2. Health and misdiagnoses
The prevalence of misdiagnoses is estimated to be up to 15% in most
areas of medicine
Most cases are due to testing (44%) or clinician assessment errors (32%).
28% of the misdiagnoses are major (i.e resulting in death, permanent
disability, or near life-threatening event) and 41% moderate (i.e
resulting in short-term morbidity, increased length of stay, higher level
of care or invasive procedure). *1
3. Status
The misdiagnosis problem is often considered to be an
individual clinician’s problem. Yet the facts and figures
presented earlier rather suggest misdiagnoses to be more
of a systemic problem.
4. What is the actual proposal?
Always the same:
Building a shared medical data repository.
(Hadoop cluster, for instance). Including
images data, records data (text mining).
Experts systems to recommend possible
diagnosis. (Bayesian thery, Dempster-Shafer
evidence theory, fuzzy logic, etc.)
Images? EEG, CT. (density clusters?)
5. So what’s new in the market to help with this?
IBM Watson? cognitive computing. Processes Natural Language.
After the physician enters the symptoms and sings the systems looks for related
information and suggests different diagnosis with a probability.
7. What do we have?
https://sisa.msal.gov.ar/sisa/
8. What are the costs of misdiagnosis?
Health system costs: people tend to retry with other physicians. Every
physician gives different drugs, that have a discount supported by their
social insurance.
Social costs: the opportunity costs of the patients (a simple measure would
be the time expenditure looking for the “right” diagnosis multiply by their
salary). Death of worsen diseases.
Negative externalities: all patients receive worse attention because every
resource comes from the same “bag”.
9. Common and yet sad dialog
Patient: -“I’ve been to almost 10 specialist and they all tell me different
things but I only have one body.”
Physician: -“You should choose the one that inspires you more
confidence.”
10. Are we missing something?
Until now we are thinking that not all physicians get to the same
diagnosis. We are also considering other’s physician studies to
enrich their current diagnosis.
Some clinics do have a unified record.
But we don’t have Watson now and not even a unified record
system.
What we do have is the patient’s appointments. All the
appointments that he booked and the doctors he consulted.
Also, if the patient has a medical secure, we also have the drugs he
bought.
11. The proposal: “path data”
Considering the minimum of the data we do have, as I said,
appointments schedules, patients and doctors, we could analyze
the patient’s path. Why a patient go to see many doctors of the
same specialty? What do they give him?
If we could add the symptoms and signs we could analyze the last
doctor he saw for that particular “illness” (considering that almost all
symptoms and signs are the same) There might be some doctors
that know better his job in some arenas.
12. But …what interactions we have in the web?
Knowledge has been democratized. Many places like
http://stackexchange.com/ allow people to ask questions and support a
platform where others can answer. In this case anyone can give an answer,
but you do have a reputation you build with every answer you gave and
the questions you asked before.
There are many medical fora for different specialties. Most of them are built
by doctors who own a clinic and want to attract patients. So the answers
are only given by them.
http://www.cechin.com.ar/objetivos-del-foro/
There are also popular forums where people who have doubts can ask
questions to be answered for others that no necessary are suitable for that.
http://grupos.emagister.com/foro_otorrinolaringologia/6792
13. The proposal- share data that is already in
the system
Let’s suppose the case when the patient doesn’t get better. A better way to
save time, money and suffering is to create a database with all the “difficult
cases”. Doesn’t need to be terminal diseases. If the patient continues seeking a
solution is because his illness is unbearable for him and the consequences of
this are still all the 3 mentioned above.
There should be uploaded all the tests performed to the patient and the
description (which has already been written in a platform). Everything must be
available for the “responders”.
Each doctor, who wants to participate, must create a profile with credentials of
his specialty. It’s important that he remains anonymous (doesn’t matter the
time he has been working, his name, his status. If we knew that, it might tend to
influence other’s answers and to contaminate the service).
14. The cases
Could be originated by the clinical center. It will get all the
information from there. Including exams and images. Remember:
most of this data is already in the system.
Could be originated by the patients. (they should be validated in
some way).
Everything must be shown anonymous.
15. What are the incentives?
For patients: it’s clear, they will have more opinions and might reach
faster the solution, avoiding pain, waste of time and money.
For doctors: They might learn from others. Not enough? Ok. The
profile is anonymous but still you could have another profile (secretly
linked by the system to the one used to answer) where you can see
only the total amount of credits achieved.
For social security: save costs, better service, get to know better the
doctors that are working for them.
For regular people: learn in a safe environment (if it is open to the
community).