See the same doctor when the diagnosis isn’t obvious

I’m thankful that I don’t have to research symptoms on the internet, diagnose myself, and come up with my own treatment.  All I have to do is go to my doctor and describe my symptoms.  My doctor can listen to my symptoms, match them up to what he knows, ask questions to pinpoint what might be going on, order appropriate tests, and recommend a treatment plan.

Odds are, when people get sick, that their very common symptoms have a very common cause.  Occasionally problems arise because medical schools teach, “When you hear hoofbeats, think horses, not zebras,” and sometimes doctors interpret this as “zebras don’t exist.”

But zebras do exist.  Good doctors don’t entirely eliminate all thought of zebras when they’re examining patients.  Good doctors recognize that statistics only apply to groups, not individuals (sorry, that’s my math background showing).  The chances that an individual sitting in the doctor’s exam room has a rare condition are 50-50.  Either the person has something rare, or not.  Those are the only two options, and they’re equally likely when you’re dealing with an individual.

The doctor thinks, “Given the symptoms and test results, this could be A, B, C, or D.”  That list of possibilities, arranged from most-likely to least-likely, is called the differential diagnosis.  The ddx is not based on preponderance of all diseases in the world.  The ddx is supposed to be based on how closely symptoms match what is known of various diseases.  For instance, if I go to the doctor because I’ve been vomiting for the past five days and have lost fifteen pounds, the differential diagnosis is not going to include “common cold”  even though colds are indeed quite common.  The symptoms have to match up, and good doctors will consider how closely symptoms match a disease, not just how common a disease is.  Rare diseases belong in the ddx if they could logically be reached based on the patient’s symptoms.

What’s a patient to do, then, when symptoms don’t improve even after we’ve followed the doctor’s recommended treatment plan?

See the same doctor when the diagnosis isnt obvious

This is where I think patients sometimes make a mistake.  Don’t ignore the symptoms.  Don’t look for a new doctor.  If it’s still a problem, go back to see the doctor again.  See the same doctor – the person who’s already given some thought to what might be wrong. Simply say, “Here’s what I tried, but it didn’t help.  Do you have any other ideas on what I can do?” This can get the doctor to thinking differently about your symptoms/diagnosis.

According to Dr. Jerome Groopman, in How Doctors Think, a good question to ask is, “What else could this be?” (Dr. Groopman is only responsible for that question, not the rest of my strange thoughts on this topic.)

If the treatment for A doesn’t work, sometimes people think, “Well, the first doctor didn’t help, so I’ll find a different doctor.”  A better solution would be to go back to the same doctor so that options B, C, D, etc. can be considered.

See the same doctor when the diagnosis isnt obvious

If the patient goes elsewhere, the doctor has no way of knowing that he was wrong.  In the above diagram, his thinking stops in the red box.

Some things are tricky to figure out.  Sometimes doctors miss things and need to take a second (or third) look.  Either the doctor left something off the differential diagnosis, or the patient didn’t think to provide a key piece of information, or maybe the problem is outside the doctor’s area of expertise.

We need purple power.  Don’t let a false “all is well” stand.  Return for a follow-up appointment so that the doctor moves into the purple boxes of his decision tree.  The incorrect diagnosis needs to be discarded.  Don’t give up.  We can ask what else could this be?

“WarmSocks” blogs at ∞ itis.

Submit a guest post and be heard on social media’s leading physician voice.

Comments are moderated before they are published. Please read the comment policy.

  • jsmith

    A generally agree, but with a slight caveat: If you go back to the same doctor three times and he or she still doesn’t know what is up, I think another opinion is indicated. This does not have to mean a formal consultation with a specialist outside the building. It could mean having the doc discuss it with a colleague or a couple colleagues or even (in a group practice) having the second doc come in for a quick look at you. We do this a fair amount in my group practice. If the second doc is also stumped, I will sometimes usually get on the phone to discuss the case with a specialist at the Univ. of Washington. This is a free service available to physicians within our state–Medcon it’s called.

    • http://warmsocks.wordpress.com/ WarmSocks

      I agree with your caveat. It didn’t occur to me that there are doctors who don’t ask for another opinion if they’re stumped or not sure. Mine does.

      Thank you for commenting.

  • Adam Rothschild, M.D., M.A.

    While I agree in general with what both you and jsmith (above) say, I must point out that you make a statement that is patently incorrect: “The chances that an individual sitting in the doctor’s exam room has a rare condition are 50-50. Either the person has something rare, or not. Those are the only two options, and they’re equally likely when you’re dealing with an individual.” Yes, the patient either has the condition or he doesn’t, but the probability that the patient has a given condition is NOT equal to the probability that he doesn’t have the condition simply because those are the only two choices. Rather, the probability that the patient has a given condition depends on the prevalence of that condition and the conditional probabilities of the patient having the condition given the findings. You might consider refreshing your math background by reviewing Bayes’ Theorem.

    • http://warmsocks.wordpress.com/ WarmSocks

      Thank you for the comments. I glossed over the stats, not realizing anyone might be interested.

      My stats profs were always adamant that statistics apply to groups, not to individuals. Another way to look at is is to picture statistics as a way to view data; stats apply to sets of data, but not to a single person.

      My 50/50 reference was suggested by Bayes’ ideas. He said that prior probability distribution looks at the state of things before experiments are performed, and that before experiments are performed, experiment populations should be assumed to be equiprobable – thus my 50/50 claim.

      Bayes’ theorem is about conditional probabilities, and as you point out, is often applied in the medical field. It’s how the probability of a false-positive is determined.

      Bayes theorem can be used to calculate the probability that a disease exists in a specific patient, given the condition that the results of a diagnostic procedure indicate that the disease is there. It’s not a stand-alone probability. Testing must be done first.

      The theorem doesn’t have anything to do with a patient sitting in an exam room who hasn’t had any testing done. Once testing is done, then you can look at the probability that the patient has the disease that was tested for. I imagine that a good history & exam could be used to rule out some conditions, but it’s not valid to rule out a disease just because it’s rare. :)

      • Adam Rothschild, M.D., M.A.

        WarmSocks, unfortunately your understanding of the application of Bayes’ Theorem in a situation that does not require a formal diagnostic test is incorrect. In medical terms, Bayes’ Theorem is defined as follows: the probability of DISEASE given FINDING equals the probability of FINDING given DISEASE times the probability of DISEASE divided by the probability of FINDING. A FINDING does not have to be a formal diagnostic test such as an MRI or lab test; it can simply be a sign or symptom. Furthermore, the mere prevalence of a given disease in a population has a major effect on the probability of a patient having that disease.

        A simple example: A healthy 10 year-old patient with no complaints comes to see me in the office. What is the probability that this patient has strep throat? Practically nil. Now let’s say that the same, otherwise healthy 10 year-old comes to see me in the office FOR a sore throat. Given that all I know is that the patient is 10 years old and has a sore throat, what’s the probability that this patient has strep? A bit higher than the first case, maybe 5%, although with the limited information that I have, it’s much more likely to be viral because viral sore throats are significantly more common than strep (i.e., they have a higher prevalence). Now I ask some more questions and examine this patient and find out that she has a fever of 38.5 C, doesn’t have a cough, has tender anterior cervical lymph nodes, and has pus all over her tonsils. Given all of this additional information (each one of which can be considered a “test” for the purpose of Bayes’ Theorem), the probability that this patient NOW has strep is approaching 50%. At this point, I’m going to prescribe this patient antibiotics, and given this situation, many doctors wouldn’t even bother doing a throat culture to confirm. Might the patient have rare-but-serious Kawaski disease? Sure, but it’s still much more probable that that patient has strep (or a virus) because the prevalence of Kawasaki disease is so low. Now if I treat this patient and the symptoms don’t resolve within a week or two, THEN it’s time to think about Kawasaki disease, which would be aided by returning to the initial doctor as you suggest.

        • jsmith

          Dr. R is correct on Bayes. Any chunk of relevant information–sign, symptom, demographic data, test result, etc. can be used in Bayesian updating. Bayesian updating is at the heart of the process of differential diagnosis, one of medicine’s most powerful tools.

  • Chris

    To simply explain Dr Rothschild’s post- if you have ankle pain after you twisted it, it is most likely a sprain. It is possibly a tear, but the treatment is the same initially. And 90% or more of the time, that holds. SO the risk is only 10% (I’m making up the percentages here on this particular diagnosis.) that you have something else. The risk is you don’t get better is not 50-50 that it’s something else; it might be that you just had a very bad sprain.The risk has to do with how often the disease occurs and your specific risk factors.
    However, if you have other complicating illnesses, things change (for example for the diagnosis of the sprained ankle, if you have osteoporosis , or have been on chronic steroids, this can affect you bones and tendons, so the risk changes, Doctors are flipping through the various risks in their heads all the time-we don’t want to over treat and test, (and you don’t want the bill ) and we don’t want to under treat. We don’t know that you didn’t get better unless you follow up.
    Returning , calmly, for follow-up is important. Trying to work with that doctor, assuming you have trust in the relationship, is a good idea, and getting a referral through him/her is a good idea. Getting a diagnosis from Dr Oz on Oprah is a bad idea.( I’m not knocking him, but he knows nothing about the specifics of your illness,He’s talking from a television screen), Getting a diagnosis from your repairman is a bad idea. (one of my patient with Parkinson’s got brought to the chiropractor’s by his repairman ) I found out about it afterward. What he needed was a spine surgeon.( He was already getting epidurals for spine issues but fell, and I wasn’t informed,What do i know? I don’t fix washing machines,,,,)

  • drpak

    “There’s an app for that”.

    There’s a new app out called Rx Bayes that allows a more intuitive understanding on Bayes’ theorem. It’s in the itunes app store.