AI for Health care
or something simpler might do too
I recently came across an article in a newspaper which said that doctors in India spend less than 2 minutes per patient. Having had to visit the hospital a couple of times myself, over the past year, I realize that this is not only true for public hospitals but also for relatively reputed private hospitals as well.
In theory, these hospitals have an appointment time. Appointments are given in intervals of 10 to 15 minutes. This does not leave a lot buffer for delays on the part of the patient or the doctor. In practice, you therefore end up waiting for a consultation well past your allocated time — sometimes well over an hour. Finally, it is almost your turn. A nurse will call you and take your vitals height, weight and BP and you are scurried into the doctors chamber just as the previous patient is being scurried out.
The doctor then hears you out for about 30 seconds and sometimes does the obligatory check with a stethoscope for another few seconds before writing out a battery of tests. You are then asked to do the tests and come back with the reports based on which the doctor decides on the course of the treatment. They tend to disregard most of the things you say and depend almost entirely on the reports.
The point is not to rant. Though, not being an AI entity (or a Vulcan for that matter), it is sometimes unavoidable and I’ll mostly edit out later.
An important part of meeting a doctor is the “healing touch”, a human perspective. Doctors today are too overwhelmed by the flood of patients to really give each patient a lot of time. The hospital assign a different doctor each time you visit unless you ask for a doctor by name. So, the patient-doctor relationship is contained to the current episode. There is no deeper, longer bond for building trust or understanding. No knowledge of family history or circumstances. No time to recognize idiosyncrasies of either the patient or the doctor.
In the absence of all the human parameters, all that is left is the data and the analysis of said data. Hence the battery of lab-tests.
The other important aspect is that lab-tests are getting cheaper, better and faster. Though hospitals still tend to be slow, expensive and bureaucratic, internet based services for health tests are available today. These services collect samples from the home to give reports faster, directly to the mailbox at a fraction of the cost.
It is therefore theoretically possible to run a larger set of screening tests on every incoming patient before they are pushed into the consultation process. Running the tests in bulk for every incoming patient would probably bring down the overall cost of the tests. This would also save time by reducing the consultation / tests / consultation / prescription cycle to test / consultation + prescription for most cases.
And then, there is automation … or AI if you will, though this can be introduced in stages.
Here is a gist of the proposal:
- Develop a questionnaire or an adaptive test (that can be administered by a computer system) to all patients for initial automatic diagnosis
- Use the results of this questionnaire to evaluate (automatically) the most likely causes (possibly a primary and secondary diagnosis)
- The system then automatically recommends all the lab-tests required to validate the diagnosis and primary and secondary diagnosis
- The reports from the lab-tests are fed back to the system to validate the initial primary and secondary diagnosis
- If the lab-test corroborate the initial diagnosis, the appropriate medicines can then be automatically recommended by the system
- If the lab tests do not corroborate the initial primary/secondary diagnosis or the system flags for a human intervention
For patients that leave the system at step 5, human doctors need not get involved. They could be met by nursing or other staff to explain the ailment, explain the medicines prescribed etc. Patients could still be charged for this service — for the automatic consultation as well as the lab-tests.
For patients that reach step 6, a qualified doctor would come in. Such a doctor would presumably have more time available since a lot of patients would have already been taken care of adequately by the system.
What I am saying is not new or revolutionary.
For me personally, what is new is the realization that there is an urgent need to have this today. Today — and not in some distant sci-fi future.
The world population has increased and is increasing. We have to take care of people that need health care. We have to take care of our doctors who are overwhelmed by the influx of patients.
Most importantly perhaps, we have to take care of our humanity.
I don’t know whether AI will eventually take over the world some day or just make it a better place. But today, from where I stand, it looks like it may just help us maintain our humanity a little while longer… if we use it right.
