Technology based medical diagnosis
I have been recently exploring health sector. Unlike education, where everyone has seen schools from inside and have experienced, medical sector is a black box.
The excellent TV Series, House M.D, along with conversations with few doctors helped me realize two key aspects of medical diagnosis
- The key job of doctor is to diagnose the disease with minimum symptoms.
- Key constraint in step 1 is the cost of medical tests.
Technology can help enhance doctors’ performance in both the steps.
Let me explain what the two aspects means, before discussing the role of technology.
1) Job of a doctor: Suppose a patient goes with headache and stomach ache. What does it mean for the doctor?
Headache can be due to numerous reasons and so can be stomach ache. Further, headache and stomach ache may not just be stand alone aspects, they can be symptoms of an underlying major disease.
The doctor’s job now is to use these signs to narrow down the possible diseases that explain these symptoms. This is essentially a mental exercise of matching these signs with diseases that the doctor is aware of! It’s called “differential diagnosis”
Currently, doctors do it using their memory. The problems with this are
a) doctors may not be aware of all possible diseases
b) even if they are aware, they can’t do the huge matching exercise
c) even if the doctors are aware, they might subconsciously discard some diseases because the probability of their occurrence is too low!
As one can see, this is an extremely inefficient process prone to high errors. It’s too much to expect doctors to remember many possible diseases, symptoms and match them instantly.
The key to improve it essentially involves two parameters — doctor’s memory and capacity of match-making process.
We may note that these exact things — memory and processing — are the two aspects that the computers are really good at! Hence, technology would be of a great help here.
A technology tool here would be one where the doctor inputs the symptoms and the tool does the matching and throws out the possible diseases!
This can improve doctors’ efficiency multiple times. It is primarily an IT challenge.
2. Cost of medical tests: As discussed above, a doctor has to make a decision looking at the signs/symptoms. Doctors use medical tests (blood tests, scans etc.) to get additional information so that they can narrow down the domain of diseases.
Ideally, if one were to be 100% precise in diagnosis, one should test the patient for all diseases across the spectrum. It means several hundred types of blood tests and whole body scans.
It isn’t currently feasible because
a) one can’t draw huge sample enough to do hundreds of tests.
b) the costs will hit the roof.
c) time taken for the results to come is long.
Hence, doctors narrow down the possible diseases even before order medical tests and order only particular blood tests. This is prone to error as the doctor might miss out something pre-medical tests.
The two binding constraints, thus, here, are
- Availability of blood for testing
- Cost of diagnosis
Technology can be of great help here.
- A machine can do precise division of blood, even into hundreds of equal parts and be efficient in doing the titrations, thus reducing the amount of blood required.
- Advances in medical technology can dramatically drive down the costs of medical tests.
As opposed to the technology solution for differential diagnosis, which is highly IT based, the technology solution to driving down medical costs is improvements in robotics and biochemistry.
A futuristic picture of tech-enabled medical diagnosis
From the above discussion, we may infer that, two key technologies can drive down the costs of diagnosis significantly and increase the accuracy of diagnosis. They are
- Blood testing devices that use minimal blood to do wide range of blood tests, in little time, at a cheaper cost; and scanning devices that can do instant scans, like security check, at minimal cost.
- Technology tool for differential diagnosis — a tool that narrows down the diseases based on the symptoms and results from diagnosis.
In an ideal world, where we have these two technology tools, the diagnostic process will be simple and accurate: a patient walks into a clinic, tells the symptoms and gives blood for testing, computer matches the results from diagnostic tests and symptoms to the possible diseases, and gives the summary to the doctor. The doctor just looks at the summary data and exercises human judgment to give the final clearance!
Such process would eliminate the judgment errors in diagnosis and dramatically improve the precision and quality of health care.
To give an idea of the extent of efficiency improvements — Imagine an engineer doing calculations without calculators OR a programmer compiling computer programs on punching machines. Before modern day compilers, programmers had to run the algorithm multiple times in their head before taking it to punching machine because they can’t afford to lose the time.
It was extremely inefficient, time taking and much energy and mental space is doing the mental calculations. Processing time and cost of processors were the two key constraints.
Calculators and modern day compilers have drastically brought down these two costs. They made processing cheaper and faster thereby easing engineers’ mind space, enabling them to take quantum leaps. Now, instead of using mind space to do the mundane calculations, programmers use it to do creative aspects. It increased the speed of iterative process, thus making it possible to build complex softwares.
Well built differential diagnostic tool, blood testing and scanning devices will have a similar effect on the efficiency of doctors. Such technology tools will also make efficient tele-medicine a reality!
The ideal technologies that can surpass human doctors may not be available tomorrow but we can at least start working on technologies that can at least make marginal improvements to accuracy of doctors’ diagnosis, by using the tech more as supplementary tool.
Of the two, the matching tool, and innovative medical testing devices, the former is easier to do because it is more of an IT task. The latter might require some time as it requires innovations in biochemistry and robotics. So, building a good differential diagnosis tool that shortlists diseases both based on symptoms, and medical tests, is a good place to start. I see that some companies have started doing it (example — though this has only symptoms as shortlisting criteria).
It might be a good idea to develop a product for Indian context that shortlists diseases initially based on symptoms, suggests possible medical tests, and after the results, shortlists diseases based on both symptoms and medical tests.
Finally, precise diagnostic tools will only increase the accuracy of diagnosis and precision of treatment. To actually translate it into patients’ wellbeing, the treatment costs should also come down, which is a challenge of its own.