We as physicians and medical providers have a love-hate relationship with the internet. I do really like my patients being more informed and being able to find information regarding their diagnoses. I do however, struggle with my patients coming in and telling me they’ve been “diagnosed” with this or that ailment, which, upon closer examination, actually is a self-diagnosis, perhaps with some assistance from well-meaning friends, TV commercials or Dr. Google.
Sometimes, it may be the correct diagnosis. But sometimes, an assumption about a diagnosis may lead us as clinicians down an incorrect path as we fail to consider the actual symptoms of you – the flesh and blood patient in front of us. We can get distracted by the facts as presented in what we learn in medical school to call the “past medical history.” I’ve always worried about this as a clinician, but now a study is here to tell me whether I should continue to worry or not.
Some people smarter than I reviewed 23 different online symptom checkers diagnoses for 45 standardized written patient scenarios provided to test medical students. The symptom checkers provided the correct diagnosis in 34% of the cases (interestingly, only 24% for emergent diagnoses only). They faired better when utilizing a top 20 possible diagnosis list, scoring the correct diagnosis within the list 58% of the time.
When the symptom checker’s purpose was to triage the symptoms (advise the patient on whether they needed emergent, non-emergent or self care), they actually did a bit better. When the situation was emergent, the symptom checker was correct with triage advice (remember – not necessarily diagnosis) 80% of the time. But the accuracy dropped off with the non-emergent (55% accuracy) and self-care (33% accuracy) scenarios and some of the programs actually always advised emergent care.
Currently, medical providers’ diagnostic accuracy rate appears to be 85-90% (hey, we’re all human!). An interesting comparison would have been to have some real life medical providers review the symptoms to see how we held up. I like to think I’m better than a computer, but can I objectively prove that I can make a diagnosis equally as well?
Also, would we humans be any better without being able to lay hands on our patient and do an actual physical exam? We humans have information the computers do not – medical history, medications, qualifying factors, and the very important physical exam. This makes a difference. As does our role in medicine – I suspect an ER physician vs a primary care physician might have similar capabilities regarding triage (I hope I know when I situation is emergent!), but different results regarding correct diagnosis depending on the nature of the ailment. We have different roles in the medical field and therefore different skill sets regarding diagnosis of common day to day maladies. Please don’t expect the ER doctor to use their brain space to know how to treat run of the mill urinary incontinence, but I would absolutely anticipate he could quickly distinguish between non urgent causes and life threatening ones (spinal compression, severe infection, seizure, etc). And if you have an acute heart attack or stroke, you definitely want their skills at hand as they will be the difference between whether you hug your kids at their next birthday.
All in all, the online symptom checkers appear to be able to give you a top 20 list of possible diagnoses when you input your symptomology. If you are willing to go through the treatment for each of those possible diagnoses in succession, not worry about missed diagnosis and want to save the copay, then the symptom checkers are worthwhile.
For me, I prefer to utilize my time with patients making sure we don’t miss something severe, then focus on the most likely cause of their symptoms. After a thorough history and physical exam, as needed. Take that, Computer!
Study Source: “Evaluation of symptom checkers for self diagnosis and triage: audit study,” BMJ, June 2015