How much can computer science help with diagnosis?

[randombetch]

---

Members do not see this ad.

I have some computer science experience (~three years of computational research with C, two years of AP Computer Science in high school, couple of college courses with the Comp Sci department), and I feel like medicine doesn't tap into the field of computer science as much as possible. For instance, let's say a patient comes in complaining of neck pain. Why can't a computer program simply ask the patient a series of questions (ie: the same questions a good doctor would ask, with the series of questions being tailored by the answers to the preceding questions) similar to Microsoft's Support Center, and then come up with a list of possible diagnoses (possibly even 20-30 diseases long if the patients' answers were a bit vague or contradictory) and a list of tests that could help narrow the list down more? From there, the doctor can start narrowing the choices while considering all of the possibilities. The list of diagnoses could even have a rough probability estimate next to each disease.

I'm probably simplifying things a bit, but it doesn't seem impossible to me. The computer program should be able to figure out (using more questions) which symptoms are relevant or irrelevant, the same way a doctor would.

 

[RogueUnicorn]

How much time would that really save, though, if the doctor has to look through the Qs, the As, and the list of diagnoses anyway? they already make a habit of asking questions when they have been asked before with the answers written in the chart. essentially this computer you refer to is what a physician's job is.

 

[jbz24]

For simple things a computer program can probably figure out the most likely diagnoses, but it won't really save any time and it would be very limited to simple things. How would a computer program help diagnose a patient who comes in unconscious? or demented? what if the pertinent finding is on physical exam and not from the history?

I think computer programs would be great for certain reminders (i.e. the patient that just came in is 50 years old - did you ask him/her about a colonoscopy?) but for actual diagnosis it won't get very far, unless you can actually build a robot that not only asks questions but can do a physical exam and analyze laboratory/imaging data.

 

[muhali3]

Although this is hypothetically possible, the computer system would depend upon the patient's own self-assessment of their condition, and they might report symptoms incorrectly. This would lead to a list of possible diagnoses that may not include what they actually have, and would actually make doctors do more work than before. Patients would also have to learn how to use the computer system, computers might freeze, etc. Imo, it would just make things much slower and inefficient.

 

[randombetch]

Wow, so stuff like this does exist! Thanks for showing me.

All the points raised seem very valid (especially how the software would rely on the patient's self-assessment).

 

[illegallysmooth]

Also:

-Signs vs. symptoms

-Acute vs. chronic effects

-baseline vs. status at presentation

-subjective medical history


Yesterday in the ER we had a 28yo female present with chest pain and hx of "EF of 40, osteoporosis, B12 deficiency, etc." Turns out she has a "very mild" B12 deficiency, a stress test showed a normal EF (after an echo showed the low EF), bone scan was normal. She was having an anxiety reaction and severely exaggerating her medical history.

I can only imagine what diagnoses a computer program would have spit out on her, LOL.