests, the black line, and have a lower level of diagnostic accuracy, the orange line, while experts, shown in the right side of the graph order far fewer tests and have a much higher diagnostic accuracy. This is because many novices or beginners use a hypothesis based approach. They take each single symptom and create a hypothesis with regards to the cause. This approach is tedious, time consuming, and requires excessive testing to exclude each possibility. This approach is rarely effective and increases the danger of misdiagnosis. Let's look at how the beginner approaches a typical case. A 34 year old female with no past medical history presents with pain on urination for four days. She has been experiencing left sided flank pain and chills for one day accompanied by nausea and vomiting. On physical exam her temperature is 40°C, blood pressure is 120/80, pulse is 110, respiratory rate 20. She is ill appearing and sweating and has left flank tenderness, and suprapubic tenderness. How did the beginner approach this case? Too often the novice breaks down the individual problems and makes a lengthly problem list including pain on urination, flank pain, nausea and vomiting, chills, fever, and suprapubic tenderness. The novice orders a set of tests for each problem in isolation. For pain on urination, he or she might order Urinalysis and Culture, GC and Chlamydia PCR, Pelvic Exam, Vaginal Wet Prep, Gram Stain and PAP smear. For flank pain, an abdominal CT scan might be ordered, amylase, a liver function test and for nausea and vomiting, might consider a GI consult, endoscopy. And for fever, might order a influenza PCR, CMV PCR, Blood Cultures, Cardiac Echo, PPD, VDRL, HIV, and Histoplasmosis Antigen. And for suprapubic pain, might also order a CT scan, Amylase, liver function tests and possibly Colonoscopy. This in an extreme example but based on my personal observations of beginners and even more seasoned diagnosticians, this shotgun approach is all too common. Experts use a very different approach called Chunking. Chunks are pieces of information that are bound together through meaning. Barbara Oakley, in her outstanding Coursera course "Learning to Learn", explores this approach in more detail. Medical experts chunk by uniting bits of information using illness scripts. They are able to compare and contrast multiple illness scripts collected through the years of clinical practice and by using these comparisons are able to accurately predict the pretest probability of each disease they are considering. The ability to develop this type of pattern recognition and pretest prediction of probably takes time, roughly 10 years. Experts start with the history and physical exam and create a patient illness script focusing on four elements. Epidemiology: What are the demographics, risk factors and exposure history. Duration: How quickly did the illness come on? Hyper acute, acute, subacute or chronic. And is this symptom constant or episodic? Third, Classic Symptoms and Signs: they focus on symptoms and signs that a patient must have for a specific disease, as well as rejecting features, that if present virtually exclude the disease they are considering. Over time, experts see more and more cases and develop an increasingly detailed and sophisticated catalog of symptoms and signs associated with each disease. Finally, they may also take into account the past medical history. This may be important but in many cases it may not be. These approaches are described in greater detail in another excellent course by Catherine Lucy, Medical Decision Making, also offered by Coursera. The expert then creates a tiered diagnosis ordered by probability. Tier 1: very likely, greater than 90 percent to likely 67 to 90 percent. This tier contains all of the key elements of the illness script and has no rejecting factors. Tier two: uncertain, 34 to 60 percent, 66 percent probability. The patient has some of the key elements of the illness script and no rejecting features. And finally, tier 3 includes unlikely diagnosis, 10 to 33 percent probability, and very unlikely, less than 10 percent the diagnostic possibility. Unlikely possibilities have only one key element of illness script and very unlikely possibilities also have a rejecting feature. This approach is nicely outlined by Dr Lucey in her paper in evidence based medicine referenced at the bottom of this slide. Now, let's return to the case we discussed earlier. Unlike the novice who evaluates each symptom and sign individually, the expert creates a summary statement that brings together the chunks, the key elements of the patient's presentation: a 34 year old female with acute onset of fever, chills, nausea and vomiting accompanied by dysuria, flank pain and tenderness as well as super pubic tenderness. The expert then tiers the possible causes by relative probability. Tier one would be pyelonephritis with an estimated probability of 95 percent. Cystitis is less likely, likely, being placed in Tier 2, having an estimated pretest probability of 30 percent because patients with cystitis rarely have flank pain, fever or chills. And Tier three, cholecystitis or pancreatitis, neither disease is associated with dysuria, making them unlikely at less than 10 percent. Because the diagnosis is most likely to be pyelonephritis, rather than using a shotgun approach, the expert is like a sharp shooter and orders focused series of tests to assess the etiology and severity of her infection. The expert would order a peripheral white cell count with differential which would reveal a white count of 20,000 with 80 percent PMNs, 15 percent bands and 5 percent lymphs. A urinalysis would reveal a 250 white cells per low power field, a urine culture which would grow E. coli, sensitive to S cefazolin and a renal ultrasound showing no structural abnormalities or hydronephrosis, and finally blood cultures, two of two growing E.coli. To summarize, the expert uses an illness script and applies pattern recognition to determine the highest probability disease and confirms this diagnosis by judicious diagnostic testing. While the novice tests for each individual symptom resulting in excessive testing, increasing the danger of false positive tests, increasing cost and possibly delaying diagnosis. This approach not only is costly, but can increase the risk of patient harm. Our cognitive approach, as compared to a procedural approach, saves money, saves time, reduces harm, and improves patient outcomes. More is not always better. Isn't it time we focused on thinking rather than doing? Thank you.
